o o t o. 1546 ART -DAYS GeophySical Interpretatiofis of the Libby Thrust Belt, ‘ Northwestern Montana - yi. U. s. DEPosnofij DEC 0 1 1997 US GEOLOGICAL SURVEY PROFEJSDVSIONAL PAPER 1546 The Library ~ UC Berseley Reéeivad on : 12-18-99 Geological Suryey firofessional paper Geophysical Interpretations of the Libby Thrust Belt, Northwestern Montana By M. Dean Kleinkopf With a section on Deep Folds and Faults Interpreted from Seismic Data By Jack E. Harrison And a section on Interpretation of Magnetotelluric Soundings By W.D. Stanley U.S. GEOLOGICAL SURVEY PROFESSIONAL PAPER 1546 UNITED STATES GOVERNMENT PRINTING OFFICE, WASHINGTON : 1997 U.S. DEPARTMENT OF THE INTERIOR BRUCE BABBITT, Secretary U.S. GEOLOGICAL SURVEY Gordon P. Eaton, Director For sale by U.S. Geological Survey, Information Services Box 25286, Federal Center Denver, CO 80225 Any use of trade, product, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government Library of Congress Cataloging-in-Publication Data Kleinkopf, M. Dean (Merlin Dean), 1926- Geophysical interpretations of the Libby thrust belt, northwestern Montana / by M. Dean Kleinkopf ; with a section on deep folds and faults interpreted from seismic data by Jack E. Harrison and a section on interpretation on magnetotelluric soundings by W.D. Stanley. p. - em. - (U.S. Geological Survey professional paper ; 1546) Includes bibliographical references. Supt. of Docs. no.: 1 19.16: 1546 1. Thrust faults (Geology)-Montana. 2. Geology, Stratigraphic-Proterozoic. 3. Geology, Structural-Montana. 4. Geophysical prospecting-Montana. I. Title. II. Title: Libby thrust belt, northwestern Montana. III. Series. QE606.5.U6K52 1995 551.8'7-dc20 94-1628 CIP 1, 2. 1, 2. CONTENTS ADSHTACE erea revere rere nevere revere rere errr rere reer reer errr errr errr ener en enne renner ener rere rere ener neenee 1 INI OGUCtON eevee eevee evere rear reer rre rere rere ene ener errr errr errr rere rere reer ener en ener renner ene 2 GEOLOGY ..... cc serre reer rere reer reaver errr errr neer errr errr rre renner nner errr err errr ner ren enne enne nn reer nen nn 2 AerOMAGNetic ANOMAIY cee errr reer errr errr rere rre renner nre neer e neer enne neenee 5 Gravity ANOMAIY DAA eevee rere errr errr reer errr rere rre errr rene nre rene enne n errr nene nner en 5 Geologic Interpretation of Aeromagnetic eee eres 6 Geologic Interpretation of Bouguer Gravity ANOM@IIG$.............c eee eee 8 Geophysical Interpretation of Deep-Seated GeoIOg1¢C FeatUIeS eee 9 Modeling Of GIaVity PFOfiI@S errr re rere errr rere reer reer nner neenee er neenee es 9 Deep Folds and Faults Interpreted from Seismic Data, by Jack E. Harrison .......... 13 Interpretation of Magnetotelluric Soundings, by W.D. Stanley ...... 13 rade ht I ete hs o ooo 19 enc e 19 PLATES [Plates are in pocket] Maps of the area of the Libby thrust belt, northwestern Montana, showing: 1. Total-intensity aeromagnetic anomalies, Bouguer gravity anomalies, and generalized geology. 2. Residual total-intensity aeromagnetic anomalies, complete Bouguer gravity anomalies, and structural features. FIGURES Maps showing: 1. - Generalized geology Of B@lt .e rere errr errr nere ener err rer rre rere neer reer renner ern reer een rrr n rennes 3 2. Structural features, lines of sections and profiles, and locations of magnetotelluric soundings, LiDDY thIUSt errr rere revere rere errr rere rrr errr rre erve errr revere reer renner errr neer never nner nner nnn nner nee nner nnn 4 Generalized lithologic and geophysical 10g$, Gibbs NO.1 eee errr errr reer neenee reer 10 Gravity block models of extended geologic cross sections: A. = cscs eer erea eer ever ever erea erea reer errr reer errr rere reer erie ere rere rere rr rere errr re rene n errr ener ene re rere rre enn en enne renee 12 §, 0 JJ .ll eraser ere er renee rrr errr nere ern n bener neer rene reer nner errr enn bere ener nnn 14 Diagram showing interpretation of seismic reflection profiles MT-I ANG ee eer revere rere renee renee es 16 Generalized resistivity log of the Gibb§ NO. 1 DOFEROIG cece errr errr errr re reer errr nnn errr reer reer re nr ener enne 17 Magnetotelluric and geologic section across the eastern part of the Libby thrust belt and the western part e 18 TABLE Lithologies and densities assumed for gravity models of extended geologic sections B-B' and J-J', Libby thrust belt, NOTtAWEStETN eee evere renner reer errr errr reer renee rere rere renner rr nner ee nnn rrr e 11 III GEOPHYSICAL INTERPRETATIONS OF THE LIBBY THRUST BELT, NORTHWESTERN MONTANA By M. Dean Kleinkopf ABSTRACT Interpretations of gravity and aeromagnetic anomaly data, as well as results from two seismic reflection profiles and five magnetotelluric soundings, were used to study buried structure and lithology of the Libby thrust belt of northwestern Montana. Gravity modeling, supplemented by structural data from the seismic reflection profiles and geo- logic contraints derived from projection of surface geology, measured sections, and lateral consistency of sills, leads to the interpretation that thrust slices of folded crystalline base- ment form the core of the Purcell anticlinorium and Sylva- nite anticline, which consist mainly of rocks of the Middle Proterozoic Belt Supergroup. Gravity anomaly data show marked correlation with major structure of the area. The Purcell anticlinorium exhibits positive anomalies in excess of 20 mGal and the Sylvanite anticline anomalies in excess of 10 mGal. In the northern part of the study area, a distinct northwest-trending high-gradient zone suggests that a buried crustal shear zone may be present just north of Libby. The Rainy Creek and Bobtail Creek stocks, which show distinct positive magnetic anomalies, lie along this trend, and their emplacement may have been influenced by the postulated zone of deformation. The most distinct magnetic anomalies in the principal study area are five positive anomalies associated with Creta- ceous or younger cupolas and stocks that either are exposed or are known from drillholes to be present in the near-sur- face. Amplitudes range from 100 to more than 3,000 nT. Short-wavelength anomalies are associated with outcrops of magnetite-bearing sedimentary rocks of the Ravalli Group of the Belt Supergroup. A strip of high magnetic gradient correlates with the Hope fault zone. Modeling of the mag- netic anomaly data was not done because of the absence of long-wavelength magnetic anomalies having likely sources in deeply buried Proterozoic crystalline basement. The mafic sills of dioritic to gabbroic composition show little or no magnetic response in outcrop, probably because the mag- netite content in the sills was reduced by chemical processes related to interaction of hydrothermal fluids with cooling magmas of the intrusions. Evidence from the magnetic and gravity anomaly data suggests that the Cabinet Mountains Wilderness is underlain by a major batholith of felsic composition that extends from north of the Dry Creek stock south almost to the Vermilion River stock. Along this trend are a number of small high-amplitude positive anomalies of 60-100 nT in areas of outcrops of the Wallace Formation of the Belt Supergroup and, in a few cases, the Ravalli Group. In the case of the Ravalli rocks, the highs probably are not related to outcrops of magnetite-bearing Ravalli group rocks that in other places cause linear magnetic highs along ridge tops; instead, because of their equidimensional shape on the map, they may indicate near-surface sources related to granitic intru- sions or cupolas of a larger batholith mass at a few kilome- ters depth beneath the Cabinet Mountains Wilderness. A basal surface of detachment is inferred from a series of seismic reflections. Depth to the basal surface is 9-18 km, and this surface dips about 15° in inferred crystalline base- ment rocks. Strong reflections at and below 4 seconds are attributed to stratification in the basement such as layered gneiss or mylonite along the detachment zone for the older tectonic folding. No seismic reflection is present for the Belt-crystalline basement contact. Just below the Pinkham thrust fault, strong reflections attributed to stratification in crystalline basement rocks image relief in the basement on the western flank of the Purcell anticline. Interpretation of the magnetotelluric data suggests a thick conductive section that is incompatible with a crystalline basement composed of granitic gneiss, mafic intrusive rocks, and high-grade metasedimentary rocks con- sidered typical for the region. Thus, there are major differ- ences between the deep crustal model developed from the magnetotelluric data and the model obtained from the gravity and seismic data. This heterogeneous type of crust is generally not low in resistivity unless partial melting has occurred. Low resistivities may be possible in granitic rocks at depths of 10-15 km and temperatures of 500°C -600°C if 1-3 percent free water is present; however, there is no evidence in the study area for these levels of temperature. A thermally related basement conductive zone could possibly be present now at depth where a low-resistivity (3-5 1 2 GEOPHYSICAL INTERPRETATIONS, LIBBY THRUST BELT, MONTANA ohm-m) zone forms the bottom layer in the western part of the magnetotelluric profile. This zone corresponds approxi- mately with the interpreted basal surface of detachment that separates basement rocks from the Prichard Formation of the Belt Supergroup. Some of the low resistivities observed likely are caused by pre-Prichard crystalline rocks (meta- morphic (metasedimentary) or igneous) that are conductive because they contain large amounts of metallic minerals. The striking variations in observed resistivities in the Prichard Formation of from 0.6 to more than 400 ohm-meters probably reflect the percentage and continuity of iron-sulfide-rich zones. Another possible line of thermal evidence is the absence of magnetic sources in pre-Belt crystalline rocks at depths of 10-15 km. The lack of mag- netic anomalies may relate to shallow Curie-point tempera- tures, above which magnetism does not exist. The magnetotelluric data have limitations in structural analysis because resistivities are mostly controlled by the percentages of metallic minerals and not by lithology. Accu- racy of the interfaces obtained from the magnetotelluric data is not great, both because of the widely spaced sounding locations and because of the very large contrasts in resistiv- ity between unmineralized and mineralized zones in the Belt Supergroup. INTRODUCTION Regional geophysical studies conducted by the U.S. Geological Survey in the northern Rocky Mountains during the past 25 years provide new insights about the geologic framework and mineral resources of the region. In this report, the emphasis is on interpretation of geophysical data compiled for the Libby thrust belt in northwestern Montana (fig. 1). Interpretations complement the results of geologic mapping described by Harrison and Cressman (1993), who also discussed the geology and structural framework of the Libby thrust belt. During the course of their studies, Harrison and Cressman constructed geologic cross sections across the Libby thrust belt on the basis of measured sections and pro- jections of surface geology. Depth cutoff of these sections was 4.3 km (14,000 ft) below sea level. In the study described herein modeling of gravity anomaly data was done along two of the geologic cross sec- tions to provide information about structure and lithology of the Purcell anticlinorium and Sylvanite anticline from depths greater than 4.3 km to at least as deep as Precambrian crys- talline basement in the middle part of the crust. No magnetic modeling was done because of the absence of magnetic anomalies having sources likely in deeply buried crystalline basement. Included in this report are sections on interpretation of seismic and magnetotelluric data. J.E. Harrison describes interpretations of data from about 70 km of COCORP (The Consortium for Continental Reflection Profiling) seismic reflection surveying along two profiles that cross the Sylva- nite anticline and Purcell anticlinorium. W.D. Stanley describes analysis of five magnetotelluric soundings along a profile that extends across the Libby thrust belt and the Pur- cell anticlinorium. The lines of section of the gravity models and the COCORP profiles and the locations of the magneto- telluric soundings are shown in figure 2. Data from 13 audiomagnetotelluric soundings in two profiles on the Syl- vanite anticline provide information about resistivities of rocks in the upper few kilometers of the crust (Long, 1988). The geology shown on plate 1 is a simplified version of the 1:125,000-scale geologic map compiled by Harrison and Cressman (1993). The western parts of geologic maps for the Wallace and Kalispell 1°x2°quadrangles (Harrison and oth- ers, 1986, 1992) provide the regional geologic context of the Libby thrust belt. Magnetic and gravity anomaly data are also shown on plate 1. To provide a broader perspective of the geophysical setting, magnetic and gravity anomaly maps (pl. 2) were compiled at a scale of 1:500,000 for the Libby thrust belt and adjacent areas, extending from near lat 47°20' N. to lat 49°00" N. and from near long 114945" W. to long 116°15'W. Previous studies of gravity and magnetic anomaly data applied to geologic framework and mineral resource investi- gations of the region are described in King and others (1970), Harrison and others (1972, 1980, 1985), Kleinkopf and others (1972, 1982, 1988), Wynn and others (1977), Kleinkopf (1977, 1981, 1983, 1984), Kleinkopf and Long (1979), Kleinkopf and Wilson (1981), Kleinkopf and Ban- key (1982), Kleinkopf and Harrison (1982), Fountain and McDonough (1984), and Harris (1985). The author thanks the many U.S. Geological Survey colleagues who contributed to this project. Jack Harrison offered many constructive suggestions and provided many stimulating discussions during various phases of the project and report preparation. The sections on seismic and magne- totelluric investigations by Jack Harrison and Dal Stanley provide substantive contributions to the conclusions of this paper. The geology was digitized and compiled under super- vision of Stanton H. Moll; the geologic map of plate 1 was completed in final digital form by Nancy Shock. Viki Ban- key and Mike Brickey collected gravity data in the field. Viki Bankey compiled and edited the gravity and magnetic data and prepared early versions of the gravity and magnetic anomaly maps, and Gerda Abrams prepared later versions of the gravity and magnetic anomaly maps used in this report. Several colleagues offered valuable constructive comments during the technical review process. GEOLOGY The Libby thrust belt (fig. 1) is in the northwestern part of the Belt basin, which formed along the western edge of the GEOLOGY 3 118° 112° 110° ag) WASHINGTON OREGON "/ 44° Figure 1. 50 100 KILOMETERS __ L ___ EXPLANATION Tertiary volcanic rocks Tertiary to Jurassic AZ C intrusive rocks Quaternary to Cambrian sedimentary rocks I Late Proterozoic rocks (Windermere System of Canada) Middle Proterozoic Belt Supergroup (Purcell Supergroup of Cana a)—Sti}3p1e shows areas of high- grade metamorphism; may include some pre- -_- Belt rocks X0 ,; ¢} Early Proterozoic pre- ' Belt metamorphic rocks UNITED STATES Contact High-angle fault Right-lateral fault -g--y- Thrust fault-Sawteeth on upper plate ——1—— Anticline + Syncline Map showing generalized geology of the Belt terrane in the region of the Libby thrust belt, northwestern Montana and sur- rounding area. Modified from King (1969) and Harrison and Cressman (1993). North American craton in Middle Proterozoic time (Harrison, 1972; Harrison and Cressman, 1993). The Belt terrane is the oldest of a series of tectonostratigraphic assem- blages that make up a wedge of supracrustal rocks along the western edge of the continental craton (Price, 1981). The Libby thrust belt is one of a series of major north-north- west-trending structural features north of the Lewis and Clark line, a major intraplate tectonic boundary (Reynolds and Kleinkopf, 1977). The Libby thrust belt was character- ized by Harrison and Cressman (1993) as a "ripped-apart syncline between two anticlinal structures," between the Purcell anticlinorium to the east and the Sylvanite anticline to the northwest. The Moyie thrust system is along the western and northwestern margin of the Libby thrust belt and overrides it on the southwest (Harrison and Cressman, 1993). The thrust belt is limited to the south by the Hope fault (fig. 1), which Harrison and Cressman described as a "crustal flaw." Some 15 km of Middle Proterozoic Belt Supergroup rocks underlies the Libby thrust belt. Rocks of the Belt Supergroup grade upward from turbidites through mar- ginal-marine, tidal-flat, and shallow-shelf deposits. The sequence consists mostly of fine-grained sedimentary rocks, mainly argillite, siltite, quartzite, and carbonate. The Pri- chard Formation, the oldest of the Belt units, is dominantly quartzite but contains beds of pyritic-pyrrhotitic argillite, GEOPHYSICAL INTERPRETATIONS, LIBBY THRUST BELT, MONTANA _CANADA__ I. “N MONTANA 4 49°r————— Line of section shown in figure 4 Composite line of section shown in NO. 1 GIBBS figure 5 BOREHOLE \ p Clark Fork 4’6‘ . mu | l Study area I \ ago I - = ! I~ 0 10 20 30 40 50 KILOMETERS L___ | | | | | EXPLANATION Normal fault-Bar and ball on downthrown side Thrust fault-Sawteeth on upper plate Strike-slip fault-Arrows show direction of apparent movement Anticline-Showing direction of plunge. Dashed where inferred T. -«--w- 4--- —fl— Overturned anticline +> ©3 e Syncline-Showing direction of plunge. Dashed where inferred Location and number of magnetotelluric sounding Line of seismic reflection profile Figure 2. Map showing major structural features in the area of the Libby thrust belt, northwestern Montana. Lines of extended geologic cross sections B-B' and J-J' (Harrison and Cressman, 1993) used in gravity modeling, lines of seismic reflection profiles MT-1 and MT-2, and locations of magnetotelluric soundings 1 through 5 and Gibbs No. 1 borehole are also shown. AEROMAGNETIC AND GRAVITY ANOMALY DATA 5 which is more dense than overlying rocks of the Ravalli Group. The thickest exposures of Prichard rocks in the map area are in the Sylvanite anticline, west of Yaak (pl. 1A), where some 5.6 km of Prichard rocks is exposed (Harrison and Cressman, 1993). Almost 1,000 m of mafic sills is recorded in measured sections in the Yaak area. Succesively overlying the Prichard Formation are the Ravalli Group, the Helena and Wallace Formations, and the Missoula Group. A number of short-wavelength positive magnetic anomalies (pl. 14) are associated with outcropping metased- imentary rocks of the Burke and Revett Formations of the lower part of the Ravalli Group. In hand specimen, particu- larly those of the Burke Formation, euhedral grains of mag- netite were observed in siltite (Kleinkopf and others, 1972). The overlying Wallace Formation consists mainly of calcar- eous and dolomitic fine- to medium-grained clastic rocks. Above the Wallace Formation is the Missoula Group, which consists primarily of interbedded red and green clastic rocks that are less dense than other Belt rocks. Rocks of the Mis- soula Group produce negative gravity anomalies where thick sequences are juxtaposed against denser Belt rocks. Only minor amounts of Cambrian sedimentary rocks are pre- served in outcrop, and no other Phanerozoic sedimentary rocks are present in the study area except for alluvial, lake-bed, and glacial deposits and other low-density sedi- ments of Pleistocene to Holocene age that have accumulated in modern stream valleys. Rocks of the Belt Supergroup exhibit an increase in metamorphic grade from east to west across the basin and with depth in the stratigraphic section. Metamorphic grade ranges from the biotite zone of the greenschist facies in the lowest exposed part of the Prichard Formation through chlo- rite-sericite rocks in middle Belt to high-grade diagenesis at the top of the Belt Supergroup (Harrison and Cressman, 1993). Because the sedimentary rocks are mostly argillite and siltite, the average rock density generally increases downsection with depth and with increased metamorphic grade. Exceptions are rocks of carbonate sequences of the Helena and Wallace Formations, which are more dense than the underlying rocks of the Ravalli Group. Although massive quartzite is present throughout the section, its volume as compared to that of the total section is low, and thus its influ- ence on gravity patterns is minimal. The Belt sequence in the Libby thrust belt has been intruded by Precambrian sills of diorite to gabbro, early Cre- taceous plutons ranging from granite to pyroxenite, and Eocene plutonic rocks of quartz monzonite porphyry to granodiorite. Middle to Late Proterozoic mafic sills are abundant in the lower part of the Prichard Formation and are less abundant in the Ravalli Group, Wallace Formation, and Missoula Group. The best exposures of sills are in cirques, particularly on the Sylvanite anticline, and on ridges (Harri- son and Cressman, 1993). The sills are altered (King and others, 1970; Bishop, 1973) and commonly exhibit little or no magnetic expression. Scattered exposures of Early Cretaceous and Eocene plutonic igneous rocks are present in the study area. These rocks, ranging in composition from granite and granodiorite to syenite and pyroxenite, have been intruded into Belt strata (Harrison and Cressman, 1993). The largest exposed plutons are the Dry Creek stock (unit Kg, pl. 1), about 15 km southwest of Libby in the Cabinet Mountains, and the Vermilion River stock (unit Kg), just east of the town of Trout Creek. Smaller exposures of granitic rocks similar to the rocks of the Dry Creek stock are present a few kilometers north, west, and south of the stock. Mafic intrusive com- plexes, principally pyroxenite and syenite (unit Kps), are exposed at Vermiculite Mountain, 10 km east-northeast of Libby. At Bobtail Creek, 12 km north of Libby, a pluton of predominantly coarse grained, porphyritic syenite (unit Kps) contains segregations of pyroxene and amphibole. AEROMAGNETIC ANOMALY DATA Aeromagnetic anomaly data for the study area were compiled from analog records obtained from two regional aerial surveys flown under contract to the U.S. Geological Survey. The elevation of most of the study area is between 1 and 2 km, and several mountain peaks are higher than 2.3 km. Both surveys were flown east-west at a nominal altitude of 2.1 km above sea level, except to clear the high peaks. The survey north of lat 48°30' N. was flown in 1972 by Scintrex Minerals Inc., at a line spacing of 3.2 km (U.S. Geological Survey, 1973). The survey south of lat 48°30 N. was flown in 1968 by Lockwood, Kessler, and Bartlett, Inc., at a line spacing of 1.6 km (U.S. Geological Survey, 196927; Kleinkopf and others, 1972). The total-intensity aeromagnetic anomaly data are shown at scales of 1:250,000 and 1:500,000 (pl. 1A, 24). Plate 1A is a mosaic of the aeromagnetic anomaly maps received from the contractors and is superimposed on a gen- eralized version of the digital geologic map of Harrison and Cressman (1993). The residual total-intensity data were reduced to a single datum, then continued to a common alti- tude of 2.1 km above sea level and merged by fitting along map boundaries using programs developed by M.W. Webring (written commun., 1981) and R.E. Sweeney (writ- ten commun., 1981). The merged data were then gridded at a spacing of 1 km and contoured at an interval of 20 nT (pl. 2A). The International Geomagnetic Reference Field (IGRF) for Epoch 1975 (Peddie and others, 1976), which is about 6.6 gammas per kilometer to the northeast, was removed from the data. GRAVITY ANOMALY DATA Gravity data for 775 stations were extracted from regional digital data sets compiled in support of U.S. 6 GEOPHYSICAL INTERPRETATIONS, LIBBY THRUST BELT, MONTANA Geological Survey programs in geologic framework and mineral resource appraisal. The data originate from a variety of sources including unpublished files of the U.S. Geological Survey, Wynn and others (1977), Kleinkopf (1981), Kleinkopf and Wilson (1981), Bankey and others (1985), and a U.S. Department of Defense, Defense Mapping Agency (DMA), data base (written commun, 1989). Measurements were made with high-sensitivity gravity meters using four-wheel-drive vehicles and, in some cases, foot traverses (Bankey and others, 1982, 1985; Brickey and others, 1980). The station spacing is variable, ranging from 2-3 km along roads to greater than 5 km in roadless areas. Gravity observations were made at locations of known, or recoverable, horizontal and vertical positions in terms of longitude, latitude, and elevations. These positions include survey bench marks, photogrammetric elevation points shown on U.S. Geological Survey 7.5- and 15-minute topo- graphic maps, and locations of low topographic relief at which elevations and horizontal positions could be estimated with confidence. The observed gravity is referenced to the IGSN-71 (International Gravity Standardization Net) datum (Morelli and others, 1974) by means of ties to U.S. Depart- ment of Defense bases ACIC-0442 at Missoula, Mont., and ACIC-4006-1 at Wallace, Idaho (U.S. Department of Defense, 1974). Gravity reduction procedures are based on equations discussed by Cordell and others (1982). All gravity stations were reduced to the complete Bouguer anomaly, assuming a mean crustal density of 2.67 g/cm*, using the 1967 gravity formula (International Association of Geodesy, 1967). Terrain corrections were made using com- puter software to access digital terrain files. Terrain correc- tions were calculated radially around each station to a distance of 166.7 km (Plouff, 1977). The final-processed data, consisting of complete Bouguer gravity anomalies, were gridded and contoured using computer routines based on minimum curvature (Briggs, 1974; Webring, 1981). GEOLOGIC INTERPRETATION OF AEROMAGNETIC ANOMALIES The dominant anomaly patterns on the 1:250,000-scale aeromagnetic anomaly map (pl. 14) consist of equidimen- sional positive anomalies, short-wavelength positive and negative anomalies elongated generally north to northwest, and zones of steep magnetic gradient. Three types of geo- logic sources that contain relatively high percentages of magnetite account for most of the anomalies. The sources mainly are in outcrop but in some cases are partly or totally buried in the near-surface. The principal magnetic anomaly sources are igneous intrusive rocks, magnetite-bearing sedi- mentary rocks of the Ravalli Group, and major fault zones. The Proterozoic basement, where drilled in western Montana, is primarily a mixture of granite and granitic gneiss, both of which typically are of low magnetization (Harrison and others, 1972). Long-wavelength anomalies that may have sources in deeply buried crystalline basement rocks at about 15 km are not known to be present in the region (Harrison and others, 1980). The lack of detectable magnetic anomalies in the study area is attributed to the relatively nonmagnetic character of the basement rocks and their deep burial. Resistivites measured at crystalline base- ment levels may be due to high-temperature crystalline base- ment rocks (see discussion on magnetotelluric data). Correspondingly, the lack of magnetic anomalies may be ascribed to a crystalline basement complex at depths of 10-15 km that has temperatures above the Curie point, at which magnetization of the rocks is lost. The mafic sills of dioritic to gabbroic composition normally would be expected to be magnetic but show little or no magnetic response in outcrop. This anomalous low magnetization is attributed to reduction of magnetite content in the sills by chemical processes related to interaction between hydrother- mal fluids and cooling magma of the intrusions (King and others, 1970; Harrison and others, 1972). Gravity anomaly data, on the other hand, show distinct anomalies that have sources, as determined from modeling, at or near crystalline basement levels. The most distinctive magnetic anomalies in the princi- pal study area are five positive anomalies (numbered 1, 2, 3, 4, and 5 on pl. 14) associated with Cretaceous or younger cupolas and stocks that are exposed or are known from drill- holes to be present in the near-surface. These anomalies range in amplitude from 100 to more than 3,000 nT. The magnetic data provide information about the subsurface extent of these igneous intrusions and give clues for locating other buried intrusions that have subtle expressions. The most prominent of the five magnetic anomalies (anomaly 1, pl. 14) is associated with a pyroxenite-syenite complex (unit Kps) that is partly exposed at Rainy Creek-Vermiculite Mountain, about 6 km east-northeast of Libby. The anomaly has an amplitude exceeding 3,000 nT, and it correlates with the main mass of ultramafic rocks. Pyroxenite is the principal ultramafic rock, and altered pyroxenite in this stock has been a major source of vermicu- lite in the United States. Various aspects of the geology of the complex and adjacent areas were studied by Larsen and Pardee (1929), Boettcher (1966), and Harrison and Cress- man (1993), who concluded that the complex is a laccolith. The complex consists of a biotitic core surrounded by biotite pyroxenite and an outer ring of magnetite pyroxenite. A large pluton of syenite and associated alkaline syenite dikes are in the southwestern part of the complex. The resulting anomaly is a combined expression of the total complex and does not distinguish syenite from pyroxenite rocks. Larsen and Pardee (1929) reported that the syenite locally contains 3-12 percent magnetite. Boettcher (1966) speculated that minor fenitization of Belt rocks along the northern contact of the complex may indicate a large body of carbonatite at GEOLOGIC INTERPRETATION OF AEROMAGNETIC ANOMALIES 7 depth. The Bouguer gravity anomaly data show a residual high resulting from a change in gradient along a nosing in the contours (pl. 1B) that is attributed to mafic rocks of the com- plex that are denser than the enclosing rocks. The residual gravity high correlates with the positive magnetic anomaly over outcrops of the complex. Northwest of the Rainy Creek complex, about 8 km north-northwest of Libby, a pluton (unit Ks, pl. 1) crops out in two places near Bobtail Creek. The associated single pos- itive magnetic anomaly (anomaly 2, pl. 1A) has an amplitude of 900 nT and is slightly elongated north-northwest. The plu- ton is mainly syenite and includes segregations of pyroxene and amphibole (Gibson, 1948). A high nosing in the gravity contours correlates with the positive magnetic anomaly over the complex in much the same character as at anomaly 1. Both the Rainy Creek and Bobtail Creek complexes are esti- mated to be about 100 m.y. old (Harrison and Cressman, 1993). A northwest-trending belt of high gradient in the Bou- guer gravity anomaly data suggests that the two features may be connected in the subsurface, possibly along a north- west-trending structural zone, although no surface indication of such a zone is evident in the geologic mapping. Boettcher (1966) stated that the syenite and ultramafic rocks at Bobtail Creek and Rainy Creek may be comagmatic. A low-amplitude positive magnetic platform extends to the south of anomaly 1 across the Kootenai River for some 15 km in the form of a wide plunging nose in the magnetic contours (pl. 14). A small, but distinct nose is superimposed on the magnetic platform anomaly (pl. 24) near Fisher Mountain. The anomaly sources may be a shallow cupola and a large related mass of moderately magnetic rock that gives rise to the platform anomaly. In the southwestern part of the map, just east of Trout Creek, a positive magnetic anomaly (anomaly 3, pl. 14) of almost 300 nT is associated with outcrops of hornblende granodiorite (unit Kg) of the Vermilion River stock. The axis of the anomaly is oriented north-northeast and is offset toward the northwest edge of the outcrops. Intrusive rocks probably are present in the subsurface beneath lower Pri- chard rocks that crop out northwest of the outcrops of the stock. The associated broad gravity high reflects the high-density hornblende granodiorite rocks that are in con- tact with rocks of the lower part of the Prichard Formation. Near the southern end of the Libby thrust belt study area and just northeast of Thompson Falls, a circular positive anomaly (anomaly 4, pl. 1A) of more than 200 nT amplitude indicates the presence of a buried intrusion. According to drill-core information provided by the staff of Noranda Exploration, Inc. (unpub. data, 1979), quartz monzonite porphyry is present about 775 m below the ground surface (Kleinkopf and others, 1988). The configuration of the anomaly and the horizontal extent of the steepest gradient on the flanks of the anomaly are consistent with depth of burial as indicated by the core-hole information. Using potassium-argon techniques, analysis of biotite and potasssium feldspar from the quartz monzonite gave radio- metric ages of 40-50 Ma (Marvin and others, 1984), younger than ages for the major intrusions to the north. An associated gravity low is discussed in the section on interpretation of the gravity anomaly data. Another well-exposed intrusive complex, the Dry Creek stock, about 15 km southwest of Libby, has less prom- inent magnetic expression (anomaly 5, pl. 1A). The expres- sion is composite and consists of three small positive anomalies of less than 100 nT in areas of granitic outcrops in high elevations of the eastern part of the stock. To the west, magnetic gradients decrease rapidly, and there is no anomaly over exposures of the stock (Kleinkopf, 1981). The lithology of the Dry Creek stock is quartz monzonite to granodiorite (Gibson, 1948), and the granitic rock here probably belongs to a 100-m.y.-old intrusive event (Marvin and others, 1984; Harrison and Cressman, 1993). The complex character of the magnetic response of the Dry Creek stock is strikingly dif- ferent than the distinct 300-nT positive anomalies associated with the Vermilion River and Liver Peak stocks. Magnetic susceptibilities of three samples of granodiorite from the eastern part of the Dry Creek complex range from 0.019 to 0.072 International System (S1) units (Kleinkopf, 1981) and generally correlate with topographic highs. The magnetic highs may relate to small stocks that formed as apophyses of the larger, relatively less magnetic granitic complex of the Dry Creek stock. Harrison and Cressman, 1993) reported that all stocks and plutons in the map area have contact meta- morphic aureoles. In particular, the Dry Creek stock has a 1,000-m-wide hornfels zone where it intrudes rocks of the Wallace Formation and Missoula Group. The stock is sheared and foliated by faulting, particularly on the west side (Wells and others, 1981) where the magnetic map shows little or no magnetic expression. I infer from the lack of mag- netic expression that little primary magnetite was present in the intrusion. Gibson (1948) reported no alteration of the stock but observed that magnetite formed after initial crystal- lization of the magma and probably is interstitial to pyroxene and other minerals. A small outcrop of granodiorite is about 12 km south of the Dry Creek stock (pl. 1A). No magnetic anomaly was detected on two flight lines that passed just to the north and to the south of this outcrop, which further indi- cates that the granodiorite is of low magnetization (Kleinkopf, 1981). Many of the elongated short-wavelength magnetic anomalies are related to magnetite-rich horizons predomi- nantly in siltite layers of the Burke and Revett Formations of the Ravalli Group (Kleinkopf and others, 1972). These anomalies are typically 50-100 nT in amplitude, and many are elongated north-northwest along outcrops of Ravalli Group rocks that are topographically high and, in some cases, structurally controlled. Field checks in the east-central part of the area show that positive anomalies are associated mainly with outcrops of the Burke Formation. Hand speci- ments of the Burke exhibit abundant euhedral magnetite 8 GEOPHYSICAL INTERPRETATIONS, LIBBY THRUST BELT, MONTANA grains. Magnetic susceptibilities for three samples collected in this area of from 0.013 to 0.035 SI units can account for the observed anomalies. In the northern part of the area at Mount Henry, and just to the south, two distinct positive anomalies correlate with outcrops of Revett Formation in high topography (Bankey and others, 1986). The Hope fault zone and faults of the Lewis and Clark line are expressed by parallel magnetic alignments (pl. 24). The Hope fault is along the southwestern margin of the Libby thrust belt and is delineated in the magnetic anomaly data by zones of steep gradients approximately parallel with the fault (pl. 1A). In general, the magnetic patterns are more variable on the northeastern side of the fault (Kleinkopf and others, 1972). Southwest of the Libby thrust belt, the magnetic anom- aly data show expressions of igneous intrusions and fault zones. A few kilometers north of Wallace, Idaho, the regional aeromagnetic anomaly map (pl. 24) shows a north-northeast-elongated, 200-300-nT, positive anomaly that correlates with quartz monzonite exposures of the Gem stock intrusive complex (Hobbs and others, 1965; Kleinkopf and others, 1988). Also in the Wallace area, distinct west-northwest linear magnetic trends reflect structural alignments of the Thompson Pass, Osburn, and Placer Creek faults, which are part of the Lewis and Clark line (Kleinkopf and others, 1988). A large, complex positive magnetic anomaly 25 km west of Thompson Falls correlates with out- crops of granitic intrusions (unit Ks) that are composed of syenite. The elongate shape of the anomaly suggests that granitic rocks are present as an extension of, or possibly as a separate, pluton in the subsurface southwest of the outcrop. GEOLOGIC INTERPRETATION OF BOUGUER GRAVITY ANOMALIES In the area of the Libby thrust belt, there is a strong cor- relation of gravity anomalies with both geologic framework and major lithologic units. The geology and principal struc- tural features are shown, together with the Bouguer gravity anomaly, to illustrate the gravity signatures and to provide a reference for discussions of the geologic interpretations of the gravity anomaly data (pls. 1B, 2B). The regional gravity field dips gently to the east across the study area and exhibits a complex of regional-scale positive and negative anomalies that correspond to broad open folds west of the Rocky Mountain trench (Harrison and others, 1980). The gravity anomaly patterns can be discussed in terms of configurations of the -130- to -140-mGal contours (pls. 1B, 2B). In the middle and southern parts of the study area, structurally deformed rocks of the Missoula Group generally are less dense than less deformed adjacent rocks of older parts of the Belt Supergroup. The -140-mGal contour delim- its a broad negative area whose source is attributed to low-density Missoula Group rocks that also have experi- enced some reduction in density due to faulting and fractur- ing associated with development of the Libby thrust belt. Much of the grain of the gravity anomaly patterns within the broad negative area is approximately parallel with the strike of thrust and normal faults of the Libby thrust belt (pls. 1B, 2B). The broad negative gravity feature has its lowest value of -160 mGal just east of Thompson Falls, where it is approximately coincident with the Liver Peak magnetic anomaly (anomaly 4, pl. 14). This deep gravity low (-160 mGal) forms the southern part of a horseshoe-shaped negative anomaly that is open to the north and has residual lows at both ends of the horseshoe. The distinct low (-160 mGal) centered on the south flank of Liver Peak may reflect low-density felsic rocks of the intrusion. In addition, the deepest part of the gravity low may relate to low-density hydrothermally altered rocks associated with molybde- num-tungsten-mineralized rock at Liver Peak. The west and northwest extension of the horseshoe anomaly may reflect a structural trough of low-density Quaternary sediments and localized fracturing of Belt rocks at the intersection of the Hope fault and south-trending thrust faults of the Libby thrust belt. Alternatively, a less magnetic phase of low-density intrusive rocks at Liver Peak may be extensive in the subsurface in the west and northwest areas. North of lat 48° N., about 25 km northeast of Trout Creek, the regional gravity low defined by the -140-mGal contour narrows and follows the north- to northeast-trending Fisher River where it cuts around geologic structure. The low gravity values may express crossfaulting that controls the river direction here. Northwest of the river, a northeast-trending gravity divide is defined by -140-mGal contours. Northwest of the divide, gravity values decrease to less than -144 mGal over the alluvial valley south of Libby, which is in the center of a broad, relatively flat bottomed low approximately defined by the -130- to -140-mGal contours. Assuming that about 6 mGal of relief is associated with the alluvial valley, the thickness of valley fill may exceed 400 m. North of Libby, north-northwest gravity trends corre- late with major thrust faults of the Libby thrust belt as it passes north into Canada. The broad, but distinct north-trending gravity low that extends south of Yaak separates the well-defined gravity highs associated with the Purcell anticlinorium on the east and the Sylvanite anticline on the west. The Sylvanite anticline and the Purcell anticlinorium exhibit prominent gravity highs that have amplitudes exceeding 10 and 20 mGal, respectively (pls. 1B, 2B). The high associated with the Purcell anticlinorium generally trends northwesterly; however, on close inspection, gravity patterns show that the broad high is composed of three sepa- rate northerly trending highs arranged en echelon along a northwest trend. The small anticlinal axis passing just west of the Gibbs No. 1 drillhole correlates with the southernmost gravity high. The gravity high associated with the Purcell GEOPHYSICAL INTERPRETATION OF DEEP-SEATED GEOLOGIC FEATURES 9 anticlinorium is broken and offset at a location about 10 km east of Libby. The offset is part of a west-northwest trend in the gravity anomaly data that passes just north of Libby and extends to the Moyie thrust fault near Troy. In addition, discontinuous trends in the magnetic anomaly data correlate with the gravity trends, notably just south of the Rainey Creek intrusive complex (anomaly 1, pl. 1A). The composite gravity high associated with the Sylvanite anticline exhibits a distinct northerly trend and subtle, superimposed north- westerly trends. The mapped axis of the anticline is offset southwest of the axis of the gravity high. Gravity modeling, discussed in the next section of this paper, indicates that the Purcell anticlinorium and the Sylvanite anticline are very likely cored by stacks of thrust slices of dense crystalline basement rocks that account for the large gravity highs across these two structures. West of the Cabinet Mountain Wilderness, the Moyie thrust fault is along the west side of a narrow troughlike gravity anomaly related to the north-trending Bull River valley (pl. 2B). At the north end of the valley, the thrust turns and follows the northwest extension of the gravity trough into Idaho. The deepest part of the trough is about 10 km northwest of Troy. Three distinct lows are along the trough associated with Bull River valley. The deepest low, which exceeds 6 mGal amplitude, is at the north end of the valley, an area of complex geologic structure (pl. 1B). This deep low is separated from the next low to the south by an east-north- east-trending gravity ridge. Contour alignments and trend projection across the Cabinet Mountain Wilderness provides some evidence that the ridge may express a buried fault zone that extends northeast as far as the gravity platform just north of Libby. Although no continuous fault zone has been mapped in the surface geology, the geologic map shows dis- continous northeast-trending faults and areas of fault inter- section along the projected gravity feature. In the Cabinet Mountains Wilderness, in the central part of the Libby thrust belt, a line of positive gravity anomalies, uniformly about 6-10 mGal amplitude above background, extends south-southeast from near the Kootenai River to as far as the latitude of Trout Creek. One of the highs is cen- tered north of the exposures of the Dry Creek stock about 15 km southwest of Libby but is much more extensive than either outcrops of the stock or the associated magnetic anom- alies described earlier in this report as well as in Kleinkopf (1981). The second gravity high is at Snowshoe Peak, and a possible correlative magnetic source is in the subsurface beneath outcrops of the Wallace and Helena Formations and the Ravalli Group. The third gravity high is at Flat Top Mountain, about 20 km south-southeast of the second; the highest part of this anomaly is about 5 km west of Flat Top Mountain. Between the second and third highs, the south-southeast trend of the highs is broken by a disconti- nous zone of gravity patterns that trend east-northeast toward Fisher Mountain; crosscutting gravity contour alignments are along the east-northeast trend. In addition, both in the Cabinet Mountains Wilderness and in the main part of the Libby thrust belt, structural complexities and crossfaulting are present along the trend (pl. 1). The anomalous gravity and the structural features may reflect a zone of weakness in the crust of unknown origin. The fourth gravity high is about 16 km south-southeast at Trout Creek and correlates with the Vermilion Creek stock of hornblende granodiorite composi- tion. Kleinkopf (1981) concluded from interpretations of the gravity high over the Dry Creek stock and from the distribu- tion of the magnetic anomalies that the Cabinet Mountains Wilderness may be underlain by an extensive granitic batholith that is relatively nonmagnetic. The gravity highs may reflect large near-surface intrusions along the feature, and the small positive magnetic anomalies may indicate near-surface sources associated with small granitic intru- sions or cupolas that may be apophyses of a possible batholithic mass at a depth of a few kilometers, as suggested by the gravity model along geologic section J-J' (fig. 5) (see following section). On the basis of surface mapping and indi- cations in the outcropping sedimentary rocks, Harrison and Cressman (1993), in their structure sections, showed the Dry Creek stock extending to the south. GEOPHYSICAL INTERPRETATION OF DEEP-SEATED GEOLOGIC FEATURES MODELING OF GRAVITY PROFILES Modeling of Bouguer gravity anomaly data for the Libby thrust belt and adjacent areas supports the interpreta- tion that the crystalline basement is elevated beneath the Pur- cell anticlinorium and the Sylvanite anticline. The model studies update previous model studies of the Purcell anticli- norium by myself, described in Harrison and others (1980), by addition of new gravity control along the sections and new subsurface lithologic information obtained from well logs in the Atlantic Richfield-Marathon Oil Gibbs No. 1 borehole, NE % sec. 2, T. 28 N., R. 27 W. (pl. 2). An impor- tant constraint on the validity of the modeling is selection of reasonable densities for the major rock types, or combina- tions of rock types, depicted in the geologic cross sections. These rock types include crystalline basement, presumed to be Proterozoic in age, Prichard Formation and sills, and post-Prichard strata composed mainly of rocks of the Ravalli Group, Helena and Wallace Formations, and Missoula Group. I selected values for density based on laboratory measurements and on supporting evidence from well logs taken in the Gibbs No. 1 deep borehole (fig. 3). Densities for a given rock unit within the study area are assumed, on the basis of the general regional consistency of rock types, to be laterally uniform. In laboratory measure- ments of metamorphic rocks, Smithson (1971) found a den- sity range from 2.7 g/cm> for granitic gneiss to 2.86 g/cm* 10 Sonic KILOMETERS 0 - 057 2.0 - 2.5 - 3.0 - 35 - 4.0 - 45 - 50 -I 5.5 - 6.0 - +--- GEOPHYSICAL INTERPRETATIONS, LIBBY THRUST BELT, MONTANA Surface section BURKE FORMATION PRICHARD FORMATION ARCO Gibbs Neutron No. 1 borehole (Informal members of Gamma Cressman, 1985) ray density 250 - 215 - 3.00 Member H L____L___ j \9 ......... Ss 9 @ 10 % 8 ‘11 P22 Member G 10 J §5 9 ~ r10 10 *~ 8 MemberF a 10 a 6 ¢ 6 6 F O R M A T I 0 N P R I C H A R D -* 5 - Member E e14 Pinkham .............. e- 12 thrust fault 12 * 10 Members A through D? &~7 «14 o 16 & 35 Little Wolf Creek KILOMETERS 0 -. Transition member 05 - Upper laminated 10-7 argillite member Argillite bed Lower part 2.0 - EXPLANATION Diorite sill Argillite Siltite Dolomitic siltite Calcareous siltite .'] Poorly sorted quartzite Well-sorted quartzite Phyllite Mica schist Mylonite Hornblende- and garnet- bearing quartz lenses Interlaminated and closely interbedded siltite and argillite in proportion shown 4 12 - Direction and amount of dip Figure 3. Generalized lithologic and geophysical logs, Gibbs No.1 borehole, Flathead County, Montana. Neutron density log is in grams per centimeter per second per second. Data courtesy of Atlantic Richfield and Marathon Oil Companies. Surface section at Little Wolf Creek is also shown. Modified from Harrison and Cressman (1993). GEOPHYSICAL INTERPRETATION OF DEEP-SEATED GEOLOGIC FEATURES 11 for granulite facies. I used 2.8 g/cm? as a reasonable estimate for the density of the crystalline basement. Density values for 13 samples of post-Prichard Belt rocks (measured in USGS laboratories) range from 2.58 to 2.76 g/cm" and average 2.67 g/cm". Earlier laboratory measurements of Pre- cambrian mafic sill samples yielded an average density value of 2.91 g/cm" (Harrison and others, 1972). The Precambrian sills in the borehole (fig. 3) are characterized by high density, low radioactivity, and rather high velocity (Harrison and oth- ers, 1985). The neutron-density log shows considerable vari- ation in density of the sills, presumably related to alteration. Density values range from about 2.76 g/cm» to spikes on the log exceeding 3.0 g/cm>. The information from these logs, together with data from the lithologic log, provides reliable thickness estimates for the sills, which total more than 1,000 m of the drilled rock. I estimated that 7.6 km or more of Pri- chard Formation underlies the borehole site. The assumption was made that the density of the Prichard Formation, without sills, is in the range of from 2.65 to 2.70 g/cm3, considering the thick sequences of quartzite and argillite. Thus, combin- ing these values with an verage value of about 2.91 g/cm" for the Precambrian sills, I obtained an average density for the total column of Prichard Formation and sills of about 2.73 g/cm". The lithologic units and corresponding densities assigned to the bodies are listed in table 1. Gravity models were calculated along the two lines of section, B-B' and J-J'"' prepared by Harrison (Harrison and Cressman, 1993), in an attempt to gain information about structure and lithology along the zone of detachment and at crystalline basement levels at depths of approximately 15 km. The sections were simplified and extended laterally (figs. 4, 5) for purposes of the gravity modeling. Geologic data in Idaho used to extend section B-B' are from a geologic map of the Eastport area (Burmester, 1985). The expansion of section J-J' in Montana is from new mapping by Harrison (Harrison and others, 1992); the extension of section J'-J'" is offset 13 km southward along strike in order to cross the Purcell anticlinorium at the location of the Gibbs No. 1 bore- hole (fig. 2). Two-dimensional modeling of the gravity data was done on a mainframe computer using the software SAKI (Webring, 1982). Control for the modeled sections was pro- vided by random-spaced gravity stations that are along or projected into the lines of section. The resulting gravity con- trol along the modeled sections varies from 1 to more than 5 km in spacing of stations. The Purcell anticlinorium and the Sylvanite anticline are appropriate for gravity model studies. Both structures exhibit prominent positive gravity anomalies that correlate with geologic structures at the surface. The gravity anomaly across the Sylvanite anticline is well defined and is as much as about 10 mGal in amplitude. The gravity expression asso- ciated with the Purcell anticlinorium has an amplitude of 15-20 mGal, and the anticlinorium is more than 50 km wide and is remarkably continuous and linear for about 100 km to the north into Canada. Several theories on source and mass Table 1. - Lithologies and densities assumed for gravity models of extended geologic sections B-B' and J-J', Libby thrust belt, northwestern Montana. [See text for explanation. Lines of section are shown in figure 2; geologic sections and gravity models are shown in figures 4 and 5] Density Body (g/cm") Lithology Model B-B' extended 1 2.80 Crystalline basement. 2 2.73 Mafic sill complex. 3 2.73 Prichard Formation and sills. 4 2.68 Ravalli Group. 5 2.70 Ravalli Group. 6 2.70 Ravalli Group and Helena and Wallace Formations. 7 2.80 Crystalline basement. Model J-J' (extended J"-J'"') 1 2.80 Precambrian crystalline basement. 2 2.73 Cretaceous pluton of quartz monzonite porphyry. 3 2.73 Prichard Formation and sills. 4 2.73 Prichard Formation and sills. 5 2.67 Mainly Ravalli Group, Wallace Formation, and Missoula Group. 6 2.75 Mainly altered Wallace Formation and Missoula Group. 7 2.69 Mainly Ravalli Group, Wallace Formation, and Missoula Group. 8 2.64 Ravalli Group. 9 2.179 Ravalli Group and Prichard Formation. distribution have been suggested to account for the promi- nent gravity high associated with the Purcell anticlinorium. In a study of electrical conductivity and gravity data, I (in Wynn and others, 1977) modeled the anomaly as a basement uplift interpreted as the result of movement of major fault blocks. In a later study, Harrison and others (1980) con- cluded that the source of the anomaly beneath the Purcell anticlinorium is not a vertical uplift but very likely is a stack of thrust slices of high-density crystalline basement rocks that are elevated above the regional basement surface. Dur- ing the early 1980's, a popular theory favored among various petroleum explorationists attributed the source of the Purcell anomaly to high-density lower Paleozoic carbonate rocks that would provide attractive exploration targets. In a regional structural analysis, Fountain and McDonough (1984) postulated that the anomaly is caused by a high-den- sity body in the form of a basement ramp anticline beneath the Purcell anticlinorium. Harris (1985) modeled the Purcell gravity anomaly and attributed the source of the anomaly to mafic sills in the lower part of the Prichard Formation above a uniformly dipping crystalline basement surface. In this report, I refine, on the basis of further analysis and studies of the data, earlier interpretations in which the source of the 12 GEOPHYSICAL INTERPRETATIONS, LIBBY THRUST BELT, MONTANA B B' MILLIGALS Observed \. Calculated -120 -140 OBSERVED AND CALCULATED GRAVITY ELEVATION {IN KILOAQETERS) SEA LEVEL - GRAVITY MODEL ELEVATION Moyie SYLVANITE LIBBY (IN KILOMETERS) _/ ANTICLINE THRUST BELT SEA LEVEL | GEOLOGY 0 25 50 KILOMETERS L i 1 i i | | EXPLANATION - Helena and Wallace Formations Ravalli Group Prichard Formation and mafic sills Basement rocks Figure 4. Gravity block model of extended geologic cross section B-B' showing inferred lithology and structure beneath the Sylvanite anticline and northwestern part of the Purcell anticlinorium, Libby thrust belt, northwestern Montana. Numbers on gravity model refer to body number and assumed den- sity (in grams per cubic centimeter) as shown in table 1; for example, for body 2, the density is 2.73 g/cm3. GEOPHYSICAL INTERPRETATION OF DEEP-SEATED GEOLOGIC FEATURES 13 anomalies is thrust slices of dense crystalline basement rocks that core the Purcell anticlinorium. DEEP FOLDS AND FAULTS INTERPRETED FROM SEISMIC DATA By Jack E. Harrison Seismic data support the interpretation of crystalline basement beneath Prichard as shown in figure 6. The seismic data identify the mafic sills, which are excellent seismic reflectors (fig. 3), and the sills clearly mimic surface struc- ture to depths of about 8 km below ground level at which seismic data lose definition (Harrison and others, 1985). Seismic data were collected by others who discussed the data with us but preferred that they not be acknowledged. Two deep seismic lines of profile across the Sylvanite anticline and the Libby thrust belt were prepared by COCORP (The Consortium for Continental Reflection Profiling) in 1984-85 (Potter and others, 1986; Yoos and others, 1991). In 1986, graphs of the unmigrated data were generously made avail- able for our use and are shown here (fig. 6). Locations of the profiles MT-1 (Montana-1) and MT-2 (Montana-2) are shown on plate 1 and in figure 2, and interpretation of the profiles (generously aided by my colleague, M.W. Reynolds) is shown in figure 6. Most of the shallow reflectors (<3 seconds) relate reasonably well to surface geology, even though the seismic data are not migrated and therefore are not precisely in the correct geographic position. The mafic sills in the lower part of the Prichard Formation (fig. 3) are known seismic reflec- tors and are visible on both of the seismic profiles. The Pinkham thrust fault, as identified on profile MT-2, projects from the surface to the logical position and dip angle shown, a position that agrees well with that obtained from other data from about 40 km farther to the south in the Libby thrust belt. Of less certain origin are the reflections recorded below the Wallace Formation and the Missoula Group on profile MT-2 near Pipe Creek. These may be reflections from sills in the Prichard Formation, or they may be unmigrated reflections from the steep faults and bedding in the area. Below 3-seconds two-way travel time, the source of reflections is more speculative. The inferred basal surface of detachment is marked on profile MT-1 by a series of reflec- tions and extends readily in profile MT-2 to separate various different domains of reflections. The depth to the basal sur- face varies from about 9 to 18 km across the area, and the surface dips about 15° in inferred crystalline basement rocks. These values do not differ drastically from the depth and dip of inferred basement rocks under the Purcell anticlinorium as shown for the Rocky Mountains of southern British Colum- bia in cross sections by Price and Fermor (1985) and by Okulitch (1984, fig. 6). The strong reflections at and below 4 seconds (fig. 6) are, therefore, attributed to layered zones in the basement, perhaps mylonite, to the detachment zone for the older tectonic folding (?), or to strongly layered gneiss. No obvious reflection is present, at least along the profiles, for the Belt-basement contact. The lack of a base- ment reflection is most obvious on profile MT-1, where any reasonable thickness of the Prichard Formation requires that such a contact be at or above 4 seconds. Because the lowest part of the Prichard Formation in mid-Belt terrane is in the garnet zone of regional metamorphism and the Middle Prot- erozoic basement where drilled in western Montana is a mix- ture of granite, granitic gneiss, and some mafic intrusive rocks, a contact between metamorphosed Belt terrane and granite or granite gneiss would produce too little contrast to cause seismic reflections. A dissimilarity in basement rock types or structures is inferred from the contrast between the few deep seismic reflections above the Roderick Mountain thrust fault on profile MT-1 and the multiple prominent reflections below that thrust fault on profiles MT-1 and MT-2. None of the reflections are necessarily at the Belt-basement contact. A consequence of these observations is that the true thickness of the present Belt Supergroup, the thickest known sedimentary sequence of Middle Proterozoic rocks in the world, can only be determined by as yet undrilled deep boreholes. INTERPRETATION OF MAGNETOTELLURIC SsOUNDINGS By W.D. Stanley Extensive magnetotelluric surveys have been con- ducted across the Purcell anticlinorium and Libby thrust belt. Phoenix Geophysics supplied the data from a profile of five soundings (fig. 2) of a proprietary survey to me for inclusion in a study of regional structures. Another profile of the proprietary data, not shown in this report, is about 32 km south of the Phoenix profile and passes through the location of the Gibbs No. 1 borehole, a convenient geological calibra- tion point. The magnetotelluric soundings from the survey are of high quality and quite unusual in that a striking low-resistivity zone was detected on all the soundings. Ear- lier audiomagnetotelluric studies (Wynn and others, 1977; Long, 1983, 1988) demonstrate that in the Prichard Forma- tion a discontinuous zone, typically at depths of 0.9-2.5 km, is very conductive and has resistivities of a few ohm-meters or less. The vertical distribution of mineralized zones is indi- cated by the low resistivities in the resistivity log from the Gibbs No. 1 borehole (fig. 7). In the borehole, moderate amounts of sulfide minerals in members F, G, and H of the Prichard Formation cause resistivities of less than 200 GEOPHYSICAL INTERPRETATIONS, LIBBY THRUST BELT, MONTANA 14 I SH313W070 05 N NOLLWA313 :.Nu WNIHONNOLLNY Av. ajoya10q /9__:m Buoje W3LSAS NI) Lon sqq! wy g1 10s. LSNYHL FIAOW | 3 N 9 16 28 1138 LSNHHL A§8811 . Bent) uP iP § P T1300W ALIAVHS OL 13A31 v3S T (SH3L3WNO70X NID NOLLWAI13 _=Nv |_ Taam vas F--- +o 4+ r NI} NOLLWAI13 aP iP P ALIAVHS a31v¥in91vd any a3AH3s80 OSL cal SIvIMIA :_\v parenojey pemuesqo SIvOMIW uP iP P GEOPHYSICAL INTERPRETATION OF DEEP-SEATED GEOLOGIC FEATURES 153 EXPLANATION \ | Cretaceous granite Cambrian sedimentary rocks Missoula Group Missoula Group and Wallace Formation - Helena and Wallace Formations Ravalli Group dnouB1adng ag Prichard Formation + + { Basement rocks Figure 5 (above and facing page). Gravity block model of extended geologic cross section J-J' showing information about inferred lithology and structure beneath the Purcell an- ticlinorium, Libby thrust belt, northwestern Montana. Num- bers on gravity model refer to body number and assumed density (in grams per cubic centimeter) as shown in table 1; for example, for body 1, the density is 2.80 g/cm3. ohm-m, and greater amounts of iron sulfide minerals at about 3 km (member E) and 5 km (members A-D?) produce resistivities of 2-20 ohm-m. The low resistivities are attrib- uted to pyrite or pyrrhotite films and disseminated mineral- ized zones in the slightly carbonaceous argillitic unit. The low-resistivity zones in the Prichard Formation and other units of the Belt Supergroup represent easily mappable features for evaluation of mineralization in the Belt basin (Long, 1983). Interpretations of the five soundings provided by Phoenix Geophysics were used to construct the resistivity cross section shown in figure 8. The soundings are one-dimensional in character (isotropic) in that the resistivi- ties measured in the two coordinate directions of the field setup are very similar. The data were interpreted using inter- active forward models of four and five layers, and the cross section was constructed by connecting the layered models from each sounding location (fig. 8). Accuracy of the inter- faces obtained from the magnetotelluric data is not great, both because of the widely spaced sounding locations and because of the very large contrasts in resistivity between unmineralized and mineralized zones in the Belt Super- group. Layer interfaces should be resolved, however, to within 10-15 percent of their true depth, particularly the top of the lowest resistivity layer that represents the deepest layer in the models shown in the cross section. Most of the magnetotelluric data can be discussed in terms of two factors controlling resistivity in rocks of the Belt Supergroup-fluids and conductive mineralized rock. The 1,200-1,600-ohm-m layer beneath soundings 1 and 2 correlates with the Missoula Group, Helena and Wallace Formations, and Ravalli Group (fig. 8). These resistivities are typical for older carbonate and metasedimentary rocks that contain fresh pore waters, and they are characteristic of most Belt rocks above the Prichard Formation (Long, 1983). The 130-400-ohm-m units at the surface at soundings 3 and 4 are associated with the more mineralized conductive Prichard Formation, as are the conductive units (0.6-5 ohm-m) at soundings 1-4. The moderately low resistivity (250 ohm-m) in the Ravalli Group for the magnetotelluric model at sounding 5 is probably related to low-grade stra- tabound copper sulfide minerals, which are commonly present in the Spokane Formation, whereas the very low resistivities of 1.5 ohm-m for the model probably reflect the highly continuous iron sulfide zones in the Prichard Forma- tion. The thicknesses of the low-resistivity units that form the basal layer on the magnetotelluric models could not be uniquely determined from the magnetotelluric soundings, even though the data extended to frequencies of 0.0005 Hz; however, estimates based on the magnetotelluric phase information suggest that conductive units must be thicker than 7.7 km (with an error of about 0.1 km). The intermedi- ate layer of 20 ohm-m required at soundings 1 and 2 proba- bly represents a part of the Prichard Formation that contains moderate amounts of iron sulfide minerals. The upper boundary of this low-resistivity layer approximately matches the interpreted location of the Pinkham thrust fault, which places highly resistive rocks of the Ravalli Group and Helena and Wallace Formations over the Prichard Formation (fig. 8). Magnetotelluric data should be effective in deter- mining the lithology and depth to basement beneath Belt rocks, given a more extensive data set that would allow more accurate interpretations. The magnetotelluric data have limitations in structural analysis because resistivities are controlled mostly by the percentages of metallic minerals and not by rock lithology. The magnetotelluric data require that conductive rocks, probably of the Prichard Formation, extend to depths of more than 10.8 km (with an error of about 1.6 km) below magne- totelluric soundings 1, 2, and 3 and that a low-resistivity (3-5 ohm-m) zone forms the bottom layer on the western part of the profile (fig. 8), corresponding approximately with the interpreted basal surface of detachment that separates crys- talline basement units from the Prichard Formation. The lithology of the basement in the study area cannot be resolved using the magnetotelluric data. Outcrops and boreholes in the region indicate basement of possible gra- nitic gneiss, mafic intrusive rocks, and high-grade metasedi- mentary rocks. The Belt basin probably formed on stretched crust of Archean age that has a Middle Proterozoic (1,730 Ma) metamorphic overprint. This type of crust is not gener- ally low in resistivity unless partial melting has occurred. Feldman (1976) pointed out that low resistivities can be produced in granitic rocks at depths of 10-15 km and 16 GEOPHYSICAL INTERPRETATIONS, LIBBY THRUST BELT, MONTANA NORTHWEST SOUTHEAST 116° YaaklRiver 0 | 0 R EXPLANATION oa --- \ -=- ----- Zone of seismic reflectors- | i p k f Closer spacing of lines shows R \ more intense reflections | \ \ P | --- Thrust fault-Showing direction of movement. Double-headed & arrow indicates backsliding 3 | B - on fault o LLJ & Zz --- Geologic contact-Approximately < ] / located or inferred a ___ 3 <--- , Z 5-7 /// [-5 LJ > _ Ro G 05> z A t 92 «J» gt ACHN‘EN PINKHAM 2 oY, ce 062 / THRUST e w | E a C a g _ FAU PROFILE MT-1 <--------- LIBBY THRUST BELT --------> NORTHWEST SOUTHEAST Yaak River M Pipe 1Creek M Carrigan Campground West Side Highway I 0 0 p w / WL Z { roperick \ P \ MOUNTAIN T THRUST FAULT P M -W jun BASAL SURFACE TWO-WAY TRAVEL TIME, IN SECONDS | 4 OF DETACHMENT meme tue -- PROFILE MT-2 10 10 0 5 10 15 20 25 KILOMETERS Locale | 1 1 ] Figure 6. Interpretation of seismic reflection profiles MT-1 and MT-2, Libby thrust belt, northwestern Montana. Unmigrated data pro- vided at H:V=1:1 at 6 km/sec. Surface geologic data are generalized from plate 1 of this report; 12 high-angle faults are not shown to avoid clutter. Geologic units: P, Prichard Formation; R, Ravalli Group; W, Wallace Formation; H-W, Helena and Wallace Formations; M, Missoula Group; S, Proterozoic mafic sills; B, crystalline basement rocks. Lines of profiles are shown in figure 2. GEOPHYSICAL INTERPRETATION OF DEEP-SEATED GEOLOGIC FEATURES 17 pop ga uu i cg cial ig a cul ig aa ig ua 0 i a cual 1 j - EXPLANATION 5 a | E w S Z Clastic rocks 0.5 - 6 A -_ fo] l B een re L é | (a?) Schist and 1.0 - d g phyllite [a) _ 6 5 = a Ett m z _| --- ic si 15- 5 Mafic sills tE T Co car LL - © a A E 2.0 - g w z |__ & 6 i - C S 2 75 - © CC v v v g Q Jrac za y Fer rad, z E & a' L7? ”ALL; £ $ ® $3.0— 2 lal lal & o m pc- -_ £ ® - .'_,_h'_.l_f 3 35 - Pike dks AYA t iAP ao - <--- a a 45 - < £ ® a £ ® 3 5.0 - mmry m---rmm 0.2 2 20 200 2,000 20,000 200,000 RESISTIVITY, IN OHM-METERS Figure 7. - Generalized resistivity log (Dual Lateralog) of Gibbs No. 1 borehole, Flathead County, Montana. Electrical data courtesy of Atlantic Richfield and Marathon Oil Companies. Lithologic log simplified from figure 3. temperatures of 500°C-600°C if 1-3 percent free water is present, perhaps from metamorphic processes. Continued extension of the crust beneath the Belt during sedimentation is indicated by periodic intrusion of mafic dikes and intru- sions that have continental tholeiite affinities (Harrison and Cressman, 1993). Regional burial metamorphism reached the garnet grade of the greenschist facies (350°C or higher) in the lowest exposed Prichard during basin fill of at least 16 km of sediment and sills. Thus, there is no evidence in Belt rocks for temperatures of 500°C-600°C required to produce the type of conductors envisioned by Feldman (1976); how- ever, a thermally related basement conductive zone could possibly now be present at depth beneath magnetotelluric soundings 1, 2, and 3. Another possible line of thermal evi- dence is the absence of magnetic anomaly sources at depths of 10-15 km, presumed to be near the top of pre-Belt crys- talline rocks. This absence of magnetic sources may relate to shallow Curie-point temperatures, above which magnetism does not exist (Mabey and others, 1978). Curie temperatures are about the same order of magnitude as those suggested by Feldman (1976), 500°C-600°C. The extensive mafic volcanism and somewhat linear nature of the Belt-Purcell trend suggest that Belt rocks may have been deposited on pre-Belt rocks in a deep continental 18 GEOPHYSICAL INTERPRETATIONS, LIBBY THRUST BELT, MONTANA <--- LIBBY THRUST BELT £ 3 s C 1 "g ® 5 G ELEVATION 5 a Banfield (IN KILOMeTeRs) 2 & Mountain $. 1 SEA LEVEL __ |- PURCELL ANTICLINORIUM --------------> & 3 & fel 3 5 3 ELEVATION .S (In Kiometens) +o + o+ 15 to d Ob 40 40 $0 $ $ $ $ g og o 5 | E. i 1986 0 5 10 15 kilometers Geology by J.E. Harrison, 198 NO VERTICAL EXAGGERATION EXPLANATION Missoula Group Helena and Wallace Formations 116° 115° Ravalli Group | ae & 9 Prichard Formation 5 x4 O8 +00 + , o e Inferred basement rocks 48°30 «& - I-- Copper sulfide zone East edge of plate 1 27" 277 Zone of intermittent iron sulfide minerals | Conductive zone between sills 0 _ 10 _ 20 - 30 _ 40 _ 50 KILOMETERS in Prichard Formation R Index map showing location of magnetotelluric --- Thrust fault-Double-headed arrow indicates backsliding Contact-Inferred below 10,000 ft 130 =-- Interpreted boundary between magnetotelluric model units- Numbers are model resistivities in ohm-meters soundings in northwestern Montana Figure 8. Magnetotelluric and geologic section across the eastern part of the Libby thrust belt and the western part of the Purcell anticli- norium. Numbers indicate magnetotelluric soundings, the locations of which are shown in figure 2. margin rift analogous to the Animike Basin of Minnesota and Wisconsin (Larue, 1981) or in a linear continental rift or aulocogen similar to the Amadeus Basin of Australia (Wells and others, 1970). Both the Amadeus and Animike basins contain large volumes of mafic volcanic rocks and thick Proterozoic anoxic shales that have high metal content (Lawler and Vadis, 1986; Wells and others, 1970). Dense high-velocity rocks that represent basement to the gravity and seismic methods may actually be dense, volcanogenic metasedimentary rocks that have high metallic mineral concentrations, similar to rocks in the Animike and Amadeus basins. In addition, I speculate that the high con- ductivity of pre-Prichard crystalline basement in this area may be related to large amounts of metallic minerals. Thus, the magnetotelluric data can be used to postulate that units below the Belt Supergroup may be mineralized crystalline basment complex, unless temperatures are greater than 500°C-600°C. SUMMARY AND REFERENCES CITED 19 SUMMARY AND CONCLUSIONS Regional geophysical studies conducted by the U.S. Geological Survey in the northern Rocky Mountains during the past 25 years provide new insights about geologic frame- work and mineral resources. Conclusions presented in this report are based on integrating interpretations of these grav- ity, magnetic, seismic, and magnetotelluric data. The gravity anomaly data show a marked correlation with major structures. The Purcell anticlinorium exhibits positive anomalies of more than 20 mGal, and the Sylvanite anticline exhibits positive anomalies of more than 10 mGal. Results of gravity modeling indicate that the Purcell anticli- norium and the Sylvanite anticline are very likely cored by stacks of thrust slices of dense crystalline basement rocks that account for the large gravity highs across these two structures. The gravity anomaly data for the Cabinet Mountains Wilderness show a string of positive gravity anomalies about 6-8 mGal amplitude above background. The highs extend south-southeast in a belt from near the Kootenai River to as far as the latitude of Trout Creek. The gravity anomaly data and the distribution of magnetic anomalies suggest that the Cabinet Mountains Wilderness may be underlain by a large, uplifted granitic batholith. The gravity highs are interpreted to reflect intrusions along this feature, and the small positive magnetic anomalies are inferred to indicate near-surface sources that may be associated with small granitic intrusions or cupolas that are apophyses of the batholithic mass at depth, as shown in the gravity model. The principal magnetic anomaly sources are igneous intrusive rocks, major fault zones, and magnetite-bearing sedimentary rocks of the Ravalli Group. Although the sources mainly are present in outcrop, in some cases they are partly or totally buried in the near-surface. The most important magnetic anomalies in the principal study area are five distinct positive anomalies associated with Cretaceous or younger cupolas and stocks that are exposed or are known from drillholes to be present in the near-surface. These anomalies range in amplitude from 100 to more than 3,000 nT. Modeling of the magnetic anomaly data was not done because of the absence of long-wavelength magnetic anomalies having sources likely located in buried crystalline basement. The lack of detectable anomalies is attributed to the rel- atively nonmagnetic character and deep burial of the base- ment rocks or to possible elevated Curie isotherms, which are compatible with one explanation of low resistivities obtained from magnetotelluric measurements. The mafic sills of dioritic to gabbroic composition show little or no magnetic response in outcrop, probably because of a reduc- tion of magnetite content in the sills by chemical processes related to interaction between hydrothermal fluids and cooling magmas of the intrusions. Interpretation of seismic reflection data for the study area shows an inferred basal surface of detachment at depths of 9-18 km. Strong reflections at and below 4 seconds are attributed to stratification in the basement such as layered gneiss or mylonite along the detachment zone for the older tectonic folding. No obvious reflection is present, at least along the profiles, for the Belt-basement contact. Just below the Pinkham thrust fault, strong reflections attributed to stratification in crystalline basement rocks image relief in the basement on the western flank of the Purcell anticline. Interpretation of magnetotelluric data shows a thick conductive basal layer that is incompatible with a crystalline basement composed of granitic gneiss, mafic intrusive rock, and high-grade metasedimentary rocks. This type of crust is not generally low in resistivity unless partial melting has occurred, perhaps at depths of 10-15 km and temperatures of 500°C-600°C if 1-3 percent free water is present. A pos- sible explanation for a heat source at these depths is the lack of magnetic sources, which may indicate shallow Curie tem- peratures, as previously postulated. An alternative explana- tion is that pre-Prichard crystalline rocks are conductive simply because they contain large amounts of metallic min- erals. This is certainly the case for the Prichard Formation because surface observations of variations in resistivity range from 0.6 to more than 400 ohm-m and most likely reflect the percentage of iron sulfide minerals and continuity of iron sulfide zones. Accuracy of the interfaces obtained from the magnetotelluric data is not great, both because of the widely spaced sounding locations and because of the very large contrasts in resistivity between unmineralized and mineralized zones in the Belt Supergroup. REFERENCES CITED Bankey, Viki, Kleinkopf, M.D., and Hoover, Donald, 1986, Geo- physical maps of the Mount Henry roadless area, Lincoln County, Montana: U.S. Geological Survey Miscellaneous Field Studies Map MF-1534-C, scale 1:50,000. Bankey, V.L., Brickey, MR., and Kleinkopf, M.D., 1982, Principal facts for 1980 gravity stations in the Wallace 1°x2° quadran- gle: U.S. Geological Survey Open-File Report 82-132. Bankey, V.L., Kleinkopf, M.D., Brickey, MR., Mancinelli, J.A., and Kulik, D.M., 1985, Gravity survey data of the Kalispell 1°x2° quadrangle, Montana: - U.S. Geological Survey Open-File Report 85-221. Bishop, D.T., 1973, Petrology and geochemistry of the Purcell sills in Boundary County, Idaho, in Belt Symposium, v. 2: Idaho Bureau of Mines and Geology Special Publication, p. 15-66. Boettcher, A.L., 1966, The Rainy Creek igneous complex near Libby, Montana: University Park, Pennsylvania State University, Ph.D. thesis, 155 p. 20 GEOPHYSICAL INTERPRETATIONS, LIBBY THRUST BELT, MONTANA Brickey, MR., Bankey, V.L., and Kleinkopf, M.D., 1980, Principal facts for gravity stations in part of the Wallace 1°x2° quadran- gle, Idaho and Montana: U.S. Geological Survey Open-File Report 81-178. Briggs, I.C., 1974, Machine contouring using minimum curvature: Geophysics, v. 39, p. 39-40. Burmester, RF., 1985, Preliminary geologic map of the Eastport area, Idaho and Montana: U.S. Geological Survey Open-File Report 85-517, scale 1:48,000. Cordell, Lindrith, Keller, G.R., and Hildenbrand, T.G., 1982, Bouguer gravity map of the Rio Grande rift, Colorado, New Mexico, and Texas: U.S. Geological Survey Geophysical Investigations Series Map GP-949, scale 1:1,000,000. Fabiano, E.B., and Peddie, N.W., 1980, Magnetic declination in the United States-Epoch 1980: U.S. Geological Survey Miscella- neous Investigation Series Map I-1283, scale 1:5,000,000. Fabiano, E.B., Peddie, N.W., and Jones, W.J., 1976, Magnetic total intensity in the United States-Epoch 1975.0: U.S. Geological Survey Miscellaneous Investigations Map 1-915, scale 1:5,000,000, 2 sheets. Feldman, I.S., 1976, On the nature of conductive layers in the Earth's crust and upper mantle, in Adam, A., ed., Geoelectrical and geothermal studies: Budapest, Akademiai Kaido, p- 721-730. Fountain, D.M., and McDonough, D., 1984, Crustal structure of northwestern Montana and southern Alberta and British Columbia based on the interpretation of gravity data: Montana Geological Society Field Conference and Symposium, North- western Montana and adjacent Canada, Guidebook, p. 253-258. Gibson, Russell, 1948, Geology and ore deposits of the Libby quad- rangle, Montana, with sections on Pleistocene glaciation by W.C. Alden and Physiography by J.T. Pardee: U.S. Geological Survey Bulletin 956, 131 p., 2 pl., scales 1:125,000 and 1:250,000. Harris, D.W., 1985, Crustal structure of northwestern Montana: Missoula, University of Montana, M.S. thesis, 59 p. Harrison, J.E., 1972, Precambrian Belt basin of northwestern United States-Its geometry, sedimentation, and copper occurrences: Geological Society of America Bulletin, v. 83, no. 5, p. 1215-1240. Harrison, J.E., and Cressman, E.R., 1993, Geology of the Libby thrust belt of northwestern Montana and its implications to regional tectonics: U.S. Geological Survey Professional Paper 1524, scale 1:1 25,000. Harrison, J.E., Cressman, E.R., and Kleinkopf, M.D., 1985, Regional structure, the Atlantic Richfield-Marathon Oil No. 1 Gibbs borehole, and hydrocarbon resource potential west of the Rocky Mountain trench in northwestern Montana: U.S. Geo- logical Survey Open-File Report 85-249, 5 p. Harrison, J.E., Cressman, E.R., and Whipple, J. W., 1992, Geologic and structure maps of the Kalispell 1°x2° quadrangle, Montana and Alberta and British Columbia: U.S. Geological Survey Miscellaneous Investigations Series Map I-2267, scale 1:250,000, 2 sheets. Harrison, J.E., Griggs, A.B., and Wells, J.D., 1986, Geologic and structure maps of the Wallace 1°x2° quadrangle, Montana and Idaho: U.S. Geological Survey Miscellaneous Investigations Series Map I-1509-A, scale 1:250,000, 2 sheets. Harrison, J.E., Kleinkopf, M.D., and Obradovich, J.D., 1972, Tectonic events at the intersection between the Hope fault and the Purcell trench, northern Idaho: U.S. Geological Survey Pro- fessional Paper 719, 24 p. Harrison, J.E., Kleinkopf, M.D., and Wells, J.D., 1980, Phanerozoic thrusting in Protozeroic belt rocks, northwestern United States: Geology, v. 8, p. 407-411. Hobbs, SW., Griggs, A.B., Wallace, RE., and Campbell, A.B., 1965, Geology of the Coeur d' Alene district, Shoshone County, Idaho: U.S. Geological Survey Professional Paper 478, 139 p. International Association of Geodesy, 1967, Geodetic reference system, 1967: International Association of Geodesy Special Publication 3, 74 p. King, ER., Harrison, J.E., and Griggs, A.B., 1970, Geologic implications of aeromagnetic data in the Pend Oreille area, Idaho-Montana: U.S. Geological Survey Professional Paper 646-D, 17 p. King, P.B., 1969, Tectonic map of North America: U.S. Geological Survey, scale 1:5,000,000, 2 sheets. Kleinkopf, M.D., 1977, Regional geophysical studies of the north- ern belt basin, Montana: Geological Society of America, Rocky Mountain Section, Abstracts with Programs, v. 9, no. 5, p. 738. 1981, Geophysical surveys of the Cabinet Mountains Wil- derness, Lincoln and Sanders Counties, Montana, in Mineral resources of the Cabinet Mountains Wilderness, Lincoln and Sanders Counties, Montana: U.S. Geological Survey Bulletin 1501, 77 p, scale 1:62,500. 1983, Gravity expression of the Rocky Mountain Trench, in Geologic and structural framework and minerals and energy resources of northwest Montana: Tobacco Root Geological Society Field Conference, Kalispell, Montana, 1983, Abstracts, p. 13-14. 1984, Gravity and magnetic anomalies of the Belt Basin, United States and Canada: Belt Symposium, 2nd, Abstracts with Programs; Montana Bureau of Mines and Geology Special Publication 90, p. 85-87. Kleinkopf, M.D., and Bankey, V.L., 1982, Gravity studies of regional structures in the northern Rocky Mountains: Geologi- cal Society of America Annual Meeting, 95th, New Orleans, Abstracts with Programs, v. 14, no. 7, p. 531. Kleinkopf, M.D., and Harrison, J.E., 1982, The application of aeromagnetic and gravity anomaly maps in mineral resource assessments of the Wallace quadrangle, Northern Rocky Mountains [abs.]: Society of Exploration Geophysicists Annual Meeting, 52nd Dallas, Technical Program Abstracts and Biographies, v. 48, p. 454. Kleinkopf, M.D., Harrison, J.E., and Bankey, V.L., 1982, Slides showing aeromagnetic and gravity anomaly maps and a metallic mineral resource favorability map of the Wallace 1°x2° quadrangle, Montana and Idaho: U.S. Geological Sur- vey Open-File Report 82-709. 1988, Aeromagnetic and gravity anomaly maps and a hydrothermal veins favorability map of the Wallace 1°x2° quadrangle, Montana and Idaho: U.S. Geological Survey Mis- cellaneous Field Studies Map MF-1354-E, 2 sheets, 1:250,000. Kleinkopf, M.D., Harrrison, J.E., and Zartman, RE., 1972, Aero- magnetic and geologic map of part of northwestern Montana SUMMARY AND REFERENCES CITED 21 and northern Idaho: U.S. Geological Survey Geophysical Investigations Map GP-830, scale 1:250,000. Kleinkopf, M.D., and Long, C.L., 1979, Geophysical studies in the Wallace quadrangle, northern belt basin, Montana and Idaho [abs.]: Northwest Mining Association Annual Convention, 85th, Spokane, Washington, Program, p. 15. Kleinkopf, M.D., and Wilson, D.M., 1981, Aeromagnetic and gravity studies of the Scotchman Peak Wilderness Study Area, Lincoln and Sanders Counties, Montana, and Bonner County, Idaho, in Mineral resources of the Scotchman Peak Wilderness Study Area, Lincoln and Sanders Counties, Montana, and Bon- ner County, Idaho: U.S. Geological Survey Bulletin 1467, 26 p. Larsen, E.S., and Pardee, J.T., 1929, The stock of alkaline rocks near Libby, Montana: Journal of Geology, v. 37, no. 2, p. 97-112. Larue, D.K., 1981, The Early Proterozoic Pre-Iron Formation Menominee Group siliclastic sediments of the southern Lake Superior region-Evidence for sedimentation in platform and basinal settings: Journal of Sedimentary Petrology, v. 51, no. 2, p. 397-414. Lawler, TL., and Vadis, M., 1986, Mineral potential of iron-rich volcanogenic metasediments, Glen Township, east-central Minnesota: Annual Institute on Lake Superior Geology, 32nd, p. 49. Long, C.L., 1983, Maps showing results of audio-magnetotelluric studies in the northwestern part of the Wallace 1°%2° quadran- gle, Montana and Idaho: U.S. Geological Survey Miscella- neous Field Studies Map MF-1354-B, scale 1:250,000. 1988, Interpretations of two audio-magnetotelluric profiles across the Sylvanite anticline, Idaho and Montana: U.S. Geo- logical Survey Open-File Report 88-658. Mabey, D.R., Zietz, Isidore, Eaton, G.P., and Kleinkopf, M.D., 1978, Regional magnetic patterns in part of the Cordillera in the western United States, in Cenozoic tectonic and regional geophysics of the western Cordillera: Geological Society of America Memoir 152, p. 993-106. Marvin, RF., Zartman, RE., Obradovich, J.D., and Harrison, J.E., 1984, Geochronometric and lead isotope data on samples from the Wallace 1°x2° quadrangle, Montana and Idaho: U.S. Geo- logical Survey Miscellaneous Field Studies Map MF-1354-G, scale 1:250,000. Morelli, Carlo, Gantar, C., Honkasla, Tauno, McKonnel, R.K., Tan- ner, J.G., Szabo, Bela, Uotila, U.A., and Whalen, G.T., 1974, The international gravity standardization net 1972 (I.G.S.N.): Paris, Bureau Central de I' Association Internationale de Geod- esie Special Publications 4, 194 p. Okulitch, A.V., 1984, The role of the Shuswap Metamorphic Com- plex in Cordilleran tectonism-A review: Canadian Journal of Earth Sciences, v. 21, no. 10, p. 1171-1193. Parasnis, D.S., 1968, A supplementary comment on R. Green's Note, "The EAEG'S Choice of Units": Geophysical Prospect- ing, v. 16, p. 392-393. Peddie, N.W., Jones, W.J., and Fabiano, E.B., 1976, Magnetic incli- nation in the United States-Epoch 1975.0; U.S. Geological Survey Miscellaneous Investigations Series Map I-912, scale 1:5,000,000, 2 sheets. Plouff, Donald, 1977, Preliminary documentation for a Fortran program to compute gravity terrain corrections based on topog- raphy digitized on a geographic grid: U.S. Geological Survey Open-File Report 77-535, 43 p. Potter. C.J., Sanford, W.E., Yoos, TR., Prussen, E.I., Keach, RW., Oliver, J.E., Kaufman, S., and Brown, LD., 1986, CORCORP deep seismic reflection traverse of the interior of the North American Cordillera, Washington and Idaho-Implications for orogenic evolution: Tectonics, v. 5, p. 1007-1026. Price, R.A., 1981, The Cordilleran foreland fold and thrust belt in the southern Canadian Rocky Mountains, in McClay, K.R., and Price, NJ., eds., Thrust and nappe tectonics: Geological Society of London Special Publication 90, p. 47-48. Price, R.A., and Fermor, P.R. 1985, Structure section of the Cordil- leran foreland thrust and fold belt west of Calgary, Alberta: Geological Society of Canada Special Paper 84-14. Reynolds, M.W., and Kleinkopf, M.D., 1977, The Lewis and Clark line, Montana-Idaho-A major intraplate tectonic boundary: Geological Society of America Abstracts with Programs, v. 9, no. 7, p. 1140-1141. Smithson, S.B., 1971, Densities of metamorphic rocks: Geophysics, v. 36, no. 4, p. 690-694. Sweeney, RE., 1990, IGRFGRID-A program for creation of a total magnetic field (international geomagnetic reference field) grid representing the Earth's main magnetic field: U.S. Geo- logical Survey Open-File Report 90-45, 37 p. U.S. Department of Defense, 1974, World relative gravity reference network, pt. 2, North America: Defense Mapping Agency Aerospace Center Reference Publication 25, v. 2, with supple- ment updating gravity values to the International Gravity Stan- dardization Net (IGSN), 1,635 p. U.S. Geological Survey, 1969a, Aeromagnetic map of the Libby and Mt. Pend Oreille quadrangles, Lincoln and Sanders Coun- ties, Montana, and Bonner County, Idaho: U.S. Geological Sur- vey Geophysical Investigations Map GP-682, scale 1:62,500. 1969b, Aeromagnetic map of the Thompson Lakes quadran- gle, Lincoln, Sanders, and Flathead Counties, Montana: U.S. Geological Survey Geophysical Investigations Map GP-683, scale 1:62,500. 1969c, Aeromagnetic map of the McGregor Lake-Tally Lake area, Flathead and Lincoln Counties, Montana: U.S. Geo- logical Survey Geophyscial Investigations Map GP-684, scale 1:62,500. 1969d, Aeromagnetic map of the Trout Creek quadrangle, Sanders and Lincoln Counties, Montana, and Shoshone Coun- ty, Idaho: U.S. Geological Survey Geophysical Investigations Map GP-685, scale 1:62,500. 1969e, Aeromagnetic map of the Thompson Falls quadran- gle, Lincoln and Sanders Counties, Montana: U.S. Geological Survey Geophysical Investigations Map GP-686, scale 1:62,500. 1969f, Aeromagnetic map of the Hubbart Reservoir-Hot Springs area, Sanders, Flathead, and Lakes Counties, Montana: U.S. Geological Survey Geophysical Investigations Map GP-687, scale 1:62,500. 1969g, Aeromagnetic map of the Kingston, Kellogg, and part of the Fernwood quadrangles, Shoshone, Benewah, and Kootenai Counties, Idaho: U.S. Geological Survey Geophysi- cal Investigations Map GP-688, scale 1:62,500. 1969h, Aeromagnetic map of the Avery quadrangle, Shos- hone County, Idaho, and Mineral and Sanders Counties, Mon- tana: U.S. Geological Survey Geophysical Investigations Map GP-689, scale 1:62,500. 22 GEOPHYSICAL INTERPRETATIONS, LIBBY THRUST BELT, MONTANA 19691, Aeromagnetic map of the Haughan and St Regis quadrangles and parts of the Simmons Peak and Illinois Peak quadrangles, Shonsone County, Idaho, and Mineral and Sanders Counties, Montana: U.S. Geological Survey Geo- physical Investigations Map GP-690, scale 1:62,500. 1969;, Aeromagnetic map of the Plains, Perma, Superior, and Tarkio quadrangles, Sanders, Mineral, and Missoula, Counties, Montana: U.S. Geological Survey Geophysical Investigations Map GP-691, scale 1:62,500. 1973, Aeromagnetic map of parts of the Okanogan and Sandpoint 1°%2° quadrangles, Washington-Idaho-Montana: U.S. Geological Survey Open-File Map 294, scale 1:250,000. Webring, M.W., 1981, MINC-A gridding program based on min- imum curvature: U.S. Geological Survey Open-File Report 81-1230, 12 p. 1982, SAKI-A Fortran program for generalized linear inversion of gravity and magnetic profiles: U.S. Geological Survey Open-File Report 82-122, 29 p. Published in the Central Region, Denver, Colorado Manuscript approved for publication January 11, 1994 Edited by Judith Stoeser Graphics by Dennis Welp, Nancy Shock, R. Randall Schumann and Gayle Dumonceaux Photocomposition by Gayle Dumonceaux ¥ U.S. GOVERNMENT PRINTING OFFICE: 1997 - 573-046 / 20080 REGION NO. 8 Wells, A.T., Forman, D.J., Ranford, LC., and Cook, PI., 1970, Geology of the Amadeus Basin, central Australia: Australian Bureau of Mineral Resources, Geology and Geophysics Bul- letin 100, 222 p. Wells, J.D., Lindsey, D.A., and Van Loenen, RE., 1981, Geology of the Cabinet Mountains Wilderness, Lincoln and Sanders Counties, Montana: U.S. Geological Survey Bulletin 1501-A, p. 9-19. Wynn, J.C., Kleinkopf, M.D., and Harrison, J.E., 1977, An audio-frequency magnetotelluric and gravity traverse across the crest of the Purcell geanticline, northwestern Montana: Geology, v. 5, p. 309-312. Yoos, TR., Potter, C.J., Thigpen, J.L., and Brown, L.D., 1991, The Cordilleran fold and thrust belt in northwest Montana and northern Idaho from COCORP and industry seismic reflection data: American Association of Petroleum Geologists Bulletin, v. 75, no. 6, p. 1089-1106. g e ogg : a ar gg grease U.S. DEPARTMENT OF THE INTERIOR U.S. GEOLOGICAL SURVEY 116°00° 115°30° 115°00° 49°00° 49°00" + Eureka % i 48°30° 48°00° 47°30 t 1 7 Lambert Conformal Conic Projection. SCALE 1:500 000 Central meridian longitude 115°30" W., base latitude 47° N. 19—4 -=- C 0( PC 2G MILES 10 0 10 20 KILOMETERS <_ -~ MAP A. RESIDUAL TOTAL-INTENSITY AEROMAGNETIC ANOMALIES AND STRUCTURAL FEATURES 47°30° EXPLANATION FOR MAPS A AND B Kg Ks Kps -la __*_ 4 sp: Felsic intrusive rocks (Cretaceous) Syenite (Cretaceous) Pyroxenite and syenite (Cretaceous) Sill (Late and Middle Proterozoic) Fault Normal fault-Bar and ball on down- thrown side Thrust faullt-Sawteeth on upper plate Tear fault-Arrows show direction of horizontal movement Anticline-Showing crestline and dips of limbs Syncline-Showing troughline and dips of limbs Aeromagnetic anomaly referred to by number in text Location of Atlantic Richfield-Marathon Oil No. 1 Gibbs borehole, NE1/4 sec. 2, T. 28 N., R. 27 W., Flathead County, Montana Color bar showing residual total-intensity aeromagnetic intensity in 10, 20, and 30 nT intervals 440.0 410.0 380.0 350.0 320.0 290.0 260.0 230.0 200.0 170.0 140.0 110.0 90.0 80.0 70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0 -10.0 -20.0 ~30.0 -40.0 -50.0 -10.0 -100.0 -130.0 -160.0 Missoula MONTANA INDEX SHOWING LOCATION OF MAPPED AREA 49°00" 48°30° 48°00° 47°30° PROFESSIONAL PAPER 1546 PEATE 1 116°00° 115°30" 115°00° 49°00° 48°30° Color bar showing complete Bouguer gravity intensity in 2-mGal intervals -110.0 =~112.0 =114.0 -116.0 | * &: C & 3. O *C No: 1 Gibbs " -118.0 ~120.0 | -122.0 £719 fie,- -124.0 -126.0 -128.0 -130.0 48°00° -132.0 -134.0 -136.0 -138.0 -140.0 -142.0 -144.0 -146.0 -148.0 150.0 | -152.0 -154.0 - ow -156.0 (8s -158.0 ban.. 160.0 -162.0 930, -164.0 ; -166.0 Placer -168.0 : -170.0 -172.0 Lambert Conformal Conic projection. SCALE 1:500 000 Central meridian longitude 115°30" W., base latitude 47° N. Ag ope R 1C o MICES 10 0 10 20 KILOMETERS E- - MAP B. COMPLETE BOUGUER GRAVITY ANOMALIES AND STRUCTURAL FEATURES MAPS SHOWING RESIDUAL TOTAL-INTENSITY AEROMAGNETIC ANOMALIES, COMPLETE BOUGUER GRAVITY ANOMALIES, GEOLOGY, AND STRUCTURAL FEATURES, LIBBY THRUST BELT, NORTHWESTERN MONTANA AND NORTHEASTERN IDAHO # U.S. GOVERNMENT PRINTING OFFICE 1997-573-066 By M. Dean Kleinkopf 1997 oC U.S. DEPARTMENT OF THE INTERIOR PROFESSIONAL PAPER 1546 ) U.S. GEOLOGICAL SURVEY cip PLATE 2 C ¥ S 49°00" 116°00 115°30 115°00 114°30° EXPLANATION FOR GEOLOGIC MAPS ( y canapa _ fc e File Aa aon .. --.. asi > .." $ 3s Alluvial, landslide, and glacial and fluvioglacial i s- alt § ra? ++, ** 31.3.0}; § a * rd B 3.0 H”:- sq .a deposits and lake sediments undivided & 5'- o as A 4 R ‘Mz- as* (Quaternary) gs tes. 'P g? az g -x-x-x- Dike (Tertiary?) \ R . f ,if ay in? | Felsic plutonic rocks (Cretaceous) * : was t as. I Téa B e | a > "t., !, to, . oan F - Pyroxenite and syenite (Cretaceous) : A ¢ " "@gs ‘L £7 A U ® '. ® .' ® .. Sedimentary rocks undivided (Cambrian) 1| * e *> Cit * A *; §. | Mafic sills (Late and Middle Proterozoic) -__- _| - \\ @., py . Tir ' *, ® B *. Mafic sill (Late and Middle Proterozoic) f : oy: . ~* ', * ** ' A 1 s Ail ts a Add y % * : * I ------ Mafic sill (Middle Proterozoic) 48°30° - © g sp et Yo of "3, *: Missoula Group of the Belt Supergroup (Middle gig 313“ i t 3s ne A ama Proterozoic) 18 _. "a * =i ff . as cls ___ _| Helena and Wallace Formations of the Belt a heil 2 % a * * e Supergroup (Middle Proterozoic) : 42; j} s F Tsk ' % Ravalli Group of the Belt Supergroup (Middle [' gee" a a) «Wt g wite eng ol r-- Proterozoic) il "y's t los is A ts 3s __| Pritchard Formation of the Belt Supergroup (Middle nosey tha, stan, - l g si $ is * f Proterozoic) s mat will si. st ous 1. A... 3. + B t/ 3 he ~. & p *; % ... #4 '.. *** . p. a .."’,; s Contact-Dashed where concealed | (" 1: ies c.,! sy ' yt * ‘1 % A tol d» « wee }* A .s . * * w $%. ** ® ------ Normal fault-Dotted where concealed; bar on Joo! L . tA : * «34 2 downthrown side } S al *~, pusu. +1 * £ s # 5, < Tear fault-Arrows show direction of apparent strike slip \ ga "o s, t uf $ . *, .t oy: *% 9... Pe .:.: jar -*-*-*- Thrust fault-Dotted where concealed; sawteeth on s 33 fa sr f "." ., maw upper plate nts \s $y 9a]? s*, "* ‘ "Js.. *A i244 & -a&-&-- Thrust fault with later normal throw-Dotted where * \ "~ \ & ** "Yy. £ *s. "2." s concealed; sawteeth on upper plate; bars on « N a % s, I LE. f downthrown side 2 td .! 'I 4 Si: s" --- Anticline-Showing crestline and dips of limbs \ s € y " in: Pac eq "*. ® oal "es --- Overturned anticline-Showing direction of dip of limbs < al Ty antl "¥" f; 94 **. *s. U g * & * ..‘ * ® ---f-- Syncline-Showing troughline and dips of limbs af to'. ""* att k L 5 *s * ** 4 & ? © * tis. s* F J ® * * --- Overturned syncline-Showing direction of dips of limbs a Tals . 8 & bd G Xs § 'i. _s « ---S-- Monoclinal fold sys _ f & MT-1 f \ «wes = > * ----t+- Lines of seismic reflection profiles MT-1 and MT-2- % 2 > f Interpretation of profiles is shown in text figure 6 " Fe , 2 Aeromagnetic anomaly referred to in text | ccf % | -l 0 10 20 30 MILES Mears eg .-. T -' 0 10 20 30 40 KILOMETERS a) afs. LOCATION OF GRAVITY STATIONS g9. c- _ a_ tirana "|" a T l 147 geal o e 2 NTA Male % :o CR |_ C 3). «-> -men| or. p P - >- Y & 4 a 4: says C ;“*»?“ i X r I I grit} \, p tik - TM UNIN I it/ Ssact \)32/)}€:33\\‘ (1 fibyiii/Ql‘x \Xg(\‘<: ib 7 3&fi1 nutay {3 ..... C at X gn A 0 0 0 00000 >>of N WDF. T° °C © FNX Vey " J $ 19 < 106° Transverse Mercator Projection haan Mgrcator Projection. I Central meridian lon ituée 115500' W wis mendmn oud t nn (eee rime at- N 9 & base latitude 45° N. EXPLANATION FOR MAP A Aeromagnetic surveys were flown east-west at 7,000 ft (2.15 km) baro- metric elevation. The survey north of lat 48°30' N. was flown in 1972 by Scintrex Minerals Inc. (U.S. Geological Survey, 1973) and had a line spacing of 3.2 km; the survey south of lat 48°30' N. was flown in 1968 by Lockwood, Kessler, and Bartlett, Inc. and had a line spacing of 1.6 km (U.S. Geological Survey, 1969a-7) Contours of magnetic intensity-Magnetic contours show total-intensity magnetic field of Earth relative to arbitrary datum. - Hachured contours enclose areas of low magnetic intensity. Contours are dashed where data is incomplete. Contour intervals 10, 20, 100, and 500 EXPLANATION FOR MAP B nanoteslas (nT) in International System of Units (SI units). - Nanoteslas are equivalent to gammas in Contours of gravity intensity-Hachured centimeter-gram-second system (Parasnis, 1968). contours enclose areas of low gravity. Parameters of Earth's main magnetic field are as follows: Contour interval 2 milligals (mGal). Regional geomagnetic gradient (approximated by Density of 2.67 g/cm assumed in International Geomagnetic Reference Field, Epoch, reducing the data to complete 1975) rises northeasterly across Libby thrust belt at Bouguer anomaly about 4.5 nT/km. Total-intensity magnetic field is about 58,000 nT, inclination is about 72°, and declination is * 1 about 19° E. (Fabiano and others, 1976; Peddie and ( others, 1976; - Fabiano and Peddie, 1980) < med v + Q is ..'.. 2":- ii" at : | ..... 4 - 47°30 < © 200g »*~ MAP B. COMPLETE BOUGUER GRAVITY ANOMALIES AND xf" (8 Geology was generalized by combining appropriate digital geologic units by Harrison < . % / and Cressman (1993); compilation and preliminary design of digital map by Stanton f "n, H. Moll using ARC/INFO version 4.2 software running on a DEC VAX 11/780 Bouguer gravity anomaly from data collected € a computer; final design and production of map by Nancy Shock and Randall and compiled by U.S. Geological Survey. Schumann, using ARC/INFO version 7.0. Line elements, structural features, and geologic units were digitized and encoded with numerical values by Gayla Evans and Mary Weinheimer. Graphic design by Denny Welp SCALE 1:250 000 5 0 5 10 15 20 MILES hema e- Ped . < 5 0 5 10 15 20 KILOMETERS E- nnn NATIONAL GEODETIC VERTICAL DATUM OF 1929 1997 MAGNETIC DECLINATION FROM TRUE NORTH VARIES FROM 18° EASTERLY FOR THE CENTER OF THE WEST EDGE TO 17.5° EASTERLY FOR THE CENTER OF THE EAST EDGE MAPS SHOWING TOTAL-INTENSITY AEROMAGNETIC ANOMALIES, BOUGUER GRAVITY ANOMALIES, AND GENERALIZED GEOLOGY, LIBBY THRUST BELT, NORTHWESTERN MONTANA By M. Dean Kleinkopf # U.S. GOVERNMENT PRINTING OFFICE 1997-573-066 1 997 £ ,V\\ ington r, Washi Sedimentology, Behavior, and Hazards of Debris Flows at inie Mount Ra 547 sika _ U.S. GEOLOGICAL SURVEY PROFESSIONAL PAPER 1547 AVAILABILITY OF BOOKS AND MAPS OF THE U.S. GEOLOGICAL SURVEY Instructions on ordering publications of the U.S. Geological Survey, along with prices of the last offerings, are given in the current-year issues of the monthly catalog "New Publications of the U.S. Geological Survey." Prices of available U.S. Geological Survey publications re- leased prior to the current year are listed in the most recent annual "Price and Availability List." Publications that may be listed in various U.S. Geological Survey catalogs (see back inside cover) but not listed in the most recent annual "Price and Availability List" may no longer be available. Reports released through the NTIS may be obtained by writing to the National Technical Information Service, U.S. Department of Commerce, Springfield, VA 22161; please include NTIS report number with inquiry. Order U.S. Geological Survey publications by mail or over the counter from the offices listed below. BY MAIL Books Professional Papers, Bulletins, Water-Supply Papers, Tech- niques of Water-Resources Investigations, Circulars, publications of general interest (such as leaflets, pamphlets, booklets), single copies of Earthquakes & Volcanoes, Preliminary Determination of Epicen- ters, and some miscellaneous reports, including some of the forego- ing series that have gone out of print at the Superintendent of Documents, are obtainable by mail from U.S. Geological Survey, Information Services Box 25286, Federal Center Denver, CO 80225 Subscriptions to periodicals (Earthquakes & Volcanoes and Preliminary Determination of Epicenters) can be obtained ONLY from the Superintendent of Documents Government Printing Office Washington, DC 20402 (Check or money order must be payable to Superintendent of Documents.) Maps For maps, address mail orders to U.S. Geological Survey, Information Services Box 25286, Federal Center Denver, CO 80225 Residents of Alaska may order maps from U.S. Geological Survey, Earth Science Information Center 101 Twelfth Ave., Box 12 Fairbanks, AK 99701 OVER THE COUNTER Books and Maps Books and maps of the U.S. Geological Survey are available over the counter at the following U.S. Geological Survey offices, all of which are authorized agents of the Superintendent of Documents. +_ ANCHORAGE, Alaska-Rm. 101, 4230 University Dr. * LAKEWOOD, Colorado-Federal Center, Bldg. 810 +_ MENLO PARK, California-Bldg. 3, Rm. 3128, 345 Middlefield Rd. * - RESTON, Virginia-USGS National Center, Rm. 1C402, 12201 Sunrise Valley Dr. * SALT LAKE CITY, Utah-Federal Bldg., Rm. 8105, 125 South State St. * SPOKANE, Washington-U.S. Post Office Bldg., Rm. 135, West 904 Riverside Ave. * _ WASHINGTON, D.C.-Main Interior Bldg., Rm. 2650, 18th and C Sts., NW. Maps Only Maps may be purchased over the counter at the following U.S. Geological Survey offices: +_ FAIRBANKS, Alaska-New Federal Bldg, 101 Twelfth Ave. +_ ROLLA, Missouri-1400 Independence Rd. * _ STENNIS SPACE CENTER, Mississippi-Bldg. 3101 U. S. DEPOSITORY Jun 2 6 1995 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON Frontispiece. View of south side of Mount Rainier near the end of the 1987 drought; showing upper Kautz Glacier, right of center, and tributary Success Glacier, at lower left, separated by Kautz Cleaver. Sunset Amphitheater, far upper left, is above heads of Puyallup and Tahoma Glaciers. Note the volcanic edifice composed of steeply outward- dipping lava and pyroclastic flow units. Hydrothermal alteration has been intense along many of the stratigraphic contacts, converting them to potential planes of failure. Photograph by R. M. Krimmel, October 5, 1987. SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON By K.M. Scott, J.W. Vallance, and P.T. Pringle U.S. GEOLOGICAL SURVEY PROFESSIONAL PAPER 1547 UNITED STATES GOVERNMENT PRINTING OFFICE, WASHINGTON : 1995 U.S. DEPARTMENT OF THE INTERIOR BRUCE BABBITT, Secretary U.S. GEOLOGICAL SURVEY Gordon P. Eaton, Director For sale by U.S. Geological Survey, Information Services Box 25286, Federal Center Denver, CO 80225 Any use of trade, product, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government Library of Congress Cataloging-in-Publication Data Scott, Kevin M., - 1935- Sedimentology, behavior, and hazards of debris flows at Mount Rainier, Washington / by K. M. Scott, J. W. Vallance, and P. T. Pringle. p. - cm. - (U.S. Geological Survey professional paper ; 1547) 1. Lahars-Washington (State)-Rainier, Mount, Region. 2. Volcanic hazard analysis-Washington (State)-Rainier, Mount, Region. I. Vallance, J. W. II. Pringle, Patrick T. III. Title. IV. Series. QE599.US5S36 1995 363.3'495-dc20 94-43162 CIP CONTENTS Abstract 1 Introduction 1 ACKNOWICOGMENES see cess ece es eee ee se ce cece ce ec ese 4 Types of flows at Mount Rainier .............. _ 4 GENETAl StATEMCNE 00sec ee cee ee ee eee se ee ee eee ee cee ee ee eee en cen oe oe cov en cen ee ce ee ene se see 4 Cohesive debris flows (more than 3 to 5 PEICENt CIAY)) 7 Noncohesive debris flows (less than 3 to 5 PELCENt CIAY)..................cccccccessescesccecss 8 Hyperconcentrated flows ........ a 9 Flow magnitude NG fEQUENCY 9 Methods of study 220 9 Flows of high magnitude and low frequency (500 to 1,000 years) ........................ 10 Greenwater Lahar and OSCEOIA 12 Paradise Lahar .. 18 Round Pass Mudflow (branch On TAROM CIEEK) 19 Round Pass Mudflow (branch On PUy@Ilup RiVveT) ................cccsessessssccccccecceces 20 Unnamed pre-Electron deposits, Puyallup River System .............................. 21 1,000-YE@T-OIG IARAT 22 Electron Mudflow..................... inisesrsrerssesse 22 Other lahars and possible lahars ......................... esses ss 24 Synthesis of the record of large, low-frequency lahars.........................ccclll... 24 Flows of intermediate magnitude and frequency (100 to 500 years) ..................... 24 White RivET SYStEM..........02s0sessseeesseeesseee scc cee cece see see ee eee ee eee ee ee cece cence enne ees 24 COWIitZ RiVEP SYStEM ..... eee esse ccc eee a tress res 26 Nisqually River SySteMm .......................... 26 Puyallup Rivet SYStEM esse 30 [eron fale y - ee 30 Flows of low magnitude and high frequency (less than 100 years) ....................... 31 Recurrence interval Of SMIL GEbDTiS fIOWS 31 GiSHWIDUNOM.........2.2 0220202020000 e00 0000 cece eee ce cee cece 000 es 32 Flow texture and fOTMAtivE WANSFOTMAMHONS ccc 000 32 FIOw GyN@MICS ANG HANSFOTMANHONS....2..220 0000200 00200000000 ee eee ee eee ev ee eee ce ce 32 Debris avalanches and the TAROM LANAL 6s ee eceee cc cece eee e ccc ce 000 38 Historical floods compared with GEDTIS fIOWS 40 Summary of flow Origins @NG tT@NSfOTMAtONS se eee eee eee eee ee ee cee 60+ 41 Risk analySi$....................... be ss Se vA res ee reuse sss sere nases 42 Flow frequency and Fisk at MOUNt RAIMIET eee seee eee ee cee ee 66+ 43 Debris flOwWs aNd SUMMit-CONE VOICANISIMM 0000 eee ee cee cece cee ccc ec 000 44 Design or planning CaSES @Nd ZONANON ees A4 DefinitiON Of CASES ece cc cecccccc cscs ccc cc ences 44 Measurements and estimates Of flIOW GYNANNCS A4 MAXIMUM cece see eee ee cee ee eee ce ee cee ce cece ee sec ee cece se ce ees 45 DESiGgN OF PIANNING CASE I ee cee eee vee ee seven ce eee ee cense ee ccc ence nene s 46 DESigNn OF CASC II ee see ee ee e00 ee 0666 ee eee cece scenes 47 DeSigN Of PIQNNING CASE seee eee ee reese es eee ee cee ee eve ee eee ence cece nne 47 HAZATG ZONAONM cesses ese ece se ece ee cece sesso cece esc sec ce ces ce oes 47 ZONE ee cece cee sec cc cn ccc ce cece sec eevee ee ence cece sen ee ces 48 ZONE II seers ees ees seee se see se eee ee ev ces carer ee ces cee cece ees ee cence cense se renee 48 ZONE III esses sc see ece cece ese ce see ee se ce ces cece sence sec ce eee ce ces ee se sense nes 48 Lateral erosion associated with RaZard ZONE III 48 VI b o _ a in L yo 11. 12. 13. 14. 15-17. 18. 19. 20. 21. 22. CONTENTS PrOb@bility Of DIECUISOT 48 Travel times Of I@h@rs @Ad POtENti@l TESETVOiT 49 TIAVEl tiME$ Of IQRATS ................22.222200222002es0e esse ses se se ser er seres ss ee. 49 RESETVOIT CFFECHS se see sesssse seres ss ees 51 WRitE seres sesse seres secs 51 COWIItZ esses cess ses se cess 51 NiSQUAIIY RiVET .......0.00002020000 2202s seres reese esses sesssseeessccccee 52 eee sss sessesecersssesessess esses sss se esses sess ses 52 REfEIEMCES reese seers sessses esses ss eee seee see ees 53 PLATE [Plate is in pocket] Cross sections of selected flows at Mount Rainier. FIGURES Maps showing location of Mount Rainier, localities of lahar cross sections, and other features of interest in southwestern Washington .. o 2 Graph showing cumulative curves of particle sizes in a typical cohesive lahar (Electron Mudflow) and NONCORESiVE I@QRAP (NQtONAL seres se ee eee se 6 Graph showing relation of sorting and mean grain size for selected cohesive and noncohesive debris flows...... 8 Photograph showing northeast side of MOUNt Rainier 20d Little TAROMQ PEAK 11 Graph showing relation of flow volume and clay content in several postglacial mudflows at Mount Rainier ..... 13 Diagrammatic composite sequence of valley-fill deposits in the White River VAIIEY 222202220202 200 sess esse cress ees 14 Photographs showing MmOund-Studded Of the OSCEOIA 15 Graph showing lahar-bulking factors for four lahars at Mount Rainier, and composition changes in the OSCEOIA MUOULIOW ... 2002 seee seres se 17 Composite columnar sections of lahars and associated deposits in the upper Nisqually River drainage ............. 19 Diagrammatic composite sequences of valley-fill deposits in the Puyallup River valley ..................................... 21 Diagrammatic composite sequences of valley-fill deposits in the Nisqually River valley .................................... 27 Photograph showing dish structure in deposits of National Lahar at the LYPE IOC@IitY ...... 28 Graph showing discharge versus distance for debris flows resulting from glacial outburst or collapse ............... 30 Graph showing seasonal distribution of debris flows, hyperconcentrated flows, and floods in glacier-fed tributaries of the Nisqually River frOM 1925 t0 1990...................... «ces errs 32 Photographs showing: 15. Active front of the South Tahoma Glacier 5 days before and 1 day after the clear-weather glacial-outburst fl0O0Od And Gebri$ flOW Of JUNE 29, 1987 33 16. Area of stagnant, moraine-covered lower part of South Tahoma Glacier 5 days before and 1 day after the f1OW Of JUME 29, 1987 erea 34 17. - Ice clast, more than 1 meter in maximum dimension, included with andesite clasts of similar size in 10bate bOUIder frONt Of flOW Of OCtODET 26, 1986 35 Graph showing cumulative curves of particle sizes within successive boulder fronts and hyperconcentrated- flow deposits formed during transformation of the Tahoma Creek debris flow of October 26, 1986 ................... 36 Photographs showing debris flow levees and underlying sole layers from flows of 1986 and 1987 along TAROMQ seres eevee se eee esse eee ee ee eee ener revs 37 Graph showing cumulative curves of particle sizes of debris flows derived bY GEWAIETIMG 38 Photograph showing debris avalanche on the surface of the Tahoma Glacier at the head of the South o eo i P 39 Photograph showing megaclast at surface of the 1963 debris avalanche below the Emmons Glacier in the TQiN fOTK Of the WRit€ RiVET .............00s 00000 seres sevesssersssssssssesessesseessesesssreese seres reese 50 o o m p w b f CONTENTS TABLES Tephra units and other indications Of vOICaNiIC ACtivity seve Mainly cohesive debris flows Of SECtOT-COIIAPSE OF OTIGiM | cee cree cence Mainly noncohesive debris flows and their runout phases .............................. sess Radiocarbon dates from hyperconcentrated-flow and normal StrE@MfIOW GEPOSitS Summary of origins and transformations Of debris flOWS Ranking of debris flows described in table 5 by magnitude, frEQUENCY, AMG ISK \ Characteristics Of design- OF PI@NNING-C@SE IARATS see ece ces ce cee ce cev seve ne Celerities and travel times of the maximum lahar, Case I lahar, and Case II lahar from Mount Rainier to the nearest downstream reservoir or the Puget SOUNG IOWIANG sess esese eee cece ce cece eevee cece cee CONVERSION FACTORS Multiply By To obtain millimeters (mm) 0.03937 inches centimeters (cm) 0.3937 inches meters (m) 3.281 feet kilometers (km) 0.6214 miles square kilometers (km) 0.3861 square miles meters per second (m/s) 3.281 feet per second cubic meters per second (m/s) 35.31 cubic feet per second VII 10 12 23 25 41 43 45 50 - SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON By K.M. Scott, J.W. Vallance, and P.T. Pringle ABSTRACT Mount Rainier is potentially the most dangerous volcano in the Cascade Range because of its great height, frequent earthquakes, active hydrothermal system, and extensive glacier mantle. Many debris flows and their distal phases have inundated areas far from the volcano during postglacial time. Two types of debris flows, cohesive and noncohesive, have radically different origins and behavior that relate empirically to clay content. The two types are the major subpopulations of debris flows at Mount Rainier. The behavior of cohesive flows is affected by the cohesion and adhesion of particles; noncohesive flows are dominated by particle collisions to the extent that particle cataclasis becomes common during near-boundary shear. Cohesive debris flows contain more than 3 to 5 percent of clay-size sediment. The composition of these flows changed little as they traveled more than 100 kilometers from Mount Rainier to inundate parts of the now-populated Puget Sound lowland. They originate as deep-seated failures of sectors of the volcanic edifice, and such failures are sufficiently frequent that they are the major destructional process of Mount Rainier's morphologic evolution. In several deposits of large cohesive flows, a lateral, megaclast-bearing facies (with a mounded or hummocky surface) contrasts with a more clay-rich facies in the center of valleys and downstream. Cohesive flows at Mount Rainier do not correlate strongly with volcanic activity and thus can recur without warning, possibly triggered by non- magmatic earthquakes or by changes in the hydrothermal system. Noncohesive debris flows contain less than 3 to 5 percent clay-size sediment. They form most commonly by bulking of sediment in water surges, but some originate directly or indirectly from shallow slope failures that do not penetrate the hydrothermally altered core of the volcano. In contrast with cohesive flows, most noncohesive flows transform both from and to other flow types and are, therefore, the middle segments of flow waves that begin and end as flood surges. Proximally, through the bulking of poorly sorted volcaniclastic debris on the flanks of the volcano, flow waves expand rapidly in volume by transform- ing from water surges through hyperconcentrated stream- flow (20 to 60 percent sediment by volume) to debris flow. Distally, the transformations occur more slowly in reverse order-from debris flow, to hyperconcentrated flow, and finally to normal streamflow with less than 20 percent sediment by volume. During runout of the largest noncohesive flows, hyperconcentrated flow has continued for as much as 40 to 70 kilometers. Lahars (volcanic debris flows and their deposits) have occurred frequently at Mount Rainier over the past several thousand years, and generally they have not clustered within discrete eruptive periods as at Mount St. Helens. An exception is a period of large noncohesive flows during and after construction of the modern summit cone. Deposits from lahar-runout flows, the hyperconcentrated distal phases of lahars, document the frequency and extent of noncohesive lahars. These deposits also record the following transforma- tions of debris flows: (1) the direct, progressive dilution of debris flow to hyperconcentrated flow, (2) deposition of successively finer grained lobes of debris until only the hyperconcentrated tail of the flow remains to continue down- stream, and (3) dewatering of coarse debris flow deposits to yield fine-grained debris flow or hyperconcentrated flow. Three planning or design case histories represent different lengths of postglacial time. Case I is representative of large, infrequent (500 to 1,000 years on average) cohesive debris flows. These flows need to be considered in long-term planning in valleys around the volcano. Case II generalizes the noncohesive debris flows of intermediate size and recurrence (100 to 500 years). This case is appropriate for consideration in some structural design. Case III flows are relatively small but more frequent (less than 100 years on average). INTRODUCTION Mount Rainier volcano is potentially the most dangerous of the periodically active volcanoes of the Cascade Range, which extends from northern California into 1 2 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON 123° 122° WASHINGTON __ Area of map in figure 1A 47° 30° Area of map in figure 1B Carboo j. Mud Mountain * Dam 47°} __ 00' >o come c mw + £, €, pra - “I I 46° | __ c_] 30" Mayfield 000‘ Reservoir EXPLANATION *> N-4 _ Cross section of debris flow (see plate 1) """"" Boundary between Puget Sound lowland and Cascade Range «-* __ Boundary of Mount Rainier ST. HELENS National Park 0 10 20 KILOMETERS 0 10 20 MILES Figure 1 (above and facing page). Location of Mount Rainier, localities of lahar cross sections, and other features of interest in southwestern Washington. All cross sections are shown on plate 1. Cross-section localities for Tahoma Lahar (T) are shown on detailed map B (facing page); all other cross-section localities are shown above. A, Major streams, dams, reservoirs, and population centers in the region. British Columbia. The mountain is second only to Mount St. __ equal in volume to that at all other Cascade Range volcanoes Helens in seismic activity, and it is the highest volcano in the _ combined (Driedger and Kennard, 1986). Cascade Range. It also has the largest mass at high altitude, The volcano is the dominant feature of Mount Rainier above 3,000 m for example, and consequently has a _ National Park, located about 70 km southeast of Seattle, perennial snow and glacier mantle that is approximately _ Wash. It is drained by five major river systems (fig. 1): the INTRODUCTION 122°00 121°30' 47° 00' 1 $" 5 Indian Henry's Hunting T-2 Ground -O Curtis Ridge ) 7 [- "o River <- \BI€\ Yakima Park \ $- Mite White River --- «gre Campground 46° i Longmi o f omen L > Reflection / . (lf Rncggfiftker Lakes re EXPLANATION w‘fiTNSO?H-Rm;m mmmmm \& Cross section of debris flow Boundary of Mount Rainier National Park State highway Wonderland Trail Road bridge Trail bridge Only bridges referenced in text are shown 5 KILOMETERS 5 MILES Figure 1, continued. -B, Enlarged map of Mount Rainier National Park showing glaciers, streams, and major roads and trails. White River on the northeast, the Cowlitz River on the southeast, the Nisqually River on the southwest, and the: Puyallup and Carbon River systems on the northwest. All but the Cowlitz River traverse at least 100 km of both the Cas- cade Range and the Puget Sound lowland before emptying into Puget Sound. The Cowlitz flows more than 140 km southward and enters the Columbia River. For more than 45 km upstream of Riffe Lake, the Cowlitz's flood plain is unusually wide, averaging about 2 km in width. Mount Rainier volcano has an extensive history of post- glacial debris flows that originated from collapse of major sectors of the mountain (Crandell, 1971). Many recent studies of similar composite volcanoes show that evolution by periodic sector collapses is a common morphologic pro- gression. Risk analysis in this report shows that the debris flows of this origin pose the greatest hazard at Mount Rainier. These flows are distinguished here as cohesive debris flows to indicate the importance of their "sticky," 4 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON relatively clay-rich matrix in determining flow behavior, and their defining characteristic is their ability to remain unchanged as debris flows to their distal ends. Evidence of the flows is traced downstream, and flow dynamics are defined where the flows left the canyons of the Cascade Range to inundate flood plains, now increasingly populated, on the Puget Sound lowland (fig. 1). A second type of debris flow, which is common at Mount Rainier, is substantially different in origin and behavior -and yields deposits distinctly more granular in texture. These relatively clay-poor flows, distinguished here as noncohesive to indicate the importance of grain-to-grain contact in determining behavior, readily transform both from and to other flow types. Noncohesive debris flows originate most commonly as water surges produced by melting of snow by volcanic heat. Although less of a general hazard than the debris flows that originate as large slope failures, noncohesive flows occur more often, and their runout phases could inundate parts of the Puget Sound lowland. Because of their shorter recurrence intervals, noncohesive flows pose a more likely threat to life and property locally on and near the base of the volcano. The recognition of the deposits of this flow type and its transformations has significant implica- tions for the reconstruction of a volcano's history. As illustrated by the forecast of Crandell and Mullineaux (1978) at Mount St. Helens, 80 km south of Mount Rainier, a volcano's history is the best guide to its future behavior. Some noncohesive debris flows reflect another factor contributing to the exceptional flow hazards at Mount Rainier: its mantle of ice and snow is subject to volcanically induced melting. A geothermally formed subglacial lake has existed in at least one of the two summit craters (Lokey and others, 1972; Kiver and Steele, 1972), and it may still exist but its presence hasn't been verified since 1978 (W. M. Lokey, Pierce County Dept. of Emergency Management; written commun., 1989). A subset of noncohesive flows con- sists of debris flows formed from glacial-outburst floods. Although these flows have been a hazard only locally, they provided important behavioral models of noncohesive flows from 1986 to 1992. The purpose of this report is to define the origins, magnitude, and frequency of debris flows and other flow types associated with debris flows, with particular reference to volcanic hazards. The probability of such flows is similar in most of the drainage basins, as indicated by the present-day symmetry of hydrothermal activity, occasional small earthquake clusters centered beneath the summit, and our inability to forecast which sector of the volcano will be affected by renewed volcanic, seismic, or geothermal activity. However, several factors affect variations in probabilities between drainages. For example, sector collapses that produce huge, cohesive debris flows can occur on any flank of Mount Rainier, but northward- or northeast- ward-directed failure may be especially likely because the summit cone is formed in an amphitheater resembling a "greased bow!" that is probably open to the north (Frank, 1985, p. 180). In analyzing flows of sediment and water in populated valleys surrounding the volcano, this study complements ongoing studies by the U.S. Geological Survey of Mount Rainier's volcanic hazards and glacially related hydrologic hazards. There is necessarily some overlap in these investigations because many large debris flows were volcanically initiated, and many small debris flows originated as glacial-outburst floods. As at Mount St. Helens, however, there is a record of more large debris flows than the number of known volcanic events, and glaciers do not exert a primary influence in forming the larger flows. This report, therefore, concentrates on the actual down- stream stratigraphic record. of flows and their dynamics, as revealed by sedimentologic and paleohydrologic evidence, rather than evidence from volcanic or glacial activity. ACKNOWLEDGMENTS Although this study mainly focused on the flow record beyond the base of the volcano, some work within Mount Rainier National Park was necessary, and the writers are indebted to the personnel of the National Park Service cited herein for their observations of flow phenomena and for information on the 20th-century flow record. W. V. Steuben and D.J. Gooding, U.S. Geological Survey, overcame the extensive problems in laboratory analysis of the cohesive deposits. The manuscript benefited substantially from review by J.C. Brice, J.E. Costa, and C.L. Driedger, all of the U.S. Geological Survey, and K.S. Rodolfo, University of Illinois at Chicago; and from review of specific topics by D.R. Crandell, U.S. Geological Survey, RR. Dibble, Victoria University of Wellington, and W.M. Lokey, Pierce County Department of Emergency Management. TYPES OF FLOWS AT MOUNT RAINIER GENERAL STATEMENT Fundamental to the analysis of subaerial sediment- gravity flows at Mount Rainier is the recognition of two distinct types of debris flows that differ significantly in both texture and origin. Thus it is possible to deduce the origin of an ancient flow deposit, even one far from the volcano, from its texture, in particular the texture of the matrix phase of the characteristically bimodal flows. An earlier investigation at Mount St. Helens (Scott, 1988b) reported two distinctive types of debris flows: (1) relatively clay-rich flows that traveled long distances as debris flows and (2) more granular flows that began mainly as streamflow, then bulked (increased volume by incorporating sediment) to form TYPES OF FLOWS AT MOUNT RAINIER 5 hyperconcentrated streamflow, and continued to incorporate eroded sediment until a debris flow was formed. Distally, the sequence of flow types was reversed. These two types of debris flows are the cohesive and noncohesive debris flows, respectively, of this report. A subdivision of only mudflows (commonly defined as silt- and clay-rich debris flows) into cohesive and noncohesive types was made by Russian investigators (Kurdin, 1973, p. 311). However, noncohesive flow was defined as containing 80 percent sediment by weight (approximately 60 percent by volume), and the particles were described as being deposited and sorted as flow velocity slowed (Kurdin, 1973, p. 315). Although the sediment content was that of debris flow, the behavior described is mainly characteristic, not of debris flow, but of hyperconcentrated streamflow as noted by Costa (1984, p. 289). The subdivision of sediment-gravity flows and deposits into the cohesive and cohesionless classes of soil mechanics (for example, Postma, 1986) is probably only theoretically useful. Several workers have implicitly or explicitly questioned the reliability of clay or matrix content (for example, Lowe, 1979, 1982; Nemec and Steel, 1984) as the main criterion for distinguishing between ancient deposits of cohesive and cohesionless debris flows. This criterion is probably not useful for deposits of subaqueous debris flows, for which only the texture may be known and nothing at all may be known about the involvement of processes such as fluidization, escaping pore fluids, and the modified grain flow of Lowe (1976). Differences in behavior between flows having different matrix properties are more obvious where, as at a volcano, the deposit of a modern or postglacial flow can be seen from beginning to end, and each textural, stratigraphic, and morphologic nuance can be known. In some modern (1982) volcanic flows, even the postdepositional changes in matrix character can be defined (Scott, 1988b). Noncohesive debris flow, as defined here, is flow that retains sufficient strength (albeit with lower matrix cohesiveness than cohesive flow) to produce the diagnostic characteristics of debris flow deposits: transversely and longitudinally convex flow fronts, lateral levees, buoyed dense megaclasts, and a texture of commonly dispersed clasts, pebble size or coarser, in a finer grained granular matrix. Despite these similarities, however, the behavior of noncohesive debris flows differs radically from that of their cohesive counterparts. Neither type of debris flow has truly irreversible sediment entrainment, one of Hooke's (1967) criteria for distinguishing a debris flow from a water flow; both cohesive and noncohesive types may leave coarse, clast-supported whaleback bars at sites of rapid energy loss (Scott, 1988b). Only noncohesive flows lose coarse sediment at a rate sufficient to cause transformation down- stream to more dilute flow types, initially hyperconcentrated flow. Some cohesive flows show a slight textural change in the direction of that transformation, probably by particle settling within the rigid central plug of the flow, and by periodic loss of coarse clasts at sites of energy loss where dispersive and other particle-impact stresses are minimal. However, cohesive flows at both Mount Rainier and Mount St. Helens traveled well over 100 km and did not transform. The net results of deposition of sediment, and its addition (bulking) in reaches of high shear stress at flow boundaries, are discussed in the subsequent section on the Osceola Mudflow. Debris flow behavior correlates strongly with particle-size distribution (size classes shown in fig. 2), especially clay content (Hampton, 1975; Middleton and Hampton, 1976; Qian and others, 1989, fig. 3; Costa, 1984; Pierson and Costa, 1987). The description of debris flows as cohesive or noncohesive is intended to reflect an important empirical difference in behavior related to clay content, and thus to matrix cohesiveness. Silt content also contributes to the cohesiveness of a flow but is normally proportional: to the clay content. Both cohesive and noncohesive debris flows have a matrix phase and a coarse-sediment phase (dispersed phase of Fisher and Schmincke, 1984). The coarse sediment is dispersed throughout the matrix phase in most cases, but not all. Figure 2 shows particle-size distributions for representative cohesive and noncohesive flow deposits and for the other flow types into which the noncohesive debris flows normally transform downstream. The existence of a spectrum of debris flow behavior is implicit (see Lowe, 1979) in the original Coulomb viscoplastic model of a debris flow (Johnson, 1965; Yano and Daido, 1965) and, regardless of theological model, a range of behavior in response to varying sediment properties and content has been observed as noted above. Some flows are clearly dominated by viscoplastic behavior resulting primarily from momentum exchange within a "sticky" fine-grained matrix (cohesive flows of this report); others are more granular flows dominated by momentum exchange between coarser particles (noncohesive flows of this report) that are, however, still part of the matrix. Sand-size sediment dominates the matrix of noncohesive flows at both Mount Rainier and Mount St. Helens. Although the two flow behaviors are distinct, basic mechanisms of particle support likely overlap. A greater abundance of particle collisions probably explains the more pronounced shear-related boundary features and cataclasis in the noncohesive flows as well as the transformations of those flows to and from other flow types. It may also explain some previously divergent assumptions of debris flow theology, and it clearly accounts for some of the difficulties in modeling their behavior. The fundamental distinction is pragmatic, however, and a view of debris flows as having only two distinct types of momentum exchange and particle support does not fully consider the evidence of other dynamic interactions between their solid and fluid constituents (Iverson and Denlinger, 1987). Nevertheless, the distinction based on clay content is highly useful and it 6 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON 0.01 I I I I I I I I I I I I I I I I I p 0.5 1 |- - 2 - 5 |- 10 |- Noncohesive debris flow (National) 20 |- 30 40 |- 50 |- 60 Cohesive debris flow (Electron) /\Streamf|ow 70 - 80 |- 90 |- \3/ / [I a ll \>/. 99 |- CUMULATIVE PERCENT 99.8 |- w 8 99.9 |- 5 5 <§>|¢— g -»~!<---Pobbles —>|<-§>}<—— Sand -----»<--- Silt ----»!<- | | | | | | I I | | | | 1 | | ules lay | | C 99.99 90-8) -70 -60 -§0-4 -8 -2 40 0 10 20 80 40 500060 70 8 9 6 512 256 128 64 32 16 $8 4 200 1 05 0.25 01250.0625 0.016 0.004 0.031 0.008 0.002 mm S | Z E PERCENT PERCENT S Streamflow i 1024 1 0.001 mm 10 -5 0 10 6 32 1 0.001 mm Figure 2. - Cumulative curves of particle sizes in a typical cohesive lahar (Electron Mudflow) and noncohesive lahar (National Lahar). Other flow types into which the noncohesive debris flows normally transform downstream are also shown. rationalizes many of the features and behavioral variations The observation that a species of debris flow of lahars. Based on observations at volcanoes around the _ transforms to and from other flow types also can add greatly Pacific Rim by the senior writer, the distinction is generally _ to our knowledge of a volcano's past. For example, the applicable. presence of large noncohesive flows and their downstream TYPES OF FLOWS AT MOUNT RAINIER 7 transformations, or of synchronous flows in more than one watershed, can indicate unrecognized eruptions (magmatic or phreatic) or shallow landslides mobilized to debris flow. Also, the identification of debris-flow-related deposits is aided by knowing that noncohesive flows attenuate more rapidly than cohesive flows and that a noncohesive debris flow upstream may be marked on flood plains downstream by deposits of hyperconcentrated or normal streamflow, which are less conspicuous than debris flow deposits (Scott, 1989). The formation of debris flows from flood surges is the dominant formative process at some Cascade Range strato- volcanoes, such as Mount St. Helens (Scott, 1988b), but apparently is less common in other environments. The process probably does not involve pure autosuspension (Bagnold, 1962; Southard and Mackintosh; 1981) and is greatly facilitated by large sediment contributions from bed and bank mobilization. The efficacy of the process is dramatically illustrated at Mount St. Helens by the huge: lahar (PC1) that consists almost entirely of stream-rounded alluvium (Scott, 19882). In constructing a conceptual model of debris flow formation, Johnson (1984, p. 331) cited one example (Jahns, 1949) in which a debris flow resulted from the bulking of sediment from channel erosion by a clay-water mixture. Costa (1984) also cited several cases in which this mechanism probably occurred, and Rodolfo (1989) documented the process for rain-induced lahars. A surge from a moraine-dammed-lake breakout quickly bulked to debris flow in the Bol'shaya Almatinka River in Russia and continued to enlarge downstream (Yesenov and Degovets, 1979). The same mechanism formed debris flows from breakouts of moraine-dammed lakes at the Three Sisters volcanoes, Oregon (Laenen and others, 1987, 1992). The requisite factor both for bulking to debris flow and for continued enlargement is an abundance of loose, poorly sorted volcaniclastic and morainal sediment on steep slopes. Usage here of the popular but variously defined term "lahar" for volcanic debris flows corresponds, with one exception, to its application by Crandell (for example, 1971) throughout the Cascade Range. Glacial-outburst floods that bulk to debris flows but lack evidence of triggering by volcanism are not here called lahars; the term is reserved for, and most usefully applied to, flows that are directly or indirectly related to volcanism rather than merely the alpine environment. Other characteristics, such as composition or angularity of debris, are not definitive. Even origin on a volcano is not a reliable criterion, for in some cases bulking may not produce a debris flow until the surge is beyond the volcanic edifice. Some details of terminology are discussed by Scott (1988b). To be consistent with most formal and informal usage in the Cascade Range, the term is applied here, as by Crandell (1971), to both the flow and the deposit. Future workers may wish to conform to the definition recommended by the 1988 Penrose Conference on Volcanic Influences on Terrestrial Sedimentation (Smith and Fritz, 1989), which is essentially the definition used here except that the flow deposit is excluded. COHESIVE DEBRIS FLOWS (MORE THAN 3 TO 5 PERCENT CLAY) The largest lahars at Mount Rainier were recognized as relatively clay-rich by Crandell (1971), who logically hypothesized that a clay content of about 5 percent or more reflected an origin directly from large landslides. The clay is an alteration product of the hydrothermal system of the volcano. These flows remained debris flows to their termini. At Mount St. Helens, the critical clay content that characterized the nontransforming cohesive flows was a minimum of 3 percent. The apparent difference in critical clay content between the two volcanoes, as it affects flow behavior, mainly reflects sampling procedure and is discussed below. Hydrothermal alteration is more intense at Mount Rainier, probably as a result of its greater age. This difference is the probable cause of the dominance of large cohesive flows at Mount Rainier, whereas large noncohesive flows dominate the record at Mount St. Helens (Scott, 19882, 1989; Major and Scott, 1988). The original and type example of a cohesive flow is the 1980 North Fork Lahar at Mount St. Helens (Scott, 1988b). The differences in behavior of cohesive and noncohe- sive debris flows correlate strongly with the texture of the matrix phase: the matrix of cohesive debris flows is a mix of sand, silt, and at least 3 percent clay; that of noncohesive debris flows is silty sand with commonly about 1 percent clay. In cohesive debris flows, (1) grain interaction is cushioned by the adhering clay aggregates, thereby reducing near-boundary shear and other particle interactions recorded by the boundary features characteristic of noncohesive debris flows; and (2) the clayey matrix retards each of the following: (a) the settling of coarse particles, (b) the differential movement of all coarse-phase particles (which produces the well-developed normal and inverse grading in noncohesive flows), and (c) the miscibility of the flow with associated streamflow. The latter effects prevent or greatly retard the transformation of a cohesive debris flow to hyper- concentrated streamflow. These conclusions are empirical; the actual physics and chemistry of clay in the matrix remain to be investigated. For example, clay content may affect the viscosity of the pore fluid and, therefore, the hydraulic diffusivity of that fluid through the granular phase (Iverson, 1989). Such movement may be slight in cohesive flows, where interparticle attractive forces can dominate, but in the noncohesive regime, the character of the medium around colliding and abrading particles of the matrix must be important. Clay minerals compose most clay-size sediment, but their proportion is variable in lahars. They compose 75 percent of the clay-size sediment in the largest lahar from 8 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON Mount Rainiee (Crandell, 1971). Clay minerals are layer-lattice silicates with powerful surface forces that can provide cohesion and strength to the entire flow. Clay aggregates in turn adhere to sand in the matrix as well as the coarse-phase clasts. Varieties of clay minerals reported from the edifice of Mount Rainier and the lahars derived from it include kaolinite, montmorillonite, smectite, halloysite, illite, and pyrophyllite (Crandell, 1971; Frank, 1985). The failed sectors of the volcano contained enough water and clay to provide uninterrupted mobility as they rapidly disaggregated, first to a debris avalanche and then to a lahar. A debris avalanche is a rapid flow of rock debris (Vames, 1978), wet or dry, commonly containing many large megaclasts. Studies of the May 18, 1980, eruption at Mount St. Helens suggest that some cohesive flows may have been derived from the surface of an immobilized debris avalanche. The surficial portion of the huge 1980 debris avalanche at Mount St. Helens was saturated by dewatering after emplacement, thereby forming a critical mass of ponded muddy debris which began flowing as a broadly peaked lahar wave several hours later. All large-scale debris avalanches recorded by known deposits at Mount Rainier mobilized directly to lahars. The only debris avalanches known to have yielded lahars secondarily were small examples of shallow origin. NONCOHESIVE DEBRIS FLOWS (LESS THAN 3 TO 5 PERCENT CLAY) These more granular debris flows commonly represent the middle segments of meltwater flood surges (either volcanically or climatically induced) that both begin and end as streamflow surges. They are generally better sorted and may be finer grained on average than the larger, boulder-rich cohesive debris flows (fig. 3). The initial water surges incorporate most sediment from stream-channel deposits that are partly depleted of fine sediment as a result of selective or hydraulic sorting by fluvial processes acting both subglacially and proglacially. Consequently, through bulking, the surges form debris flows that have an average clay content of only about 1 percent. The original and type example of a noncohesive flow is the lahar of March 19-20, 1982, at Mount St. Helens (Scott, 1988b). On the volcano, glacial-outburst flows are characterized by a relatively high lahar-bulking factor (LBF; the percent- age of sediment added to the flow beyond the point of origin as revealed from clast roundness or composition; Scott, 1988b). Noncohesive debris flows originating as slides of relatively unaltered volcaniclastic debris have a distinctively lower LBF. This origin resembles, at smaller scale, the pro- cess by which the large cohesive lahars are formed from deep-seated failures. 6 EXPLANATION @ Cohesive debris flow C Noncohesive debris flow O Osceola Mudflow E Electron Mudflow F Lahar near Fryingpan Creek 0 63 P Paradise Lahar O N N National Lahar F T Tahoma Lahar @ - (peak flow facies) oN 63 1963 debris flow on surface of debris 08" avalanche 86 1986 Tahoma Creek, matrix .O 5 |- 0 E - o= “.o [ 13 m I | SORTING, IN 6 UNITS to I =f O ose o" Poorer sorting Coarser 87 1987 Ta_homa Creek, q | | matrix 1 0 -1 -26 0.5 1.0 2.0 4.0 mm MEAN GRAIN SIZE Figure 3. Sorting versus mean grain size for selected cohesive and noncohesive debris flows. Mean grain size and sorting values correspond to M, and G,; of Folk (1980). In comparison with the deposits of cohesive lahars, those of noncohesive lahars document more intense particle interaction, especially near flow boundaries, where the group of boundary features described by Scott (1988b) records the effects of shearing on particles and their size distributions. These features include a distinctive sole layer, inverse graded bedding, a lahar-abraded pavement, truncated size distributions, and grain cataclasis. Such features are clearly best developed in the noncohesive flows, but are not exclusive to them. A common but not consistent distinction between the debris flow types is a generally higher rate of attenuation of the noncohesive flows. The granularity of noncohesive flows increases their miscibility with overrun streamflow, a factor leading to their downstream transformations to hyperconcentrated flow. The effect is illustrated by the increase in transformation rate at sites of significant tributary inflow (Scott, 1988b, fig. 37). In effect, the flow becomes diluted and loses strength, and the fluid phase progressively outruns the sediment phase. The sediment component of the noncohesive debris flows is more readily deposited than that of the cohesive flows. Once transformation occurs, the peak sediment concentration characteristically lags behind peak discharge (Scott, 1988b), as in some storm-flood peaks (Guy, 1970). In cohesive debris flows, the entire mixture remains coherent and relatively constant in texture, although systematic down- stream change can occur (Scott, 1988b). FLOW MAGNITUDE AND FREQUENCY 9 HYPERCONCENTRATED FLOWS Hyperconcentrated flow is an important flow type at most Cascade Range volcanoes. The history of its recognition and the criteria by which its deposits are recognized are described elsewhere (Scott, 1988b). The most obvious feature that differentiates these deposits from debris flow deposits is their undispersed, entirely clast-supported texture. They are distinguished from flood-surge and normal streamflow deposits by poor development of stratification, sorting in a range intermediate between those of debris-flow and flood-surge deposits, locally well-developed inverse or normal grading, and the local presence of dewatering structures such as the dish structure of Wentworth (1967) and Lowe and LoPiccolo (1974). Dish structure is previously reported only from deposits of sediment-gravity flows in the deep marine environment. Bulking of a flood surge to a debris flow commonly occurs on the steep flanks of the volcano in confined channels. In this setting, the granular deposits of the hyper- concentrated flow interval are thin and rarely preserved. Channel steepness and an abundance of unstable detritus commonly result in rapid bulking in a short increment of channel. The debulking, in contrast, may occur over a long distance. In the streams draining Mount Rainier, the longest documented intervals of hyperconcentrated transport in single flows occurred in the Nisqually River from Longmire to below Alder Reservoir, more than 40 km, and in the White River from near the base of the volcano to beyond the boundary of the Puget Sound lowland, over 70 km. A distinctive facies distinguishes the interval where debris flow transforms to hyperconcentrated flow (Scott, 1988b, fig. 10). This "transition facies" begins to form as the front of the flood wave transforms, continues as the change works its way progressively back through the debris flow, and ends at the point where the entire wave becomes hyperconcentrated flow. The preserved record of this transition interval, therefore, consists of downstream-thick- ening hyperconcentrated flow deposits overlain by downstream-thinning debris flow deposits. This transition facies thus documents the origin of hyperconcentrated flow from an upstream debris flow. At Mount St. Helens, the hyperconcentrated flows were described as lahar-runout flows and interpreted as evidence of upstream lahars, based on the presence of the transition facies (Scott, 1988b). The record of ancient and modern flows at Mount Rainier confirms that most, if not all, of the significant hyperconcentrated flows there had such an origin. At most Cascade Range stratovolcanoes, the steep slopes and abundance of volcaniclastic sediment assure that any significant flood surge will bulk to debris flow. FLOW MAGNITUDE AND FREQUENCY METHODS OF STUDY The distribution of past flows in time and space is an excellent guide to the probability and extent of future flows. This study amplifies the landmark work of Crandell (1971) by likewise focusing on lahars formed during postglacial time at Mount Rainier. Older lahars, however extensive (Iahar deposit along lower Cowlitz River; Bethel, 1981), are not included because of their possible disturbance by, and potential confusion with, extensive glacial deposits. (See Dethier and Bethel, 1981.) Also, not only is the record of older lahars probably incomplete, but the conditions of their formation probably differed from present conditions. Certain flow deposits that are distinctive and correlative over significant distances are named informally in accordance with the North American Stratigraphic Code for key or marker beds (North American Commission on Stratigraphic Nomenclature, 1983). Crandell's (1971) nomenclature is retained for the units he recognized. Given the textural differences in lahar and lahar-runout facies (Scott, 1988b, fig. 10), the older flow units are dated and correlated based on tephra units, radiocarbon dating, soil formation, and clast characteristics. Table 1 summarizes the prominent tephra units and other major postglacial volcanic events at Mount Rainier. The dynamics of flows are integral to any discussion of flow behavior and hazards. Original flow-wave volumes are estimated from deposit volumes as discussed for specific flows. Ancient flow discharges are determined from flow velocities, as calculated from paleohydrologic techniques using vertical runup or superelevation around bends (see Costa, 1984, p. 304-305; Johnson, 1984, p. 305-309), and from cross-sectional areas derived from levels of flow deposits on valley-side slopes. Recently, discharges calculated using superelevation around bends were called into question by Webb and others (1989, p. 22, table 10), who noted large differences in cross-sectional areas calcu- lated for flows in straight and curved reaches. We believe such discrepancies result because flow surfaces, particularly in very sharp bends, become markedly concave, as is well shown in photographs of 1990 debris flows at Jiangjia Ravine in Yunnan Province, China, by R.J. Janda and K.M. Scott (U.S. Geological Survey). Most cross sections at Mount Rainier (pl. 1) are not from comparably sharp bends, but the observation of Webb and others still stands as a cautionary note for any discharge calculated from a section in a bend (such as section T-1, pl. 1). Post-flow valley erosion is generally small and, in dealing with flow depths of many tens of meters, is generally not more than several percent of the flow cross-section. The 10 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON Table 1. - Tephra units and other indications of volcanic activity at Mount Rainier. [Most data from Mullineaux (1974, 1986) and Crandell (1971). All events originated at Mount Rainier unless otherwise indicated) Tephra set or layer, or other volcanic event Age! Remarks Probable geothermal melting of South Tahoma Glacier... Late 1960's See Crandell (1971, p. 62). Tephra frOM SUMMit CONE (IAYEP Mid-19th century SCt W (MANLY WBM) 000000200 000 seres A.D. 1480 (layer Wn) From Mount St. Helens. Dated by Yamaguchi (1983). Pyroclastic surge 1,080 Identified locally on east side by RP. Hoblitt, U.S. Geological Lava flOWS fOIMiDG SUMMIt Layer C 2,200 Block-and-ash flow in Puyallup River valley. ................. 2,350 Set P. 2,500-3,000 SEt Y (MAINIY 1@Y@T YB) ......02022200 000000000 ver sevsrecerssessccees 3,400 (layer Yn) Layer B 4,500 Layer H. 4,700 Layer F (POSSibIG DIASt iM 5,000 Bomb-bearing flows in White River valley. .. 5,700-6,600 L@yer S (POSSIDIG DIASt) 00000000 eee 5,200 Layers N, D, L, A 5,500-6,500 Layer O 6,800 Layer R >8,750 Post-layer C, pre-set W Survey. Age estimated as from 2,100 to 1,200 absolute years by Crandell (1971, p. 14). From Mount St. Helens. From Mount St. Helens. Most prominent tephra deposit. Only layer common throughout Park. See Mullineauxz (1974, p. 19-20). See Crandell (1971, (p. 23). See Mullineaux (1974, p. 20). Also interpreted as possible blast by David Frank and Harry Glicken, U.S. Geological Survey, written commun., 1987. From Mount Mazama (Crater Lake). Latest data by Bacon (1983). Years before 1950 in radiocarbon years, except as otherwise indicated. dated tephra layers (table 1) define the levels of valley bottoms at successive postglacial intervals, as well as the erosion of valley-side slopes through time. Cycles of aggradation and degradation related to Neoglacial advances and retreats are critical only in defining the cross-sectional areas of the relatively small glacial-outburst flows. In general, the accuracy of discharges determined with these techniques, proportional to the size of the flows, compares with that of other indirect-discharge determinations. Some velocity measurements determined from runup and superelevation are suspect, however, because the techniques are unverified for debris flows and because of some locally high values (Costa, 1984), especially those above 30 m/s. Some high values measured for 1980 flows at Mount St. Helens (Fairchild, 1985; Pierson, 1985; Scott, 1988b) resulted from the lateral momentum provided by cat- astrophic ejection of wet debris that settled and flowed as a lahar. The most relevant velocities in this study are those determined near the points where flows left the confined valleys of the Cascade Range and inundated the Puget Sound lowland. Some aspects of flow rheology can be inferred from texture and fabric comparisons with modern flows (Scott, 1988b). The texture of the deposits is determined from a combination of field measurements of the b axes of gravel-size (>2 mm) particles at a level or in a grid-defined area (Wollman, 1954), and laboratory measurements of the sand, silt, and clay fractions by sieve and pipet. This combination of techniques is statistically valid (Kellerhals and Bray, 1971), and a technique for combining the two is described by Scott (1988b). The approach thus incorporates the complete spectrum of sizes in the deposit, whereas the size analyses of lahars reported in Crandell (1971) do not include sediment in the coarse cobble and boulder-size ranges. Crandell's approach could result in a reported clay content, for example, of twice the actual clay content for a lahar that contains 50 percent coarse cobbles and boulders. Differences in results were assessed by comparing his analyses with field counts of the coarser fractions at many localities. The analyses reported by Crandell (1971) are useful in showing the relative but not the absolute differences in clay content The grain-size measures reported here are graphically determined, following the method of Folk (1980). FLOWS OF HIGH MAGNITUDE AND LOW FREQUENCY (500 TO 1,000 YEARS) Most of the lahars in this category originated directly from the large debris avalanches known as sector collapses (fig. 4). The major flows are listed in table 2, along with smaller flows of the same general origin. Many of the deposits were recognized and named by Crandell (1971). Size distribution measurements show that most are the deposits of cohesive debris flows, containing at least 3 to 5 percent clay. Consequently, most of the flows did not transform as they flowed downvalley for long distances. FLOW MAGNITUDE AND FREQUENCY “Little Tahoma Peak\ 7 avalanche Valley of White River 11 Summit cone Figure 4. View of northeast side of Mount Rainier (right) and Little Tahoma Peak (left center). Note large embayment, now partly filled by the snow-clad summit crater, which yielded the sector collapse that formed the Osceola Mudflow. The flow diverged across Steamboat Prow, the apex of the partly barren triangle of rock at the right side of the photograph, into the main fork of the White River (center), now the site of the Emmons Glacier, and northward into the West Fork White River (to right of photo). Dark rubble on surface of the lower part of the Emmons Glacier is from the 1963 debris avalanche originating from Little Tahoma Peak. Some do not correlate with known episodes of volcanic activity (table 1), which has implications for hazards planning, as discussed later in the report. As noted above, Crandell (1971) concluded that the critical clay content defining flows of avalanche origin is approximately 5 percent, whereas Scott (1988b) found that the clay content defining this origin at Mount St. Helens, as well as the limit below which the flow behavior was noncohesive, is close to 3 percent. However, this difference results largely from differences in sampling and analysis procedures. Therefore, rather than modify Crandell's data, we describe the critical limit here generally as 3 to 5 percent. Probably no absolute limit exists, but the actual separation of behavior types at both Mounts Rainier and St. Helens correlates best with a clay content of about 3 percent. The observed behavior-limiting clay content will vary, depending on overall size distribution and clay mineralogy as well as analytical technique. Field "pebble counts" can be strongly influenced by the presence of a few large clasts, and the results of laboratory analyses of cohesive deposits are extremely sensitive to technique. An interesting correlation exists between flow volume, and thus avalanche volume, and the clay content of the resulting lahars. The larger the flow, the greater the clay content (fig. 5). The implication is that the larger the flow, the greater the penetration of the collapse into the hydrothermally altered core of Mount Rainier. Hydrother- mal alteration and, consequently, clay content logically increase toward the center of the volcanic edifice. Clay mineralogy supports this conclusion (Frank, 1985, p. 144-145). Near-surface clay-rich areas can yield small cohesive flows, such as the modern debris avalanche on the Tahoma Glacier (Crandell, 1971, p. 17); that flow (of relatively small volume) would be an exception if plotted on figure 5. A practical use of the overall relation is that in a downstream sequence of Mount Rainier lahars, those with the highest clay content tentatively can be inferred to have been the largest. Much of the scatter in figure 5 is caused by 12 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON Table 2. Mainly cohesive debris flows of sector-collapse or avalanche origin at Mount Rainier. [Ranges in clay content include data from Crandell (1971). Leaders (- - -) indicate data unknown; km? = cubic kilometers] Flow (pg;);n) Age' Drainage vats? Extent Broadly peaked flows that traveled a significant distance from the volcano Central part of Tahoma Lahar ...... 3-4 Post-set W Tahoma Creek; <0.15 Probably to Elbe. Nisqually River. Electron MudfIOw ......................... 6-11 530-550 Puyallup River 0.26 Puget Sound lowland. 1,000-yr-old lahat ......................... 5-12 1,050-1,000 Puyallup River Possibly At least to Mowich River; possibly >0.30 to Puget Sound lowland. Unnamed lahar (possibly same as 4-5 Same as below? Puyallup River ~~~ Puget Sound lowland. Round Pass Mudflow). Round Pass Mudflow (main part). 4-5 2,170-2,710 Puyallup River --> Puget Sound lowland. Osceola Mudflow (probably 2-15 4,500-5,000 White River (main 3 Puget Sound lowland. includes Greenwater Lahar). fork and West Fork). Greenwater Lahar (probably part <3 --> Main fork White River --> Puget Sound lowland. of Osceola Mudflow). More sharply peaked flows that attenuated rapidly on or beyond the volcano Lahar from main avalanche from <2-4 A.D. 1963 White River 20.01 White River Campground. Little Tahoma Peak. Avalanche on Tahoma Glacier --- 3 A.D. 1910-1927 Puyallup River 40.01 Below glacier terminus. (small derivitive debris flow). Round Pass Mudflow (part in 2-8 2,610-2,790 Tahoma Creek; <0.1 Unrecognized below Tahoma Creek. Tahoma Creek). Nisqually River. Pre-Y lahar at Round Pass ............ >3,400 Puyallup River; -~~ Unknown. Tahoma Creek. Paradise Lahar (probably synchro- 1-6 4,500-5,000 Paradise River; 3 ~0.1 At least to National. nous with Osceola Mudflow. Nisqually River. * Years before 1950 in radiocarbon years, except as otherwise indicated. 2 Avalanche volume, from Crandell and Fahnestock (1965). 3 Slightly modified from Crandell's (1971) estimate of 1910-1930. * Avalanche volume. 5 Estimated. the effects of scattered large clasts on the size distributions; clay content of only the matrix of each flow is more uniform. In the following sections the flows relevant to hazard analysis are discussed in stratigraphic order of their deposits from oldest to youngest. The focus in each case is on aspects that are critical to analyzing risk: flow behavior, dynamics, and age. (Age is relevant chiefly for what it reveals about flow frequency.) Details of stratigraphy and exposure localities are not given here unless they have been changed or reinterpreted from Crandell (1971) on the basis of new exposures. GREENWATER LAHAR AND OSCEOLA MUDFLOW The Greenwater Lahar was described by Crandell (1971) as a hummocky, relatively low-clay lahar that filled the White River valley to levels not surpassed at most locations by the subsequent, high-clay Osceola Mudflow (pl. 1). Backwater deposits of the Greenwater Lahar form mounds in many tributary valleys in the river system as mapped by Crandell (1971, fig. 6), and were described as undifferentiated lahar deposits by Frizzell and others (1984). The mounds are the surface expression of blocks of the volcanic edifice that were grounded or stranded in backwater areas. The unit has been interpreted as a debris avalanche (Siebert, 1984; Siebert and others, 1987). Accord- ing to Crandell (1971), the Greenwater Lahar is overlain by the Osceola Mudflow and is differentiated from it by surficial mounds and a lower clay content of the matrix. A deposit typical of the Greenwater Lahar contains 3 percent clay, without megaclasts included in the size distribution. Mean grain size (M, of Folk, 1980) is -1.7 ¢ (3.2 mm), sorting (G,; of Folk, 1980) is 5.1 $, and skewness (Sk, of Folk, 1980) is +0.04. The matrix of the Osceola is remarkably clayey; the composite deposit (matrix and coarse phases excluding megaclasts) contains 2 to 15 percent clay, with a mean of 7 percent (13 samples). Mean grain size is -0.3 to -5.7 $ (1.2 to 51.0 mm), sorting is 4.6 to 6.2 $, and skewness is -0.05 to +0.48, with all but one value positive, indicating the excess of fine material typical of lahars. Longitudinal change in size distribution is variable, FLOW MAGNITUDE AND FREQUENCY 13 10.0 T T T -T-T T 8.0 |- Noncohesive . ; debris flows --m-«s--- Cohesive debris flows - 4.0 |- -I o c o e eee . e» - 0 0 Oscaoia Mudliow 2.0 |- - 1.0 |- - 0.8 |- in 0.6 |- - o e e 1,000-year-oid lahar ® e - ® ® Electron Mudtiow 0.2 |- - 8 ® Tahoma Lahar do C 0 0.1 0.08 0.06 goo ® ® ® Paradise Lahar - I 1 DEBRIS FLOW VOLUME, IN CUBIC KILOMETERS @ 1947 Kautz Creek debris flow - 0.04 T 1 0.02 |- - 0.01 |- 8 @ Lahar from Little Tahoma Peak avalanche - 0.008 1 LOU U I 1 2 4 06 O8 10 20 40 CLAY CONTENT, IN PERCENT Figure 5. - Flow volume versus clay content in several postglacial debris flows at Mount Rainier. but a general downstream increase in mean grain size probably is the result of bulking of bed and bank materials coarser than clay, and from which the clay had largely been removed by hydraulic (or selective) sorting. Four radiocarbon dates reported from the Osceola Mudflow by Crandell (1971) range from 4,700+250 to 5,040+150 radiocarbon years. Materials collected by us yield ages of 4,455+355, 4,980+200, and an older, age-limiting date of 5,230+235 radiocarbon years. The locations and implications of the younger dates are discussed below; the older date is from charcoal fragments above layer O but at the base of two pre-Osceola lahars near Buck Creek (fig. 6A). Those two flows probably are part of the pre-Osceola lahar assemblage described by Crandell (1971, p. 23) from upstream exposures, mainly on the south valley bank of the White River near Fryingpan Creek. Charcoal fragments from a third, older lahar, which is also near Buck Creek and is stratigraphically above layer O and below the Osceola, were 6,075+320 radiocarbon years in age. Both of the dated pre-Osceola deposits are noncohesive; the young- est pre-Osceola deposit is cohesive. Pre-Osceola lahars are also present in the valley of the West Fork White River (fig. 6C). A new interpretation of the Greenwater Lahar and Osceola Mudflow in the White River system is suggested, largely based on new exposures, especially those near the confluence of Huckleberry Creek with the White River. Much of what is mapped as the Greenwater Lahar (Crandell, 1971, fig. 6) may be a peak-flow facies of the Osceola Mudflow with a high megaclast concentration, as inferred from the following evidence: the deposits of both lahars are present at the same level at several localities, such as Crandell's cross section A-A' near Buck Creek (Crandell, 1971, pl. 2), but, at the localities where the base of the younger Osceola Mudflow is exposed, it most commonly overlies lahar-runout deposits and Pleistocene glacial drift rather than Greenwater deposits. If the Greenwater Lahar is a separate unit, much time must have elapsed for it to have been so extensively eroded before occurrence of the Osceola Mudflow. A cohesive lahar locally underlies the Osceola near Buck Creek and at a locality near the confluence with the West Fork White River where there is no evidence of a significant time break between the two lahars. Two terraces north of the Mud Mountain Reservoir, near the point of discharge to the Puget Sound lowland, are underlain by a mounded lahar. Moreover, 90 percent of the mounds, which are as much as 15 m high and 60 m across, have cores composed of Mount Rainier rocks, indicating an origin from the slopes of the volcano. Less than 1 km upstream along Scatter Creek, a small tributary of the White River on the north side of the Mud Mountain Reservoir, the Osceola and underlying Pleistocene drift crop out nearly continuously without an intervening Greenwater equivalent, indicating that the mounded deposits downstream are Osceola rather than Greenwater. We infer that Osceola megaclasts were grounded here and formed mounds as peak discharge passed downstream, allowing the more fluid matrix to drain away. Other mound-bearing deposits along the West Fork of the White River must be Osceola because the Greenwater Lahar is thought to have flowed down only the main fork of the White River (Crandell, 1971, fig. 6). Two levels of mounded deposits downstream from the confluence of the west and main forks of the White River are explained by pre-existing topography or by slightly different arrival times of the peak from each fork at the confluence. The textural difference between Greenwater and Osceola deposits corresponds to that between the lateral mound-bearing deposits and the clay-rich valley deposits of the Osceola into which the White River is incised. If these are deposits of the same flow, the textural variation is explained by figure 5. The clay-rich valley deposits 14 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON W in section Lahar of Dead Man Flat lahar assemblage C and Y in section Osceola Mudflow Lahar Lateral moraine of Evans Creek age /—W, C. Y, and O in section /—Inferred peak stage of Osceola Mudflow /—W, C, and Y in section Hummocks Lahar Date: 5,230 + 235 N radiocarbon years ~ / 0 __ -as * « Vertical scale is approximate Diamict “tiff." t Lazhar-runmljt flow * * aude apparently post-W) A Osceola Mudflow £55, Peak stage of Osceola Mudflow HummOCkSF \ 2024 I l & Osceola Mud Iow\' _ came: N=) : \ ge- " g ® 00. 00 : ”U/ "* * Vashon age \ METERS h 50 —— Dead Man Flat lahar assemblage-- 25 two lahar-runout flows (both contain C tephra) 0 Vertical scale is approximate B -~.» --- Alluvial sand and gravel Inferred peak stage of Osceola Mudflow //:W and Y in section W in section Alluvium Two lahars of Dead Man Flat Post-W lahar / lahar assemblage (oldest MEEERS contains C tephra) Y in section Osceola Mudflow 25 * " nz Two lahars Three lahars of Dead Man Flat \ Bedrock o lahar assemblage Osceola Mudflow C Vertical scale is approximate (oldest contains C tephra) (base not exposed) Figure 6. - Diagrammatic composite sequence of valley-fill deposits in the White River valley (successive downstream sections). Vertical scale is shown to indicate approximate thicknesses. Horizontal scale is variable for better portrayal of stratigraphic relationships and is not shown. However, vertical exaggeration ranges approximately between 5x and 10%. A, White River between Buck and Huckleberry Creeks. B, White River at Mud Mountain Reservoir. C, West Fork White River between 25 km from Mount Rainier and confluence with main fork of White River. FLOW MAGNITUDE AND FREQUENCY 15 Figure 7. - Mound-studded surface of the Osceola Mudflow in the embayment formed by the tributary valley of Huckleberry Creek. A, View upstream across the cleared surface of the deposit. Note the gentle slope of the lateral deposit toward the White River, behind and to the left of the photographer. Largest mound visible, left of center, is approximately 3 m high. B, Aerial view of part of same area on April 20, 1982. Note that some masses that are possible megaclasts are not topographic mounds. represent a late stage of flow, assuming the normal behavior of a lahar flood wave. Thus, these deposits were probably derived from fluidization of a deeper and more altered, clay-rich part of the volcano. This hypothesis is, in essence, not greatly different from the sequence of two flows proposed by Crandell (1971, p. 23) in which the first (Greenwater), formed from relatively unaltered surficial rock, exposed the underlying, altered part of the edifice for the sector collapse that produced the second (Osceola). Separate peaks in the same huge flow are possible, as in the sequence of slide blocks that produced the 1980 debris avalanche at Mount St. Helens (Glicken, 1986), but would yield deposits revealing their separation in time, however brief. At the confluence of the White River and Huckleberry Creek, almost the entire backwater fill of mounded material, mapped as the Greenwater Lahar (Crandell, 1971, fig. 6), has been cleared of vegetation (fig. 7A). Aerial photographs of the clearing (fig. 7B) show, in addition to topographic mounds, abundant ghost-like shapes, some of which we interpret as disintegrated megaclasts. The disaggregation of blocks of relatively unaltered material, both during flow and after deposition, may partially explain the mounded, relatively low-clay backwater deposits and thus their contrast with the clay-rich deposits in the central valley. In addition, size analysis of material between megaclasts in an B 100 METERS extreme backwater position shows a prominent mode in the sand range, as in a typical noncohesive lahar, and it is likely that some of this sand originated by bulking during passage of the locally erosive peak of the flow. Calculation of the overall proportion of eroded sediment in the deposits at this point, by means of the lahar-bulking factor (LBF), shows that material derived by bulking of stream and valley sediment makes up at least 20 percent more of the sediment (excluding megaclasts) in the mounded deposit near the lateral margin than in the cohesive central-valley deposits. Our preferred hypothesis thus is that these two types of deposits (noncohesive with abundant megaclasts and 16 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON cohesive with fewer megaclasts) are facies resulting from a single large flow wave. Also at Huckleberry Creek, newly exposed Osceola Mudflow deposits overlie a thick glaciofluvial section in a gravel pit within 180 m of the nearest mound that, along with many others, protrudes from a level identical with the surface of the Osceola Mudflow. Wood from the basal 50 cm of a 3.1-meter section of the Osceola Mudflow at the quarry yields the date of 4,455+310 radiocarbon years, evidence that the deposit is not of pre-Osceola age (table 2). Wood from the upper meter of the quarry section yields the date of 4,980+200 radiocarbon years, a date likewise conformable with an age of 4,500 to 5,000 radiocarbon years for the Osceola. Excavation of the mound nearest the quarry to a depth greater than 4 m shows it to be enclosed by typical deposits of the Osceola Mudflow. The existence of two flows or two separate peaks requires, therefore, the unlikely coincidence that mounded deposits of the Osceola filled the valley to exactly the same level as did mounded deposits of the Greenwater Lahar. The lateral, but not central or distal presence of the mounds is explained both by megaclast disintegration during flow, and by the similarity of a debris flow path to the path of a tractor tread: "The coarsest clasts remain in the front and are deposited along the sides of the moving flow, but the finer debris is recycled * * *," continuing downstream because "* * * debris at the center and top of a channel moves faster than debris along the sides and bottom * * *" (Johnson, 1984, p. 287). This interpretation does not change Crandell's interpretations of the volume of the Osceola or of its distribution underlying the Puget Sound lowland. It does indicate, however, that mounds are not diagnostic of a debris avalanche in the Cascade Range. The other two notably mounded deposits, the 1980 rockslide-debris avalanche deposit at Mount St. Helens (Glicken, 1986) and the huge hummocky landslide north of Mount Shasta (Crandell and others, 1984; Crandell, 1988), are both debris avalanches. Overall, the mound-bearing deposits in the White River system, be they Osceola or two separate flows, are best characterized as those of a lahar. The fill of backwater areas with muddy, horizontally surfaced deposits (excluding the mounds) indicates fluidity and strength in the range typical of lahars. The flow originated by collapse of a hydrothermally altered, water-rich edifice. The contrast between the presumed Greenwater deposits, with the hummocky surfaces of a debris avalanche, and the Osceola Mudflow deposits is the contrast between relatively clay-poor sources with less water, and clay-rich sources with more water. The deposits represent, respectively, the initial part of the flow wave (the "Greenwater Lahar"), composed of surficial, relatively unaltered rocks, followed by flow derived from highly altered, deep-seated materials (the Osceola Mudflow). The Round Pass Mudflow, Electron Mudflow, and Tahoma Lahar, described below, also contain megaclasts expressed as mounds in lateral facies. Each of these, like the Osceola, is a lahar inferred to have formed by mobilization of a debris avalanche. The Osceola Mudflow contains weak megaclasts of unconsolidated, stratified gravel, sand, and silt that represent flood-plain deposits. Some such megaclasts were rotated from the horizontal; others were deformed. To document the proportion of eroded and disaggregated flood-plain sediment in the flow, we calculated the lahar-bulking factor from clast roundness, and we also compared the roundness and lithology of various clasts. The combination of LBF and rock type yields estimates of the proportion of material derived from the volcano, from valley-side slopes, and from the flood plain and channel. The results for all or part of the size range of -3.0 to -5.0 $ (8 to 32 mm) (fig. 8) indicate that bulking was substantial and progressive, but slightly less than indicated by the composition analysis of Glicken and Schultz (1980). This general degree of bulking is probably applicable to most of the gravel-size sediment (-1.0 ¢ and coarser; 2.0 mm and coarser). Lesser but substantial amounts of bulking probably occurred in the sand (4.0 to -1.0 ¢; 0.0625 to 2.0 mm) and silt (8.0 to 4.0 $; 0.004 to 0.0625 mm) size ranges. Bulking did not add to the already high clay content of the flow; the bed material and channel deposits that compose most flood-plain sediment generally contain less than 1 percent clay. Variations in clast roundness with composition indicate substantial bulking from steep valley sides between the volcano and Greenwater, a distance of about 40 km. Farther downstream, the same comparisons indicate the increasing incorporation of more highly rounded valley-bot- tom sediment, especially on the drift plain of the Puget Sound lowland. The roundness of Mount Rainier rock types downstream indicates the incorporation of already rounded alluvial clasts. Bulking, although substantial, was not sufficient to increase the peak discharge of the lahar wave downstream. However, the flow volume obviously would have declined much more rapidly downstream had it not been for the effects of bulking, which accounted for at least 50 percent of sediment in the -3.0 to -5.0 ¢ (8 to 32 mm) size range after 60 km of flow. Clasts in the -3 to -5 $ (8 to 32 mm) range with partially stream-rounded surfaces were examined for evidence of abrasion or breakage (cataclasis) during flow. Between 20 and 35 percent of the rounded clasts were broken, but only about half of these showed clear evidence that the breakage occurred during flow, rather than before or after. Coarse abrasion or grinding on clast faces was interpreted as the result of impacts with other particles during flow rather than glacial abrasion. Overall, cataclasis and grinding were nearly as abundant in the Osceola Mudflow as they were in the largest noncohesive lahar at Mount St. Helens (Scott, 1988a). This result occurred in spite of the probable cushioning effect of the more cohesive matrix in the clay-rich Osceola Mudflow. IN PERCENT ROCK TYPES OTHER THAN RAINIER ANDESITE, 100 80 60 40 20 100 80 60 40 20 100 LAHAR-BULKING FACTOR, IN PERCENT U Figure 8. - Lahar-bulking factors for four lahars at Mount Rainier, and composition changes in the Osceola Mudflow. 80 60 40 20 100 FLOW MAGNITUDE AND FREQUENCY I I I I I T TPU I I I I I.’ TOT d Osceola Mudflow s --- 10-100 millimeter size range (Glicken and Schultz, 1980) » [~ _ --+ 16-32 millimeter size range; -" ® T, top; M, middle; and Fo [- B, bottom of unit o" -~" M x~" I _- a" F w* =~ N pp p p Ub U p p p p U I I T OT T TP OP Pd U I T OT T T T Td Osceola Mudflow and Greenwater Lahar (G) |- -* bat __._-__-.—--.’ meme === {~~ I L L_ 1 _L L L I 1 1 1 _ T T T --T-T-T-T-TT T T I. ® LA OTT Electron Mudflow "* o L 1 1 __ 1 _ L J J J L L L L __ L_ 1 _J I I T OT OT T OTP d I I T OT OT OT T TU National Lahar and runout phases |_ Lahar + Lahar-runout flow o _ 2 - # |__ ""% __ __ -s" |-» Pebbles in bars 1 mmm 1 * L L __ 1 OJ L J J G 1 1__ L_ 1 J J L I I I T OT OT T TUT I I T OT T TOT Td Tahoma Lahar |_ 'ad _._.,w we emma ** come " TI L L __ J OL UOJ OJ J 1 L __ L_ J_ L L L G 1 1 2 5 10 20 50 100 DISTANCE FROM SUMMIT OF MOUNT RAINIER, IN KILOMETERS 200 17 18 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON PARADISE LAHAR An initially huge, mainly noncohesive lahar (fig. 5, table 2) originated on the upper flank of the volcano above Nisqually Glacier and overran Paradise Park, the subalpine meadow between the Nisqually and Paradise Rivers. This flow, the Paradise Lahar (Crandell, 1971), also spilled across the demarcating Mazama Ridge to enter the Cowlitz River system. Radiocarbon dates within the same range as those of the Osceola Mudflow (4,500 to 5,000 years) indicate that the Paradise Lahar probably reflects the same edifice collapse as the Osceola Mudflow. What was most unusual about the Par- adise Lahar was that it attenuated rapidly and left only thin deposits, in spite of having reached depths of more than 300 m in canyons on the side of the volcano. The Paradise Lahar contains from less than 1 to as much as 6 percent clay with an average of 2.5 percent (four new composite samples). Mean grain size ranges from 0.6 to -2.8 ¢ (0.7 to 6.9 mm); sorting falls between 2.5 and 4.9 ¢. Skew- ness is generally slightly positive, ranging between -0.03 and +0.20. Fine sediment is not as abundant as in more cohe- sive lahars like the Osceola. The Paradise Lahar is noteworthy for several character- istics: (1) The size distributions are strongly influenced by the presence or absence of widely distributed, hydrother- mally stained clasts, including large boulders 2 to 3 m in intermediate diameter, set in a thin (less than 1.0 m) layer. (2) Crandell's measurements (1971, p. 33) of very large flow depths on the flank of the volcano were substantiated; consequently, the ratio of deposit thickness to original flow depth is very low. (3) These flow depths declined rapidly downstream, and thus the rate of flow wave attenuation, as in the Round Pass Mudflow on Tahoma Creek, is uncom- monly high. (4) The deposit is also remarkable for its occur- rence directly above layer O, revealing its inability to erode this very thin layer (less than several centimeters) of fine-grained tephra and forest duff where flow depths were more than 100 m. The above four features lead us to infer that the flow began, essentially as interpreted by Crandell (1971, p. 35, 36), as one or more huge avalanches. We interpret this as having been only a single avalanche, which had a degree of sudden, laterally directed momentum to create a sharply peaked flow wave in which the sediment was in part dis- persed. Although these features could be explained by a large vertical drop (Crandell, 1971, p. 36), we infer that the Paradise Lahar is more likely to have been initiated by vol- canic or phreatic explosive activity than the more broadly peaked flows originating as sector collapses. Such an origin also reconciles any difficulties in comparing the apparent flow volume and the volume of deposits (Crandell, 1971, p. 36). A similar explosive origin was proposed for some lahars originating with the 1980 lateral blast at Mount St. Helens (Major, 1984; Pierson, 1985; Scott, 1988b). The low clay content of the Paradise Lahar may reflect loss of fines by the process of explosively induced sorting (Scott, 1988b) but could, of course, simply be evidence of formation from less altered, less clayey rock. If the Paradise Lahar and Osceola Mudflow are the same age, their relations and origin are comparable to those of the 1980 lahars on the South Fork and North Fork Toutle Rivers, respectively (Scott, 1988b). In that case, a major sec- tor collapse was associated with explosively initiated lahars in peripheral watersheds. The probable synchroneity of the Paradise Lahar with the Osceola Mudflow clearly suggests that both flows had a comparable explosive origin. Most likely, the Paradise Lahar resulted from concomitant failure of part of the rim of the crater formed by the sector collapse that produced the Osceola Mudflow. The Paradise Lahar attained its maximum downstream depth of 240 m near the Ricksecker Point locality described by Crandell (1971, p. 36 and fig. 13), 1.2 km upstream from the confluence of the Nisqually and Paradise Rivers. Depos- its at this locality contain charcoal yielding an age of 4,625+240 radiocarbon years. The depth of the Paradise Lahar declined rapidly down- stream, but the flow was still at least 70 m deep near Long- mire, where the deposit also overlies layer O, has only 1 percent clay, and is as much as 1.2 m thick. At that location, wood from just above layer O yields an age of 4,955+585 radiocarbon years (fig. 9). The Paradise Lahar is known to extend to Ashford (Crandell, 1971, plate 3), but it probably inundated the deeply incised flood plain of the Nisqually River beyond National. At least 1.0 m of a noncohesive lahar with hydro- thermally stained clasts occurs below tephra set Y on the main valley floor at National (fig. 9); charcoal fragments in the upper part of the unit yield an age of 4,730+320 radiocar- bon years. This probable distal part of the Paradise Lahar is evidence of a flow volume near the upper limit of the range proposed by Crandell (1971, p. 36). Although the flow was generally noncohesive, a runout phase has not been identi- fied, possibly because of the age of the flow and the conse- quent loss of the deposit by erosion or burial by later, post-set-Y aggradation. No dates for the Paradise Lahar were reported by Cran- dell (1971) who noted, however, that the flow occurred between the times of tephra layers O and D, a range between about 6,800 and 6,000 radiocarbon years ago (table 1). The reason for the discrepancy with the three radiocarbon dates reported above was investigated. Interestingly, Crandell (1963a, p. 138; 1971, p. 35) originally thought the Paradise Lahar and Osceola Mudflow were the same age before the tephra evidence apparently negated the possibility. Our radiocarbon dates support Crandell's original inference; the new radiocarbon dates conform in near equivalence to the age of the Osceola, and conflict with the tephra evidence. One possible explanation for this conflict is that the FLOW MAGNITUDE AND FREQUENCY 19 Longmire (0.4 kilometers upstream from town of Longmire) Set W at surface Lahar, locally with clay-rich megaclasts, Inset Sequence National (Near town of National) Main Flood Plain Runout phase of the National Lahar, with dish structure Set W at surface noncohesive matrix . . . Fluvial sediment, colluvium Lahar-runout flow and { alluvium Set Y Pe Paradise Lahar, with __ \-} hydrothermally . @ stained clasts Date: 4,955 + 585m tan, noncohesive Date: 2,285 + 155 -__ radiocarbon years National Lahar; yellowish- { is Fie: Megaclasts with set Y Soil r _/— Set P =-- Ash-rich soil, colluvium --- Set Y, highly bioturbated, over soil 4 ~- Date: 3,655 + 245 radiocarbon years ~* Date: 4,730 + 320 radiocarbon years" radiocarbon years Lahar-runout flow {.~:= Layer 0 iad Lahar, noncohesive Alluvium { t Fluvial sand, gravel reworked Y tephra Fluvial sediment, probably post-Evans Creek in age; possibly correlative with Winthrop Creek Glaciation Glacial-outwash sediment, probably Evans Creek in age Soil on surface with --" tp: ; Paradise Lahar (?) Inset Sequence in Gorge Lahar _ 7 Date: 410 + radio- carbon years Lahar-runout flow Fluvial sediment Figure 9. - Composite columnar sections of lahars and associated deposits at Longmire and National in the upper Nisqually River drainage. "Paradise Lahar" deposit on which Crandell based his tephra stratigraphy may actually be a nearly identical but somewhat older flow. He reports (1971, p. 33) a thin Paradise Lahar deposit underlying layer D on the east side of Paradise Valley near Sluiskin Falls. The tephras there confirm an older age, as does a date of 6,950+355 radiocarbon years obtained by us from wood collected from the lahar at the site. Thus, the main Paradise Lahar is, as originally surmised by Crandell, close to or synchronous in age with the Osceola, and the tephras correctly date an older, smaller lahar near Sluiskin Falls. ROUND PASS MUDFLOW (BRANCH ON TAHOMA CREEK) A large lahar is exposed at Round Pass from which flow diverged into both the Puyallup and Nisqually River drainages, the latter by way of Tahoma Creek (fig. 1B). Crandell (1971) showed that the branch on Tahoma Creek attenuated rapidly, is post-set Y in age, and has a radiocar- bon age of 2,610+350 years. A date obtained by us from wood in valley-bottom deposits was 2,790+130 radiocarbon years. The texture is highly variable at the three main exposures, ranging from cohesive to noncohesive as fol- lows: Round Pass, 2-8 percent clay; valley bottom, unit 3 of Crandell's section 9, 3-5 percent; and Indian Henrys Hunt- ing Ground, 3 percent. Mean grain size is between +0.5 and -3.5 $ (0.7 to 11.0 mm); sorting ranges from 3.5 to 5.0 6 (five samples). Skewness is not uniform in direction, vary- ing between -0.10 and +0.18. Rounded clasts, especially in the coarse fractions, characterize the unit at the type locality. A partial but less-than-satisfactory explanation is that valleys on both sides of Round Pass were filled with outwash gravel or debris-rich glacial ice to a higher level than they are now, and thus that the lahar was shallower than the present topography would suggest. In this event the lahar could have transported entrained material to the height of Round Pass. An early Neoglacial advance culminated in the time interval of 2,600 to 2,800 years ago (Porter and Denton, 1967), and 20 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON a synchronous advance probably occurred at Mount Rainier (Crandell and Miller, 1974). However, the lahar contacted the modern valley bottom only 3 km away from Round Pass, and Crandell (U.S. Geological Survey, written commun., 1989) does not believe that the Neoglacial advance could have filled the valleys significantly near Round Pass. Some rounded material in the lahar may have come from Evans Creek Drift in the vicinity (Crandell, 1969, pl. 1), and some undoubtedly came from local deposits of outwash gravel, which were probably more extensive and abundant than they are now. The most likely origin of the flow was as a debris avalanche from the Sunset Amphitheater mainly confined to the watershed of the South Puyallup River. The part spilling over into the Tahoma Creek valley, although possibly more than 300 m deep in the headwaters of Tahoma Creek, apparently attenuated rapidly and was a sharply peaked distributary of the flow, subsidiary to the main part that traveled a longer distance in the Puyallup River valley. The pattern of erosion of tephra set Y by the Round Pass Mudflow clearly reveals the dynamics of the flow and shows that Crandell's (1971, fig. 25) interpretation of rapid attenuation is the most likely explanation of its limited extent along Tahoma Creek. The unit's most distal exposure in that valley is 3 km upstream from the mouth of Tahoma Creek, where it supported a forest inundated by the younger Tahoma Lahar. ROUND PASS MUDFLOW (BRANCH ON PUYALLUP RIVER) - The behavior of the Round Pass mudflow in the Puyallup River system was substantially different from that of the sharply peaked distributary of the flow to the south. The initial flow was a broad lahar wave that probably reached the Puget Sound lowland. Ages of 2,710+250 and 2,170+200 radiocarbon years were obtained by Crandell (1971) from wood near the confluence with the Mowich River. A piece of wood collected by us at the same location yielded an age of 2,440+290 radiocarbon years. The flow overran a forest 4.5 km upstream from the confluence of the North and South Puyallup Rivers, just outside the park boundary. The outermost 10 rings of a buried tree from that forest yielded an age of 2,600+155 radiocarbon years. Three of the four dates are, therefore, consistent with the age of the distributary of the flow in the Tahoma Creek drainage. Clay content is variable, as in the deposits along Tahoma Creek, but the Puyallup River deposits are mainly in the cohesive category if megaclasts are excluded from the size distributions. Two upstream samples contain 4 and 5 percent clay, and downstream the deposits become more cohesive. Mean grain size of the upstream samples is -1.6 and -1.3 ¢ (3.0 and 2.4 mm), respectively, sorting is 4.2 and 4.9 $, and skewness is slightly positive. Where the unit is as much as 16 m thick near the confluence of the Mowich and Puyallup Rivers, large exposed megaclasts and the presence of many mounds farther upstream establish a slope-failure origin (as a debris avalanche) that is consistent with the generally cohesive downstream texture. That these mounded deposits probably are not those of an untransformed debris avalanche is indicated by their texture, both on Tahoma Creek and at the site of the buried forest mentioned above. At the latter locality the deposits between megaclasts have the character of a lahar, are generally cohesive in texture, and represent an upstream part of the flow distributed on the flank of the volcano as shown by Crandell (1971, fig. 25). The deposits in the upper Puyallup valley are more likely a mounded facies of a large lahar, similar in origin to the deposits mapped as the Greenwater Lahar. Both flow depth and velocity were remarkable at the site of the buried forest near the park boundary. There the megaclast-bearing flow knocked down trees 240 m above the valley bottom on the south, outer side of a broad, north- erly valley curve that begins upstream from Round Pass. The peak flow level is defined by a terrace with a mounded sur- face and a small, ephemeral lahar-margin lake. Estimates of minimum runup on lateral ridges strongly suggest a peak velocity of at least 40 m/s. Round Pass, only 3.0 km upstream from the buried forest, is 170 m above the valley bottom in a more confined reach; consequently, flow in the South Puyallap River valley was certainly deep enough to send a major distributary across Round Pass into Tahoma Creek as shown by Crandell (1971, fig. 25). Flow across divides farther upstream is even more likely. Concomitant with these findings is a probable hydraulic explanation of the high attenuation rate of the flow in the Tahoma Creek valley. Flow across the divides would have occurred only during the relatively brief passage of the peak of the high velocity flow, resulting in exactly the highly peaked, rapidly attenuating flow(s) recorded by the texturally variable deposits in Tahoma Creek valley. A noncohesive lahar and - lahar-runout deposit, which have bounding ages of 840+190 and 2,740+230 radiocarbon years, and an older runout deposit 3,530+255 radiocarbon years in age, occur 6.0 km downstream from the boundary of the Puget Sound lowland (fig. 10B). The youngest two of the three units, described below, could be distal correlatives of the Round Pass Mudflow based on their ages. However, we believe that neither of the units is a likely correlative because of their noncohesive texture. A more probable origin was as meltwater surges resulting from volcanism as in the case of the younger two deposits exposed upstream, which formed near or following the time of the block-and-ash flow in the South Puyallup River valley. The deposit of the block-and-ash flow is noncohesive (with only 1 percent clay), and it is 2,350+250 radiocarbon years old (Crandell, 1971). The Round Pass Mudflow probably extended to the Puget Sound lowland intact as a cohesive debris flow, untransformed to a hyperconcentrated runout. FLOW MAGNITUDE AND FREQUENCY 21 Peak stage of Electron Mudflow----------; Date from probably correlative unit: --- / 548 10,175 +365 radiocarbon years Au Electron Mudflow Monolithologic lahar Brown cohesive lama—r} Electron Mudflow Y in section Lahar with megaclasts Glacial deposits METERS 10 Date: 2,170 + 200 radiocarbon 5 years (Crandell, 1971) Electron Mudflow 0 Date: 1,050 + 350 radiocarbon years (Crandell, 1971) Vertical scale is approximate METERS | _ ap ccs Peak stage of Electron Mudflow 10 Date: 550 + 190 radiocarbon years Electron Mudflow-"~ Lahar-runout flow gga) _ --Date: 840 +190 radiocarbon years (base of Electron Mudflow) 5 Noncohesive lahar- - ~> .e fgatei 2,740 +230 radiocarbon years (base of anncohesive lahar) Lahar-runout flow ate: 3,530 + 255 radlocargon zears (charcoal in runout) Alluvium ———'"T\J: gay 0 Glacial deposits rod o oe o Vertical scale | Alluvium is approximate B Figure 10. - Diagrammatic composite sequences of valley-fill deposits in the Puyallup River valley (successive downstream sections). Vertical scale is shown to indicate approximate thicknesses. Horizontal scale is variable for better portrayal of stratigraphic relationships and is not shown. However, vertical exaggeration ranges approximately between 5x and 10%. A, Puyallup River at Mowich River con- fluence. B, Puyallup River 6.0 km downstream from boundary of Puget Sound lowland. UNNAMED PRE-ELECTRON DEPOSITS, PUYALLUP RIVER SYSTEM The oldest postglacial lahar recognized during this study occurs above glacial drift and below set Y in the Puyallup River system (fig. 104). The gray, generally noncohesive unit contains wood yielding an age of 10,175+365 radiocarbon years. Because tephra layers show that the valley configuration did not change greatly in postglacial time, the height to which the deposit extends above the present channel, about 100 m near the Mowich River confluence, indicates the flow was large. Deposits of brown or gray cohesive diamicts of uncertain age and origin occur locally in the Puyallup River system (fig. 104). Some exposures probably are of a non-megaclast-bearing facies of the Round Pass Mudflow; others are pre-Y in age, and their correlation is not certain. In any case, the units record at least one cohesive pre-Y, postglacial lahar that attained levels approaching, but below, the peak stage of the younger Electron Mudflow. South of the Mowich-Puyallup confluence (fig. 104), a brown cohesive diamict underlies a strikingly monolithologic and noncohesive debris flow deposit consisting of clasts of black vitric Rainier andesite. However, at the upstream site of the buried forest, a megaclast of an identical deposit is incorporated in the Round Pass Mudflow, the identification of which is verified by a nearby radiocarbon date. Thus, the brown unit below the black lahar near the confluence is probably not the Round Pass Mudflow but is most likely a pre- Y lahar. The monolithologic lahar reaches at least 43 m above the valley bottom near the Mowich River confluence. The 22 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON lahar is granular and resembles a lithic pyroclastic flow, but it is distinct from the deposit of the block-and-ash flow exposed on the volcano (table 1). The unit very likely represents the cooled downstream continuation of such a flow, and it is closely similar to a pre-Osceola unit in the White River valley. The flow was probably larger than that producing the block-and-ash flow deposit seen on the volcano; that unit has not been traced downstream with certainty, although a granular, lithologically similar lahar with a pronounced content of prismatically jointed clasts is locally present beneath the Electron Mudflow in low, channel-bank exposures. A large noncohesive lahar, untransformed to a runout phase, is recorded by deposits on the Puget Sound lowland in the Puyallup River system (table 3, fig. 10B). The age of the unit and its probable lack of correlation with the Round Pass Mudflow are discussed above in the section on the Round Pass Mudflow (Puyallup River branch). The lahar, which has an approximately 20-cm-thick, Type I sole layer (sandy, without a dispersed coarse phase), has inverse and normal grading identical to that of many granular lahars at Mount St. Helens (Scott, 1988b). The unit crops out at a level at least 6 m above the present river, 0.1 km upstream from the valley constriction at the first bridge upstream from Orting, and 6.0 km downstream from the lowland boundary. Clasts in the coarse mode are no larger than pebbles in size. The loss of coarser gravel and a locally intact framework indicate that transformation occurred not far downstream. Runout sands underlie and overlie the unit. The lack of weathering during the hiatus between the flows indicates that litle time separated the lahar and the overlying runout deposit. 1,000-YEAR-OLD LAHAR This clayey lahar extends down the Puyallup River system at least as far as a point 1.6 km below the mouth of the Mowich River (Crandell, 1971) but may have extended much farther. Wood yielding a radiocarbon date of 990+130 years was collected from beneath a clayey lahar, which is exposed in a roadcut 1.0 km up Fox Creek on the Puget Sound lowland. This lahar flowed into a reentrant, Vashon-age (Crandell, 1963b) hanging valley about 1 km upstream from the lowland and then into Fox Creek before reentering the Puyallup River. Because the height of this deposit above the present river (37 m) is typical of Electron deposits at this point, it is more probably the Electron Mudflow than the 1,000-year-old lahar. The possibility exists, however, that a cohesive lahar extended to the lowland about 1,000 years ago. ELECTRON MUDFLOW This cohesive lahar in the Puyallup valley (table 2) is like the Osceola Mudflow, although it is smaller in volume. Nevertheless, the Electron is still large, and its volume is more typical of the group of "large but infrequent" cohesive lahars that have formed at Mount Rainier in postglacial time. Like the Osceola, the Electron is relatively clay-rich (table 2, fig. 5) and has a significant LBF (fig. 8), which acted to reduce the rate of downstream attenuation and volume decline. The volume of deposits as measured by Crandell (1971, p. 57) on the Puget Sound lowland is accepted, but an additional volume representing an estimate of postdeposi- tional erosion and dewatering losses is included in the value given in table 2. The lost volume was estimated by assuming an even, nearly horizontal original surface and then determining the volume represented by the difference between that surface and the present lower, dissected surface. The clay content of the Electron Mudflow ranges from 6 to 11 percent with a mean of 8 percent (four samples). Its mean grain size is finer than that of the Osceola, in part because nearly all exposures are of thin deposits on steep valley side slopes; values range from +1.7 to -2.2 $ (0.3 to 4.5 mm). Sorting ranges from 4.2 to 5.1 $, and skewness values are positive, from +0.06 to +0.38. The unit contains only a few scattered megaclasts, forming mounds on lateral deposits in upstream reaches; the deposits are notable for a relative scarcity of mounds compared to those of other flows believed to have had the same origin of sector collapse and mobilization of the consequent debris avalanche. Crandell (1971) described large coherent boulders on the lowland, as opposed to the less coherent mound-forming megaclasts representing pieces of the failed edifice, and similar boulders were observed upstream. A wood sample obtained by Crandell (1971) near Electron yielded an age of 530+200 radiocarbon years. A sample collected by us downstream, 4.0 km below the town of Electron, yielded a date of 550+190 radiocarbon years (fig. 10B). Forest duff at the base of the deposit, 5.5 km downstream from Electron, yielded a date of 840+190 radiocarbon years. As noted by Crandell (1971), no volcanic activity has been recorded at Mount Rainier near the time the Electron Mudflow occurred. Although set W is not well developed in the Puyallup River valley (Mullineaux, 1974, fig. 18), we have tentatively identified a distal version of the tephra, which overlies the Electron on the surface of the terrace described by Crandell (1971, p. 57) near the mouth of St. Andrews Creek. This stratigraphy corresponds to the most probable absolute ages of the events. These ages are critical because of the closeness in time of both events to the Tahoma Lahar, a smaller flow in the valley of Tahoma Creek (described subsequently) which is clearly of debris-avalanche origin. Set W was deposited just before that flow and can be documented to underlie it in new exposures. The Tahoma Lahar correlates with unit 9 of Crandell's measured section 9 (1971, p. 58), where he recognized set W at the base of the deposit. FLOW MAGNITUDE AND FREQUENCY 23 Table 3. - Mainly noncohesive debris flows and their runout phases at Mount Rainier. [Many small flows excluded] Drainage and flows Age' Extent (as flow large enough to inundate flood plain) White River (including West Fork) Large gravel-rich debris flow and flood gravel extensively aggraded ~A.D. 1550 At least to Mud Mountain Reservoir. present channel; nonvolcanic in origin. At least one lahar-runout flow in both main and West Fork, the latter Post- W At least 5 to 10 km outside park boundary. valley-wide below park boundary. f Lahar in West Fork 5 km above confluence of the forks ... Post-C, pre-W _ Unknown; at least to confluence of forks. Transition facies in West Fork, 6 km above CORfIUENCE Post-C, pre-W _ Unknown; possibly to Puget Sound lowland. Dead Man Flat lahar assemblage--transition facies filled valley from near Post-C, pre-W _ At least 11 km on Puget Sound lowland as large runout Fryingpan Creek to Buck Creek; runout flow farther downstream. flows from both main fork and West Fork; flows probably reached Puget Sound. At least five lahar-runout flows inset in channel incised in Osceola, near Post-Osceola, _ Most flows in this group probably reached the margin confluence of forks of White River. pre- W. of the Puget Sound lowland. At least two lahar-runout flows, near Buck Creek, Greenwater, and also Pre-Osceola Puget Sound lowland. near Mud Mountain Reservoir. Cowlitz River At least two runout flows Post-W At least 10 km downstream from Packwood. Lahar-runout flow Post-C, pre-W _ Packwood. At least three lahars Probably post- _ At least to park boundary in Muddy Fork. Y, pre-W. Nisqually River Debris flow and runout flow of glacial-outburst origin in Kautz Creek .... A.D. 1947 Only locally overbank below confluence with Nisqually River. Lahar and lahar-runout flow Post-W At least to Elbe. Lateral parts of Tahoma Lahar Post-W At least to Elbe. Lahar-runout flow Post-P, pre-W _ At least to National. Lahar-runout flow Probably To Elbe. post-P. National Lahar (runout phase inundated all valley bottoms above Alder Post-C, pre-W _ Puget Sound. Reservoir to a depth of at least 3 m). Lahar-runout flow Post-Y, pre-W _ At least to Ashford. Large lahar and lahar-runout flow pre- Y Probably to Puget Sound lowland. Puyallup River Lahar-runout flow Post-Y, pre- _ Puget Sound lowland. Electron." Lahar Post-Y, pre- Puget Sound lowland, untransformed to runout flow. Electron." Lahar-runout flow Immediately _ Puget Sound lowland. pre- Y. Carbon River Lahar-runout flow Post-W , At least 5 km below glacier terminus. Lahar-runout flow Pre-W 8 to 10 km below glacier terminus. e Ages of tephras shown in tables 1 and 2. 2 Flows closely related in time. A cohesive diamict caps a terrace along a logging road on the north side of the Mowich River about 2 km upstream from its mouth. The unit was mapped as Round Pass Mudflow by Crandell (1971). A log at the base of the unit yielded a radiocarbon date of 530+90 years (fig. 104), indicating that the unit is the Electron Mudflow, which was more than 25 m deep near this locality. The Electron Mudflow, as interpreted by Crandell (1963b, p. 69), dammed the drainage of Kapowsin Creek to form Lake Kapowsin. The lake has a maximum depth of 9 m (Crandell, 1963b), which is deep for a lahar-margin lake (Scott, 1989) and suggests that the strength of the flow was significant (Johnson, 1984, equation 8.6¢). The original flow margin may or may not have had that much relief, however; 24 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON according to local residents, the lake level has risen substantially in the last century due to outlet blockage, either natural or constructed. OTHER LAHARS AND POSSIBLE LAHARS Other cohesive lahar deposits were reported by Crandell (1971) or observed by us. These observations were mainly at single localities on the volcano, but because the deposits are cohesive, it is possible that some of the flows were large enough to have extended long distances. Some were early in postglacial time, and their deposits on down- stream valley-side slopes may have been eroded. The deposits listed here were observed at altitudes or with thicknesses indicating that the flows were of significant size. Deposits noted by Crandell (1971) include unnamed deposits along the South Puyallup River (pre-Y in age); the relatively young unit 3 in section 10 of Crandell (1971); deposits at Round Pass (pre-Y in age); a lahar at Van Trump Park (pre-O in age); and a lahar older than the Paradise Lahar at Paradise Park. Additional clay-rich lahars discovered during this study include a post-Paradise, pre-Y lahar on Mazama Ridge above Reflection Lakes; a post-0O, pre-Osceola Mudflow lahar locally preserved near Buck Creek and Greenwater; and a post-Osceola Mudflow, pre- Y lahar along the White River near Fryingpan Creek. The pre- Y lahar on the South Fork Puyallup River may record a flow approaching 180 m in depth in that tributary (Crandell, 1971). A possible correlative consists of slumped, clay-rich deposits near Electron that yielded wood fragments with a date of 3,760+350 radiocarbon years. This deposit suggests that yet another cohesive lahar in the Puyallup valley reached the Puget Sound lowland, but it may also be interpreted as a slumped glacial deposit that incorporated younger wood. SYNTHESIS OF THE RECORD OF LARGE, LOW-FREQUENCY LAHARS At least six lahars can be documented to have inundated parts of the Puget Sound lowland or can reasonably be inferred to have done so. A seventh, the Greenwater Lahar, is not interpreted as a separate flow. These lahars occurred after deposition of layer O, 6,800 radiocarbon years ago. At least seven other postglacial flows were cohesive and, therefore, possibly large enough to have reached the lowland. The recurrence interval of these largest debris flows therefore is in the range of 500 to 1,000 years. FLOWS OF INTERMEDIATE MAGNITUDE AND FREQUENCY (100 TO 500 YEARS) Debris flows in this intermediate range were analyzed for the time following the deposition of tephra layer Yn (3,400 radiocarbon years) or C (2,200 radiocarbon years), depending on drainage. Erosion or burial by the inset deposits of younger flows prevents a complete analysis of the intermediate-size flows older than these tephras. This span of 2,200-3,400 years is a sufficient period of record for these flows because the types of debris flows that occur at Rainier have not changed significantly in postglacial time. Intermediate-size flows are dominated by noncohesive lahars and their runouts (table 3). The runout phases consisted primarily of hyperconcentrated streamflow, which extended to the Puget Sound lowland or to a large down- stream flood plain at least several times in all drainages except the Carbon River. WHITE RIVER SYSTEM Sequences of noncohesive lahars and their runout phases occurred at the following times: before the Osceola Mudflow; after the Osceola and before set C; between sets C and W; and after set W (table 3). Pre-Osceola flow deposits are seen only at the few places where the base of the Osceola Mudflow is exposed. All of these sequences are exposed near the Mud Mountain Dam and, therefore, at least locally inundated the downstream Puget Sound lowland. The largest flows are between tephra sets C and W in age. They probably resulted from the volcanism that was responsible for construction of the summit cone during part of this time interval (table 1). Because the flows were dominantly noncohesive, most of them probably originated as meltwater flood surges that bulked with sediment on the side of the volcano and then debulked, in most cases beginning near the base of the volcano. The melting may have been the result of lava flows, pyroclastic flows and surges (both mainly lithic in composition), steam eruptions, or extensive geothermal heating. Where approximately synchronous flows occurred in more than one watershed, they are probable evidence of significant episodes of volcanic or geothermal activity. Tephra-producing eruptions were not a general cause of the flows. Only pumice of tephra set C, which was erupted about 2,200 radiocarbon years ago, is found in significant amounts in any flow deposit, and radiocarbon dating shows that this pumice was mainly entrained through erosion. Some flows originated from shallow landslides, indicated by a high content of lithologically similar, hydrothermally stained clasts. The largest noncohesive flows in the White, Nisqually, and Puyallup River systems probably formed about the same time, if not synchronously, from summit-cone volcanism. Crandell (1971) did not describe the runout phases of lahars, but he noted that the extensive aggradation in the White and Nisqually River systems, which resulted largely from runout flows, could be ascribed to summit-cone volcanism. That volcanism may have begun near the time of the block-and-ash flow in the South Puyallup River valley, 2,350 radiocarbon years ago (Crandell, 1971). 'Table 4. - Radiocarbon dates from hyperconcentrated-flow and normal streamflow deposits in the Mount Rainier area. FLOW MAGNITUDE AND FREQUENCY [Data mainly from lahar runout-flow deposits, but may include deposits of runout flows from debris flows not of volcanic origin} 25 1 Location Stratigraphy and significance Age White River (including West Fork) 4.8 km upstream from confluence of Buck Creek 'Trees killed by burial by flood deposit of nonvolcanic origin 400 + 75 and White River. 2.3 km upstream from White River Campground Wood in upper part of lahar-runout deposit, 1.0 m thick ...................... 810+ 75 Confluence of Fryingpan Creek and White River Wood in Dead Man Flat lahar assemblage; unit includes layer-C pumice. 1,120 + 80 In West Fork, 6 km upstream from confluence Stump on large lahar-runout flow deposit, more than 1.8 m thick; overlain 1,255 £130 with White River. by 7 layers of flood deposits. 1.6 km upstream in Clearwater River .................. Wood below lahar-runout flow from Mount Rainier; a maximum age. 3,005 + 230 1.3 km upstream from confluence of Buck Creek Charcoal fragments from base of noncohesive lahar beneath Osceola 5,230 £235 and White River. Mudflow. 1.3 km upstream from confluence of Buck Creek Charcoal fragments from noncohesive lahar underlying above unit and 6,075 £320 and White River. overlying layer O. Cowlitz River South bank of active channel 1.8 km upstream Wood in lowest of four silty sand overbank deposits, 0.2 to 0.5 m thick, 325 £180 from Randle (river mile 104.3 on Randle representing floods occurring after youngest lahar-runout flow. 15-minute quadrangle). North side active flood plain 10 km downstream Wood from base of 0.5 m silt layer over lahar-runout flow with layer-C 440 + 70 from Randle (river mile 117.5). pumice; dates major flood. South bank of abandoned meander 5.5 km Charred wood bioturbated with set W, in upper part of lahar-runout $15 +120 upstream from Randle (near river mile 108.0). deposit, below 1.0 m of silt representing large post- W flood. Nisqually River 12.5 km upstream from main bridge crossing Bark from silt-rich unit burying cedar forest; dates flow that killed 220 £ 70 river near Yelm. flood-plain forest. 1.2 km upstream from boundary of Nisqually Wood from same unit as above 240+ 60 Indian Reservation. 0.5 km downstream from National ...................... Wood from lahar-runout flow younger than National; inset sequence in 410 +75 gorge cut in valley bottom. 1.4 km upstream from boundary of Nisqually Outermost wood of 250- to 350-yr-old cedar buried by two silt-rich flood 585 +125 Indian Reservation. deposits; tree grew on deposit correlated with the National Lahar. 6.2 km downstream from Alder Reservoir .......... Wood within upper part of 3 m of hyperconcentrated-flow deposits 790 £ 205 correlated with the National Lahar. 1.6 km downstream from bridge crossing Wood from top of fluvial unit underlying transition facies of National 2,285 £155 Nisqually River in National Park. Lahar; maximum date of that flow. Puyallup River 0.1 km upstream from main highway bridge Charcoal fragments from contact between lahar-runout flow and 840 + 190 below Electron. underlying large cohesive lahar. Do Charcoal fragments from contact between lahar-runout flow and 2,740 £230 overlying large noncohesive lahar. Do. Charcoal fragments in lahar-runout flow under large non-cohesive lahar. 3,530 £ 255 Carbon River 4.8 km downstream from glacier terminus .......... Probable lahar-runout deposit in low terrace (2-4 m above active 650 + 120 channel). ! Years before 1950 in radiocarbon years. The largest post-C, pre-W noncohesive flows in the White River system are informally designated as the Dead Man Flat lahar assemblage (fig. 6), believed to consist of several nearly synchronous flows, at least one from each fork of the White River. The flows of the assemblage are lahar-runout flows over most of their longitudinal extent. Wood from the flow in the main fork at Fryingpan Creek yields a date of 1,120+80 radiocarbon years (table 4). Although layer-C pumice is abundant in this flow deposit, the date, from a limb segment with bark that was completely contained within the deposit, is probably an accurate measure of flow age. The flow deposit is immediately over- lain by a unit interpreted as a blast deposit, with a radiocar- bon age of 1,080+25 years (table 1) determined by RP. Hoblitt (U.S. Geological Survey, oral commun., 1994). The flow apparently came down the main fork and did not 26 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON originate in Fryingpan Creek. Downstream from the National Park boundary near Buck Creek, the transition facies of this flow locally overtopped the mound-bearing Osceola surface as much as 60 m above the White River (fig. 6A). Downstream from the confluence with the West Fork, the runout deposits of the flow are interbedded with those of a similar flow, also containing set-C pumice, which probably originated in the West Fork White River at about the same time. The outermost wood of a small stump on a large runout deposit along the West Fork yielded a date of 1,255+130 radiocarbon years (table 4). Even though the tree was 55 years old, its corrected age (1,310+130) still overlaps with the radiocarbon date (1,120+80) for the Dead Man Flat assemblage on the main fork. Total thickness of the assemblage, bounded by sets Y and W, is locally more than 3 m. From 6 km below Greenwater to the Mud Mountain Dam (fig. 6B), deposits of the assemblage occur at least 30 m above the White River. Even 11 km beyond the Cascade Range front, the flows reached nearly 20 m above the present river. The dated occurrences of noncohesive flows show a concentration near but well beyond the end of the period of summit cone volcanism. This period was interpreted by Crandell (1971, p. 14) to be between 2,100 and 1,200 absolute years ago, about the same as the range in radiocar- bon years (Stuiver and Becker, 1986). Dating of the noncohesive flows in the White and other river systems indicates that the lahar-producing volcanism continued to at least 800 years ago. Wood fragments from deposits of a small flow upstream from the White River Campground yielded a date of 810+75 radiocarbon years (table 4). A flood deposit that extensively aggraded the present White River channel and killed many trees contains wood yielding an age of 400+75 radiocarbon years. The deposit was only locally emplaced as a debris flow and is dominated by non-Rainier and reworked Rainier rock types. It is probably of nonvolcanic origin. COWLITZ RIVER SYSTEM Typical lahar-runout flow deposits occur throughout the Cowlitz River system, from the headwaters of the Ohanapecosh River to cutbanks in the flood plain near Packwood. The watershed was notably less affected by lahars and lahar-runout flows than either the White River system to the north or the Nisqually River system to the west. This difference probably reflects the lack of major, deeply incised valleys in the sector of the volcano drained by the Cowlitz River. Although the three separate flows observed (table 3) were overbank, they were not markedly greater than historical floods in the watershed. An overbank thickness of about 0.5 m is typical for the lahar-runout deposits upstream and downstream from Packwood and is comparable to the thicknesses of interbedded flood deposits (table 4). The lahars observed in the upper Muddy Fork (table 3) average about a meter in thickness, a value also comparable to the thickness of younger flood deposits in that area. Except for one locality, radiocarbon dates could be obtained only from flood deposits overlying the runout flows. A date from wood within a runout deposit was 815+120 radiocarbon years (table 4). Other minimum ages of runout flows, and the probable actual radiocarbon ages of major-flood deposits occurring above them, are 440 +70 and 325+180 radiocarbon years. The dated runout deposit is consistent in age with others in the White, Nisqually, and Puyallup River systems (table 4). NISQUALLY RIVER SYSTEM The oldest noncohesive lahar and runout flow to be recorded at multiple locations is pre- Y in age. It occurs at scattered exposures upstream of Longmire, indicating an origin from the part of the watershed headed by the Nisqually Glacier. Although not seen in direct contact with the Paradise Lahar, it is probably younger. Other pre-Y flows certainly existed, but their deposits were seen at only a single locality and could not be correlated. Characteristics used to correlate flow deposits in the absence of age information include soil development, matrix color, alteration products on and in clasts, and deposit texture. A series of noncohesive lahars and lahar-runout flows occurred between the deposition of tephra sets Y and W (table 3). The flows were part of the aggradational cycle that resulted from summit-cone volcanism, as described above for the White River system. At some localities, set P (3,000 to 2,500 radiocarbon years in age) can be distinguished for further age refinement (figs. 9 and 112). Because none of the Rainier tephras are present on the west side of the volcano, which faces prevailing winds, further tephra-based dating was not possible. The best exposed lahar of this series, comparable in size with those in the White River system, can be traced to Puget Sound and is here informally named the National Lahar after the town of National (figs. 9 and 11B), a designation incorporating the runout phase. The National Lahar and its runout phase have textures typical of a noncohesive lahar and its hyperconcentrated runout-flow deposits (fig. 2). Mean grain size of the lahar is -1.2 ¢ (2.3 mm); sorting is a Figure 11 (facing page). Diagrammatic composite sequences of valley deposits in the Nisqually River valley (successive down- stream sections). Vertical scale is shown to indicate approximate thicknesses. Horizontal scale is variable for better portrayal of strati- graphic relationships and is not shown. However, vertical exagger- ation ranges approximately between 5% and 10x. A, Nisqually River upstream from Longmire. B, Nisqually River between Ashford and National. C, Nisqually River downstream from Yelm. FLOW MAGNITUDE AND FREQUENCY 27 /—W on surface A---Y in section W on surface /—Soil Colluvium @@. -Alluvium -a [National Lahar, transition facies /— Date: 2,285 + 155 radiocarbon years Lahar- -~ - runout flow := METERS Lahar-runout flow 4A Alluvium == Soil with reworked Y pumice a @ @ --Glacial outwash; possibly Winthrop Creek @ W on surface o 9... Debris] Runout flows Vertical scale is approximate ‘. Post- W debris flows x/~ 1955 flood alluvium Peak Stage of runout LLL ,: n... Terrace of glacia'- flow of National Lahar outwash sediments, Evans Creek in age Runout flow of National Lahar; with dish structure METERS A Lah A o Y in section [* *" \ Date: 3,655 + 245 radiocarbon years ; See figure 8 9 WQ _ A ®sz" Date: 4,730 + 320 radiocarbon years /’ Paradise Lahar (?); base not seen 0 Alluvium Date: 410 +75 radiocarbon years Vertical scale is approximate B Live tree, 292 years old Tree date: 585 + 125 radiocarbon years Glacial deposits a | \ /—Elood slainhnotdinundated § v - modern floods y- by \Date: 240 + 60 radiocarbon years METERS \ *e s" ~<<-eo Zul - "a Silty flood deposits 4 \ regerntp rrr Qy __ Nei - 8 o*\~-- F \ eff CZ hha -**0*."\ Channel alluvium ~~ Rave fere or ftm 2 Runout flow of Y- Nisqually River National Lahar o with cross-bedding Vertical scale is approximate C 28 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON relatively low 2.0 ¢, barely within the range characteristic of flows with sufficient strength to support dispersed coarse clasts (Scott, 1988b). Mean grain size of the runout deposits is +0.5 $ (0.7 mm); sorting is 1.3 $, within the range common to runout flow deposits (1.1 to 1.6 ¢; Scott, 1988b) and reflecting the progressive downstream loss of strength. The lahar transformed to hyperconcentrated flow near the National Park boundary, as indicated by flow deposits con- sisting of the transition facies (Scott, 1988b, fig. 10) above Longmire: The flow may have originated in more than one tributary of the Nisqually River, but a significant part of the flow was derived from the headwaters of the main stream. The inclusion of erosionally derived clasts of tephra layer C, which was distributed mainly to the east of the volcano but also into the upper Nisqually River watershed (Mullineaux, 1974, fig. 24), supports this probability. Downstream from the National Park boundary, the deposits are largely those of hyperconcentrated flow. The further transformation to normal streamflow is marked by the appearance of well-defined stratification and of sorting below 1.1 ¢. Mean grain size of the deposits near Yelm is 3.1 $ (0.1 mm); sort- ing is 0.6 ¢. A noteworthy feature of the runout phase of the National Lahar is the presence of dewatering structures (fig. 12). Dish structure, named by Wentworth (1967) and correctly interpreted by Lowe and LoPiccolo (1974), is well developed in the longitudinal interval representing the upper part of the hyperconcentrated range of sediment content. This interval corresponds to the approximately 30 km of flow downstream from the distal end of the transition facies, the point at which the transformation from debris flow to hyperconcentrated flow was complete. Dish structure in the runout deposits of the National Lahar consists of concave-upward, strata-like concentrations of darker, commonly finer sediment that truncate each other laterally and locally resemble cross strata. Concavity increases upward within 2- to 3-m sections of the runout deposits, locally producing closed, concretionary forms near the top of the unit. The structure clearly is not antidune cross-bedding as interpreted by Wentworth. The basic mechanism is, as interpreted by Lowe and LoPiccolo, expulsion of water from the deposit shortly after deposition. However, the more pronounced development of the struc- ture in older runout flows, such as that of the National Lahar, indicates additional progressive development with time as phreatic processes further concentrated fine sediment at the interfaces. Pillar structure (Lowe and LoPiccolo, 1974) is present along with the dish structure in some runout Figure 12. Dish structure in deposits of National Lahar at the type locality. Note that the concavity of the "dishes" increases upward in the section. Structures shown are about 1 m below the top of a 3-m section. Scale at left is in inches. FLOW MAGNITUDE AND FREQUENCY 29 deposits, but is less common than in the deep-water marine sequences from which both structures have been previously recorded. The type section of the flow deposit consists mainly of the lahar-runout phase in a quarry near National (fig. 9). Most attempts to date "wood" fragments in the unit yielded ages either too old (inconsistent with ages of underlying deposits, as well as known tephra ages) or too young (inconsistent with the age of trees growing on the surface). Some of the fragments that resemble carbonized wood were found to be inorganic during analysis; others are low-grade coal derived from Tertiary bedrock. Two dates establish a probable age, however. Wood within the upper, possibly reworked part of the unit below Alder Dam (table 4) yielded a radiocarbon age of 790+205 years. Near the end of the river system, a cedar tree growing on the correlative deposit died 585+125 radiocarbon years ago. This date is from its outermost wood and corresponds to an absolute age of about 550 to 650 years (Stuiver and Becker, 1986). Inasmuch as the tree was 250 to 350 years old when it died, the age of the deposit is at least 800 to 1,000 years. Upstream, trees growing on the sequence containing the flow deposits are at least 800 years old (Crandell, 1971, p. 42). In the White River valley, the most likely correlative surface is probably at least 1,000 years old (Sigafoos and Hendricks, 1961, p. 16). The National Lahar may be a close or synchronous correlative of the Dead Man Flat lahar assemblage in the White River. The lahar of that assemblage in the main fork of the White River has a radiocarbon age of 1,120+80 years. Both the National Lahar and the Dead Man Flat lahar assemblage contain abundant clasts of erosionally derived layer-C pumice. The carbon content of the National Lahar is high, indicating that it occurred after a large forest fire. Evidence from other parts of the Mount Rainier area shows that at least one such fire did occur in the interval between tephra sets Y and W. Crandell (1971, p. 57) found forest-fire debris in a bank of Tahoma Creek, and he tentatively correlated it with the block-and-ash flow, 2,350 radiocarbon years in age, in the South Puyallup River valley. A remarkable layer of forest-fire debris was found in the banks of Kautz Creek upstream from the Wonderland Trail bridge, and wood from this layer yielded a date of 1,625+70 radiocarbon years. The National Lahar, if associated with a fire, expectedly would contain carbonized wood, but none was found. This suggests either that the flow occurred much later than the fire, after flushing of charred wood from the watershed, or earlier than the fire. The presence of carbon-impregnated rock suggests the former possibility. The maximum age of the National Lahar is limited by included clasts of tephra layer C, with an age of 2,200 radiocarbon years; set P, which underlies the unit and has an age of 3,000 to 2,500 radiocarbon years; and a date of 2,285+155 radiocarbon years from wood below the unit upstream from Longmire. The maximum possible age is more closely bracketed by samples from beneath the unit, which yield ages of 1,820+300 and 1,970+250 radiocarbon years (R.P. Hoblitt, U.S. Geological Survey, oral commun., 1994). Another noteworthy feature of the runout phase of the National Labar is the distance over which the flow was hyperconcentrated, a total of more than 40 km from near National to downstream of Alder Reservoir. Long distances of hyperconcentrated flow were common to most lahar runouts at Mount Rainier. Younger noncohesive lahars, debris flows, and their runout phases in the Nisqually system (table 3) include: (1) flows occurring shortly before and after the National Lahar (fig. 114, B); (2) lahar-runout-flow deposits interbedded with flood-plain gravel near the National Park headquarters between Ashford and Elbe; (3) a lahar-runout flow exposed in the gorge upstream from National; and (4) the runout phase of the largest of the 1947 debris flows, which originated as glacial-outburst floods in response to precipitation. This last flow, although recent, is a probable example of a flow with an intermediate magnitude and recurrence interval (100 to 500 years). This conclusion is based on its inundation area (it locally overlies set W and inundated a valley width of 0.9 km at the highway crossing) as well as the magnitude of its discharge (fig. 13) and deposit volume. In terms of estimated peak discharge, the flow was at least 10 times larger than any other 20th-century flow in Kautz or Tahoma Creeks. The 1947 flow was the largest of a series of flood and debris flow surges that occurred mainly on October 2-3 in response to an intense cloudburst that caused the lower 1.6 km of the Kautz Glacier to collapse. Areas of stagnant ice resulting from long-term Neoglacial recession are major fac- tors contributing to the formation of these modern flows, which are discussed more fully in the section on smaller, more frequent flows. Inasmuch as the 1947 flows are well described by Grater (1948a and 1948b), Erdmann and Johnson (1953), Richardson (1968), and Crandell (1971), only a general description is given here, along with any new details relevant to our topic. Each 1947 debris flow was clearly noncohesive; Erdmann and Johnson (1953) recorded a "more or less complete absence of clay." Two composite samples of com- plete flow units contained 1 percent or less of clay. Crandell (1971) reported 4 percent clay in two matrix samples. A well-developed runout flow evolved from the largest 1947 debris flow; its deposit near the confluence of Kautz Creek and the Nisqually River is now a source of sand and granule gravel for aggregate, as are many other runout-flow deposits. Downstream, the flow was overbank only locally but can be traced to the western park boundary, 4.5 km downstream. Farther downstream at the gaging station at National, the flood wave had transformed from hyperconcentrated flow to normal streamflow; at that point the flow had attenuated to only 42 m*/s above base flow (Nelson, 1987). 30 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON 20,000 T T T T T T T o g Confluence of Kautz Creek & and the Nisqually River _ m 15,000 |- a Luu & A O \\ > & Kautz Creek debris flow \ a October 2, 1947 I o 0 \ - q 10,000|- e - o E \\——' Estimated s \ o_ a \\ 3 5,000 |- \ -- [®] \\\ z Tahoma Creek debris flow ~- ~- October 26, 1986 /_——-——-~\/Estlmated 0 I | _ _*= t 4 } | I 8 10 12 14 16 18 20 22 24 DISTANCE FROM SUMMIT, IN KILOMETERS Figure 13. Discharge versus distance for two debris flows resulting from glacial outburst or collapse. Rapid decline of discharge down- stream reflects debulking of sediment in response to valley expansion, slope reduction, or confluence with streamflow. The 1947 runout deposits are easily distinguished from older lahar-runout deposits of volcanic origin by their darker color, contrasting with the lighter, generally more yellow colors, typical of volcanic alteration, of the flows originat- ing as lahars. The darker color is ascribed to the dominance of morainal streambed sediment as the main 1947 source materials. This sediment originated as ice-shattered debris scoured and eroded from the surfaces of relatively unaltered lava flows. Headwalls above Kautz Glacier expose flows (frontispiece) less altered than those above Tahoma Glacier in the Sunset Amphitheater, source of several flows in the Tahoma Creek valley. As much as 18.3 m of net channel erosion accompanied formation of the flows, deposits of which totaled 38 million m* in volume along Kautz Creek (Grater, 19482). The 1947 runout deposits have size distributions typi- cal of hyperconcentrated streamflow derived by direct trans- formation from debris flow, as shown in figure 2 for the National Lahar and its runout flow. The main debris flow deposit has a bimodal size distribution, mean grain size of -2.1 $ (4.2 mm), and sorting of 3.8 $. More than 80 percent of grains in two samples of the runout flow are in the sand size range; the mean grain sizes in these samples are 1.0 and 0.9 ¢ (0.5 mm), and their sorting values are 1.3 and 1.6 ¢, respectively. The behavior of the main 1947 flow was similar to that of smaller, more frequent flows. Its peak discharge attenuated rapidly at places of rapid energy loss as sediment debulked from the flow (fig. 13), transforming the debris flow surges to hyperconcentrated streamflow. Unlike the smaller modern flows on both Kautz and Tahoma Creeks, however, the place where debulking was most rapid was not at the main slope inflection at the base of the volcano, but near the confluence with the Nisqually River. The energy loss was in response to a great increase in valley width and spreading of the flow across the maturely forested fan. The hazard implications of the rapid attenuation and debulking of noncohesive flows are discussed in the section on risk analysis. PUYALLUP RIVER SYSTEM Flows in this drainage are dominated by large, mainly cohesive lahars, the valley-bottom deposits of which would have buried most runout deposits of the generally smaller, noncohesive flows. Several deposits of noncohesive flows were seen, including the large noncohesive lahar described in the section on large, infrequent flows. That flow extended as debris flow for at least 6.0 km beyond the lowland boundary and is the most far-reaching noncohesive lahar known from Mount Rainier. It is directly overlain and directly underlain by lahar-runout-flow deposits (tables 3 and 4), and the younger, overlying runout unit is evidence of an upstream lahar nearly synchronous with the large noncohesive lahar. CARBON RIVER SYSTEM Most of this river system consists of a deeply incised bedrock gorge from which any volcaniclastic flow deposits have been eroded. A ridge extending north of the summit may have diverted some noncohesive flows originating at FLOW MAGNITUDE AND FREQUENCY 31 the summit to the White or Puyallup River systems (Crandell, 1971). Crandell observed a lahar in the headwaters, and we noted two valley-wide lahar-runout deposits that are pre- and post-W in age. Set W was not found by Mullineaux (1974, fig. 18) to extend significantly into the watershed, but a distal facies of that tephra is locally recognizable. Layer Yn is more widespread (Mullineaux, 1974, fig. 16). Its preservation at the surface on low valley-side slopes shows that no large flows have originated in the river system in the last 3,400 radiocarbon years. However, neither the diversionary effect of the ridge noted above nor the paucity of previous debris flows in the watershed changes its susceptibility to a future cohesive lahar originating as a sector collapse (Frank, 1985). FLOWS OF LOW MAGNITUDE AND HIGH FREQUENCY (LESS THAN 100 YEARS) The smallest, most frequent debris flows and their derivative runout flows are common in a few river systems at Mount Rainier, but rare in others. These flows have several general characteristics: (1) They tend to occur in clusters within periods of several years (such as those that occurred in the periods of 1967-70 and 1986-92), and decades may separate the clusters; (2) the debris flows that originate as glacial-outburst surges have been historically most common in late summer and fall; (3) the flows are uniformly noncohesive, forming from flood surges and in most cases transforming downstream through hyperconcen- trated flow to normal streamflow; (4) this transformation is rapid, occurring at the base of the volcano, and so the flows attenuate rapidly (fig. 13) and are typically contained within stream channels beyond that point; and (5) the flows have a variety of glacier-related origins and interactions; the largest flows occur during or just after periods of precipitation, which may trigger collapse of the stagnant terminal ice resulting from Neoglacial recession. Walder and Driedger (1994) have prepared a detailed analysis of the effects of outburst floods and the debris flows formed by them. Lakes dammed by terminal Neoglacial moraines are not a large hazard at Mount Rainier. Unlike the numerous moraine-dammed lakes on some Oregon volcanoes (Laenen and others, 1987, 1992), the lakes on Mount Rainier either are cirque lakes with bedrock sills or are dammed by old moraines and have highly stable outlets, having broken out long ago. However, a landslide into a lake, as has occurred at Lake George in the Tahoma Creek watershed, could catastrophically displace enough water to create a significant surge that may bulk to debris flow. A second and more hazardous type of frequent flow is a debris avalanche, which is not likely to extend far from the volcano, unlike the large debris avalanches that most commonly transform to lahars. This flow type and its possible mobilization to a lahar are discussed in a later section. RECURRENCE INTERVAL OF SMALL DEBRIS FLOWS Many small streamflow surges originate on the volcano. They have occurred at a rate of at least one per year between 1986 and 1992 (Walder and Driedger, 1993). Most are glacier-related, either as subglacial outbursts or supraglacial outbursts of ponds dammed by saturated modern moraines. Others are the result of temporary impoundment of streams by landslides, commonly in Neoglacial lateral moraines. Some of the surges are not large enough to erode the coarse bed material and thus do not bulk to debris flows. The larger surges of this type, especially those triggered by precipitation, are competent to erode all bed material, including boulders several meters in diameter. These surges bulk to debris flows and may be divided into precipitation-induced events and clear-weather events (fig. 14). The best-known examples are along the Nisqually River and its tributaries, Tahoma and Kautz Creeks. At least 20 such events between 1925 and 1990 were of sufficient size to inundate areas of inactive flood plain in those watersheds and pose a local hazard to hikers. (Flows for which only the year is known are not shown in figure 14.) Walder and Driedger (1993) have prepared a guide to their hazards and occurrences for park visitors. Between 1986 and 1988 there were eight major flows from Mount Rainiet-five on Tahoma Creek and three on Kautz Creek-and at least two smaller flows (on Tahoma Creek in late August 1987). A similar cluster of flows between 1967 and 1970 on Tahoma Creek was ascribed by Crandell (1971, p. 60) to possible geothermal activity. Although no increase in geothermal activity was known to accompany the latest cluster, Frank (1985) reported the presence of heated ground and sub-boiling-point fumaroles ° on the South Tahoma and Kautz Glacier headwalls. Study of the Tahoma Creek deposits before the initial 1986 flow verified that the lapse in reported flows between 1970 and 1986 represents a true lack of significant flows in that drainage, not just a lack of observations. Some flows of this type have occurred unseen and unrecorded in other drainages, even since construction of the Wonderland Trail, which circumnavigates the volcano near its base. Deposits are covered or eroded by those of later flows of similar noncohesive texture. The record of 20 significant, flood-plain-inundating debris flows since 1925 is clearly a minimum in the Nisqually River headwaters. An appropriate composite recurrence interval for planning purposes in those watersheds is approximately two years. At present (1994), however, even though the cluster of flows on Tahoma Creek that began in 1986 may be tapering off, it is reasonable to assume that at least one flow can be anticipated 32 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON | I I I | I | I I I c - Clear weather r _ Associated with precipitation 1 (r) in October 1926 | ? - Day not known; midpoint of week plotted <2 r (same day, year) NUMBER OF FLOWS 9 |_ c C ? _| 1 |- 2 2 r ce cero o ; r |r _ 0 1b f j L L IL | _ | | k _ 1 | | 15 1 15 31 15 31 15 30 15 31 15 30 JUNE JULY AUGUST SEPTEMBER OctosER NovemBER Figure 14. - Seasonal distribution of debris flows, hyperconcentrated flows, and floods in glacier-fed tributaries of the Nisqually River from 1925 to 1990. Only flows with known dates are shown; many others, smaller or of unknown date, are not shown. Some data from Crandell (1971); Erdmann and Johnson (1953); Richardson (1968); J. J. Major (U.S. Geological Survey, written commun., 1985); and M. Carney, S. V. Scott, and D. J. Sharlow (National Park Service, oral commun, 1986-1987). each year. Each flows area of inundation is likely to extend only locally outside that of the previous members of the cluster. However, a detailed study focusing on origins and valley responses of these flows concludes that long-term predictions of flow frequency in the watershed are not possible (Walder and Driedger, 1994). SEASONAL DISTRIBUTION A significant factor mitigating the hazards of these flows is that they tend to occur in late summer and fall (fig. 14) after most back-country tourist use. The mean date of the known flows in figure 14 is September 7. In addition, the largest flows recorded historically in each of the three major valleys in the Nisqually headwaters have occurred in October-October 2 for Kautz Creek, the 25th for the upper Nisqually River, and either the 15th (highest volume) or 26th (highest discharge) for Tahoma Creek. Each of those flows began as a precipitation-induced surge, and two were amplified by the collapse of areas of stagnant ice. Although these surges probably were amplified by subglacial water, the glaciers served mainly as conduits and temporary reservoirs of storm runoff. Many clear-weather flows are logically ascribed to subglacial storage of meltwater. These flows tend to occur earlier in the year than the precipitation-induced flows (fig. 14). The previous cluster of flows, from 1967 to 1970, occurred in the relatively narrow time interval of August 20 to September 23 (Crandell, 1971). This time of occurrence suggests an origin as glacial-outburst floods induced by warm-weather melting. Other evidence cited by Crandell (1971) suggests a geothermal origin, and we assume that possibility exists. FLOW TEXTURE AND FORMATIVE TRANSFORMATIONS Whether their origin is from precipitation or meltwater, the flows bulk rapidly through hyperconcentrated flow to debris flow. These transformations have occurred on the moraine-covered surface of the glacier for surges that exited above the terminus (fig. 15), or in the unvegetated, proglacial valleys for subglacial surges that emerged at the terminus. The proglacial valleys contain vast amounts of reworked sediment of morainal and volcaniclastic origin. This sediment readily bulks, mainly by mobilization of unstable bed and bank material, into the surges from the glaciers to yield debris flows that are uniformly noncohesive in texture and contain 1 percent or less of clay in 10 examples, including the 1947 flows on Kautz Creek. FLOW DYNAMICS AND TRANSFORMATIONS Two flows in the Tahoma Creek valley were studied as models to analyze the dynamics and transformations of a precipitation-induced flow and a clear-weather flow: the former occurred October 26, 1986, and the latter, June 29, 1987. By chance, the watershed and stream channel had been studied immediately before each event, and they were restudied afterwards. Both flows were noncohesive, although the deposits of the clear-weather flow contain slightly more clay. The mean grain size and sorting of the flow matrixes are shown in figure 3. A significant part of the October 1986 flow originated from a sinkhole-like collapse near the active glacier terminus, which is presently just below a crevassed ice fall at an altitude of about 1,830 m (6,000 ft) to 2,260 m (7,400 ft). The dimensions of the collapse were estimated from an aircraft as 9 by 15 m (R. Dunnagan, National Park Service, oral commun. 1986). The precipitation-induced surge bulked as it flowed across the top of the stagnant, moraine-covered lower portion of the glacier. Part of the flow may have entered a small sinkhole (fig. 164), and the remainder was apparently dammed temporarily on the surface before cutting a channel along the west side of the glacier. The flow probably was already a debris flow at that point, as indicated by boulder levees and deposit texture on FLOW MAGNITUDE AND FREQUENCY 33 Figure 15. Active front of the South Tahoma Glacier 5 days before (A) and 1 day after (B) the clear-weather glacial-outburst flood and debris flow of June 29, 1987. Arrow in B points to dark areas of collapse, source of at least part of the flow. 34 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON Figure 16. Area of stagnant, moraine-covered lower part of South Tahoma Glacier 5 days before (A) and 1 day after (B) the flow of June 29, 1987. Location 0.9 km downstream from active front of glacier shown in figure 15. In A, note flow of part of supraglacial stream into sinkhole (left arrow) and fresh scarps and fractures in alluvium (extending from lower left toward sinkhole), corresponding to crevasses in underlying stagnant ice. Boulder berms delineate flow of October 26, 1986 (right arrow). B shows incision into the debris-rich stagnant ice by the June 29 flow. FLOW MAGNITUDE AND FREQUENCY 35 the glacier surface (fig. 16A). As the two distributaries (subglacial and supraglacial) rejoined below the stagnant-ice terminus at 1,510 m (4,960 ft) altitude, bulking continued to enlarge the flow in the channel incised in the Neoglacial moraine. Bulking was amplified by the collapse of debris-rich ice at the front of the stagnant part of the glacier, and the resulting material may have dammed the main channel. The 1986 flow had the highest discharge of any flow in the 1986-88 period (fig. 13). The flow volume, however, was exceeded by the flow or series of flows that occurred on October 15, 1988. The deposits of that flow (or flows) inundated the entire valley floor, 0.2 km in width at the site of the former picnic area, and occurred in greatest volume at a point farther downstream than any other flow in the 1986-88 group. This observation suggests a correlation between flow size and the distance of the locus of deposition from the mountain, which is also suggested by the data shown in figure 13. Although peak discharge was higher on October 26, 1986, the volume of sediment transported on October 15, 1988, was greater, accomplished by either a broader flow wave or by multiple flows. The June 1987 flow originated from the base of the icefall (fig. 15B), at the end of a week of completely clear weather that marked the beginning of a severe drought period. (This drought also resulted in small debris flows along Tahoma Creek on August 28 and 31 and one of moderate size on September 23.) Lateral deposits of the June flow were silt-rich as the surge issued from the ice fall, and bulking to debris flow occurred on the surface of the stag- nant ice above the site of the previously existing sinkhole. Lateral erosion of stagnant ice triggered an ice-block ava- lanche into the channel, and blocks of mixed ice and rock several meters in diameter were transported. Figure 16 shows the channel several days before and the day after the flow. Below the lateral ice avalanche, the flow triggered a spectacular collapse of the stagnant glacier surface from above the sinkhole to the terminus. Rapid incision into the debris-rich ice then led to further bulking and enlargement of the flow wave. The depositional patterns of the 1986 and 1987 flows were nearly identical. The deposits were thickest within 0.5 km of the inundated picnic area (a campground before inundation in 1967) along Tahoma Creek. Boulder fronts as much as 3.5 m high (eroded or buried in 1988) represented the "frozen" termini of convex lobes of the coarse front of the flow. As movement of each lobe ceased, its deposits diverted flow from the following segment of the wave to one side. Each new surge successively stopped, diverting the following portion of the wave, and so on in a chain reaction. Distal surges in the flow were thereby created from a single flood wave, as shown by the existence of only a single berm of deposits upstream. The coarsest boulder fronts of each flow contained as much as 10 percent clasts of ice and frozen ground (fig. 17). Figure 17. Ice clast, more than 1 m in maximum dimension, in- cluded with andesite clasts of similar size in lobate boulder front of flow of October 26, 1986, Tahoma Creek drainage. Each flow front was lower and finer grained than the preceding lobe. At a point in this progressive longitudinal "sampling" of both the 1986 and 1987 flows, the transformation to hyperconcentrated flow was reached, and the successive deposition of debris flow lobes ended. The point in each case was about 0.5 km downstream from the former picnic area. The pattern documents the progressive fining, improvement in sorting, and decline in strength (shown by loss of dispersed large clasts) longitudinally within the flow wave (fig. 18). The tail of the flow wave clearly consisted of hyper- concentrated flow. Deposits having the texture characteris- tic of that flow type accreted to the sides of the debris flow channels and distributaries at levels lower than those achieved by the debris flow levees. Both the continuity in the successively finer and lower debris flow fronts and the textural transformation to hyperconcentrated flow indicate fractionation of a single flow wave. Some of these events have been interpreted as a series of separate flows because the differences in flow within a single flood wave, as well as the creation of distal surges, were not recognized. With the exception of the 1947 Kautz Creek flows, which were clearly separated, most of the flows in this category of mag- nitude and frequency began as single flood waves. The dis- tal surges described above are variants of the surges resulting from temporary damming of a confined channel by 'the coarse boulder front of a flow. (See Pierson, 1980; and Costa, 1984.) For both the 1986 and 1987 flows, the hyperconcen- trated flow deposits of the receding flood wave overlie the sole layers of distributary debris flow channels (fig. 19). 36 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON 0.01 Debris flow lobe 1 12.3 kilometers Debris flow lobe 2 12.4 kilometers - ’¢ / *A eC L 12.6 kilometers / 6g =3.65 9G = 3.79 _0G = 4.08 CUMULATIVE PERCENT I | | I | I | | 99.99 Debris flow lobe 3 |- Boulders t- Cobbles -t-<---- Pebbles ---» §H<———— Sand ------»<---- Silt ---» ' # po Hyperconcentrated flow 13.5 kilometers e / U / 6G =1.80 es 5 | | I I I | I | | -10 1024 -9 512 -8 256 =7 128 -6 64 -4 16 -5 32 olE -10 1024 -10 1024 PERCENT 10 6 0.001 mm 10 6 1 1 0.001 mm -1 0 1 1 0.5 2 0.25 3 0.125 4 05 0.0625 0.031 6 7 0.016 0.008 86 0.004 mm Debris flow lobe 3 -10 1024 0 1 0 1 10 6 0.001 mm 10 6 0.001 mm 1024 Figure 18. - Cumulative curves of particle sizes within successive boulder fronts and hyperconcentrated-flow deposits formed during trans- formation of the Tahoma Creek debris flow of October 26, 1986. After deposition of the lobate fronts, only the hyperconcentrated tail of the flow continued downstream. Down-channel distance from the peak of Mount Rainier is shown for each deposit. These highly compacted layers of pebbles dispersed in a silty sand matrix are identical to the Type II sole layers at the bases of lahars formed in 1980 at Mount St. Helens (Scott, 1988b). After the hyperconcentrated tail of the main flow wave had passed, a small secondary debris flow was formed through dewatering of the coarse debris-flow deposits. Pore fluid draining from the coarse flow fronts contained sufficient silt (15 percent of deposits) and clay (2 percent of deposits) to yield a 1-cm-thick deposit in downstream channel thalwegs. The elevated deposit margins indicate strength in the range of debris flow. Most of the deposit is sand (75 percent, fig. 20) and, like a sole layer in a subsequently active channel, is unlikely to be preserved. This deposit is a smaller version of the large lahar formed from the 1980 debris avalanche in the North Fork Toutle River at Mount St. Helens; it is likewise similar to the lahar formed from the main 1963 debris avalanche from Little Tahoma Peak into the White River drainage (fig. 20). The deposit textures of the 1963 debris avalanche, the 1987 debris flow, and the flows derived from each by dewatering are illustrated in figure 20. The slope of the cumulative curve of each derivative flow is very similar to that of the finer part of the source flow. The dewatering process thus removes part of the matrix of the primary deposit but, unlike the more common direct transformation of the debris flows to hyperconcentrated flows, produces another, relatively small debris flow. The ability of the dewatering process to produce large flows is documented by the 1980 lahar in the North Fork Toutle River at Mount St. FLOW MAGNITUDE AND FREQUENCY 37 Figure 19. Debris-flow levees and underlying sole layers from recent flows along Tahoma Creek. A, Distributary channel of the October 1986 debris flow as it appeared in May 1987. Darker, compacted sole layer, in middle, overlies lighter channel bed material. Sole layer is 30-50 cm thick. Flow lines formed by recessional hyperconcentrated flow are visible just below the coarse debris at the tops of the levees. B, Main channel of the June 1987 debris flow at same site in July 1987. Sole layer, accreted to channel sides, is being eroded. Bank topped by levee is 3.5 m high. 38 0.01 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON 0.05 1963 debris avalanche _GG = 52;- % G2 X~ 1987 debris flow | - -* |_ CG = 3.50 CUMULATIVE PERCENT 99.8 99.9 - $ |« Boulders -+- Cobbles-»t-<---- Pebbles 99.99 | I | 1 I | | 1 P | —H<-§->i<———Sand F 7° \_ Derived from 1987 debris flow Derived from 1963 debris avalanche Silt | #10 -9 -8 0-7 0-6 -4 1024 512 256 128 64 16 -5 32 -3 0 -2 4 S | Z E 1963 debris avalanche 1963 secondary debris flow 20 20 10} 10 PERCENT 0 -10 1024 0 1 106 10 0.001 mm 6 0.001 mm 0 1 -1 2 20 | 7 - 8 6 0.004 mm 0.008 | 4 5 6 0.0625 0.016 0.031 | 2 3 0.25 0.125 0 1 1 0.5 1987 primary debris flow 1987 secondary debris flow T 20 10 "o 1024 10 1024 0 1 0 1 10 6 10 6 0.001 mm 0.001 mm Figure 20. - Cumulative curves of particle sizes of debris flows derived by dewatering from the main 1963 debris avalanche in the White River valley and the June 1987 debris flow in Tahoma Creek, compared with cumulative curves of the primary deposits. Helens. No such origin, however, can be established at Mount Rainier for any debris flow larger than the relatively small 1963 example. This conclusion confirms our belief (and that of Crandell, 1971) that the large sector collapses at Mount Rainier continued directly as debris flows for long distances, rather than yielding thick, hummocky masses immobilized nearer the volcano. DEBRIS AVALANCHES AND THE TAHOMA LAHAR A small debris avalanche derived from a shallow slope failure is a second type of flow that is best included in the final general category, that of the smallest but most frequent flows and avalanches. Although less frequent than glacial-outburst debris flows, several historical debris avalanches have occurred. The two largest examples traveled (1) from the Sunset Amphitheater onto the Tahoma Glacier in the early 20th century (fig. 21) and (2) from Little Tahoma Peak onto and beyond the Emmons Glacier in 1963 (fig. 4). Smaller debris avalanches fell onto the Winthrop Glacier in 1974 (Frank, 1985, p. 138), onto the Cowlitz Glacier in 1975 (Frank, 1985, p. 138-139), and onto the Winthrop Glacier in 1989. The most recently documented debris avalanche originated August 16, 1989, from upper Curtis Ridge, as did the 1974 flow, and descended from 3,600 to 3,700 m (11,800 to 12,100 ft) to 1,950 m (6,400 ft) in altitude. Runout occurred over a horizontal distance of 4.1 km. The flow FLOW MAGNITUDE AND FREQUENCY 39 Figure 21. Debris avalanche on the surface of the Tahoma Glacier, at the head of the South Puyallup River. Note the lighter, hydrother- mally altered debris originating in the Sunset Amphitheater, contrasting with the darker morainal sediment, foreground, on and lateral to the Tahoma Glacier. The origin of this flow was similar to that of the Case III flow, the Tahoma Lahar. deposits were noncohesive in texture; deposit thickness is surprisingly thin, as are the thicknesses of the deposits of the debris avalanches on the Tahoma and Emmons Glaciers. The total volume of the flow is probably in the range of 0.1 to 0.5 million m*, based on an average thickness of about 20 cm. The main seismic record of the avalanche consisted of complex, high-amplitude signals at 1706, 1714, 1715, and 1721 hours UTC on August 16, with the 1721 event lasting for 9 minutes (Norris, in press). The length of the signal is too great to reflect the velocity of the flow and probably reflects either continuing failure or the continued rolling of house-size boulders. Multiple rock avalanches originated in pre-Rainier rocks that form the ridge known as Mount Wow and inundated the bottom of the Tahoma Creek valley with several lobes of debris, and another extended from the east end of the ridge of pre-Rainier terrane known as Mother Mountain almost to the Carbon River. The Mount Wow avalanches might have been triggered by the April 13, 1949, Olympia earthquake (M 7.1), as suggested by the decomposition stage of killed trees (decay sequence in Franklin and others, 1981). These avalanches warrant serious attention because of their extremely rapid, catastrophic emplacement and their known frequency and hazard at Mount Rainier and many other stratovolcanoes. Therefore, we focus on one large avalanche-derived flow that is typical of several young flows known at Mount Rainier. The Tahoma Lahar is the case history most suitable for planning within Park boundaries. It is distinct from the large sector-collapse debris avalanches and landslides but like those flows, also mobilized to a downstream lahar. Based on the record of all known flows, the smaller avalanches will not pose a large hazard outside the Park. The Tahoma Lahar is interpreted as a variably disaggregated debris avalanche mainly transformed to a lahar (tables 2, 3). Its deposits form a distinctive unit in the Tahoma Creek watershed; they are mainly cohesive but are locally noncohesive in some lateral exposures. Like the Paradise Lahar, the unit is characterized by a yellow color and hydrothermally stained clasts. It is post-set W in age and thus much younger than the Paradise Lahar. The Tahoma Lahar, named here, is 0.5 to 2.0 m thick on valley-side slopes; more than 20 m thick in cross section near the base of Neoglacial deposits about 0.5 km upstream from the Wonderland Trail bridge across Tahoma Creek; and at least 4.3 m thick in the valley bottom as seen in exposures only 4.8 km upstream from the Highway 706 bridge. Most 40 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON deposition probably occurred in this area, filling the expand- ing lower valley of Tahoma Creek, where deposits are now covered by those of glacial-outburst floods. Clay content is variable but is characteristic of cohesive debris flow in most exposures. Because of further disaggregation of megaclasts in the flow and a more clayey recession phase (as we also propose for the Osceola Mudflow and other cohesive lahars), clay content is highest downstream where deposits were seen locally near the center of the valley. The- flow deposits overlie set W and compose the uppermost unit of most stratigraphic sections downstream from the Neoglacial terminus (about 100 m upstream from the trail bridge), as reported by Crandell (1971, p. 58) and verified in new exposures. The older Round Pass Mudflow supports trees as much as 700 to 800 years old, some of which were killed by the Tahoma Lahar and others which were killed at least 100 years ago by flows from Mount Wow. Significant attenuation of the Tahoma Lahar began as the flow left the confined channel upstream from the former picnic area. Although deposition was pronounced between the ex-picnic area and a point about 3 km downstream, the distal configuration is estimated (pl. 1) based on levels revealed by peak-stage deposits on valley-side slopes where they are not covered by younger deposits. The stratigraphic relation of the Tahoma Lahar to Neoglacial morainal deposits and the estimate of tree ages on the lahar surface by Crandell (1971, p. 58) establish the time of the flow as shortly following the A.D. 1480 deposition of layer Wn, or about 400-500 years ago. Radiocarbon dates are variable, however. The outermost 25 rings of a tree that grew on the Round Pass Mudflow and that possibly was killed by the Tahoma Lahar provided an age of 560+75 radiocarbon years. A radiocarbon date from the outermost wood from an apparently similar tree near the site of the picnic ground was 200+50 years. That date is in a time interval for which the correlation between radiocarbon and calendar years is poor, and it could correlate with a calendar age of A.D. 1665 to 1955 (Stuiver and Becker, 1986). No volcanic activity is recorded from Mount Rainier near the probable time of the lahar. The coloring of the surficial unit, which might be taken to indicate soil formation, instead reflects an origin as a mobilized debris avalanche of hydrothermally altered rock from the Sunset Amphitheater. The source is probably a different sector of the Sunset Amphitheater than that yielding the modern clay-rich debris avalanche on the Tahoma Glacier (fig. 21; see Crandell, 1971, p. 17). The trend of the distinctively colored Tahoma deposits within the Neoglacial moraine (incised by the branch of Tahoma Creek draining South Tahoma Glacier) suggests an origin above the Tahoma Glacier rather than from the South Tahoma Glacier. A debris avalanche above the Tahoma Glacier, however, should have created a correlative lahar in the South Puyallup River downstream from the Tahoma . Glacier, and no such unit has yet been indisputably identified. A highly likely correlative, however, is unit 4 of Crandell's measured section 8 (1971, p. 57), which is younger than the Electron Mudflow, as is the Tahoma Lahar, and is texturally similar to the Tahoma Lahar. The Tahoma Lahar locally has a hummocky surface. Megaclasts form mounds in forested backwater areas of the fanhead downstream from the former picnic ground. The megaclasts are similar in composition (but with less clay) and color of alteration products to those in the modern debris avalanche on the surface of Tahoma Glacier. Many mound-forming megaclasts were eroded or buried by the glacial-outburst flood and debris flow of October 15, 1988. The strength of the Tahoma Lahar is indicated by a lake dammed by the lateral levee of the peak flow about 0.5 km upstream from the trail head. The lake had a maximum depth of about 2 m, a width of 30 m, and a length of approx- imately 100 m in 1989. The peak flow of the Tahoma Lahar probably was too cohesive for a runout phase to have formed. A lahar-runout deposit of similar age occurs in the Nisqually River near National (fig. 9, tables 3, 4). That deposit contains wood with an age of 410+75 radiocarbon years, corresponding to a true age of about 540 years (before 1994). It is more likely that the debris avalanche did not transform beyond a lahar and that the runout flow is a separate event. HISTORICAL FLOODS COMPARED WITH DEBRIS FLOWS Glacial-outburst debris flows and some smaller examples of both cohesive and noncohesive lahars are all likely to be less destructive than some historical floods have been. The largest floods of record were caused by intense precipitation on snow during prolonged warm periods, and they are described for comparison with the smallest, most frequent volcanic flows. Historical floods in the Nisqually River have been analyzed by Nelson (1987); peak annual discharges in the Puyallup River at Puyallup from 1915 to 1986 were compiled by Prych (1987, table 2). Probably the largest post-settlement flood occurred in 1867 (described by Summers, 1978, p. 235). In early December, much of a heavy snowpack on Mount Rainier was melted by four days of warm rain, causing a major flood in at least the Cowlitz River system. The city of Monticello at the mouth of the Cowlitz was completely destroyed on December 17. Upstream flooding was not reported, because settlement there had not begun. However, several other large floods between 1886 and 1911 probably inundated the entire valley bottom of the upper Cowlitz River. The valley downstream from Packwood, which is as much as 3 km wide, was apparently inundated in March 1907 and possibly again in 1909 (Packwood History Committee, 1954; Superintendent's Reports, Mount Rainier National Park, 1907 and 1910; U.S. Geological Survey stream-gaging FLOW MAGNITUDE AND FREQUENCY 41 Table 5. - Summary of origins and transformations of debris flows at Mount Rainier. Flow me har Planning or design No.! Origin Example Transformation case Cohesive debris flows (>3-5 percent clay) 1 Mobilization of deep-seated debris avalanche (sec- Osceola Mudflow, Elec- Commonly none; type C Case I. tor collapse). tron Mudflow. possible. Noncohesive debris flows (<3-5 percent clay) 2 Melting of snow or ice by pyroclasts (flow, surge, National Lahar and many Type A common ............ Case II. fall) or lava, or by geothermal heat, steam erup- similar flows. tions. 3 Mobilization of shallow debris avalanches Tahoma Lahar (lateral Type A probably com- Case III. part, mon; type C possible. 4 Relatively small, shallow debris avalanches that do 1963 Little Tahoma Peak _ Type C possible .............. None.) not disaggregate to lahars. 5 Mobilization of debris avalanche possibly caused by Paradise Lahar ................ Type A probable ............. None explosion. 6 Glacial-outburst floods bulked to debris flows: (a) Precipitation-induced flOWs ........................... 1947 Kautz Creek .......... Type B common; type C Hazard zonation. probable. (b) CleAr- Wether fIOWS ............cccsseessseessesesseesces 1987 Tahoma Creek ...... _ ...... Do. "Numbers used for comparison with table 6. *Transformation types: A. progressive transformation of wave front to B, Deposition of successive flow fronts; Chewaledngofcmmdepoduwykuneonduy Risk is much lower than in the three described planning cases. "Site-specific hazard-zone mapping based on techniques described in the text. records). Subsequent high flows occurred on the Cowlitz River at Packwood in 1933, 1959, and 1977, but cannot be directly compared. Inundation of the Cowlitz valley to a depth of approximately 2 m is described in several undated early accounts. The Nisqually River drainage was also flooded early in 1910, when flood waters from a drainage to the south overflowed into that river (Bretz, 1913, p. 27). In general, historical flood inundation has been similar in depth to that by the most recent lahar-runout flows (table 4). Historical flood data for the Nisqually River near and downstream from Longmire (Nelson, 1987) show that floods having recurrence intervals of 25 to 500 yr generally have smaller areas of inundation is than do lahars with similar recurrence intervals. This is especially true as recurrence intervals reach and exceed 100 years, because the more frequent volcanic and glacial-outburst debris flows attenuate rapidly at the base of the volcano (fig. 13), whereas rainfall floods amplify downstream with increased tributary inflow. A 500-yr flood will locally affect flood plains outside the active channel (Nelson, 1987, pls. 1 and 2), whereas a 500-yr volcaniclastic flow, like the National or even the Tahoma Lahar, could be catastrophic at a location like Longmire. While dating the younger noncohesive lahars and their runout phases, we also dated some flood deposits and groups of such deposits (table 4). Some were probably local; others were the deposits of floods affecting all drainages of the mountain. Yet others may have been the distal flood waves evolved from upstream lahar-runout flows. In assessing risk, no presumption of a debris flow is made from fluvial ted flow. bypassed by dilme tail of hyperemcenlmed flow. sediment unless a direct correlation is possible. The distal streamflow deposits of the National Lahar (fig. 11C) are an example of such a correlation. SUMMARY OF FLOW ORIGINS AND TRANSFORMATIONS In order to rank the flows according to magnitude and frequency, the preceding discussion focused on the relative sizes of the various flow types and the evidence for their ages. To assess risk, the size and frequency of flows must be known, but other factors, such as the probability of a warming (Costa, 1985), are also important. Table 5 extracts the general flow origins that can be recognized, as well as the transformations that occur with each type. For simplic- ity, the formation of a secondary debris flow from the sur- face or interstices of a primary debris flow is treated as a transformation (type C in table 5), but the original concept of flow transformations invoked a fundamental change in flow behavior (Fisher, 1983). Because the change is from debris flow to debris flow, albeit accompanied by a change in texture, there is no change in theology. Unlike cohesive debris flows, noncohesive flows undergo the complete transformation of the entire flood wave to hyperconcentrated streamflow, which then evolves to normal streamflow with sediment content below hyper- concentration. Both these distal transformations involve fundamental changes in flow behavior and grain interaction 42 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON (Scott, 1988b, table 9; Pierson and Costa, 1987). Each also involves the progressive loss of sediment, which, combined with the commonly more peaked flood wave of the noncohe- sive flows, can produce greater attenuation. Conversely, during the formative transformations to debris flow, as sediment bulks into the flow, the flow wave may be amplified in volume many times. These differences between cohesive and noncohesive debris flows are tendencies, not laws of behavior. For example, a cohesive flow can be more peaked, and can lose sediment rapidly by deposition on a wide flood plain. However, the overall tendencies are consistent with flow type. Three types of flow transformations are represented in figures 2, 18, and 20. At Mount Rainier, the first (fig. 2) is the complete, progressive transformation of the entire flood wave (Scott, 1988b, fig. 37). The second (fig. 18) is the repeated, successive deposition of the flow front as a series of lobes, until only the hyperconcentrated tail of the flow remains to flow downstream. The third (fig. 20) is the creation of a secondary debris flow by dewatering and slumping of the surface of a debris avalanche, or by drainage of the matrix from coarse, clast-supported debris flow deposits. Avalanche dewatering has produced significant lahars elsewhere, but is not known to have produced any but small flows at Mount Rainier. Although large cohesive lahars have occurred in the post-Y time period, we found no upstream debris avalanches that corresponded in size and age and from which they could have been derived secondarily. RISK ANALYSIS Risk analysis is a generic term for methods that support decision-making by quantifying consequences (magnitude and extent of lahars, for example) and the probabilities of their occurrence (frequency of lahars) (National Research Council, 1988). Which of the types of flows in table 5 pose sufficient risk to influence downstream hazards planning? An initial premise is that volcanic debris flows and their transformations can be treated like other hydrologic hazards. That is, flow events of equivalent frequency require the same planning awareness, whether the flow wave consists of sediment moving interstitial water (debris flow, and the upper range of hyperconcentrated streamflow) or water moving sediment (floods, and the lower range of hypercon- centrated flow). Floods and volcanically induced flows can be treated as separate components of a mixed population with a minor overlap in their scales of magnitude: that is, the high end of the flood scale overlaps the low end of the volcanic flow scale. Pragmatically, the risks are additive. The chief practical differences between inundation by floods and inundation by lahars are the destructive impact forces of a lahar and the long-term effects of its deposits as contrasted with the ephemeral inundation by a flood. "Hazard" refers both to the agent and to the potential for harm posed by that agent. Also, risk can be said to exist when something of value is at jeopardy. Thus, in the general case of volcanic hazards (Dibble and others, 1985): (1) RISK = HAZARD xVALUE xVULNERABILITY, where HAZARD is an event of known probability, VALUE is the economic assessment of loss, and VULNERABILITY reflects susceptibility for harm, which may vary for different things affected by the same hazard. The inclusion of the latter term is extremely valuable in assessing volcanic flow hazards. A similar approach to the dangers of volcanic flows is: (2) RISK = FLOW MAGNITUDE xFLOW FREQUENCY XVALUE xVULNERABILITY, where each flow subpopulation can be treated separately and ranked by the risk it poses. Although the results (table 6) are qualitative, they clearly separate the differing risk of each flow type and provide a logical basis for the quantitative analysis of individual case histories of flows that represent the flow types that pose the greatest risk (pl. 1). In this initial ranking, MAGNITUDE is replaced by a convenient surrogate, area of inundation, which is based on the extent of the flows as established by their deposits (tables 2, 3). FREQUENCY is the probability of each flow type, or the inverse of the recurrence interval. VALUE is also proportional to inundation area, but its inclusion is necessary to assess the relative risks of different size flows. At Mount Rainier, population and property values increase downstream - in - each _ watershed, approximately exponentially, but with a large increase as flow reaches the Puget Sound lowland (data from Pierce and Thurston Counties, Washington). Consequently, including a VALUE term correctly emphasizes the catastrophic potential of the larger flows. The VULNERABILITY factor in equation 2 signifi- cantly affects the danger of certain flow types. That is, vulnerability to a flow type is reduced if there is the probability of a warning in the form of volcanic activity precursory to the flows. People and movable objects in the path of rapid debris avalanches at Mount Rainier are far more vulnerable than those near the attenuating debris flows of glacial-outwash or rainfall origin. Vulnerability also depends on probable reservoir levels and whether they can be drawn down in the event of a warning. For example, vulnerability is reduced by the fact that Mud Mountain Reservoir on the White River is solely a flood-control structure and is thus normally empty. No single flow type and origin will pose the greatest hazard throughout an entire river system. On the highly populated Puget Sound lowland, the huge sector-collapse debris avalanches mobilized as lahars (flow 1, table 6) pose a - A. A. - te iti RISK ANALYSIS 43 Table 6. Ranking of debris flows described in table 5 by magnitude, frequency, and risk. Rank from greatest to least Magnitude Frequency Risk (Inundation area) Flow 1 Flow 6 Flow 1 Flow 2 Flow 4 Flow 2 Flow 4 Flow 3 Flow 3 Flow 5 Flow 2 Flow 5 Flow 3 Flow 5 Flow 4 Flow 6 Flow 1 Flow 6 the greatest danger. In valleys on and immediately adjacent to the volcano, noncohesive lahars (flows 2, 3, or 5, table 6) and debris avalanches (flow 4, table 6) pose the greatest danger. And, for hikers along proglacial streams on the volcano, a debris flow formed from a glacial-outburst flood (flow 6, table 6) is the greatest statistical risk. FLOW FREQUENCY AND RISK AT MOUNT RAINIER This discussion focuses on flows of the frequencies most commonly used in - long-term - hydrologic planning-100 and 500 years (Brice, 1981). These recurrence intervals correspond to probabilities of 1 percent and 0.2 percent per year. By contrast, Latter and others (1981) believe it is "desirable" to incorporate events with recurrence intervals of 1,000 and perhaps 10,000 years when assessing volcanic risk. Although practice is variable, design frequency for bridges on primary roads is commonly 50 years, with some states using a 50-yr flood for the bridge superstructure and a 100-yr flood for the substructure (Brice, 1981). Flow frequencies for structures such as reservoirs and power plants are commonly lower (that is, return periods are higher) than these values and are commonly controlled by economic factors (Linsley and others, 1958). In occurrence, lahars at Mount Rainier differ from those at Mount St. Helens in an important way. The latter have a significant tendency to cluster in groups, and their time distribution can be analyzed both in an eruptive period, as at present, or over any other time interval. At Mount Rainier, in contrast, both volcanism and lahars are scattered throughout postglacial time (tables 2-4). Therefore, the occurrence of one large lahar does not increase the odds of a second, as it does during the modern eruptive period-at Mount St. Helens. The assumption of basically random occurrence, as in flood analysis, is probably valid at Rainier. All recurrence intervals discussed here are based on mountain-wide occurrences over undivided intervals of postglacial time. This dispersion of risk, rather than its definition within each river system, reflects the uncertainty in knowing what river system or systems will experience the next major lahar. For example, the Carbon River system records the lowest frequency of lahars. However, consider- ing the modern topography and structure of the volcano, that river system may have substantial risk of conveying part or most of a huge, sector-collapse lahar. The river system also contains a large volume of glacial ice that, although covered with insulating rockslide debris, is subject to melting and thus to the formation of noncohesive lahars. The example illustrates the need to reassess risk once the location of any precursor intrusive activity is evident. For example, volcanic activity affecting the Carbon River sector will pose an extreme risk of large debris flows. Conversely, the White River system illustrates the possible temporary reduction in risk of a second large lahar following a significant sector collapse and before edifice reconstruction. The crater remaining after the Osceola Mudflow is now largely infilled, however, and the original failure plane could facilitate renewed failure. Correlations between changes in risk and the occurrence of flows are complicated, perhaps hopelessly so, by the lack of knowledge of hydrothermal alteration and structure within the edifice. Supporting evidence of a temporary risk reduction is not definitive and, at Rainier, cohesive flows have recurred in the same drainage. Other factors also support a volcano-wide risk assessment: (1) large cohesive flows have recurred from a single drainage; (2) a single flow has affected more than one drainage (Osceola and Round Pass Mudflows); (3) three of the river systems, the White, Puyallup, and Carbon Rivers, join downstream within range of Rainier lahars; and (4) a major explosive eruption like that at Mount St. Helens in 1980 would produce lahars in all of the main drainages. The situation is largely analogous to arid-zone flood-hazard mapping where, although only one sector of an alluvial fan will probably be affected by any given flood, all parts must be considered potentially prone to inundation (Scott and others, 1987; Scott, 1992). The length of time needed to evaluate frequency depends on flow size-the smaller the flow type, the shorter the time span needed to establish recurrence interval statistically. The time intervals selected, such as the post-Y time interval used for the definition of noncohesive lahars, are in part a function of geological convenience, but each is sufficient to define flow probability. Even if older and smaller flows are eroded or obscured (a possibility given the number of postglacial episodes of aggradation and degrada- tion in entire river systems), the analysis is not affected substantially. Lahars are far more numerous than episodes of known volcanic activity (producing juvenile eruptive products) at Mount Rainier. Neither cohesive nor noncohesive lahars correlate well with volcanism, and many of the latter probably resulted from geothermal heat flux and steam eruptions. The noncohesive lahars that formed by bulking of meltwater surges are not obviously linked to the most clearly 44 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON recorded eruptions, those producing tephra. Prevailing west winds have distributed Rainier tephras on the east side of the volcano (Mullineaux, 1974), yet the Cowlitz River (east side) has a sparser record of lahars than all other Rainier drainages except the Carbon River (northwest side), and the large number of flows in the Nisqually River (southwest side) is similar to that in the White River (northeast side). The lack of a clearly recorded association of the large cohesive lahars with known volcanism is discussed by Crandell (1971), and his conclusion is reinforced by the ages of the additional cohesive lahars reported here. DEBRIS FLOWS AND SUMMIT-CONE VOLCANISM Along with the small glacial-outburst flows, a previously unrecognized grouping of lahars is an exception to the general lack of time-clustering of flows at Mount Rainier. Noncohesive lahars and derivative lahar-runout flows (tables 3 and 4) occurred throughout the post-Y time interval, as described in the section on flows of intermediate size and frequency. The deposits of these flows form much of the fill in the White and Nisqually River valleys recognized by Crandell (1971) and then believed to have had a normal fluvial origin. The last such flow occurred in 1947. There is, however, a clear concentration of flows late within the post-C, pre-wW time interval. The radiocarbon dates in table 4 define flow activity that peaked between about 2,200 and 800 radiocarbon years ago, and particularly in the last 500-600 radiocarbon years of that interval. The interval is bounded by calendar ages of about 2,250 to 710 years (Stuiver and Becker, 1986). This interval of flow activity does not coincide with an eruptive period as defined at Mount St. Helens (Mullineaux, 1986). Rather, it overlaps the assumed end of lava and pyroclastic flow activity during building of the summit cone above the east rim of the volcano (fig. 4; Fiske and others, 1963, p. 80). Summit-cone lava flows are believed to have occurred between about 2,100 and 1,200 calendar years ago (Crandell, 1971, p. 14). Either lahar-producing activity associated with the construction of the summit cone contin- ued later than believed, or later pulses of geothermal heat or steam eruptions created major meltwater surges. Geothermal activity at the modern summit (Frank and Friedman, 1974) produces only local melting. DESIGN OR PLANNING CASES AND HAZARD ZONATION DEFINITION OF CASES The dynamics of debris flows are described here in hydrologic and hydraulic terms, because that nomenclature is appropriate, and because it is familiar to land-use planners and civil engineers dealing with structures subject to inundation. What we believe is the best example of each of the flow types that pose risk (table 6) is described in this section, its dynamics are presented (table 7), and its flow cross sections are portrayed (pl. 1). There is no substitute for the description of real-world flow behavior, through the case-history approach when dealing with engineering problems involving complex phenomena. No rheologic model deals with the spectrum of debris flow behavior at Mount Rainier. MEASUREMENTS AND ESTIMATES OF FLOW DYNAMICS Velocities (table 7) are based on measurements of runup on obstacles to flow, or on superelevation of flow around bends. Johnson (1970), Costa (1984), Fairchild (1985), Pierson (1985), and Scott (1988b) have analyzed or commented on field applications of this method and the accuracy of the results. The behavior of the large cohesive lahars near the boundary of the Puget Sound lowland is critical, so special attention was given to the measurements near that point. Several sets of measurements defined flow above 15 m/s and approaching 20 m/s. No runup measure- ment was ideal, as in the case of a steep bare surface (a fric- tionless surface is assumed) normal to flow. We believe that 20 m/s is a conservative minimum velocity for the cohesive lahars at the lowland boundary. A velocity in the range of 25 to 30 m/s was obtained from runup of the branch of the Osceola Mudflow in the West Fork White River where it entered the White River valley at a high angle, and the estimate of velocity of at least 40 m/s for the Round Pass Mudflow near the base of the volcano was noted in the section on that lahar. In general, runup measurements were more readily obtained than measurements of superelevation in bends for both cohesive and noncohesive lahars. Cross sections were defined by the distribution of deposits. Unlike floods, debris flows leave deposits accreting on valley sides to the level of peak flow. Delineation of the highest peak flow deposit, equivalent to the high water mark of a flood, was commonly confirmed at multiple points, and the cross sections of flows were measured only where the valley-bottom deposit of the same flow was known. In a few instances, the thickness of valley fill of a flow was extrapolated longitudinally. As noted above (under "Flow magnitude and frequency"), markedly concave flow surfaces in sharp bends may have the effect of exaggerating both cross-sectional areas and the discharges calculated from them (Webb and others 1989, p. 22, table 10). However, of the sites used for calculating the discharges shown in table 7, none has a radius of curvature sufficient to cause this effect. Flow wave volumes were estimated from the volumes of their deposits. The.volume was in most cases not increased tes ade nlin eee P Weve ate a DESIGN OR PLANNING CASES AND HAZARD ZONATION 45 Table 7. Characteristics of design- or planning-case lahars. [Characteristics determined as described in text; N. A. = not applicable] Characteristic Maximum lahar Case I Case II Case III Debris flOW tYPE .........c.cccccscees> Cohesive Cohesive Noncohesive Usually cohesive.'! Recurrence interval (yrs) ......... ~10,000 500-1,000 100-500 <100 Volume at lowland >3,000 230 60 (Puyallup R.) NA. boundary m*). 65 (Carbon R.) Mean flow velocity (m/s): Base of volcano ............... 2 >40 2 >30 10 2 >30 Lowland boundary .......... >20 ~20 ~1 NA. 1 km on lowland .............. ~10 ~8 ~3-4 N.A. Cross-sectional area of flow at ~90,000 ~16,000 1,000 (Puyallup R.) N.A. lowland boundary (m?). 1,200 (Carbon R.) Peak discharge at lowland >1,800,000 ~320,000 7,700 (Puyallup R.) NA. boundary (m/s). 8,400 (Carbon R.) Flow depth (m): Base of volcano ............... ~200 ~50 15 55 Lowland boundary .......... ~100 22 8 N.A. 1 km on lowland .............. <30 ~10 1-3 NA. Sediment concentration at low- >60 >60 ~40 (Puyallup R.) N.A. land boundary (percent by ~45 (Carbon R.) volume) Extent (or inundation area) ...... To Puget Sound Inundation of 36 km" All active flood plains (except Runout phases of or Columbia R. (Electron) to ~50 Cowlitz R.) above reservoirs, noncohesive lahar (in Cowlitz R. km? (modern recur- if present; otherwise could extend an drainage). rence of same flow). upstream of Puyallup. additional 10 km. 'Some may be partly or entirely noncohesive depending on source area. *Estimated by comparison with similar flows at other volcanoes. *Estimated using the linear relation observed between sorting and concentration at Mount St. Helens (Scott, 1988b). to account for loss of water because, in the case of the cohe- sive debris flows, a large proportion of the deposits probably remained saturated, and in the case of the noncohesive debris flows, the water content of the flow wave was largely inter- stitial between grains in nearly continuous contact. Volumet- ric comparisons worked well for the Electron Mudflow where the distal end is defined and the deposits were exten- sively augered (Crandell, 1971). It is less exact where flows continued into Puget Sound or, in the case of the runout phases of the noncohesive lahars, where the flow wave was diluted eventually to streamflow. Corrections for loss of deposits by erosion are an additional source of error, but reconstructions of the original depositional surfaces were possible for some flows. MAXIMUM LAHAR The term "maximum lahar" is substituted for the "worst-case flow" of hydrologic analysis, because there can always be a flow worse than that defined as the worst case. We also prefer the term to "most-extreme lahar," used for a moraine-dammed-lake breakout in which the most-extreme case is displacement of an entire lake by a snow or debris avalanche (Laenen and others, 1992). The true "worst-case" or "most-extreme" analog at Rainier is the improbable removal of the entire edifice. "Maximum lahar" is analogous to the "maximum mudflow" or "maximum credible mudflow" used in forecasts of lahars at Mount St. Helens (for example, U.S. Army Corps of Engineers, 1985). The term is intended to imply that, although larger flows are possible, they are so unlikely they need not be considered. The Osceola Mudflow (Crandell, 1971), is the maximum lahar at Mount Rainier. Cross sections of this flow are shown on plate 1, and its dynamics are described in table 7. The inundation area of the actual flow is easily discernible on the Puget Sound lowland (Crandell, 1963b, 1971), although the flow was difficult to define upstream (noted in sections O-1 and O-2, pl. 1). The inundation area of a modern cohesive lahar of the same size could extend to Puget Sound, through Tacoma along the Puyallup River and through Seattle by way of the Green River system and the Duwamish Waterway. The lower resistance to flow of the modern unforested river valleys would allow a recurrence of this flow to go farther and faster than did the original flow approximately 5,000 radiocarbon years ago. Relative sea level in the Duwamish Embayment of Puget Sound was higher at that time, and the flow entered the sound farther upstream. In a well 6 km northwest of Auburn, deposits of the flow occur 85 m beneath present sea level and are 7 m thick (Luzier, 1969, p. 14); submarine deposition is probable. The Osceola Mudflow had a volume many times that of the next largest cohesive lahar. We accept Crandell's (1971) estimate of 2-3 km} for the present volume, but the original volume may have been as much as twice that amount if 46 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON subsequent erosion and a possibly larger original submarine extent are taken into account. A lahar this size has occurred only once in postglacial time, within the last 10,000 years. When compared with all other large cohesive lahars, it is a statistical outlier. It is tentatively assigned a recurrence interval of 10,000 years. Thus, for illustrative purposes, its probability approximates that of a return of glaciation to the Puget Sound lowland. One or more events of at least this size have a 1 percent chance of occurring within a century (Reich, 1973). An event of this frequency is not normally considered in hazards planning, but Latter and others (1981) propose that it should be. In modern risk analysis, such an event is described as one of "low probability and high consequences," with the implication that the risk may be unacceptable at even very small probabilities. The record at Mount Rainier indicates that the 'most probable recurrence of a maximum lahar will be a debris avalanche that transforms directly to a lahar on or near the volcano. Primary transformation is not a certainty, however, and Crandell (1988, fig. 18) calculates the probable runout distances in each river system at Mount Rainier of untransformed debris avalanches with a volume of at least 1 km°. The hazards of untransformed debris avalanches are discussed by many (including Crandell, 1988; Scheidegger, 1973; Siebert and others, 1987; and Francis and Self, 1987); the risks from debris avalanches are generally greater than those from lahars, mainly because of higher flow velocities. A debris avalanche from Mount Rainier would probably be at least partially saturated; such a flow would have the potential to yield a large secondary lahar as did the 1980 example at Mount St. Helens. If a primary debris avalanche occurs, downstream warnings of a subsequent lahar would be necessary. The lag time at Mount St. Helens from avalanche emplacement to lahar initiation was about five hours. This sequence of events can be regarded as a much less probable variant of both the maximum lahar and Case I, below. DESIGN OR PLANNING CASE I Case I is a large cohesive debris flow having a recur- rence interval of 500 to 1,000 years and is the appropriate case for long-term planning in the watersheds draining Mount Rainier. Even one event (or more) equal to or greater than a flow with a 1,000-yr recurrence interval has a 9.5 percent probability of occurring at least once in the next century (Reich, 1973). If the Osceola Mudflow is excluded on the grounds that it is a statistical outlier of this flow type, several smaller cohesive lahars form a discrete population. The most recent and best defined of these flows is the Electron Mudflow. The importance of this flow to hazard analysis has long been recognized (Crandell, 1971; Cullen, 1977; and Cullen Tanaka, 1983). The lahar is here assigned a magnitude and frequency, and its dynamics are specified at the margin of the Puget Sound lowland, where risk increases greatly (table 7, pl. 1). The volume of the Electron Mudflow deposits on the Puget Sound lowland was satisfactorily determined by Crandell (1971, p. 57) as slightly more than 183 million m*. The flow deposit is overlain by reworked deposits of the flow. Its original volume (table 7) is estimated by assuming deposition near the levels of the highest medial flow deposits. This assumption is based on the downstream behavior of the cohesive lahar originating in the North Fork Toutle River at Mount St. Helens in 1980. The risk of this type of flow surpasses that of all smaller but more frequent flows. Moreover, the risk is increased by the lack of a clear association with major episodes of volcanic activity which could provide a warning (Crandell, 1971; Scott and Janda, 1987). Such flows may be triggered by nonmagmatic seismicity, by steam eruptions, or just by gravity in places where a failure plane has been lubricated by clay and geothermal pore fluids. No assumption of precursor volcanic activity can be made in planning for these flows. This is a conservative approach that is consistent with the available evidence. Sector collapses of the size that produce cohesive lahars can occur on any side of the volcano (Frank, 1985, p. 181). Given the lack of evidence that one flow of this type will stabilize the affected sector of the volcano thereafter, this is the best assumption. Potential effects on downstream areas differ only slightly among watersheds. The main complicating factor is the presence of reservoirs in three of the five major watersheds. A modern recurrence of a large cohesive lahar will inundate a larger area of the Puget Sound lowland than did the prehistoric flows because of the greatly reduced friction on deforested flood plains. The distribution of a modern flow can be predicted by estimating the deposit thickness on unforested flood plains and distributing the design volume at the mountain front over the corresponding area. On the basis of the behavior of the 1980 cohesive lahar at Mount St. Helens, which traversed clearcut and forested flood plains, the modern thickness would be close to 70 percent of the prehistoric thickness. Some additional bulking of the flow on the cleared flood plains will reduce its natural rate of attenuation, but erosion will probably be concentrated in active channels as it was under forested conditions. Thus the inundation area of a modern flow of the same type and same original volume as the Electron Mudflow could increase to approximately 50 km> (compared to the 36-km> area of the Electron). A similar but somewhat larger volume will just be spread over a larger area. The flow record indicates that the most probable recurrence of Case I will be a debris avalanche that transforms to a lahar on or near the volcano. As noted for the "maximum labar," this origin is not a certainty. Untransformed debris avalanches (Crandell, 1971, fig. 18) A &... che...... Apo... DESIGN OR PLANNING CASES AND HAZARD ZONATION 47 can be regarded as a much less probable variant of Case I. Because such flows have not occurred at Mount Rainier, it is impossible to specify probable magnitudes or frequencies except by means of examples at other volcanoes, as Crandell has done. DESIGN OR PLANNING CASE II Case II is a noncohesive flow represented by the National Lahar, which with its runout phases is a suitable example of this category that can be extrapolated to all watersheds. The recurrence interval of noncohesive flows in the size range of the National is near the lower end of the 100- to 500-yr range and thus is analogous to the 100-yr flood, one widely considered for structure design and flood-plain management. Flow cross sections (pl. 1) can be applied upstream from reservoirs. Comparison shows that this design debris flow will be larger than design water floods in upstream reaches, but will be smaller downstream. This difference is explained by the continuous attenuation of a lahar or lahar-runout flow, as compared with the typical amplification of a meteorologic flood as tributary inflows increase downstream. Measured flood-carrying capacities of the Puyallup, White, and Carbon Rivers on the Puget Sound lowland illustrate this trend (Prych, 1987). Nonetheless, the noncohesive debris flows increase the risk of flood-plain inundation throughout a river system without reservoirs. Upstream, the lahar subpopula- tion presents more risk than meteorologic floods. Inundation levels can be estimated by adjusting the cross-sectional areas (pl. 1) for the attenuation, as shown, due to distance from the volcano. The flow wave in this design and planning case consists of hyperconcentrated flow during much of the flow interval beyond the base of the volcano. Hyperconcentrated flow probably will persist to the boundary of the Puget Sound lowland, but will transform to normal streamflow rapidly beyond that point because of rapid loss of sediment from the flow wave on flood-plain surfaces. The runout phases of the National Lahar (figs. 2, 11B, and 11C) are representative of changes expected in future noncohesive flows. If Case I presents the greatest total risk, should Case II be considered as well as Case I in any part of a drainage? The answer here is affirmative, because of the distinction between planning for the longest term that is cost-effective, as in a land-use evaluation contingent on Case I, and designing for a flow with a high degree of probability during the life of an individual structure. For example, a flow equal to or greater than the event with a recurrence interval of 100 years has a 64 percent probability of occurring at least once in the next century (Reich, 1973). An additional rationale for the application of Case II is its probably greater association with precursory volcanic activity than Case I. In the event of impending eruptive activity forecast by a monitoring network, Case II is the minimal flow event that logically can be expected. Each river system contains enough glacial ice to provide meltwater capable of producing a noncohesive lahar of this size. DESIGN OR PLANNING CASE III Case III is a relatively small debris avalanche, originating as a landslide, that probably will transform to a debris flow. Two moderate-sized and several small debris and rock avalanches have occurred since 1900. The largest of these came within a kilometer of the White River Campground in 1963, albeit at a time (December) when the campground was closed (Crandell and Fahnestock, 1965). Neither moderate-sized flow transformed directly to a lahar, but both produced small debris flows by dewatering and slumping of their surfaces. The origin of the Case III flow is the same as that of both Case I and the maximum lahar, but the smaller Case III examples occur much more frequently on the volcano. They probably will recur without warning and certainly will move at high velocity. As the case history best exemplifying this flow, the debris avalanche yielding the Tahoma Lahar (pl. 1) is the most appropriate example. At least part of that debris avalanche transformed to a hummocky lahar with a flow depth of as much as 55 m in confined canyons on the vol- cano and 20 m shortly beyond the base of the volcano. Sub- sequent attenuation was rapid. Greatly adding to the risk from these flows is their high velocities, almost certainly well in excess of 30 m/s (67 mph). The velocity of the largest avalanche from Little Tahoma Peak was at least 35 to 40 m/s (Crandell and Fahnestock, 1965). A velocity of about 50 m/s, which was increasing at the point of measurement, was reported for the 1980 debris avalanche at Mount St. Helens (Voight and others, 1981). The best outcome of planning for such a rapid flow can only be to minimize exposure. Risk associated with a runout flow like that possibly developed from the Tahoma Lahar will be much less than that of Case II. Within the Park, however, consideration can be given to siting new campgrounds and facilities above the flow depths of the Tahoma flow wave (pl. 1), and in other drainages, above its extrapolated cross-sectional area adjusted for distance from the summit. HAZARD ZONATION Hazard-zone analysis is another approach to assessing the composite risk of all flows, including those smaller and more frequent than the above cases. The approach involves the determination of past inundation levels from tephra 48 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON layers, vegetation, and fan and flood-plain morphology. The combination of criteria yields areas of inundation over several time intervals useful for land-use planning. Hazard zones delineated by deposits and dendrochronology have effectively defined the risks of small, high-frequency flows at Mount Shasta (Osterkamp and others, 1985). A similar analysis can be useful in siting individual facilities in Mount Rainier National Park. Hazard zones, however, are not calculated or presented in this report. The zones can be determined, as needed, for specific locations according to the following guidelines. ZONE I Tephra set W, deposited over most of the National Park, can be treated as a paleohydrologic crest-stage gage by measuring the height to which the layer has been truncated by flow against valley side slopes. Set W consists of two layers, deposited in A.D. 1480 and 1482 (Yamaguchi, 1983 and 1985). The first of the two layers predominates at Mount Rainier. Its eroded lower margin records the highest level of flow since deposition in at least the upper part of each drainage surrounding the mountain. Where a drainage did not convey a significant flow like the Tahoma Lahar in the last 500 years, the level to which set W is eroded provides an estimate of the inundation potential without a major eruption or sector collapse. Zone I is thus defined. Because zone I defines only the most recent time interval, preference in planning should be given to the planning and design case histories selected with the perspective of longer time priods. ZONE II The area inundated since the beginning of the 20th century can be established by the historical record and by dendrochronology; such data can provide a good approxi- mation of the level to which flow has extended in the last century. In most cases, glacial-outwash flow pattems subsequent to the start of Neoglacial recession about 60 to 200 years ago (Sigafoos and Hendricks, 1961 and 1972; Burbank, 1981) are discernible from vegetation patterns on aerial photographs. ZONE III This area encompasses the modern alluvial fans at the base of the volcano, active flood plains, and marginal areas subject to lateral erosion. The boundaries would seem obvi- ous except on the alluvial fans, whose surfaces are broad and convex and locally support a mature forest. The fans record debris flow deposition triggered by the decreasing slope and expansion in reaches at the base of the volcano. Debulking is rapid in such areas, which explains why the smaller flows attenuate rapidly (fig. 13). LATERAL EROSION ASSOCIATED WITH HAZARD ZONE III Lateral (or bank) erosion is an additional hazard down- stream along the streams with normal flood plains. The high terrace bordering the Nisqually River upstream from Longmire (figs. 9 and 11A; set W on surface) is being cut by normal fluvial erosion, requiring some resiting of trails. Progressive lateral erosion tends to be localized and is controlled by factors such as channel pattern (Brice and Blodgett, 1978, chap. 4), bank material (Schumm, 1960, 1961), and vegetation (Scott, 1981). Channel pattern normally will be the determining fac- tor in localizing erosion at banks cut against a terrace. A meandering pattern will result in erosion at the outsides of bends, for example. Braided streams in the proglacial environments may regularly impinge against bedrock valley walls, and where streams are confined by steep-sided Neoglacial moraines, erosion can be large, episodic, and unpredictable in location. Trails and climbing routes along crests of Neoglacial moraines are subject to mass failures triggered by lateral erosion at the base of the moraine. Most of the major streams fall in the category of "streams wider at bends," which, according to Brice and Blodgett (1978), have greater lateral instability than either equiwidth streams or those with random variation. The White River shows the effects of a cohesive valley fill, the deposits of the Osceola Mudflow, in reducing lateral erosion. Where the active channel is incised into the valley fill, an uncommonly low width/depth ratio results (in the range of 4-7 at locations downstream from the national park). This low ratio is a function (inverse) of the high silt and clay content of the central-valley facies of the Osceola and is consistent with the findings of Schumm (1960) elsewhere. PROBABILITY OF PRECURSOR VOLCANIC ACTIVITY Detectable volcanic activity may precede the largest cohesive and noncohesive lahars, but this is not a premise on which the planning process can rely. As demonstrated by the flow record over the last thousand years (a period including lahars that exemplify each of the planning cases), no flows correspond to the single known episode of major activity during that time (tables 1 through 4). A correlation with precursor events is suggested by the concentration of large flows, such as the Osceola Mudflow and Paradise Lahar, during the mid-Holocene when many tephra-producing as a s 4A caca te tbe TRAVEL TIMES OF LAHARS AND POTENTIAL RESERVOIR EFFECTS 49 events occurred. This association, noted by Crandell (1971) and Mullineaux (1974, p. 17), does not relate the flows directly to volcanism but is evidence of a linkage that may lack better definition only because of the confidence limits on radiocarbon dating. The potential nonvolcanic causes of both the cohesive and noncohesive debris flows and of the relatively small debris avalanches at Mount Rainier include (1) regional, nonmagmatic seismicity; (2) edifice effects; and (3) several phreatic effects of the active hydrothermal system, including rapid ice or snow melting, steam eruptions, failure in response to increased pore pressure, and lubrication of potential slip surfaces such as those of previous deep-seated failures again buried by edifice construction. The first two effects are discussed further here. Mount Rainier is the site of occasional small earthquakes, the largest two of which may have been due to a strike-slip fault on the south side of the volcano (Crosson and Frank, 1975; Crosson and Lin, 1975). Nevertheless, the general areal distribution of historical earthquakes, such as a cluster of seven earthquakes in 1987 with magnitudes of 0.8 to 2.1 at depths less than 5 km, shows a clear association with the volcano (University of Washington Geophysics Program, written commun., 1988). These earthquakes may _ be the result of edifice effects. Like many subduction-related stratovolcanoes, Mount Rainier is noteworthy for the large mass of layered material at high altitude, leading to gravitational stresses such as those described as edifice effects in a study of Hawaiian volcanoes (Fiske and Jackson, 1972). Some microearthquake activity at Rainier was ascribed to these crustal-loading effects (Unger and Decker, 1970), a view later modified (Unger and Mills, 1972). Some low-frequency tremors recorded at Longmire may result from glacier or debris flow movement in the Tahoma Creek, Kautz Creek, or upper Nisqually River drainages. The Rainier area is subject to large regional earth- quakes (Gower, 1978). If the Cascadia subduction zone off- shore is storing the elastic energy characteristic of other subduction zones, several great earthquakes are necessary to fill the seismic gap represented by the zone (Heaton and Hartzell, 1987). The deep, plate-boundary earthquakes of 1949 (M 7.1) and 1965 (M 6.5), both with epicenters on the east side of Puget Sound, caused local incidences of rock- falls, slope failures, and flood-plain liquefaction throughout much of western Washington (Schuster and Chleborad, 1989). As many as five great earthquakes have occurred in the last 3,100 years, the latest about 300 years ago, as sug- gested by the stratigraphy of buried wetlands in southwest Washington (Atwater, 1988). Major slope failures can occur, however, in response to minor seismic accelerations at times of high susceptibility; for example, the largest his- toric landslide in Canada was probably triggered by a small (M 3.1) seismic event (Evans, 1989). Many similar volcanoes have experienced collapses yielding large debris avalanches; Mount Rainier is appar- ently unusual because the large collapses have transformed directly to lahars, but more detailed study may show this also has occurred at some other volcanoes. Major sectors of the mountain are composed of steep, outward-dipping lava flows (frontispiece) between which hydrothermal water has infused and altered material to clay-rich zones that are potential slip surfaces (fig. 22). Where exposed at the sur- face, these zones range from 0.2 to several meters in thick- ness. At depth, hydrothermal alteration is doubtless more intense. Consequently, volcaniclastic flows beginning as large debris avalanches from Mount Rainier may have little correlation with the warning signs of an eruption. Only one of the modes of collapse described by Francis and Self (1987) requires precursory activity. Moreover, Siebert (1984) found a uniform relation between the height of origin and the runout distance of debris avalanches, regardless of whether the initial slope failure was induced by an explosion or merely by gravity, indicating that the energy from an explosive initiation does not increase the runout distance and thus does not increase the hazard. TRAVEL TIMES OF LAHARS AND POTENTIAL RESERVOIR EFFECTS The Puyallup and Carbon River systems presently (1994) have no reservoirs. The White and Nisqually River systems each have one reservoir, and the Cowlitz has two of greatly differing capacities. All but one of the reservoirs are far enough from the volcano that they are not at direct risk of a debris avalanche, except for secondary or transformation phases. All are at some risk from a Case I lahar; the travel time of such a flow from the volcano to the upstream end of each reservoir would be no more than several hours. Other possible reservoir impacts include the beneficial, hazard-reducing effects of impounding or attenuating volcanic flows. The importance of a warning of dam failure is dramatically illustrated by the much greater numbers of survivors where warning was received (Costa, 1985). The importance of a warning system is equally applicable to Case I and Case II lahars whether or not a reservoir is involved. So rapid and localized are Case III flows that only advance planning and minimizing exposure will be effective. TRAVEL TIMES OF LAHARS The time it takes a debris flow wave to travel from point to point involves uncertainties, but a range of values can be estimated. A major source of uncertainty is the difference between mean flow velocity (table 7) and the flow-wave velocity (celerities in table 8). The first of these is the observed (or estimated) forward speed of mud and water in 50 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON Figure 22. Megaclast at surface of the 1963 debris avalanche below the Emmons Glacier in the main fork of the White River. Note the lighter colored zone of hydrothermal alteration (outlined) developed along a flow contact or fracture zone in volcanic breccia. The zone is about 1 m thick and is typical of flow contacts (frontispiece), fractures, and fault zones at Mount Rainier. the flow; the latter is the rate at which the form of the flow wave progresses down the channel (celerity). A pronounced difference is apparent from a comparison of the velocity data of Fairchild (1985) and the celerity data of Cummans (1981) for the same 1980 lahars at Mount St. Helens. In gen- eral, the velocity of the material in the flow at the time of peak stage was faster than the speed with which the flow wave moved. For example, the average peak flow velocity of the 1980 North Fork lahar was about 9 m/s (Fairchild, 1985, fig. 4.4) over a channel interval where celerity of the maximum stage was only 2.1 m/s (Cummans, 1981, p. 485). The factors causing this difference are summarized below. The 1980 North Fork lahar was texturally similar in behavior to the Case I lahar, and each was derived from a debris avalanche. The 1980 lahar originated by slumpin g of the surface of the 1980 avalanche, whereas the Case I example was mobilized directly from such a flow. The behavior of the two cohesive lahars was probably similar with the exception that the 1980 flow probably was more broadly peaked at the point of origin. Thus, the celerity of the Case I lahar was likewise less than the actual flow velocity. Although a flood wave of water commonly travels Table 8. Celerities and travel times of the maximum lahar, Case I lahar, and Case II lahar from Mount Rainier to the nearest down- stream reservoir or the Puget Sound lowland. [Maximum and minimum values of celerity calculated as described in text; R., River; Res., Reservoir] Property Maximum - CaseI - Case II lahar lahar lahar LRAT LYDE 22200202002 200 server Cohesive Cohesive Nonco- hesive Celerity of peak stage! (m/s): Approximate maximum......................... 25 22 8 Approximate minimum ........................ 6 5 3 Range of possible travel times® (hours): White R. to Mud Mtn. Res. (56 km) 06-26 0.7-3.1 1.9-5.2 Cowlitz R. to Riffe Lake (77 km) ...... 0.9-3.6 1.0-4.3 2.7-7.1 Nisqually R. to Alder Res. (45 km) ... 0.5-2.1 06-25 16-42 Puyallup R. to lowland (38 km) ......... 0.4-1.8 - 0.5-2.1 1.3-3.5 Carbon R. to lowland (38 km) ........... 0.4-1.8 _ 0.5-2.1 1.3-3.5 ! Probably equivalent to peak discharge. ? Equal to estimated mean peak velocity between the volcano and either the first reservoir or the Puget Sound lowland. 3 Values in parentheses show distance along valley axis from volcano to head of reservoir or to boundary of Puget Sound lowland. TRAVEL TIMES OF LAHARS AND POTENTIAL RESERVOIR EFFECTS 51 faster than its constituent water and sediment particles, this is not possible in a debris flow. The highest possible celerity of a lahar is the peak flow velocity. The minimum celerity may be estimated by the relation of celerity to peak flow velocity for the 1980 North Fork lahar. That ratio is 0.23. By such reasoning, the celerity of the Case I lahar varies from a minimum of 5 m/s to a maximum of about 22 m/s. This range of possible values is used to estimate travel times of the Case I lahar to the reservoirs (or the Puget Sound lowland) in river systems draining the volcano (table 8). Because the actual behavior of comparable flows is utilized, values intermedi- ate within this range but tending nearer the longer travel times (Iower celerities) can be considered for planning emer- gency response. Using an intermediate value, rather than the longest travel time, will adjust for the more rapid arrival of the leading edge of the flow. For example, though the possible travel times for the Case I lahar down the Puyallup or Carbon Rivers to the lowland range from 0.5 to 2.1 hours (table 8), a probable range for planning purposes would be 1.0 to 1.5 hours. The remaining uncertainties, however, require that these figures be treated as approximations. By similar reasoning, the behavior of the noncohesive South Fork lahar at Mount St. Helens was probably similar to that of the noncohesive Case II flow. The ratio of celerity to peak flow velocity over a 40-km interval for the 1980 lahar was about 0.35 (Cummans, 1981; Fairchild, 1985). Consequently, the possible values of Case II celerity could range from a minimum of about 3 m/s to a maximum of about 8 m/s. These values are used to compute possible travel times (table 8). Several factors contribute to this apparent discrepancy between the peak flow velocity and the wave celerity of debris flows: (1) the velocity measurements use relations unverified for debris flows and may be too high (Costa, 1984); (2) material may be continuously recycled through the peak to a point of temporary storage lateral to or beneath the flow, and that material then may re-enter the flow by draining back after passage of the peak, again to move forward into the peak (the caterpillar-track-path analogy; Johnson, 1984, p. 287); (3) velocity-measurement sites may be concentrated in narrow reaches where flow is faster than the average rate over many reaches; and (4) the flow waves commonly broaden during movement (Fairchild, 1985; Pierson and others, 1990), increasing the distance between the leading edge and the peak stage of the flow, and thereby reducing peak stage celerity. RESERVOIR EFFECTS WHITE RIVER The Mud Mountain Dam (fig. 1), near the boundary of the Cascade Range and the Puget Sound lowland, is entirely a flood-control structure and is normally drawn down to negligible storage. The maximum-impoundment level extends to within 56 km downvalley from the summit of Mount Rainier. Capacity above dead storage is 131 million m*, about 57 percent of the Case I volume (table 7). The actual proportion of the Case I volume retained would be greater because the depositional surface of the debris flow deposits would approach the slope of the stream channel, whereas the calculated reservoir capacity assumes a horizontal water surface. In addition, the backwater effects of the debris flow deposits could trigger an additional, unknown volume of upstream deposition. The dam is a rock-and-earthfill structure with a rolled impervious core and probably could safely contain at least its capacity, so that any flow but the maximum lahar would be significantly attenuated. No wave of translation is likely to be generated by a lahar entering the reservoir, and any volcanically induced flow across the spillway would be debris flow. A concern is spillway abrasion by intralahar impact forces, described by Scott (1989). Runouts of Case II and III flows would be contained by the reservoir. Probable routing of any uncontained flow, as interpreted by Crandell (1971), would be within the White River valley unless (1) valley walls near Buckley were overtopped, sending part of the flow into the Carbon River drainage, or (2) flow extended to Auburn and crossed into the Green River drainage. COWLITZ RIVER The large, concrete-arch Mossyrock Dam (Riffe Lake, fig. 1) is a multipurpose impoundment. The upstream end of the reservoir at capacity is 77 km downvalley from the summit of Mount Rainier. Usable storage capacity is 1,600 million m*, a value 7 times the volume of the Case I example but only half that of the maximum lahar (table 7). Not all of this capacity is available because levels and releases are subject to flood-control and power-generation agreements. The normal operational goal is for annual refill by July 31. Gradual releases are normally maintained through the rest of the summer, increasing after late October in preparation for maximum flood-control drawdown between December 1 and March 1. Refill then begins with snowmelt runoff. The Case I example would be contained during the winter draw- down period and much of the remaining period. Runouts of Case II and III flows are not of concern at any time. The risk of dam failure resulting from a wave generated by a volcanically induced inflow is greatly reduced by the improbability of a large debris avalanche extending to the reservoir. A lahar inflow would be far more gradual than either a landslide inflow, as at Vaiont, Italy (Kiersch, 1964), or a hypothetical debris avalanche like the one that could possibly enter Swift Reservoir near Mount St. Helens (Major and Scott, 1988). Given the large attenuation that would occur in the lahar flood wave through deposition on 52 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON the Cowlitz River flood plain upstream, it is unlikely that a lahar-generated wave by the Case I example could cause failure. The prehistoric Case I inundation area is, in fact, similar to the flood-plain area upstream of Riffe Lake. Riffe Lake could be drawn down in response to an emergency if precursor volcanic activity has occurred and the risk of a lahar is high. Once a lahar has been initiated, though, time would be too short for any significant draw- down. The maximum drawdown rate, without presently causing inundation downstream at Castle Rock, is in the approximate range of 1,100 to 1,400 m*/s, or about 0.3 percent of usable storage in 1 hour with zero inflow (data from Department of Public Utilities, City of Tacoma). Probable travel times of both the maximum lahar and the Case I lahar are less than approximately 4 hours (table 8). NISQUALLY RIVER Alder Dam (fig. 1) is a small version of Mossyrock Dam, likewise of concrete-arch design and with similar operational modes. Usable capacity, corrected for the most recent sediment survey, is about 198 million m. This value is a negligible portion of the Maximum lahar but is approx- imately 86 percent of the Case I example (table 7). However, sedimentation rate and trap efficiency are high, and several additional percent of capacity have probably been lost. As at Riffe Lake, this capacity is never entirely available. The dam could clearly be destroyed by the maximum lahar. It is also the most vulnerable to a Case I flow of the three reservoirs potentially affected by volcanic debris flow. It is the reservoir most vulnerable to failure caused by a wave of translation, because the relatively confined valley upstream is capable of conveying a large lahar without great volume loss. Consequently, a high priority needs to be given to drawing down this reservoir in the event of probable volcanic activity. Inflowing lahars will generate long-period waves that will translate through the reservoir to affect the dam. In a model of the behavior of Swift Reservoir in response to rapid inflow (Pacific Power and Light, 1980), peak wave action in the reservoir occurred before significant water level rise. That reservoir is closer to Mount St. Helens (13 km by the most direct route) than Alder Reservoir is to Mount Rainier (45 km). Thus, the time to peak of a lahar wave is probably greater at Alder Reservoir, resulting in a more gradual rise. However, given the variety of attenuation rates and wave shapes possible for different flows at Mount Rainier, this is not certain. What is certain, however, is that the riskiest potential lahar reaching Alder Reservoir from Mount Rainier is far larger than that reaching Swift Reser- voir from Mount St. Helens and that volumetric impacts probably will dominate as hazards at Alder Reservoir. Cran- dell (1988, fig. 18) shows Alder Reservoir to be within range _ of a debris avalanche that has not transformed to a lahar. Loss of capacity in Alder Reservoir, because of the high "background" sedimentation rate, will probably make the structure uneconomic in a fraction of the recurrence interval of the Case I flow. When the reservoir is filled with sediment, decisions either to modify the structure and increase capacity or to abandon it need to consider the possibility of a Case I lahar. The same aspects of emergency response that apply to Riffe Lake also apply to Alder Reservoir. Total streamflow at the highway bridge near Yelm cannot exceed 227 m*/s without causing flooding of habitation there (data from Department of Public Utilities, City of Tacoma). If a release of 150 m*/s from the reservoir is possible (depending on downstream tributary inflows), drawdown with zero inflow is at a rate of 0.3 percent of total capacity per hour. The travel time of a Case I flow from the volcano is less than about two hours (table 8). CONCLUSIONS Two types of debris flow have occurred periodically throughout postglacial time at Mount Rainier: (1) cohesive debris flows, containing at least 3 to 5 percent clay, which have flowed untransformed more than 100 km from the volcano; and (2) noncohesive debris flows, containing less than 3 to 5 percent clay, which commonly transform down- stream to more dilute flows, passing through the range of hyperconcentrated flow. The noncohesive debris flows form most commonly by bulking of a flood surge with volcaniclastic and morainal sediment, and during formation they also pass through the range of hyperconcentration. Three subgroups of a mixed population of lahars and glacially related debris flows were studied over time intervals related to their frequency, and from this spectrum of case histories an example of each was selected for consideration in flow-hazard analysis. Case I, a large cohesive lahar formed by mobilization of a deep-seated landslide, is capable of inundating parts of the Puget Sound lowland or the Cowlitz River valley. It is suitable for consideration in hazards planning in lowland areas. Case II is a noncohesive lahar of intermediate size and sufficient frequency that it may be applicable to the design of some structures such as dams and power plants around the volcano. Case III originates as a debris avalanche of a typical size observed at Mount Rainier, and which probably will mobilize to form a lahar. It poses risk primarily to local areas within Mount Rainier National Park. The maximum lahar is typified by the largest flow in the postglacial history of Mount Rainier. It is a statistical outlier of the group of large cohesive lahars. The smallest and most frequent flows are dominated by glacial-outburst floods that bulk to debris flows and provide behavioral models of the larger noncohesive debris flows. They commonly attenuate rapidly at the base of the volcano through the rapid debulking of sediment, yielding hyperconcentrated streamflow and REFERENCES CITED 53 secondary debris flow by any of three types of transforma- tions. Their inundation potential can be assessed and hazard zones can be established based on the level of erosion of tephra set W, on dendrochronology, and on numerous his- toric flows. . Each of the five major river systems draining the volcano has a record of lahars. Although the records indicate that flows differ in size and frequency among river systems, the risk of future lahars may not correlate highly with the record in each individual system. Rather, extrapolation of the entire volcano's past history to the future is more appro- priate, with the risk dispersed among the individual drain- ages. Sites of future instability of the type producing major areas of collapse and the large cohesive lahars cannot be forecast. This is likewise true of the sites where intensified geothermal activity will produce noncohesive lahars by melting of ice and snow, and the places where renewed explosive volcanism will greatly increase the risks of all types of flows. However, the types of flows that will recur, as well as their approximate sizes and probabilities, can be forecast for the volcano as a whole. This is the most essential element of flow-hazard analysis. REFERENCES CITED Atwater, B.F., 1988, Buried Holocene wetlands along the Johns River, southwest Washington [abs.]: Proceedings of the Symposium on Holocene Subduction in the Pacific Northwest, May 1988, Quatemary Research Center, University of Washington, p. 4. - Bacon, C.R., 1983, Eruptive history of Mount Mazama and Crater Lake caldera, Cascade Range, U.S.A.: Journal of Volcanology and Geothermal Research, v. 18, p. 57-116. Bagnold, R.A., 1962, Auto-suspension of transported sediment- Turbidity currents: Proceedings of the Royal Society of London, v. A225, p. 315-319. Bethel, J.P., 1981, Volcanic mudflow deposits in the Cowlitz River valley, Washington [abs.]: Geological Society of America Abstracts with Programs, Cordilleran Section, March 25-27, 1981, Hermosillo, Sonora, Mexico, v. 13, no. 2, p. 45. Bretz, J H., 1913, Glaciation of the Puget Sound region: Washington Geological Survey Bulletin 8, 244 p. Brice, J.C., 1981, Stability of relocated stream channels: Federal Highway Administration Report, no. FHWA/RD-80/158, 177 p. Brice, J.C., and Blodgett, J.C., 1978, Countermeasures for hydraulic problems at bridges-Volume 1, Analysis and assessment: Federal Highway Administration Report, no. FHWA-RD-78-162, 169 p. Burbank, D.W., 1981, A chronology of late Holocene glacier fluctuations on Mount Rainier, Washington: Arctic and Alpine Research, v. 13, p. 369-386. Costa, J.E., 1984, Physical geomorphology of debris flows, in Costa, J.E., and Fleisher, P.J., eds., Developments and applica- tions of geomorphology: Berlin, Springer-Verlag, p. 269-317. 1985, Floods from dam failures: U.S. Geological Survey Open-File Report 85-560, 54 p. Crandell, D.R., 1963a, Paradise debris flow at Mount Rainier, Washington, in Short papers in geology and hydrology: U.S. Geological Survey Professional Paper 475-B, p. 135-139. 1963b, Surficial geology and geomorphology of the Lake Tapps Quadrangle, Washington: U.S. Geological Survey Professional Paper 388-A, 84 p. 1969, Surficial geology of Mount Rainier National Park, Washington: U.S. Geological Survey Bulletin 1288, 41 p. 1971, Postglacial lahars from Mount Rainier Volcano, Washington: U.S. Geological Survey Professional Paper 677, 73 p. 1988, Gigantic debris avalanche of Pleistocene age from ancestral - Mount - Shasta - volcano, - California, - and debris-avalanche hazard zonation: U.S. Geological Survey Bulletin 1861, 32 p. Crandell, D.R., and Fahnestock, RK., 1965, Rockfalls and avalanches from Little Tahoma Peak on Mount Rainier, Washington: U.S. Geological Survey Bulletin 1221-A, 30 p. Crandell, D.R., and Miller, R.D., 1974, Quaternary stratigraphy and extent of glaciation in the Mount Rainier region, Washington: U.S. Geological Survey Professional Paper 847, 59 p. Crandell, DR., Miller, C.D., Glicken, H.X., Christiansen, RL., and Newhall, C.G., 1984, Catastrophic debris avalanche from ancestral Mount Shasta volcano, California: Geology, v. 12, p. 143-146. Crandell, D.R., and Mullineaux, D.R., 1978, Potential hazards from future eruptions of Mount St. Helens volcano, Washington: U.S. Geological Survey Bulletin 1383-C, 26 p. Crosson, R.S., and Frank, David, 1975, The Mount Rainier earth- quake of July 18, 1973, and its tectonic significance: Seismological Society of America Bulletin, v. 65, p. 393-401. Crosson, R.S., and Lin, J.W., 1975, A note on the Mt. Rainier earthquake of April 20, 1974; Seismological Society of America Bulletin, v. 65, p. 549-556. Cullen, J.M., 1977, Impact of a major eruption of Mount Rainier on public service delivery systems in the Puyallup Valley, Washington: Seattle, Wash., University of Washington, M.S. thesis, 202 p. Cullen Tanaka, Janet, 1983, Volcanic hazards assessment for Mt. Rainier, Washington: Olympia, Wash., Washington State Department of Emergency Services, 28 p. Cummans, John, 1981, Chronology of mudflows on the South Fork and North Fork Toutle River following the May 18 eruption, in Lipman, P.W., and Mullineaux, D.R., eds., The 1980 eruptions of Mount St. Helens, Washington: U.S. Geological Survey Professional Paper 1250, p. 479-486. Dethier, D.P., and Bethel, John, 1981, Surficial deposits along the Cowlitz River near Toledo, Lewis County, Washington: U.S. Geological Survey Open-File Report 81-1043, 11 p. Dibble, RR., Nairn, L.A., and Neall, V.E., 1985, Volcanic hazards of North Island, New Zealand-Overview: Journal of Geodynamics, v. 3, p. 369-396. Driedger, C.L., and Kennard, P.M., 1986, Ice volumes on Cascade volcanoes-Mount Rainier, Mount Hood, Three Sisters and Mount Shasta: U.S. Geological Survey Professional Paper 1365, 28 p. Erdmann, C.E., and Johnson, A., 1953, Preliminary report, flood problem on Nisqually River at Longmire, Washington: 54 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON National Park Service unpublished report [available at National Park Service library, Longmire, Wash.], p. 2-11. Evans, $.G., 1989, Earthquake-induced landslides in the Canadian Cordillera [abs.]: Geological Society of America Abstracts with Programs, v. 21, no. 5, p. 77. Fairchild, LH., 1985, Lahars at Mount St. Helens, Washington: Seattle, Wash., University of Washington Ph.D. thesis, 374 p. Fisher, R.V., 1983, Flow transformations in sediment gravity flows: Geology, v. 11, p. 273-274. Fisher, R.V., and Schmincke, H.U., 1984, Pyroclastic rocks: Berlin, Springer-Verlag, 472 p. Fiske, R.S., Hopson, C.A., and Waters, A.C., 1963, Geology of Mount Rainier National Park, Washington: U.S. Geological Survey Professional Paper 444, 93 p. Fiske, R.S., and Jackson, E.D., 1972, Orientation and growth of Hawaiian volcanic rifts-The effect of regional structure and gravitational stresses: Proceedings of the Royal Society of London, ser. A, v. 329, p. 299-326. Folk, RL., 1980, Petrology of sedimentary rocks: Austin, Tex., Hemphill Publishing Company, 182 p. Francis, Peter, and Self, Stephen, 1987, Collapsing volcanoes: Scientific American, v. 256, p. 90-97. Frank, D.G., 1985, Hydrothermal processes at Mount Rainier: Seattle, Wash., University of Washington, Ph.D. thesis, 195 p. Frank, D.G., and Friedman, J.D., 1974, Newly detected infrared anomalies at Mount Rainier, Washington [abs.], in Symposium on applications of satellite and airplane remote sensing of natural resources in the Pacific Northwest, Tumwater, Washington, May 2, 1974: Washington Department of Natural Resources, p. 22. Franklin, J.F., Cromack, Kermit, Jr., Denison, William, McKee, Arthur, Maser, Chris, Sedell, James, Swanson, Fred, and Juday, Glen, 1981, Ecological characteristics of old-growth Douglas-fir forests: U.S. Forest Service, Pacific Northwest Forest and Range Experiment Station, General Technical Report PNW-118, 48 p. Frizzell, V.A., Jr., Tabor, R.W., Booth, D.B., Ort, K.M., and Waitt, RB., Jr., 1984, Preliminary geologic map of the Snoqualmie Pass 1:100,000 Quadrangle, Washington: U.S. Geological Survey Open-File Map 84-693. Glicken, Harry, 1986, Rockslide-debris avalanche of May 18, 1980, Mount St. Helens volcano: Santa Barbara, Calif., University of California, Ph.D. thesis, 303 p. Glicken, Harry, and Schultz, P.H., 1980, Martian channel erosion-The lahar analogy: Planetary Science, v. 11, no. 5. Gower, HD., 1978, Tectonic map of the Puget Sound region, Washington, showing locations of faults, principal folds, and large-scale Quaternary deformation: U.S. Geological Survey Open-File Report 78-426, 21 p. Grater, RK., 1948a, A report on the Kautz Creek flood studies: National Park Service unpublished report [available at National Park Service library, Longmire, Wash.], 9 p. 1948b, The flood that swallowed a glacier: Natural History, v. 57, p. 276-278. Guy, H.P., 1970, Fluvial sediment concepts, chapter C1 of Book 3, Applications of hydraulics: U.S. Geological Survey Techniques of Water-Resources Investigations, TWI 3-C1, 55 p. Hampton, M.A., 1975, Competence of fine-grained debris flows: Journal of Sedimentary Petrology, v. 45, p. 834-844. Heaton, T.H., and Hartzell, S.H., 1987, Earthquake hazards on the Cascadia subduction zone: Science, v. 236, p. 162-168. Hooke, R. LeB., 1967, Processes on arid-region alluvial fans: Journal of Geology, v. 75, p. 438-460. Iverson, R.M., 1989, Pore-water effects in mass-movement mechanics: International Geological Congress, 28th, Washing- ton, 1989, Abstracts, v. 2, p. 104-105. Iverson, R.M., and Denlinger, R.P., 1987, The physics of debris flows-A conceptual assessment, in Beschta, R.L., Blinn, T., Grant G.E., Ice, G.G., and Swanson, F.J., eds., Erosion and sed- imentation in the Pacific Rim: International Association of Hydrological Sciences Publication 165, p. 155-165. Jahns, RH., 1949, Desert floods: Engineering and Science Monthly, California Institute of Technology Contribution No. 499, p. 10-14. Johnson, A.M., 1965, A model for debris flow: University Park, Pa., Pennsylvania State University, Ph.D. thesis, 305 p. 1970, Physical processes in geology: San Francisco, Freeman, Cooper, and Co., 577 p. 1984, Debris flow (with contributions by Rodine, J.R.), in Brunsden, Denys, and Prior, D.B., eds., Slope instability: New York, John Wiley and Sons, Ltd., p. 257-361. Kellerhals, Rolf, and Bray, D.., 1971, Sampling procedures for coarse fluvial sediments: Proceedings of the American Society of Civil Engineers, Journal of Hydraulics Division, v. 97, no. HY8, p. 1165-1180. Kiersch, G. A., 1964, The Vaiont River disaster: Civil Engineering, v. 34, p. 32-39. Kiver, E.P. and Steele, W.K., 1972, Summit firn caves, Mount Rainier, Washington: Science, v. 173, p. 320-322. Kurdin, R.D., 1973, Classification of mudflows: Soviet Hydrology, no. 4, p. 310-316. Laenen, Antonius, Scott, K.M., Costa, J.E., and Orzol, LL., 1987, Hydrologic hazards along Squaw Creek from a hypothetical failure of the glacial moraine impounding Carver Lake near Sisters, Oregon: U.S. Geological Survey Open-File Report 87-41, 48 p. 1992, Modeling flood flows from a hypothetical failure of the glacial moraine impounding Carver Lake near Sisters, Ore- gon, in Subitzky, Seymour, ed., Selected papers in the hydro- logic sciences 1988-92: U.S. Geological Survey Water-Supply Paper 2340, p. 151-164. Latter, J.H., Houghton, B.F., and Hackett, WR., 1981, Volcanic risk assessment at Ruapehu volcano, chap. 4 of Houghton, B.F., and Smith, LE.M., eds., New Zealand Volcanological Workshop, Turangi, New Zealand, November 27-29, 1981, Handbook and proceedings: unpaginated. Linsley, RK., Jr., Kohler, M.A., and Paulhus, J.LH., 1958, Hydrology for engineers: New York, McGraw-Hill Book Company, Inc., 340 p. Lokey, W.M., Mack, Ross, Miller, M.M., Prather, B.W., and Kiver, E.P., 1972, Project Crater-Mt. Rainier glacio-volcanological research, 1970-72, in Arctic and Mountain Environments Symposium, Michigan State University, April 22-28, 1972, Program with Abstracts: East Lansing, Mich., Glaciological and Arctic Science Institute, Michigan State University [unpaginated]. Lowe, D.R., 1976, Grain flow and grain flow deposits: Journal of Sedimentary Petrology, v. 46, p. 188-199. y- - y --+~ REFERENCES CITED 55 1979, Sediment gravity flows-Their classification and some problems of application to natural flows and deposits: Society of Economic Paleontologists and Mineralogists, Special Publication 27, p. 75-82. 1982, Sediment gravity flows-II. Depositional models with special reference to the deposits of high-density turbidity currents: Journal of Sedimentary Petrology, v. 52, p. 279-297. Lowe, D.R., and LoPiccolo, R.D., 1974, The characteristics and origins of dish and pillar structures: Journal of Sedimentary Petrology, v. 44, p. 484-501. Luzier, J.E., 1969, Geology and ground-water resources of southwestern King County, Washington: Washington Department of Water Resources, Water-Supply Bulletin 28, 260 p. Major, J.J., 1984, Geologic and rheologic characteristics of the May 18, 1980, southwest flank lahars at Mount St. Helens, Washington: University Park, Pa., Pennsylvania State University, M.S. thesis, 225 p. Major, J.J., and Scott, K.M., 1988, Volcaniclastic sedimentation in the Lewis River valley, Mount St. Helens, Washington- Processes, extent, and hazards: U.S. Geological Survey Bulletin 1383-D, 38 p. Middleton, G.V., and Hampton, M.A., 1976, Subaqueous sediment transport and deposition by sediment gravity flows, in Stanley, D.J., and Swift, D.J.P., eds., Marine sediment transport and environmental management: New York, Wiley, p. 197-218. Mullineaux, D.R., 1974, Pumice and other pyroclastic deposits in Mount Rainier National Park, Washington: U.S. Geological Survey Bulletin 1326, 83 p. 1986, Summary of pre-1980 tephra-fall deposits erupted from Mount St. Helens, Washington State, U.S.A: Bulletin of Volcanology, v. 48, p. 17-26. National Research Council, 1988, Estimating probabilities of extreme floods-Methods and recommended research: Washington, National Academy Press, 141 p. Nelson, L.M., 1987, Flood characteristics for the Nisqually River and susceptibility of Sunshine Point and Longmire facilities to flooding in Mount Rainier National Park, Washington: U.S. Geological Survey Water-Resources Investigations Report 86-4179, 18 p. Nemec, W., and Steel, R.J., 1984, Alluvial and coastal conglomer- ates-Their significant features and some comments on gravelly mass-flow deposits, in Koster, EH. and Steel, RJ., eds., Sedimentology of gravels and conglomerates: Canadian Society of Petroleum Geologists Memoir 10, p. 1-31. Norris, RD., in press, Seismicity of rockfalls and avalanches at three Cascade Range volcanoes-Implications for seismic detection of hazardous mass movements: Seismological Society of America Bulletin. North American Commission on Stratigraphic Nomenclature, 1983, North American Stratigraphic Code: American Associ- ation of Petroleum Geologists Bulletin, v. 67, p. 841-875. Osterkamp, W.R., Hupp, C.R., and Blodgett, J.C., 1985, Magnitude and frequency of debris flows, and areas of hazard on Mount Shasta, northern California: U.S. Geological Survey Open-File Report 85-425, 35 p. Pacific Power and Light, 1980, Study of effects of potential volcanic activity on Lewis River Projects: Portland, Oreg., Board of Consultants Report, paginated sections. Available from Pacific Power and Light Co., 920 Southwest 6th Ave., Portland, OR 97204. Packwood History Committee, 1954, Packwood on the march: Unpublished compilation of early historical accounts, unpaginated. Available at Packwood Branch, Lewis County Public Library, Packwood, WA 98361. Pierson, T.C., 1980, Erosion and deposition by debris flows at Mount Thomas, North Canterbury, New Zealand: Earth Surface Processes, v. 5, p. 227-247. 1985, Initiation and flow behavior of the 1980 Pine Creek and Muddy River lahars, Mount St. Helens, Washington: Geological Society of America Bulletin, v. 96, p. 1056-1069. Pierson, T.C., and Costa, J.E., 1987, A rheologic classification of subaerial sediment-water flows, in Costa, J.E., and Wieczorek, G.F., eds., Debris flows/avalanches-Process, recognition, and mitigation: Reviews in Engineering Geology, v. 7, p. 1-12. Pierson, T.C., Janda, R.J., Thouret, J.-C., and Borrero, C.A., 1990, Perturbation and melting of snow and ice by the 13 November 1985 eruption of Nevado del Ruiz, Colombia, and consequent mobilization, flow and deposition of lahars: Journal of Volcanology and Geothermal Research, v. 41, p. 17-66. Porter, S.C., and Denton, G.H., 1967, Chronology of Neoglaciation in the North American Cordillera: American Journal of Science, v. 265, p. 177-210. Postma, George, 1986, Classification for sediment-gravity deposits based on flow conditions during sedimentation: Geology, v. 14, p. 291-294. Prych, EA., 1987, Flood-carrying capacities and changes in channels of the lower Puyallup, White, and Carbon Rivers in western Washington: U.S. Geological Survey Open-File Report 87-4129, 69 p. Qian Yiyang, Yang Wenhai, Zhao Wenlin, Cheng Xiuwen, Zhang Longrong, and Xu Wengui, 1980, Basic characteristics of flow with hyperconcentration of sediment, in Li Boning, chairperson, Proceedings of the International Symposium on River Sedimentation, March 24-29, 1980, Beijing, China: Beijing, Gaunghua Press, p. 175-184. Reich, B.M., 1973, How frequently will floods occur?: Water Resources Bulletin, v. 9, p. 187-188. Richardson, Donald, 1968, Glacier outburst floods in the Pacific Northwest, in Geological Survey Research 1968: U.S. Geological Survey Professional Paper 600-D, p. D79-D86. Rodolfo, K.S., 1989, Origin and early evolution of lahar channel at Mabinit, Mayon Volcano, Philippines: Geological Society of America Bulletin, v. 101, p. 414-426. Scheidegger, AE., 1973, On the prediction of the reach and velocity of catastrophic landslides: Rock Mechanics, Springer-Verlag, v. 5, p. 231-236. Schumm, S.A., 1960, The shape of alluvial channels in relation to sediment type: U.S. Geological Survey Professional Paper 352-B, 30 p. 1961, Effect of sediment characteristics on erosion and deposition in ephemeral-stream channels: U.S. Geological Survey Professional Paper 352-C, 70 p. Schuster, RL., and Chleborad, AF., 1989, Historic earthquake-induced landslides in western Washington: Geological Society of America Abstracts with Programs, v. 21, no. 5, p. 141. Scott, K.M., 1981, Erosion and sedimentation in the Kenai River, Alaska: U.S. Geological Survey Professional Paper 1235, 35 p. 56 SEDIMENTOLOGY, BEHAVIOR, AND HAZARDS OF DEBRIS FLOWS AT MOUNT RAINIER, WASHINGTON ----1988a, Origin, behavior, and sedimentology of catastrophic lahars at Mount St. Helens, Washington, in Clifton, H.E., ed., Sedimentologic consequences of convulsive geologic events: Geological Society of America Special Paper 229, p. 23-36. 1988b, Origins, behavior, and sedimentology of lahars and lahar-runout flows in the Toutle-Cowlitz River system: U.S. Geological Survey Professional Paper 1447-A, 76 p. 1989, Magnitude and frequency of lahars and lahar-runout flows in the Toutle-Cowlitz River system: U.S. Geological Survey Professional Paper 1447-B, 33 p. 1992, Risk analysis of sediment gravity flows at Cascade Range volcanoes-Approaches and analogies with alluvial fans, in Jones, ME., and Laenen, Antonius, eds., Interdiscipli- nary approaches in hydrology and hydrogeology: Minneapolis, Minn., American Institute of Hydrology, p. 222-230. Scott, KM., and Janda, R.J., 1987, Multiple origins of catastrophic lahars at Cascade Range volcanoes-Probabilities of precursor events, and hazards in downstream areas: Hawaii Symposium on How Volcanoes Work, Hawaiian Volcano Observatory, Hilo, Hawaii, January 19-25, 1987, Abstract Volume, p. 229. Scott, K.M., Laenen, Antonius, and Kresch, D.L., 1987, Hydrologic hazards downstream from volcanoes-Techniques of risk analysis, in Accurti, Dante, chairperson, Realistic approaches to better floodplain management: Association of State Flood- «plain Managers, 11th Annual Conference, Seattle, Wash., June 8-13, 1987, Proceedings; Natural Hazards Research and Appli- cations Information Center Special Publication 18, p. 286-291. Siebert, Lee, 1984, Large volcanic debris avalanches-Characteris- tics of source areas, deposits, and associated eruptions: Journal of Volcanology and Geothermal Research, v. 22, p. 163-197. Siebert, Lee, Glicken, Harry, and Ui, Tadahide, 1987, Volcanic hazards from Bezymianny- and Bandai-type eruptions: Bulle- tin of Volcanology, v. 49, p. 435-459. Sigafoos, R.S., and Hendricks, E.L., 1961, Botanical evidence of the modern history of Nisqually Glacier, Washington: U.S. Geological Survey Professional Paper 387-A, 20 p. 1972, Recent activity of glaciers of Mount Rainier, Wash- ington: U.S. Geological Survey Professional Paper 387-B, 24 p. Smith, G.A., and Fritz, WJ., 1989, Penrose Conference report-Volcanic influences on sedimentation: Geology, p. 375-376. Southard, J.B., and Mackintosh, ME., 1981, Experimental test of autosuspension: Earth Surface Processes and Landforms, v. 6, p. 103-111. Stuiver, Minze, and Becker, Bernd, 1986, High-precision decadal calibration of the radiocarbon time scale, AD 1950-2500 BC: Radiocarbon, v. 28, p. 863-910. Summers, C.G., 1978, Go to the Cowlitz, Peter Crawford: Longview, Wash., Speedy-Litho Press, 324 p. * U.S. GOVERNMENT PRINTING OFFICE: 1995- 673-046/86096 Unger, J.D., and Decker, R.W., 1970, The microearthquake activity of Mt. Rainier, Washington: Seismological Society of America Bulletin, v. 60, p. 2033-2035. Unger, J.D., and Mills, K.F., 1972, Microearthquakes at Mt. Rainier-1969: Seismological Society of America Bulletin, v. 62, p. 1079-1081. U.S. Army Corps of Engineers, 1985, Toutle River sediment- retention structure-Mudflow design criteria: Portland, Oreg., Portland District Letter Report, 7 p. Vames, D.J., 1978, Slope movement types and processes, in Schuster, RL., and Krizek, R.J., eds., Landslides-Analysis and control: National Research Council, Transportation Research Board, Special Report 176, p. 11-35. Voight, Barry, Glicken, Harry, Janda, R.J., and Douglas, P.M., 1981, Catastrophic rockslide avalanche of May 18, in Lipman, P.W., and Mullineaux, D.R., eds., The 1980 eruptions of Mount St. Helens, Washington: U.S. Geological Survey Pro- fessional Paper 1250, p. 347-377. Walder, J.S., and Driedger, C.L., 1993, Glacier-generated debris flows at Mount Rainier: U.S. Geological Survey Open-File Report 93-124 (Volcano Hazards Fact Sheet), 2 p. 1994, Geomorphic change caused by outburst floods and debris flows at Mount Rainier, Washington, with emphasis on Tahoma Creek valley: U.S. Geological Survey Water Resources Investigations Report 93-4093, 93 p. Webb, R.H., Pringle, P.T., and Rink, G.R., 1989, Debris flows from tributaries of the Colorado River, Grand Canyon National Park, Arizona: U.S. Geological Survey Professional Paper 1492, 39 p. Wentworth, C.M., 1967, Dish structure, a primary sedimentary structure in coarse turbidites [abs.]: American Association of Petroleum Geologists Bulletin, v. 51, p. 485. Wolman, M.G., 1954, A method of sampling coarse river-bed material: American Geophysical Union Transactions, v. 35, p. 951-956. Yamaguchi, D.K., 1983, New tree-ring dates for recent eruptions of Mount St. Helens: Quaternary Research, v. 20, p. 246-250. 1985, Tree-ring evidence for a two-year interval between recent prehistoric explosive eruptions of Mount St. Helens: Geology, v. 13, p. 554-557. Yano, Katsumasa, and Daido, Atsuyuki, 1965, Fundamental study on mud-flow: Bulletin of the Disaster Prevention Institute (Japan), v. 14, p. 69-83. Yesenov, U. Ye., and Degovets, A.S., 1979, Catastrophic mudflow on the Bol'shaya Almatinka River in 1977: Soviet Hydrology, Selected Papers, v. 18, p. 158-160. PROFESSIONAL PAPER 1547 RZ U.S. DEPARTMENT OF THE INTERIOR 75 U.S. GEOLOGICAL SURVEY PEATE I| [9k , U. )S 4 OSCEOLA MUDFLOW ELECTRON MUDFLOW NATIONAL LAHAR (and runout phases) TAHOMA LAHAR Phi # log! 1,600 T T T T T T T s T T T T T T T T T T T 1,000 | | | 1,350 F ay 0 c)., § a a a - - at f 4 1,500 |- (from Crandell, 1971) - 8 u u u“, (p= % ~ . 5 peal ~ F ~ Sa we t EecvonMudtiawm __ __ _ ._ __ _"" 3 3 a oS 17 meters ""% fe & sso |- oom brk san tom. ante man ame. Aad Soe BONO IS NETL ie wee Bo wee met mee ons S & % Tr oe sn am as am us ap 1450 |- € 5 foi @ aso |- - @ 1300 |- i C < J: E National Lahar 5 LJ 22. § ¢ } c 0 c l ys sul oe oN Z ¢ ays s 0 ae , aloe [t ty a., ive o s Ge (111 ~ CTT " rT r t tot l [- Sl of f- 5 a -> 3 g 1,400 g 500 § s Z 2 5 [- § A 925 (- # [) 125 - E ul 1,350 |- & a 3 3 1 2 S £ < 450 1 1 1 | | 1 | | | 1 1 E _____ Yd E c i 0 400 800 1,200 1,600 2,000 2,400 Inferred pre-National -1 2 ' DISTANCE, IN METERS $174 hens Gol s £ < 1,300|- = f 900 | | 1 | 1,250 1 | | | § I SECTION E-1 0 200 400 600 0 200 400 600 te f y | j | f . (27.6 kilometers downstream) DISTANCE, IN METERS DISTANCE, IN METERS x. 0 400 800 1,200 1,600 ; SECTION N-1 SECTION T-1 DISTANCE, IN METERS 450 ; ; ; ; ; (13.2 kilometers downstream) (9.0 kilometers downstream) SECTION O-1 Us f (10.4 kllometer§ downstream) E E (é) 575 I | | | [ | | [ 1,250 | | [ | [ f- led a g . . gy -=" § FC a W -> 950 | | T T T T | T T 2 % 4 pS 2 National Lahar heat o f e a me aeon de ie ae anl te os.. mk ive ant, am n em maen eam mane cael mie. ne limed Taak! pai mand she monn Ton! mnt agus L Landslide younger-*y = u fe- ; Fak LI 350 | than Electron Mudflow * *! Lg $ 550 Nllsgrearlly ________ g 1,225 |- Lg _ Ashomauahs: _ __ . aon .._ p 900 |- ~I i= 8 f sa s.. ¢ ~ J fT -t ~~ ~-Inferred pre-National Lahar profile “>-' 5 ta E had E 8 Measured thickness of deposits is 1-2 meters 8 Ci ~> "3 < u§J 4 g50 |- "Maximum lahar" = < | | | | I | | | | y { i | | | | | 525 W 1,200 |- atk 2g > ~~ Mudtion A tao oo as ag 00 yoo T 200 0 200 400 600 800 1,000 © ! - f Pe é © sol- a a DISTANCE, IN METERS DISTANCE, IN METERS g 0 3 a I> . f SECTION N-2 =» E m lnffi/fifilgfi 5231018 $ flaundaltiopAlea/fil SECTION E-2 (30.0 kilometers downstream) = Z < } we So firon'msccigsd ”U 193V1VL (36.4 kilometers downstream) ' 8 1,175 |- =I 150 |- Bedrock exposed nearby at this level & 2 300 T T T T T T T p00 I I I F Aver f ~ "" o Er preistionsl (32.8 kilometers downstream) C 2 C= 8 450 |- Lahar profile RM 1125 If Inferred pre-Electron --__X 9 < Measured thickness of * | | | Mudflow profile X= «-J aas T I T T T i O ~-===2z n ~~ bad f E “<3 650 |- Inferred thickness in well ~~ ~ / «-I ( 200 | | | g a is about 7 meters \7\_, 5 d 3 Inferred pre-Osceola Mudflow profile i: > C 44 - 150 |- =I u Tahoma 600 1 | 1 L I | | I 1 1 1 a -A p a 1,050 |- Crick =I 0 400 800 1,200 1,600 2,000 2,400 2G _| . m-ccar mur tt - fe _ 5 ~ 0 Puyallup t & p DISTANCE, IN METERS 4 us fhe 7 5G, ~~ « F Pd Tn ai gn aa e ha e ea w* .- < _ SECTION O-3 . E Q 100 __ Inferred pre-Electron ~ T z « (39.4 kilometers downstream) E % Mudflow profile u_I &, s Nevanal Later. " 1,025 | | I -, 8 u, ree NR 0 200 400 20 1s0|- pule pte -I DISTANCE, IN METERS 650 [ 1 ; ; [ | ; ; ; | ; ; I 50 1 1 1 1 1 - 9 Inferred pre-National | 4 0 1,200 3 < Lahar profile 3 SECTION T-3 0 00 80 o I S & (11.5 kilometers downstream) P I i L aa tou onl sut Ou | O y inundetion levello! Oscaoin __. uy Su uel ane cence ms DISTANCE, IN METERS §~§ m i 600 |- al SECTION E-4 $35 | | ( = | LE” Ly Whflimlglfiwiguidigol Ii/eLofjliz—igh wiegfwffiigeglayggffiwm ys (49.5 kilometers downstream) 0 200 400 600 850 | | | | | | 2 $ Greenwater Greenwater 2% =-- —\ pax i DISTANCE, IN METERS w" | "f .% o rrr n meal. 2 "y 3 150 SECTION N-4 w S "us. Z T T T T .| E fo 500 - w u 825 = T Well logs at Greenwater; 0-6 meters of lahars on Osceola Mudflow; (5 Tr 3 k . a megaclasts at 9-12 meters; alluvium at 15-18 meters 5 to 190 |- §§ a P, 15 T | I o -t Tahoma Lahar 450 1 1 1 1 1 1 1 1 1 U 1 1 1 a -A a Electron Mudflow CC _j g w' "|- PCT TGTTGT rT TT ~~ Tc uTTt 3 0 400 800 1,200 1,600 2,000 2,400 2,800 Z L4” wwwwwwwwwwwww ( rie eal aman ia e. sn in ales E ‘->“ 3 c>> Megaclasts protruding from g W . | 00 Se wee cam wns seer ene mate mes mon md pre mnt ser he joan fm Soe s wen soe rw owe rans mom e maw mae mw mass San mn n has an aon (e Ie f= DISTANCE, IN METERS § \~-——————Z ——————————————————— u & C E§ sw : hsurface of lahar 4 SECTION O-4 E 8 50 |- Inferred pre-Electron Mudflow profile - £ L. < P al |_ Pl P Aan a ae -_. a National Laher, _ Z. |_. _. __. EQ |. ..' . >mmecrclc~ 1175 I I 1 | 1 I I 650 ; ; ; ; ; ; ; 0 1 | | 1 1 | 1 I | 5 < Inferred pre-National Lahar profile-" ___ ~~~---==-=-- 0 200 400 600 800 0 400 800 1,200 1,600 2,000 # 25) . ml) f . a n 5% DISTANCE, IN METERS 2 ¥ s , f DISTANCE, IN METERS SECTION T-4 p a 600 |- Maépggag‘lgnaitgf’nge' = SECTION E-5 DISTANCE, IN METERS (15.4 kilometers downstream) g «/. ' ~!* -~ (53.4 kilometers downstream) SECTION N-5 a> of (105.4 kilometers downstream) H w o L CC _j 2 3 68 l— R@ so0|- - > f sod TI \\ o DISTANCE, IN METERS Puysliup ® 5 < e Greenwater < | | | SECTION 0-5 s E 700 1 (53.2 kilometers downstream) McMillin @ Mud Mountain Dam CROSS SECTIONS 0 200 400 GOO O-1 to 0-6 OSCEOLA MUDFLOW - Maximum lahar (largest lahar in post- DISTANCE, IN METERS t glacial history of Mount Rainier); regarded as a statistical SECTION T-5 / I T I I I I I I I I I I I I outlier in a group of large cohesive lahars. Downstream (17.8 kilometers downstream) ak inondetoniebe) ct 0 is Magn distances of cross sections are from summit along White River ane ae as on mal me re hey be tep ai ho on on mane mine tas an haf all am del eine mp all ;.. PINE - rs ~> . valley _________ -- 47°00 - 5 f amma e e ‘ E-1 to E-5 ELECTRON MUDFLOW --- Planning or Design Case I (selected 7 a Thickness of Osceola Mudflow is about 4 meters A as the most typical large cohesive lahar); useful for estimating >. -). 400- 5 a % potential future inundation of Puget Sound lowland or 2 Y > Cowlitz River flood plain, a flow that could affect downstream uf (LL) 8 reservoirs g > ood |E S = & 8 | f N-1 to N-5 NATIONAL LAHAR (and runout phases) --- Planning or , < tro " - J# Design Case II (selected as most typical noncohesive lahar); s 300 |- - Inferred pre-Osceola Mudflow profile = A MQWLBAINIER smaller but more frequent than above flows. Downstream a distances of cross sections are from summit along Nisqually River valley 250 1 1 1 1 1 1 1 3 1 | 1 1 1 1 | 0 400 800 1,200 1,600 2,000 2,400 2,800 3,200 s j e T-1 to T-5 TAHOMA LAHAR -- Planning or Design Case III (selected as the ational * t hazardous of the most frequent flows); flow represents a DISTANCE, IN METERS es k , P 46°45" |- wE-FERS * rapid debris avalanche transformed to a lahar; mainly a hazard _ SECTION 0-6 o 5 10 MILES within Mount Rainier National Park. Downstream distances of (71.2 kilometers downstream) | 1 cross sections are from summit along Tahoma Creek valley CROSS SECTIONS OF SELECTED FLOWS AT MOUNT RAINIER, WASHINGTON By K. M. Scott, J. W. Vallance, and P. T. Pringle 1995 je _ Hydrogeology of Jurassic and Triassic Wetlands - 75 __ in the Colorado Plateau and the Origin of i545 - Tabular Sandstone Uranium Deposits + U.S.}QEOLOGICAL sURVEY PROFESSIONAL PAPER 1548 Vy. S. EEPOSITORY AVAILABILITY OF BOOKS AND MAPS OF THE U.S. GEOLOGICAL SURVEY Instructions on ordering publications of the U.S. Geological Survey, along with prices of the last offerings, are given in the current-year issues of the monthly catalog "New Publications of the U.S. Geological Survey." Prices of available U.S. Geological Survey publications re- leased prior to the current year are listed in the most recent annual "Price and Availability List." Publications that may be listed in various U.S. Geological Survey catalogs (see back inside cover) but not listed in the most recent annual "Price and Availability List" may no longer be available. Reports released through the NTIS may be obtained by writing to the National Technical Information Service, U.S. Department of Commerce, Springfield, VA 22161; please include NTIS report number with inquiry. Order U.S. Geological Survey publications by mail or over the counter from the offices listed below. BY MAIL Books Professional Papers, Bulletins, Water-Supply Papers, Tech- niques of Water-Resources Investigations, Circulars, publications of general interest (such as leaflets, pamphlets, booklets), single copies of Earthquakes & Volcanoes, Preliminary Determination of Epicenters, and some miscellaneous reports, including some of the foregoing series that have gone out of print at the Superintendent of Documents, are obtainable by mail from U.S. Geological Survey, Map Distribution Box 25286, MS 306, Federal Center Denver, CO 80225 Subscriptions to periodicals (Earthquakes & Volcanoes and Preliminary Determination of Epicenters) can be obtained ONLY from the Superintendent of Documents Government Printing Office Washington, DC 20402 (Check or money order must be payable to Superintendent of Documents.) Maps For maps, address mail orders to U. S. Geological Survey, Map Distribution Box 25286, Bldg. 810, Federal Center Denver, CO 80225 Residents of Alaska may order maps from U.S. Geological Survey, Earth Science Information Center 101 Twelfth Ave., Box 12 Fairbanks, AK 99701 OVER THE COUNTER Books and Maps Books and maps of the U.S. Geological Survey are available over the counter at the following U.S. Geological Survey offices, all of which are authorized agents of the Superintendent of Docu- ments. +_ ANCHORAGE, Alaska-Rm. 101, 4230 University Dr. * _ LAKEWOOD, Colorado-Federal Center, Bldg. 810 +_ MENLO PARK, California-Bldg. 3, Rm. 3128, 345 Middlefield Rd. * _ RESTON, Virginia-USGS National Center, Rm. 1C402, 12201 Sunrise Valley Dr. * SALT LAKE CITY, Utah-Federal Bldg., Rm. $105, 125 South State St. * _ SPOKANE, Washington-U.S. Post Office Bldg., Rm. 135, West 904 Riverside Ave. * _ WASHINGTON, D.C.-Main Interior Bldg., Rm. 2650, 18th and C Sts., NW. Maps Only Maps may be purchased over the counter at the following U.S. Geological Survey offices: U FAIRBANKS, Alaska-New Federal Bldg, 101 Twelfth Ave. ROLLA, Missouri-1400 Independence Rd. STENNIS SPACE CENTER, Mississippi-Bldg. 3101 Hydrogeology of Jurassic and Triassic Wetlands in the Colorado Plateau and the Origin of Tabular Sandstone Uranium Deposits By Richard F. Sanford U.S. GEOLOGICAL SURVEY PROFESSIONAL PAPER 1548 UNITED STATES GOVERNMENT PRINTING OFFICE, WASHINGTON : 1994 U.S. DEPARTMENT OF THE INTERIOR BRUCE BABBITT, Secretary U.S. GEOLOGICAL SURVEY Gordon P. Eaton, Director Published in the Central Region, Denver, Colorado Manuscript approved for publication March 10, 1994 Edited by Judith Stoeser Graphics prepared by R.F. Sanford and Carol A. Quesenberry Photocomposition by Carol A. Quesenberry For Sale by U.S. Geological Survey, Map Distribution Box 25286, MS 306, Federal Center Denver, CO 80225 Any use of trade, product, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government Library of Congress Cataloging-in-Publication Data Sanford, Richard F. Hydrogeology of Jurassic and Triassic wetlands in the Colorado Plateau and the origin of tabular sandstone uranium deposits / by Richard F. Sanford. p- cm.-(U.S. Geological Survey professional paper ; 1548) Includes bibliographical references. Supt. of Docs. no. : I 19.16: 1548 1. Paleohydrology-Colorado Plateau. 2. Geology, Stratigraphic-Jurassic. 3. Geology, Stratigraphic-Triassic. 4. Geology-Colorado Plateau. I. Title. II. Series QE39.5.P27826 1995 553.4'932'097913-dc20 94-15974 CIP CONTENTS ADSHTACE. ..... eave revers seres erve erver errr erin errr er errr reer ener eee eres er nene reer errr nere errr reer nner ene rene 1 e 1 Description of Tabular Sandstone UTANIUM DePOSitS eee eee reer errr ere neenee eee es 1 Hydrogeological Principles US@d iN eee eee verre errr rere ener neenee neenee nees 2 CONtrOIS ON GFOUNAG-WAtEL FIOW revere rer errr errr reer neer errr ener ense ener nene renee 2 ae 2 ecus e 2 l tim eae 3 e 3 Features Of GOUAU-WAtEL FIOW severe rer vere vere er erin errr errr renee ener ener ener ees 4 Scale of GrOUund-WAter-FIOW SYStEMS reer errr errr errr errr reer errr eee 4 COMPOSitiON@l seee eee ee eevee rarer errr errr rer naren rer rere rene rene nene 4 Recharge and Discharge in Fluvial-Lacustrine ENViroAMENtS ...... 4 Transient Effects and Water-TAbI@ VATIAtONS rere rer errr rere reer eee} 4 Paleohydrogeol0gy in the COIOF@QO PIAtEAU eee raver rere reer rere rere rrr renee enne enn ees 5 Major Controls On Late JUASSi¢ WAtET eee evere errr errr errr rere rere renee eines 5 ae 5 ea ae 5 COMPOSitONM errr revere rere errr errr ese rr seer eres errr ere renee errr nere errr ener e renee rene 7 Salt Wash Sandstone Member of the MOrriSON FOMMAtION erea 7 Recapture Member Of the MOITISON eee reer errr rere renee errr rere eres 11 Westwater Canyon Member of the FOFM@tION eee ermm rere eee} 13 Brushy Basin Member of the MOITiSON eee ere rere rere rere renee rere 15 Jackpile Sandstone Member of the MOrTiSON eee} 17 Lower Part Of the CRhiAIG@ eee revere revere reaver errr ever renee reer ener ees 19 Topographic Controls ON GrOUNU-WAtET FIOW eee errr errr errr reer eee ree 23 Geologic Controls ON GrOUNU-WAt@L FIOW evere revere rer rere rere rer ern nere} 23 Diagenesis @Ad UTANIUM DEPOSitION ere evere verre errr errr renner nene anes 25 Origin of Tabular S@NdStONE UTANIUM DEPOSItS eee ever ever rer rere rre errr errr renee ene} 26 Nfo Graas in ae 26 ONE OF TWO SOIUtIONS? averse revere vere errr errr rere rere r errr renee renner rere rene nen en en ees 27 Seepage, COMPACtION, @AQ D@NSItY eevee rere rere reer ere nere rere renner neer ere 27 Transient, Depression-Focused Ground-Water eee ere renee eee 29 onine ra de ull emia hs to oe 31 RefeLEACE@S eave evere rer errr errr rere rere rere serene nner reer errr e rere nere renee rere nere renee rene ne nes 34 FIGURES Diagram showing classification of topographic, geologic, and compositional controls on ground water showing conditions faVOr@bIG fOf ere errr rre errr err rere rere renee rere errr ener en enne renner ene nene nere ener ns 3 Map showing major geologic features that influenced ground-water flow in the Late Jurassic in the (elo etal a Ee sso aes 6 Simplified map and cross section of the Colorado Plateau for Late Jurassic time at end of deposition of the Salt Wash Sandstone Member Of the MOITiSON errr rarer ere rer er errr errr rere rere rere en enne nenas 8 Generalized map showing source rocks for solutes in ground water in the Colorado Plateau region ........................ 9 Schematic cross section showing influences on ground-water flow in the Colorado Plateau region during the Late Jurassic at the end of deposition of the Salt Wash Sandstone Member of the Morrison Formation.................... 10 II IV 10. 11. 12. 13. 14. 15. 16. 17. 18. CONTENTS Map showing areas of alteration in mudstone underlying uranium-bearing sandstone and location of uranium deposits in the northern end Of the UTAVAN MiNETAl ress rvs reer errr reese reece eee 11 Simplified map and cross section of the San Juan Basin for Late Jurassic time at the end of deposition of the Recapture MEMber Of the MOFTiSON FOTMAtION rere rere rere rrr rere rre 12 Map showing contours of sandstone to mudstone thickness ratio in the Westwater Canyon Member of the MOITiSON FOTMAtON erve esses seres esr esses esr eer err err enes 14 Simplified map and cross section of the San Juan Basin for Late Jurassic time at the end of deposition of the Westwater Canyon Member Of the MOITiSON FOTMAtION revers rere rere rr errr rere rre 16 Detailed schematic cross section of the Westwater Canyon Member of the Morrison Formation and underlying units for Late Jurassic time at the end of WestwWater CANYON G@POSItION rere rere 18 Map showing contours of the interval of total ilmenite-magnetite dissolution in the upper part of the Westwater Canyon Member Of the MOFTiSON FOTMAtION verre verse rer ere rr errr rer rere rer esse ress errr ne rere rere nees 19 Maps showing interpretations of the diagenetic patterns in the Brushy Basin Member of the MOITISON FOFM@tON eres eres eres eres er sere eres rere errr errr ere nes 20 Schematic map of Colorado Plateau area showing diagenetic alteration in Brushy Basin Member of the Morrison Formation, zone of mixed local and regional discharge at distal edge of alluvial plain, and location of tabular sandstone uranium deposit Clusters in the MOFTiSON FOTMAtON rere revere rere rere rer rere reeves 21 Detailed schematic cross section of Brushy Basin Member and underlying units for Late Jurassic time at the end of Brushy Basin deposition in the COIOFAGQO PIAtEAU ATCA seee reese rere reese rere rer rere rere rere} 22 Isopach map of the thickness of the Jackpile Sandstone Member of the Morrison Formation showing facies of underlying Brushy Basin Member, location of uranium deposits, and directions of 23 Map showing relationship between uranium deposit clusters and ground-water discharge in the Colorado Plateau during deposition of the lower part of the Triassic Chinle FOFMAtION seee eee reer erea reeks 24 Map showing relationship between uranium deposits, syndepositional synclines, and organic-rich lacustrine mudstone in part of the Chinle FOTMAtiON iN SOUtREASt UIA reer rere rrr rere rere rer er es 25 Schematic cross section across a topographic depression occupied by a channel or lake showing reI@ti0NShip$ Of RYUTOIOGY AMQ sess sree eres rere rere rer ere 30 Hydrogeology of Jurassic and Triassic Wetlands in the Colorado Plateau and the Origin of Tabular Sandstone Uranium Deposits By Richard F. Sanford ABSTRACT During parts of the Jurassic and Triassic Periods, fluvial- lacustrine sediments were deposited in the area of the Colo- rado Plateau, and tabular sandstone (tabular-type) uranium deposits formed at a density-stratified ground-water interface in areas of regional ground-water discharge. The typical effects of topographic, geologic, and compositional controls on ground-water flow, together with modern ground-water analogs, can be used to reconstruct ground-water flow during and shortly after sedimentation. Diagenetic effects of the pas- sage of ground water can also provide further constraints. In the Upper Jurassic Morrison Formation and lower part of the Upper Triassic Chinle Formation, tabular sandstone uranium deposits are in lenticular, arkosic, fluvial, channel sandstone overlain by tuffaceous overbank and lacustrine mudstone and underlain by marine rocks, especially evaporites. The thin, subhorizontal, tabular-shaped bodies float in the sandstone with no apparent lithologic control. They are in reduced sand- stone within dominantly redbed sequences of sandstone and mudstone. The deposits favor transitional facies of interbed- ded sandstone and mudstone between dominantly fluvial sandstone and dominantly overbank and lacustrine mudstone. They tend to be concentrated in syndepositional synclines. Organic matter, either as detrital coalified plant fragments or structureless impregnations, is closely associated with ura- nium ore. Decreases in paleotopographic slope are indicated by lithofacies changes, typically from distal alluvial plain to mudflat or floodplain. Geologic controls focused discharge in zones of abrupt thinning and pinching out of Lower Jurassic and upper Paleozoic aquifer systems. Tabular sandstone ura- nium deposits formed where topographic controls favored mixed local and regional discharge. Transient, depression- focused recharge of humic-acid-bearing ground water at wet- lands in paleotopographic depressions may have provided a mechanism for downward and downdip transport of humic acid that precipitated at a subhorizontal, density-stabilized interface between relatively fresh, shallow ground water and discharging, saline regional flow. Uranium was precipitated during and after humate precipitation. INTRODUCTION Previous studies suggest that tabular sandstone (tabular- type) uranium deposits formed in areas of ground-water dis- charge (Sanford, 1982, 1988, 1990a, b, 1992), and quantita- tive models for selected areas support the hypothesis (Sanford, 1982, 1990a, 1994). In this study, I reconstruct the hydrogeology of the Colorado Plateau during parts of the Jurassic and Triassic Periods when fluvial-lacustrine sedi- ments and tabular bodies of uranium ore were being depos- ited. I examine additional evidence pertaining to ground- water flow, extend the earlier studies to other areas of ura- nium deposits on the Colorado Pl&ceau, and reconcile evi- dence for ground-water paleodischarge and evidence of ground-water paleorecharge at the site of uranium deposition. Hydrologic principles, observation of modern ground- water-flow systems, and results of quantitative modeling help predict the flow of ancient ground water. Hydrologic controls are then combined with stratigraphic, sedimento- logic, paleoenvironmental, and structural data from the Col- orado Plateau to construct conceptual models. The effects of the passage of ground water based on diagenetic evidence further constrain the model. Previous hydrogeologic work on the Colorado Plateau typically focused on surface water or shallow ground water but neglected the large-scale flow of ground water in the basin as a whole. Paleotopography was not fully utilized in paleohydrogeologic reconstruction. In this paper, I incorpo- rate concepts previously proposed to provide an integrated model for basin- and local-scale ground-water flow in the Colorado Plateau as a whole. DESCRIPTION OF TABULAR SANDSTONE URANIUM DEPOSITS Tabular sandstone uranium deposits are tabular, origi- nally subhorizontal bodies entirely within reduced fluvial sandstone of Late Silurian age or younger. Tabular sand- stone uranium deposits constitute the largest uranium 1 2 JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS resource type in the United States. They account for approx- imately 65 percent of the total United States production and reserves of uranium (Chenoweth and McLemore, 1989). Throughout this paper, the terms "reduced" and "oxidized" refer to iron unless otherwise noted. The general features of tabular sandstone uranium deposits are summarized elsewhere (Finch, 1967; Fischer, 1974; Adams and Saucier, 1981; Nash and others, 1981; Thamm and others, 1981; Finch and Davis, 1985; Granger and Finch, 1988; Shawe and others, 1991; Sanford, 1992). Typically, tabular sandstone uranium deposits are in lenticu- lar, arkosic, fluvial, channel sandstone overlain by tuf- faceous overbank and lacustrine mudstone and underlain by marine rocks, especially evaporites. The thin, subhorizontal, tabular-shaped bodies "float" in the sandstone with no appar- ent lithologic control. They are in reduced sandstone within dominantly redbed sequences of sandstone and mudstone. The deposits favor transitional facies of interbedded sand- stone and mudstone between dominantly fluvial sandstone and dominantly overbank and lacustrine mudstone. They tend to be concentrated in syndepositional synclines. Organic matter, either as detrital coalified plant fragments or structureless impregnations, is closely associated with ura- nium ore. The age of the host rocks ranges from Late Silurian to Tertiary, the older age limit corresponding to the first appearance of woody land plants. The deposits probably formed shortly after deposition of the host sandstone. In the Colorado Plateau, the largest tabular sandstone uranium deposits are in sandstone of the Upper Jurassic Morrison Formation and Upper Triassic Chinle Formation. HYDROGEOLOGICAL PRINCIPLES USED IN RECONSTRUCTION Here I consider both the controls that govern the flow of ground water and the effects of the passage of ground water. The controls are described by the general principles of phys- ical hydrogeology, and they lead to predictions that can be tested by looking for evidence of the effects of the passage of ground water. CONTROLS ON GROUND-WATER FLOW Controls on ground water flow are topographic, geo- logic, and compositional. Topographic controls are imposed by the configuration of the water table and, indirectly, the ground surface. Geologic controls are imposed by the distri- bution of permeability and thickness of units in the saturated zone. Compositional controls are imposed by masses of water having different density due to salinity. Darcy's Law provides a framework for classifying these controls (fig. 1) (Hitchon, 1969; Winter, 1988). The volumetric flow rate (Q) in saturated porous media is the product of three parameters: hydraulic conductivity (K), cross-sectional area normal to flow (A), and gradient in hydraulic potential (dh/d/) (Hub- bert, 1940, 1953, 1956). Although distinguishable conceptually, ground-water controls may be correlated. Finer sediments commonly are associated with both lower topographic slope and lower per- meability. Thinning of aquifers may be accompanied by a decrease in permeability owing to an increase in the propor- tion of fine-grained sediments. TOPOGRAPHY In mature, topographically uplifted basins, such as those of the Colorado Plateau, gravity-driven flow predomi- nates; recharge occurs at higher elevations, whereas dis- charge occurs at lower elevations (Kreitler, 1989). For an unconfined aquifer at the surface, a decrease in the slope of the water table results in a lower potential gradient and smaller flow rate (Q) through the downstream end of the medium, which causes discharge at the decrease in slope. The most common expression of this case is at decreases or breaks in slope, typically where an alluvial fan or plain meets an almost flat valley bottom (Freeze and Wither- spoon, 1967; McLean, 1970; Habermehl, 1980; Allison and Barnes, 1985; Duffy and Al-Hassan, 1988; Straw and oth- ers, 1990; Winter and Woo, 1990). A special case of a con- cave break in slope that focuses discharge is at the shoreline where the water table meets a standing body of water (see, for example, Pfannkuch and Winter, 1985; Cherkauer and Nader, 1989). Wetlands typically are present in areas of perennial discharge. In this study, I assume that paleotopographic highs were recharge areas, that paleotopographic depressions were discharge areas, and that discharge occurred at major decreases in slope due to a decrease in dh/d/ (fig. 1). Thus, the largest scale topographic controls on ground-water flow are indicated by the major highs in orogenic zones and the major depressions in sedimentary basins. Following the the- ory of Toth (1962, 1963), it is assumed that the deepest regional flow discharges at the lowest elevations and that shallower flow discharges farther upslope. The distal edge of the alluvial plain, or transition from alluvial plain to mudflat, is expected to be a zone of discharge and mixing of shallow local and deeper regional ground water (McLean, 1970; Duffy and Al-Hassan, 1988; Straw and others, 1990). GEOLOGY Both a facies change resulting in lower permeability and a thinning of hydrostratigraphic units in the downstream direction favor discharge of ground water and the presence of wetlands (fig. 1) (Freeze and Witherspoon, 1966, 1967; Garven and Freeze, 1984a, b; Bethke, 1989; Kreitler, 1989). The thinning may involve just one aquifer within a basin or HYDROGEOLOGICAL PRINCIPLES USED IN RECONSTRUCTION 3 GROUND-WATER FAVORABLE FOR DARCY'S LAW: Q = KA dh/dl CONTROLS DISCHARGE Decrease in: Topography Concave change in slope dh/dl Geology Decrease in transmissivity Permeability Decrease in permeability K Thickness Thinning of aquifer A Composition Thinning of fresh-water lens A Figure 1. - Classification of topographic, geologic, and compositional controls on ground water showing conditions favorable for discharge. Q is volumetric flow rate; K is hydraulic conductivity; A is cross-sectional area normal to flow; dh/d! is gradient of hydraulic potential (head) with distance. the basin as a whole. For the Colorado Plateau in Mesozoic time, geologic controls on ground-water flow are recon- structed from permeability data and stratigraphic evidence. AQUIFERS AND CONFINING UNITS Ground water in horizontally stratified sedimentary basins tends to flow laterally through aquifers and vertically through confining units (Hubbert, 1940; Toth, 1978; Kreitler, 1989). The notion that fine-grained or clay-rich units are impermeable continues to be expressed by some geologists, even though it was understood more than 90 years ago to be a "time-honored delusion" (Munn, 1909). Flux across con- fining units can be significant; the lower permeability (K, fig. 1) of a confining unit is compensated by the cross-sectional area (A), which is orders of magnitude greater when flow is perpendicular to bedding, and by a higher potential gradient (dh/d/) that may exist across the confining unit. For example, upward flux through the bed of Lake Frome, Australia, is 170-180 mm/yr (Allison and Barnes, 1985). COMPOSITION A gravitationally stable body of saline water may force fresher water flowing over it to discharge (fig. 1). (By defi- nition, saline water has 1,000-35,000 mg/L total dissolved solids, and brine has >35,000 mg/L total dissolved solids (Robinove and others, 1958).) Some surface water and much ground water is saline (Hanor, 1983; Hammer, 1986; Kreitler, 1989). Saline bodies of water, including salt lakes and the oceans, typically receive inflow from more dilute surface and ground water. Buoyancy forces cause less dense fresher water to overlie more dense salt water in the subsur- face. At equilibrium, the fresh water-salt water interface is horizontal. When the upper fluid is moving, the interface slopes upward in the direction of movement (Cooper and others, 1964). The salt-water wedge acts as a barrier to the flow of relatively fresh water, which tends to discharge at the surface more or less at a shoreline or other break in slope (McLean, 1970; Duffy and Al-Hassan, 1988; Dutton and others, 1989; Fee and others, 1992; Herczeg and others, 1992; Hines and others, 1992; Long and others, 1992; Macumber, 1992). In topographic depressions where the water table slopes inward and ground-water flow converges, the hydraulic potential for shallow ground water is least at the depression (Hubbert, 1940). The salt water-fresh water interface must therefore rise toward the depression, even if salt water fails to discharge at the surface. Compositional controls on ground-water flow cannot be reconstructed in ancient basins with as much certainty as topographic and geologic controls because the original fluids have disappeared. The chemical evolution of the ground water can be inferred, however, from the composition of the rocks through which the water passes. Because evaporite minerals dissolve rapidly, ground water that passes through evaporites or evaporite-cemented sediments becomes saline, as demonstrated by salinities as high as 439,000 ppm in mod- ern basins, including the Paradox Basin (Mayhew and Heyl- man, 1965; Hanor, 1983; Kreitler, 1989). The contribution of evaporites to saline ground water is well documented (Gal- laher and Price, 1966; Boswell and others, 1968; van Everd- ingen, 1971; Anderson and Kirkland, 1980; Johnson, 1981; Fisher and Kreitler, 1987; Banner and others, 1989; Dutton and others, 1989). Evaporites are not required, however, for the presence of brines; brines can evolve by interaction of rock and normal or evaporatively concentrated sea water originally trapped during sedimentation (Fisher and Kreitler, 1987; Dutton, 1987) or by reverse chemical osmosis (Brede- hoeft and others, 1963; Coplen and Hanshaw, 1973; Graf, 1982). Thus, in reconstructing ancient ground-water flow, the fact that ground water flowed through or around evapor- itic horizons is strong evidence for saline ground water, but flow through marginal-marine rocks deposited in an arid or sabkha environment also suggests saline water. A close 4 JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS modern analog to ancient saline water movement is the dis- solution of the Middle Jurassic Carmel Formation evapor- ites, transport of saline water downdip in the underlying Lower Jurassic Navajo Sandstone, and upward discharge from the Navajo (Taylor and Hood, 1988). FEATURES OF GROUND-WATER FLOW The features discussed here are commonly observed characteristics of ground-water flow in sedimentary basins analogous to ancient basins such as those of the Colorado Plateau. SCALE OF GROUND-WATER-FLOW SYSTEMS The features of ground-water-flow systems vary with the scale of the system, which is generally classified as regional, intermediate, or local (Toth, 1962, 1963). Regional flow recharges at major drainage divides and discharges at basin floors. Intermediate flow recharges and discharges in areas that are separated by one or more local topographic highs and lows. COMPOSITIONAL VARIATIONS Toth (1962, 1963) predicted that chemical contrasts would be greatest across the boundaries between flow sys- tems of different magnitude as a result of the different flow paths. In fact, numerous modern examples demonstrate that diverse types of ground water from local, intermediate, and regional flow systems converge in mixed discharge zones (Counts, 1957; Toth, 1963; Freeze and Witherspoon, 1966; Gallaher and Price, 1966; Boswell and others, 1968; van Everdingen, 1971; Foreman and Sharp, 1981; Mono Basin Ecosystem Study Committee, 1987; Sharp, 1988; Swanson and others, 1988; Banner and others, 1989; Dutton and oth- ers, 1989; Whittemore and others, 1989; Huff, 1990; Fee and others, 1992; Herczeg and others, 1992; Hines and oth- ers, 1992; Long and others, 1992; Macumber, 1992; Strobel, 1992). The local flow system typically discharges dilute water, whereas the intermediate and regional flow systems discharge saline water of diverse compositions, commonly within a relatively small area. For example, in the Cascade Range, north-central Oregon, two springs only meters apart vary from less than 100 to 8,000 mg/L total dissolved solids (Ingebritsen and others, in press). Both springs are in a regional discharge area, but the saline spring is fed by the regional flow, and the fresh-water spring is fed by local flow. Such compositional contrasts at the surface clearly indicate the presence of steep compositional gradients, or ground-water interfaces, in the subsurface. Such interfaces may be widespread in the shallow subsurface (Runnells, 1969) and are observed in modern coastal environments (Cooper and others, 1964), in geothermal areas (Williams and McKibben, 1989) in unconsolidated alluvium beneath large rivers (Counts, 1957; Gallaher and Price, 1966; Boswell and others, 1968; Foreman and Sharp, 1981; Sharp, 1988), and where rivers erode exposed evaporites (Warner and others, 1985). A variety of diagenetic reactions occur at such inter- faces, including precipitation of humate (Swanson and Pala- cas, 1965; Hair and Bassett, 1973; Sholkovitz, 1976; Ortiz and others, 1980; Fox, 1983; Thurman, 1985, p. 26ff and 394ff), precipitation of dolomite (Ward and Halley, 1985), precipitation of calcite (Mono Basin Ecosystem Study Com- mittee, 1987), dissolution of calcite (Plummer, 1975), and other reactions (Magaritz and Luzier, 1985; Randazzo and Bloom, 1985; Hines and others, 1992). In the Colorado Pla- teau, tabular layers of humate, uranium, and dolomite sug- gest former interfaces (Fischer, 1947, 1974; Shawe, 1955, 1962, 1966, 1976; Granger and others, 1961; Granger, 1968; Wood, 1968; Melvin, 1976; Sanford, 1982, 1990a, b, 1992; Granger and Santos, 1986; Northrop and Goldhaber, 1990; Wanty and others, 1990; Hansley and Spirakis, 1992). RECHARGE AND DISCHARGE IN FLUVIAL-LACUSTRINE ENVIRONMENTS Recharge and discharge in complex present-day lake- groundwater systems (Meyboom, 1967; Lissey, 1971; Stephenson, 1971; Winter, 1976, 1978, 1986, 1988) are likely analogs for such systems in the ancient Colorado Pla- teau. Lakes always are in topographic depressions, and the water table normally slopes toward the lake, causing dis- charge of ground water to the lake. Lakes at higher eleva- tion may recharge the ground-water system, but the lake at the lowest elevation typically only receives ground-water discharge and contributes no recharge (Winter, 1976). A flow-through lake may have discharge on the upgradient side and recharge on the downgradient side. Greater width of the lake relative to the thickness of the underlying aquifer favors increased focusing of ground-water discharge around the lake margins (Pfannkuch and Winter, 1984). At playa lakes, relatively fresh water discharges around the shoreline, and saline water recharges or discharges in the center (Friedman and others, 1982; Allison and Barnes, 1985; Spencer and others, 1985; Duffy and Al-Hassan, 1988; Winter and Woo, 1990). TRANSIENT EFFECTS AND WATER-TABLE VARIATIONS The water-table level fluctuates as a result of fluctua- tions in precipitation (Lissey, 1971; Winter, 1983). Three zones are distinguishable: (1) the permanently saturated zone beneath the steady-state position of the water table, (2) the transiently saturated zone above the steady-state water-table level but below the permanently unsaturated zone, and (3) the permanently unsaturated zone. Water-table fluctuations PALEOHYDROGEOLOGY IN THE COLORADO PLATEAU 5 leave traces in the authigenic mineralogy of the sediments. The most conspicuous and easily mappable feature in ancient sedimentary rocks is color variation due to the oxidation state of iron. Red, yellow, and buff indicate oxidized iron, whereas green, gray, and black indicate reduced iron. Oxidized and reduced sedimentary rocks have been related to paleo-water- table level in the Colorado Plateau and elsewhere (Walker, 1967; Reading, 1978, p. 48-49; Dodson and others, 1980; Huber, 1980; Dubiel, 1983, 1989; Davis, 1988; Ghiorse and Wilson, 1988; Dubiel and others, 1991). Water-table level has been related to redox conditions in analogous modern systems (Jackson and Paterson, 1982; Fee and others, 1992; Hines and others, 1992; Long and others, 1992; Macumber, 1992). In this study, I assume by analogy that most of the per- manently saturated zone tends to be reducing where organic matter was deposited, the transiently saturated zone and upper part of the permanently saturated zone have mottled or variegated sediments, and the permanently unsaturated zone has oxidized sediments. Transient, depression-focused recharge to the ground- water system may occur during high-water periods and may constitute a major source of shallow, relatively fresh ground water at topographic depressions (Gallaher and Price, 1966; Lissey, 1971; Winter, 1983; Wood and Petraitis, 1984; Fet- ter, 1988, p. 45ff; Logan and Rudolph, 1992). During periods of high water, topographic depressions fill with surface water that then migrates into unsaturated, porous sediments of the stream bank or lake shore. This newly infiltrated ground water is slowly released to the ground-water system. In the Southern High Plains of Texas, for example, most of the recharge is focused at such topographic depressions (Wood and Petraitis, 1984). In the coastal plain of Argentina, fresh water recharges in marshes that are only 0.5-1.0 m below the surrounding land surface and overlies regional saline ground water (Logan and Rudolph, 1992). This phe- nomenon provides an explanation for apparent downward flow of ground water in areas of regional discharge in the ancient Colorado Plateau, as discussed following. PALEOHYDROGEOLOGY IN THE COLORADO PLATEAU MAJOR CONTROLS ON LATE JURASSIC GROUND WATER TOPOGRAPHY The drainage divides marking the southern and western margins of the drainage basin (Mogollon and Elko high- lands, respectively, fig. 2) (Peterson, 1984, 1986, 1988a, in press; Thorman and others, 1990, 1991) during deposition of the Morrison Formation must have been beyond the present- day margins of the Colorado Plateau, perhaps in southern Arizona or northern Mexico (Billodeau, 1986) and Nevada. Paleocurrent directions of streams in the Morrison Forma- tion (Craig and others, 1955; Mullens and Freeman, 1957; Young, 1978; Dodson and others, 1980; Tyler and Ethridge, 1983; Peterson, 1984, 1986, in press; Turner-Peterson, 1986) indicate that the regional topographic slope, and probably the regional ground-water flow, was mainly northeastward. Intermediate and local flow systems were undoubtedly affected by tectonically active structures within the Colorado Plateau. Throughout deposition of the Morrison Formation, the Elko and Mogollon highlands probably were regional recharge areas (fig. 2). GEOLOGY The most important geologic controls on ground-water flow in the Late Jurassic were buried Precambrian blocks and Phanerozoic eolian sandstone and carbonate aquifers (fig. 3) (Jobin, 1962; Sanford, 1982, 1990a; Stone and oth- ers, 1983; Freethey and Cordy, 1991; Geldon, in press). Northeastward from the middle of the Colorado Plateau, in the direction of ground-water flow, sedimentary rocks beneath the Morrison Formation thin from as much as 5,000 m (15,000 ft) to a thin veneer over the Uncompahgre and San Luis blocks, remnants of the Pennsylvanian-Permian ances- tral Rocky Mountains, which formed a hydrologic barrier. Northeast of these blocks, sediments again thicken in the Eagle and Piceance Basins. The ancestral Front Range block formed a second barrier behind these basins. Ground water must have been forced around the ends and over the top of the barriers where it discharged. Similar flow of ground water around the Uncompahgre block takes place today in Devonian and Mississippian carbonate rocks (Taylor and Hood, 1988). Major aquifer systems in the Colorado Plateau are the lower Paleozoic, upper Paleozoic, and Lower Jurassic aqui- fer systems (fig. 3) (Jobin, 1962, 1986; Sanford, 1982, 1990a; Taylor and others, 1986; Freethey and Cordy, 1991; Geldon, in press). The lower Paleozoic aquifer system includes the karstic Mississippian Leadville Limestone and correlative Redwall Limestone. The upper Paleozoic aquifer system includes the eolian Cedar Mesa Sandstone Member of the Cutler Formation, Coconino Sandstone, De Chelly Sandstone, Glorieta Sandstone, Meseta Blanca Sandstone Member of the Yeso Formation, Weber Sandstone, and White Rim Sandstone Member of the Cutler Formation. The Cedar Mesa Sandstone Member is the most transmissive of these units, and the arkosic Cutler Formation, with which the eolian sandstone units intertongue, is only an aquifer locally (Geldon, in press). The Lower Jurassic aquifer system includes the eolian Lower Jurassic Wingate and Navajo Sandstones. In general, the eolian sandstone aquifers thin and the percentage and thickness of confining units thicken to the northeast. Of the eolian sandstones, the thickest parts of Lower Jurassic Navajo Sandstone were the most trans- missive and most variable hydrostratigraphic unit on the JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS 109° 37° -_ _ WYOMING 1 1 L_ - __ L_ COLORADO NEVADA UTAH - PICEANCE AND EAGLE BASINS UNCOMPAHGRE WATERPOCKET UPLIFT ELKO HIGHLANDS i MoGoLLonN HIGHLANDS £13;sz 10 Plateau «|Q Z 6 I o 0 50 100 MiLES N |a &1z 0 - 50 _ 100 KILOMETERS L] |2 I I 1 Figure 2. Map showing major geologic features, both active tectonic structures and topographic nonstructural features, that influ- enced ground-water flow in the Late Jurassic in the Colorado Plateau region. Shaded area is area of present-day Colorado Plateau. Outward-pointing arrows are anticlines, inward-pointing arrows are synclines, and dashed arrows indicate generalized direction of sur- face- and ground-water flow. Modified from Craig and others (1955), Mallory (1972), Peterson (1984, 1986, 1988a, in press), and Turner-Peterson (1986). PALEOHYDROGEOLOGY IN THE COLORADO PLATEAU 7 Colorado Plateau. The decrease in thickness of the Navajo Sandstone from more than 600 m (2,000 ft) to a feather edge across the Colorado Plateau (Jobin, 1962; Blakey and others, 1988) probably determined the major discharge zone during Late Jurassic time. The pinchout of the Navajo Sandstone, from 150 m (500 ft) to zero (Jobin, 1962; Blakey and others, 1988), forms an arcuate band with its main axis oriented north-south and bowed to the east (fig. 3). The underlying Wingate Sandstone and overlying Entrada Sandstone also tend to thin to the northeast; however, the change in their thicknesses is much less than that of the Navajo Sandstone (Jobin, 1962; Blakey and others, 1988). The upper Paleozoic aquifer system pinches out west of the Navajo pinchout in the Paradox Basin area but not in the San Juan Basin area, where the Navajo is absent. The highly transmissive Lead- ville and Redwall Limestones probably were important for regional flow but may have been too deeply buried under confining units to influence near-surface ground-water flow in Triassic and Jurassic time. Thickness variations suggest that northeast-flowing regional ground water generally flowed from the upper Pale- ozoic aquifer system upward into the base of the Lower Juras- sic aquifer system (fig. 3). This flow then discharged upward from the Lower Jurassic aquifer system where it thinned to the northeast. Because the Navajo Sandstone thins so dramat- ically, it probably was a major source of upward ground- water discharge. Thinning of the Wingate and Entrada Sand- stones, as well as thinning of the whole sedimentary section over the buried Uncompahgre block, contributed to discharge beyond the pinchout of the Navajo Sandstone. North and east of the Uncompahgre block, the eolian Middle and Upper Pennsylvanian and Lower Permian Weber Sandstone of the upper Paleozoic aquifer system favored recharge and focus- ing of ground water in a northerly direction around the Front Range block into central Wyoming. COMPOSITION Judging from the widespread distribution of evaporites and evaporitic clastic rocks throughout the Colorado Plateau (fig. 4), regional and intermediate ground water was proba- bly saline toward the distal ends of flow paths. The deepest regional flow penetrated Pennsylvanian and Permian evaporites. This ground water probably approached halite saturation and contained significant Na+, K*, Ca"+, CIT, and $042” from the dissolution of halite, sylvite, anhydrite, and gypsum. Analysis of modern ground water whose solutes came from these evaporites indicates that the ancient brines were probably Nat-CI~ type (Hanshaw and Hill, 1969; Thackston and others, 1981; Sanford, 1990a). Sulfur iso- topes from authigenic barite cement are consistent with the transport of Pennsylvanian sulfate up to the Upper Jurassic Morrison Formation (Breit and others, 1990). Ground water at intermediate depths flowed through Jurassic and Triassic evaporites (fig. 4). Gypsum in the Lower and Middle(?) Triassic Moenkopi Formation (Blakey, 1974), Middle Juras- sic Curtis Formation (Peterson, 1988a), Middle Jurassic Todilto Limestone Member of the Wanakah Formation (Ridgley, 1989), and Upper Jurassic Tidwell Member of the Morrison Formation (Peterson, 1988a) suggests that inter- mediate ground water below and in the lower part of the Morrison Formation was saline. Local ground water that remained in the uppermost layer of alluvium was probably a dilute Na+-Ca"+-HCO;- solution, judging from analysis of modern ground water in the Colorado Plateau and elsewhere (Hanshaw and Hill, 1969; Thackston and others, 1981; Davis, 1988; Sanford, 1990a). This relatively fresh water probably rested on the saline water formed by passage through Mesozoic and older evaporites and marine sediments. SALT WASH SANDSTONE MEMBER OF THE MORRISON FORMATION Owing mainly to local paleotopographic variations, the different members of the Morrison Formation exhibited dis- tinct differences in ground-water flow. The Salt Wash Sand- stone Member was deposited on a broad fan-shaped alluvial plain that included, from southwest to northeast, dominantly low sinuosity, sand-dominated stream deposits; high-sinu- osity, mud-dominated, floodplain deposits; and lacustrine mudstone-limestone deposits in the undifferentiated Morri- son Formation (fig. 3) (Craig and others, 1955, Mullens and Freeman, 1957; Young, 1978; Galloway, 1979; Dodson and others, 1980; LC. Craig, U.S. Geological Survey, unpub- lished data, 1982; Tyler and Ethridge, 1983; Peterson, 1984, 1986, in press; Peterson and Tyler, 1985; Peterson and Turner-Peterson, 1987). LC. Craig (unpublished data, 1982) estimated the slope of the alluvial plain at from 1:1,000 to 1:2,000. During deposition of the Salt Wash Sandstone Member, ground water probably discharged at wetlands in topo- graphic depressions in the Henry Basin, Uravan mineral belt, and Carrizo Mountains area (figs. 3, 5). A variety of evi- dence, including the transition from low- to high-sinuosity channels, decrease in thickness of sandstone, decrease in sandstone as a percent of total thickness, increase in mud- stone thickness and percentage, transition from sandstone- mudstone to claystone-limestone facies, and transition from fluvial sheet gravels to mudflat-lake muds (Craig and others, 1955; Mullens and Freeman, 1957; Peterson and Tyler, 1985), suggests a major decrease in topographic slope at the distal edge of the alluvial plain. The greatest concentration of uranium deposits in the Salt Wash Member is where stream deposits in the Salt Wash Member thin from 60 to 30 m (200-90 ft) (Craig and others, 1955) and where the Navajo Sandstone thins from 150 m (500 ft) to zero (fig. 3). JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS 109° UTAH ELKO HIGHLANDS (recharge) xt i- . ~& ~*~ Line of cross section MOGOLLON HIGHLANDS (recharge) % 1 1 O < I 9 51 5 Edge of Nn! z Colorado ml Plateau JF | Z 1 1 0 50 100 MILES I }_,_A_r_|__r_—l I 0 50 100 KILOMETERS Evaporites of the Carme! Formation SOUTHWEST N Evaporites of the Kaibab F tio EXPLANATION Major eolian and carbonate aquifers Evaporitic rocks Other rocks Evaporites of the Todilto Limestone Member Lower Paleozoic aquifer system FEET 3000 1500 0 HENRY BASIN PARADOX BASIN NORTHEAST METERS 1000 Evaporites of the 500 Paradox Formation 0 Sandstone aquifer Upper_ UNCOMPAHGRE P2132} ' PRECAMBRIAN Saline BLOCK system PALEOHYDROGEOLOGY IN THE COLORADO PLATEAU 9 EXPLANATION Salt Wash Sandstone Member Navajo Sandstone Precambrian block p Zone of mixed discharge hl Uranium deposit cluster Thickness of channel deposits in Salt Wash Member (in feet) ————— Thickness of Navajo Sandstone (in feet) ------------ Pinchout of Wingate Sandstone (east edge) wlc _> Edge of Precambrian block Figure 3 (above and facing page). - Simplified map and cross sec- tion of the Colorado Plateau for Late Jurassic time at end of deposi- tion of the Salt Wash Sandstone Member of the Morrison Formation. Thickness of stream deposits in the Salt Wash Member, total thick- ness of the underlying Lower Jurassic Navajo Sandstone, and loca- tion of buried Precambrian blocks indicate areas of ground-water discharge. Arrows on map and cross section show direction of ground-water flow. Areas in cross section delineated by boxes are shown in detail in figure 5. Dotted line in cross section is dilute-saline interface. The Lower Jurassic aquifer system is composed of the Wingate and Navajo Sandstone aquifers. Modified from Craig and others (1955), Baars (1962), Jobin (1962), Mallory (1972), Peterson (1984, 1986, 1988a, in press), and Blakey and others (1988). The Henry Basin was an area of decreased topographic slope as indicated by thicker sediments, more channel sand- stone, higher sinuosity channels, increased upper flow regime horizontal laminations, and lacustrine mudstone (Peterson, 1984, 1986). Topographic depressions within the Henry Basin are suggested by thick sediments, evaporite deposits, lacustrine deposits, repetition of facies, and coinci- dence of synclines with present-day synclines (F. Peterson, 1980, 1984). Tabular sandstone uranium-vanadium deposits are associated with topographic depressions, actively subsid- ing synclines, and reduced, carbon-bearing lake deposits, and they slope upward stratigraphically in the downstream flow direction of surface water (F. Peterson, 1980; Northrop and Goldhaber, 1990). The tabular ore layers and diagenetic alterations have been interpreted as suggesting two fluids with a mixing zone between them (Northrop and Goldhaber, 1990; Wanty and others, 1990). Although some of the spe- cific reactions proposed by Northrop and Goldhaber (1990) and Wanty and others (1990) are questionable (Spirakis, 1991; Hansley and Spirakis, 1992), the tabular nature of the uranium-vanadium bodies strongly suggests a density- stabilized interface. The tabular shape, upward slope, and association with paleotopographic depressions are consistent with a saline water interface that arches upward at the topo- graphic depression (fig. 5A). "~"!~ coLonabo ot n mons ___ _L ~" NEVA ..__.___ £ 4 37° : _‘ F Edge of Colorado Plateau ~ & 0 50 100 MILES 0 50 100 KILOMETERS EXPLANATION [j --------- Area and boundary of Jurassic source rocks -«~.«_~~ Area and boundary of Triassic source rocks ————— Area and boundary of Pennsylvanian and Permian source rocks 2 Figure 4. Generalized map showing source rocks for solutes in ground water in the Colorado Plateau region. Source rocks include halite-sylvite-bearing evaporites, anhydrite-gypsum evaporites, and clastic rocks containing anhydrite-gypsum beds. Pennsylvanian and Permian source rocks include Paradox Formation, Supai Formation, Cedar Mesa Sandstone Member of the Cutler Formation, San An- dres Limestone, and Yeso Formation. Triassic source rocks include Moenkopi Formation. Jurassic source rocks include Carmel Forma- tion, Todilto Limestone Member of the Wanakah Formation, and Tidwell Member of the Morrison Formation. Data from Baars (1962), Mallory (1972), Blakey (1974, 1980), J.A. Peterson (1980), F. Peterson (1984, 1988a), and Ridgley (1989). 10 SOUTHWEST JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS NORTHEAST Summerville Formation ’*-—4 ________ --A Saline regional flow Fresh-water lake _- i> Morrison Formation I~ me me e ee ee n Brackish-water lake (?) Saline regional flow . E,",Na\}ajo Sandsionei") o Wingate Sandstone © Precambrian >, Uncompahgre :, 11 /f. block (Cc,! Salt anticline EXPLANATION Aquifer ys and siltstone and siltstone Tabular uraniun-vanadium deposit Oxidized (red-yellow) mudstone Reduced (gray-green) mudstone 1 Anhydrite-gypsum rock < Steady-state water table —————— Flow system divide (interface) o Stagnation point Figure 5. - Schematic cross section showing influences on ground-water flow in the Colorado Plateau region during the Late Jurassic at the end of deposition of the Salt Wash Sandstone Member of the Morrison Formation (not to scale). Location of areas are delineated by boxes on cross section of figure 3. No vertical scale implied. Flow lines, flow-system divides, and stagnation points drawn according to the theory of Toth (1962), Hitchon (1969), Winter (1976, 1983, 1988), and others, as constrained by oxidized and reduced rocks suggestive of recharge and discharge. Principle sources of dissolved solids are evaporitic rocks. A, Henry Basin. B, Paradox Basin. In the Uravan mineral belt, tabular sandstone uranium deposits are associated with distributary channels, thick host sandstone, higher sandstone to mudstone ratio, lenticu- lar rather than flatbedded sandstone, scour-and-fill bedding, and overlying conglomeratic sandstone (Weir, 1952; Phoe- nix, 1958; Shawe, 1962; Motica, 1968; Thamm and others, 1981), all of which suggest topographic depressions. The deposits typically are within gray reduced sandstone that contains carbonized wood and overlies a green reduced mudstone bed (fig. 6) (McKay, 1955). As in the Henry Basin, the shape, upward slope, and associated reduced rocks are consistent with a saline water-fresh water inter- face inclined upward toward an area of perennial discharge (fig. 5B). An absence of uranium deposits in the Four Corners area between the Uravan mineral belt and the Carrizo Moun- tains deposits coincides with an inferred local recharge area (fig. 3). This area, characterized by an abundance of eolian sandstone (Bluff Sandstone Member of the Morrison Forma- tion) and a lack of channel sandstone, has been interpreted as a local topographic high (Peterson and Tyler, 1985; Blakey and others, 1988; Peterson, 1988b, in press). PALEOHYDROGEOLOGY IN THE COLORADO PLATEAU 11 South of the recharge area, in the Carrizo Mountains and vicinity (fig. 3), uranium deposits are in a relatively thick part of the Salt Wash Sandstone Member where the sedimentary structures and thickness data indicate an east- ward-trending, syndepositional structural depression (Hilp- ert, 1969). This area also is characterized by an abrupt decrease in the sandstone to mudstone ratio and by abrupt changes in rock color (Chenoweth and Malan, 1973). Modeling of regional ground-water flow in the Colo- rado Plateau during Late Jurassic-Early Cretaceous time indicates that local dilute ground water and regional saline fluids converged at the topographically and geologically controlled discharge zones (Sanford, 1982, 1994). Recharge was in highland areas to the southwest, and deep regional flow gained solutes through interaction with evaporites and other rocks along the flow path (fig. 3). At the more local scale, hydrologic theory and modern analogs suggest the presence of an interface that would have been near the surface in topographic depressions (fig. 5). In some areas, such as the Henry Basin, the saline water inter- face may have remained at shallow depth and not reached the surface (fig. 5A). In other areas, such as the Uravan mineral belt, the interface may have reached the surface, allowing relatively fresh water and saline water to dis- charge together (fig. 5B). RECAPTURE MEMBER OF THE MORRISON FORMATION Similar to the Salt Wash Sandstone Member, the Recapture Member of the Morrison Formation was depos- ited on a broad, roughly fan shaped alluvial plain sloping to the northeast (fig. 7) (Craig and others, 1955; L.C. Craig, unpublished data, 1982; Peterson and Turner-Peterson, 1987; Peterson, in press). A sandstone facies between conglomeratic and clay- stone-sandstone facies (Craig and others, 1955; LC. Craig, unpublished data, 1982) marks a decrease in paleotopo- graphic slope where local and regional ground-water sys- tems probably discharged. Syndepositional subsidence (Hilpert, 1969) favored local and regional ground-water discharge. Geologic controls favoring discharge from the Recap- ture Member include thinning of the Lower Jurassic and upper Paleozoic aquifer systems. Dramatic thinning of the Navajo Sandstone favored discharge beneath the northwest- ern part of the Recapture. The Wingate Sandstone, which is the major aquifer east of the pinchout of the Navajo Sand- stone, is thickest in the Four Corners area and thins from there toward the northeast (Jobin, 1962; Blakey and others, 1988). Thinning of the Wingate to the northeast also favored discharge. The Entrada Sandstone has a relatively constant thickness (Jobin, 1962; Blakey and others, 1988). 0 10 MILES 0 10 KILOMETERS UTAH 38° 30' |- COLORADO «£ EXPLANATION El Greenish-gray alteration Mottled alteration - Red-brown alteration m Uranium deposit discovered after 1950 e Uranium deposit discovered through 1950 Figure 6. - Map showing areas of alteration in mudstone underly- ing uranium-bearing sandstone and location of uranium deposits in the northern end of the Uravan mineral belt. Uranium deposits are closely associated with reduced mudstone in areas of inferred ground-water discharge. Mudstone alteration and uranium deposits discovered through 1950 from McKay (1955); deposits discovered after 1950 from Umetco Company (unpublished map, 1976). The Lower Permian Glorieta Sandstone, Meseta Blanca Sandstone Member of the Yeso Formation, San Andres Limestone, and De Chelly Sandstone are today the major upper Paleozoic aquifers in the San Juan Basin (Stone and others, 1983). The sandstone aquifers thin and pinch out toward the north and east (Baars and Stevenson, 1977; Blakey and others, 1988), but, because the present trans- missivity of the San Andres Limestone is related to dissolu- tion near the outcrop, it may not have been an aquifer in Jurassic time. Tabular sandstone uranium deposits in the Recapture Member in northwest New Mexico (Hilpert, 1969) are at the transition between the conglomeratic and sandstone faces and near the pinchout of the Navajo and Wingate Sandstones. An eastward-trending troughlike structure traverses the main part of the mining district (Hilpert, 1969). Anhydrite cement in the Recapture Member (Hans- ley, 1990) and the upper part of the underlying Middle Jurassic Wanakah Formation (Ridgley, 1989) suggests a source of saline water in the underlying Middle Jurassic Todilto Limestone Member of the Wanakah Formation or the Permian evaporitic rocks of the Black Mesa uplift (figs. 2, 4). 12 37° --= JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS Limit of recognizable Recapture Member \: ')coLoRraDo. __ _L e_ ls :\ NEW MEXICO SAN LUIS PRECAMBRIAN BLOCK (hydrologic barrier) Colorado 50 100 MILES Plateau 0 50 100 KILOMETERS EXPLANATION Conglomeratic facies © Uranium deposit cluster Sandstone facies nninfoninionts Thickness of Navajo Sandstone (in feet) laystone-sandst faci Claystone-sandstone facies ,...... .., Thickness of Wingate Sandstone Zone of mixed discharge (in feet) SOUTHWEST Navajo Sandstone and Wingate Sandstone aquifers NORTHEAST Zone of mixed discharge De Chelly Sandstone aquifer Evaporites of the Yeso Formation Evaporites of the lower Cutler Formation Evaporites of the Todilto Limestone Member of the Wanakah Formation EXPLA ON XPLANATI FEET - METERS Major eolian aquifers 3000 1000 - Evaporitic rocks 1500 500 E: Other rocks 0 -o Figure 7. Simplified map and cross section of the San Juan Basin for Late Jurassic time at the end of deposition of the Recapture Member of the Morrison Formation. Discharge is favored by decrease in topographic slope in sandstone facies and by pinchout of the Navajo Sandstone and Wingate Sandstone aquifers. The De Chelly Sandstone (not shown) also thins abruptly just east of the pinchout of the Wingate Sandstone. Arrows on map and cross section show direction of ground-water flow. Dotted line in cross section is dilute-saline interface. Data from Craig and oth- ers (1955), Baars (1962), Jobin (1962), Mallory (1972), Baars and Stevenson (1977), and Blakey and others (1988). PALEOHYDROGEOLOGY IN THE COLORADO PLATEAU 13 WESTWATER CANYON MEMBER OF THE MORRISON FORMATION The Westwater Canyon Member was deposited on a fan-shaped alluvial plain traversed by braided streams flow- ing generally toward the northeast and locally to the east and southeast (Craig and others, 1955; Hilpert, 1969; L.C. Craig, unpublished data, 1982; Condon and Huffman, 1984; Con- don and Peterson, 1986; Turner-Peterson, 1986; Peterson and Turner-Peterson, 1987). Local and regional ground-water discharge in the West- water Canyon Member probably occurred along a north- west-trending zone where a transition from an alluvial-plain facies to a mudflat facies suggests a major change in slope (figs. 8, 9). A map of sandstone to mudstone ratio indicates the regional paleotopographic slope (fig. 8). Sandstone to mudstone ratios vary from 300: 1 in the southwest to 0.5:1 in the northeast (Robert Lupe, U.S. Geological Survey, unpub- lished data, 1982; Kirk and Condon, 1986). Discharge prob- ably occurred in a zone of major decrease in topographic slope, approximately from the >10: 1 to 4:1 contours of sand- stone to mudstone ratio (fig. 8). Tectonic activity affected the distribution of facies and the paleoslope on a more local scale, as indicated by locally thick sediments, high sandstone to mudstone ratios, and smaller numbers of sandstone-mud- stone interbeds (Kirk and Condon, 1986). Stream channels were concentrated in subsiding structures, whereas overbank muds were deposited on paleohighs. Thus, local ground- water flow tended to recharge in the local highs dominated by overbank mudstone and discharge in the depressions dominated by channel sandstone. The most significant geologic control on ground-water flow in the San Juan Basin probably was the thinning of the De Chelly Sandstone and the Meseta Blanca Member of the Yeso Formation (fig. 9). The De Chelly Sandstone thins abruptly from 240 to 100 m (800-300 ft) and then pinches out gradually toward the east (Baars and Stevenson, 1977). The Meseta Blanca Member thins abruptly from 150 to 100 m (500-300 ft) and then pinches out gradually toward the north (Baars and Stevenson, 1977). Because the change in thickness (rather than the thickness itself) is the critical geo- logical control on ground-water flow, the region of abrupt thinning (rather than the pinchout) was most likely the zone of maximum geologically controlled discharge for regional ground water. Tabular sandstone uranium-humate deposits in the western part of the Grants uranium region were a major source of uranium (Kelley, 1963; Hilpert, 1969; Rautman, 1980; Adams and Saucier, 1981; Turner-Peterson, Santos, and Fishman, 1986; Finch and McLemore, 1989). The deposits are in the Westwater Canyon Member and cluster in a belt from just upslope of the 10:1 contour of sandstone to mudstone ratio to the 4:1 contour, in the zone of abrupt thinning of the Meseta Blanca Member (fig. 9). The ura- nium region itself consists locally of two or more local belts of deposits that are 3-5 km (2-3 mi) apart (Granger and oth- ers, 1961). The main deposits tend to be in the centers of the thickest sandstone masses, which are in syndepositional synclines (Hilpert and Moench, 1960; Hilpert, 1969; Kirk and Condon, 1986). The general position of the Grants ura- nium region probably was controlled by ground-water dis- charge at the maximum change in topographic slope, indicated by the major facies change, and by the zone of maximum thinning of the Meseta Blanca Member. Local clusters of uranium deposits within the region were proba- bly controlled by local paleotopographic depressions. Discharging regional ground water was probably saline owing to dissolution of anhydrite-bearing evaporites of the Middle Jurassic Todilto Limestone Member of the Wanakah Formation (figs. 4, 10). Dissolution is shown by the numerous breccia pipes that terminate downward in the Todilto (Moench and Schlee, 1967; Hilpert, 1969; Hunter and others, 1992). The oxidation state of iron in the sedimentary rocks suggests local recharge and discharge areas within the gen- eral area favorable for ground-water discharge in the Grants uranium region. For example, a syndepositional syncline contains reduced rock and uranium deposits, whereas the adjacent paleohighs contain oxidized rock and lack ura- nium deposits (Turner-Peterson, 1985). The combination of paleotopographic lows and reduced rock suggests that high water table, ground-water discharge, and preservation of organic matter are associated with the deposits. Detrital ilmenite and titaniferous magnetite in sandstone of the Westwater Canyon Member show variable degrees of alteration to anatase, leucoxene, hematite, and pyrite (Adams and others, 1974; Adams and Saucier, 1981; Reynolds and others, 1986). Alteration probably occurred by incongruent dissolution of detrital iron-titanium oxide minerals by ground water that was reducing, neutral to weakly acid, and humic acid rich (Adams and others, 1974). Dissolution of iron-tita- nium oxide minerals is diagenetic evidence for shallow, rel- atively fresh, ground water, and the pattern of alteration indicates the extent of the shallow ground-water system (figs. 10, 11). In general, the intensity of iron-titanium oxide dis- solution decreases downward and from southwest to north- east (Adams and others, 1974; Adams and Saucier, 1981; Turner-Peterson, 1985, 1986; Reynolds and others, 1986). Some authors (Adams and Saucier, 1981; Turner-Peterson, 1985; Turner-Peterson and Fishman, 1986) have plotted the total thickness of the zone of iron-titanium oxide dissolution. This measure does not truly represent the intensity of iron- titanium oxide mineral alteration because the thickness of the host Westwater Canyon Member also decreases to the south- west. Plotting the same data in terms of the percentage of the Westwater Canyon Member that is altered better indicates 14 37° JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS 109° I, I Limit of recognizable Westwater Canyon Member Q~ i Edge of [3 © 1 Colorado + o I / Plateau 24 UTAH 1 had COLORADO ARIZONA 5 NEW MEXICO MCKINLEY COUNTY 1 Albuquerque 50 T 100 KILOMETERS 100 MILES ] EXPLANATION Westwater Canyon Member of the Morrison Formation Topographically controlled discharge- Sandstone/mudstone decreases from >10 to 4 Sandstone/mudstone data point Figure 8. Map showing contours of sandstone to mudstone thickness ratio in the Westwater Canyon Member of the Mor- rison Formation. Decrease in ratio from southwest to northeast indicates decrease in topographic slope. Ground-water dis- charge and mixing is favored where sandstone to mudstone ratio is from >10 to 4. Data from Kirk and Condon (1986; area in New Mexico outlined by dotted curve) and from Robert Lupe (U.S. Geological Survey, unpublished data, 1982; data points shown by solid circles). PALEOHYDROGEOLOGY IN THE COLORADO PLATEAU 15 the intensity of alteration in the Westwater Canyon (fig. 11). The contours indicate a wedge-shaped lens of alteration that includes the entire Westwater Canyon Member in parts of the south and west (100 percent alteration) and thins to a feather edge toward the northeast. Pockets of less alteration are present in the middle of the southern edge of the Westwater Canyon Member but probably are secondary effects super- imposed on the regional trend. Although considerable lati- tude is possible in contouring the data, there is no reason to interpret a pinching-out of the alteration zone to the south, as indicated by previous authors (Adams and Saucier, 1981; Turner-Peterson, 1985; Turner-Peterson and Fishman, 1986). The pattern of iron-titanium oxide alteration has been interpreted to indicate downward flow of fluid and is a major argument in support of the "lacustrine-humate model" of ura- nium ore formation (Turner-Peterson, 1985; Turner-Peterson and Fishman, 1986). This evidence alone is not compelling because the alteration pattern can equally well be interpreted as a result of downdip flow of dilute ground water that tended to be confined to the upper part of the sandstone by saline ground water. BRUSHY BASIN MEMBER OF THE MORRISON FORMATION The Brushy Basin Member and the lithologically simi- lar undifferentiated Morrison Formation consist of discon- tinuous sandstone beds in a predominantly mudstone and claystone matrix with limestone beds and nodules (Craig and others, 1955; Dodson and others, 1980; Bell, 1983, 1986; Turner-Peterson, 1985, 1987; Lockley and others, 1986; Peterson and Turner-Peterson, 1987; Turner and Fishman, 1991). Lithofacies suggest a low-gradient alluvial plain and scattered lakes or playas. A large lake having well-devel- oped diagenetic facies similar to those in modern playas such as Searles Lake (Smith, 1979) is thought to have covered 150,000 km mainly in northwest New Mexico and south- west Colorado (figs. 12-14) (Bell, 1983, 1986; Turner-Peter- son, 1985, 1987; Turner and Fishman, 1991). The general direction of streams was northeastward. The generally fine grain size of the Brushy Basin Mem- ber suggests a low and uniform topographic slope that would not have favored significant recharge or discharge over much of the area. Lithofacies vary from continuous sand- stone (sandstone to mudstone >1.0), to discontinuous sand- stone (sandstone to mudstone <1.0 and >0.6), to dominantly mudstone (sandstone to mudstone <0.6) (fig. 124) (Bell, 1983, 1986). The main change in topographic slope was probably in the southern part of the depositional area of the Brushy Basin Member in northwest New Mexico where mostly sandstone was deposited (continuous sandstone, fig. 12A, and sandstone, fig. 12B). Quantitative modeling of ground-water flow during deposition of the Brushy Basin Member indicates that dis- charge occurred throughout a large topographic depression (Sanford, 1990b, 1994). A lake in the depression would have focused ground-water discharge at the shoreline. Most of this discharge presumably took place on the upstream or southwest side of the lake because of the regional hydraulic gradient (fig. 13). Both regional and local ground-water sys- tems probably discharged around the playa. In the absence of a lake, ground water probably discharged in the topo- graphically lowest parts of the closed depression. The low permeability of the Brushy Basin Member, compared to the Westwater Canyon Member, suggests that ground-water flow through the Brushy Basin Member may have been mainly vertical. Geologic controls on ground- water flow during deposition of the Brushy Basin Member were similar to those for the rest of the Morrison Formation, as discussed above (figs. 3, 7, 9). Deep, saline ground water discharged where Jurassic and upper Paleozoic aquifer sys- tems thinned abruptly in the subsurface. The Uncompahgre and San Luis blocks forced most of the remaining ground water to discharge or flow around the blocks. Diagenetic studies of the Brushy Basin Member have focused on the alteration associated with a large alkaline- saline playa lake (figs. 12, 13) (Bell, 1983, 1986; Turner- Peterson, 1985; Turner-Peterson and Fishman, 1986; Turner and Fishman, 1991). Turner and Fishman proposed well- defined facies boundaries. TE. Bell (oral commun., 1989) considered the facies boundaries too poorly defined to allow identification of discrete boundaries. A calcite facies (Bell, 1983, 1986) is important evidence for fluid mixing. The mar- ginal mudflat of the Brushy Basin playa corresponds to the smectite-discontinuous sandstone and calcite facies (figs. 12A, 14) (Bell, 1983, 1986), alternatively termed the smec- tite facies (fig. 12B) (Turner-Peterson, 1985; Turner-Peter- son and Fishman, 1986; Turner and Fishman, 1991). The smectite-discontinuous sandstone facies contains isolated limestone nodules, and the calcite facies contains abundant laterally continuous limestone beds (Bell, 1983, 1986). By analogy with modern playas and saline lakes (Eugster and Hardie, 1978; Mono Basin Ecosystem Study Committee, 1987; Duffy and Al-Hassan, 1988), this carbonate formed as a result of discharge of bicarbonate-bearing ground water at the edge of the playa. The increasing amount of carbonate going from smectite to calcite facies suggests increasing car- bonate concentration and deeper flow depth of the discharg- ing ground water. Diagenetic minerals such as analcime, zeolites, and potassium feldspar occupy the central part of the playa and are interpreted as evidence for an alkaline-saline brine (fig. 12) (Bell, 1983, 1986; Turner-Peterson, 1985; Turner- Peterson and Fishman, 1986; Turner and Fishman, 1991). Other diagenetic indicators are more difficult to interpret. 16 JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS 109° LTG san luis '!, <»... PRECAMBRIAN . : , BLOCK |; ©.. Edge of Colorado Plateau 32" --- COLORADO NEW MEXICO i Albuquerque 1 l 0 50 100 MILES : Line of i 4 ! I | I I cross section 0 50 100 KILOMETERS | SOUTHWEST Area of detail Dilute ~---- 3% ra Entrada EXPLANATION Major eolian aquifers - Evaporitic rocks I: Other rocks NORTHEAST Sandstone Evaporites aquifer of the o- Todilto Limestone g Member of the Wanakah Formation b Meseta Blanca Evaporites of the Member aquifer Yeso Formation Saline j FEET - METERS 1000 3000 1500 -|- 509 o -o PALEOHYDROGEOLOGY IN THE COLORADO PLATEAU 17 For example, the alteration of smectite to illite, which increases downdip toward the center of the basin, has been interpreted as evidence both for upward flow of deep ground water (Whitney and Northrop, 1987) and for down- ward flow of playa lake water (Owen and others, 1989; Turner and Fishman, 1991). Modern playa analogs (Allison and Barnes, 1985; Duffy and Al-Hassan, 1988) and computer modeling (San- ford, 1990b, 1994) show that flow can be downward or upward depending on hydrodynamic conditions. If the ground water had a constant density throughout, it would discharge at a topographic depression; however, in a playa setting where surface water is highly concentrated, buoy- ancy may cause ground water to sink beneath the playa and recirculate by convection back updip to discharge near the shoreline or margin of the playa (Duffy and Al-Hassan, 1988). Downward flow of playa lake water is suggested by zeolite alteration in the Westwater Canyon Member beneath the Brushy Basin Member (Hansley, 1986, 1990). One pos- sible interpretation that is consistent with hydrologic theory, modern analogs, model predictions, and diagenetic alter- ation is shown in figure 14. Although the flow of shallow ground water in the alluvial plain probably was consistently downdip, the flow of ground water in the playa probably EXPLANATION [Z] Westwater Canyon Member of Morrison Formation Upper Paleozoic aquifers >300 feet Precambrian block Zone of mixed discharge Uranium deposit cluster ————— Thickness of De Chelly Sandstone (in feet) --------- Thickness of Meseta Blanca Member of Yeso Formation (in feet) Contour of sandstone to mudstone ratio in Westwater Canyon Member of Morrison Formation Figure 9 (above and facing page). - Simplified map and cross sec- tion of the San Juan Basin for Late Jurassic time at the end of depo- sition of the Westwater Canyon Member of the Morrison Formation. Transition in sandstone to mudstone ratio from >10 to 4 indicates topographically controlled discharge. Abrupt thinning of the De Chelly Sandstone and Meseta Blanca Sandstone Member aquifers indicates geologically controlled discharge. Where both coincide is an inferred zone of mixed local and regional discharge. Arrows on map and cross section show principle direction of ground-water flow. Area in cross section delineated by box is shown in more detail in figure 10. Dotted line in cross section is dilute-saline interface. Data from Craig and others (1955), Baars (1962), Jobin (1962), Peterson and others (1965), Mallory (1972), Baars and Stevenson (1977), and Blakey and others (1988). varied with time, alternately moving downward or upward as the density of the lake and pore fluid varied. According to one version of the lacustrine-humate model, uranium deposits in the Westwater Canyon Member and Jackpile Sandstone Member are genetically related to the mudflat (smectite) facies of the Brushy Basin Member (Turner-Peterson, 1985; Turner-Peterson and Fishman, 1986). Although a fair correspondence between deposits and the mudflat facies exists in the San Juan Basin, the cor- respondence completely breaks down for other parts of the Colorado Plateau (fig. 13). In the Uravan mineral belt, most deposits in the Salt Wash Member are beneath the anal- cime-potassium feldspar facies of the Brushy Basin Mem- ber. Furthermore, the Salt Wash Sandstone Member may be separated from the upper part of the Brushy Basin Member by an unmineralized lower part that has been interpreted as equivalent to the Recapture Member (Turner and Fishman, 1991). In the Henry Basin, uranium deposits are beneath the alluvial or continuous-sandstone facies. Thus, the corre- spondence of Brushy Basin mudflat with tabular sandstone uranium deposits is probably a fortuitous coincidence lim- ited to the San Juan Basin rather than a general characteris- tic having genetic significance. JACKPILE SANDSTONE MEMBER OF THE MORRISON FORMATION The Jackpile Sandstone Member overlies and interfin- gers with the Brushy Basin Member in the southeastern part of the San Juan Basin (Moench and Schlee, 1967; Owen and others, 1984). It consists mostly of fluvial sandstone depos- ited in an active northeast-trending syncline. The upper part of the Jackpile Member was removed in places by erosion prior to deposition of the Upper Cretaceous Dakota Sand- stone (Moench and Schlee, 1967; Adams and others, 1978). The presence of zeolite facies mudstone of the Brushy Basin Member beneath the thickest part of the Jackpile Sandstone Member indicates that a low area existed prior to Jackpile deposition (Bell, 1983). Isopachs and interfingering relations with the Brushy Basin Member indicate active sub- sidence during deposition (Moench and Schlee, 1967). Pres- ervation of the Jackpile Member suggests that low topography persisted while surrounding areas were uplifted and eroded. Paleocurrent directions and isopachs of the Jackpile Sandstone Member (Moench and Schlee, 1967; Adams and others, 1978) suggest that shallow, relatively fresh ground water flowed northeast along the axis of the syndepositional syncline (fig. 15). Because the area was a topographic depression, deeper ground water probably discharged into the base of the unit (Sanford, 1990b, 1994). Diagenetic evidence indicates the timing, relative extent, and movement of the shallow and deeper ground water. Iron-titanium oxide minerals were extensively 18 JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS SOUTHWEST NORTHEAST Sandstone to mudstone ratio 10:1 4:1 1:1 Zone of complete iron-titanium oxide dissolution Tabular uranium- humate deposit Spri waa al prings Dilute local flow Carbonate mudflat Saline regional flow Figure 10. - Detailed schematic cross section of the Westwater Canyon Member of the Morrison Formation and underlying units for Late Jurassic time at the end of Westwater Canyon deposition. Location of area of cross section is delineated by box on cross section of figure 9. No vertical scale implied. Arrows show direction of ground-water flow. Dotted line is dilute-saline interface. Flow is primarily gravity driven and downdip. Discharge occurs where topographic surface has decrease in slope. Breccia pipes, formed by dissolution of anhy- drite-bearing Todilto Limestone Member of the Wanakah Formation, were conduits for upward flow of saline water. Shallow fresh water forms a lens resting upon deeper regional saline water. Unit abbreviations: Je, Entrada Sandstone; Jwt, Todilto Limestone Member of the Wanakah Formation; Jwh, Beclabito Member of the Wanakah Formation; Jcs, Cow Springs Sandstone; Jmr, Recapture Member of the Morrison Formation; Jmw, Westwater Canyon Member of the Morrison Formation. dissolved in "hydrologically continuous" sandstone of the Jackpile Sandstone Member (Adams and Saucier, 1981). This alteration was paragenetically very early and probably occurred during or shortly after deposition (Adams and oth- ers, 1978). As discussed earlier for the Westwater Canyon Member, iron-titanium oxide dissolution suggests alteration by dilute, reducing, humic-acid-bearing ground water. Other alteration suggests later incursion of brackish or saline ground water from the underlying Brushy Basin Member or lower units into the deepest parts of the syn- cline (Adams and others, 1978; Bell, 1983). Albitization of feldspar, which decreases in intensity upward, suggests that the later ground water was sodium-rich fluid similar to that inferred for pore water in the zeolite facies of the Brushy Basin Member directly beneath the Jackpile Mem- ber (Bell, 1983). Alternatively, saline ground water may have discharged upward from units such as the Todilto Limestone Member of the Wanakah Formation. Tabular uranium-humate deposits suggest deposition at an interface between relatively fresh and saline ground water within the Jackpile Sandstone Member. The presence of the deposits in the thickest part of the sandstone above the deepest part of the structure is consistent with ponding of saline water at depressions within the sandbody (Bell, 1983). Breccia pipes that extend up from the Todilto Lime- stone Member and to the Jackpile Sandstone Member (Moench and Schlee, 1967; Hilpert, 1969; Hunter and oth- ers, 1992) may have been conduits through the Brushy Basin Member for the upward-moving, deep regional ground water. Although such conduits may have aided upward ground-water flow, they would not have been nec- essary for significant upward flow. Discharge through modern playa lakebeds is a common phenomenon (Mey- boom, 1967; Lissey, 1971; Stephenson, 1971; Winter, 1976, 1978, 1986, 1988; Friedman and others, 1982; Alli- son and Barnes, 1985; Spencer and others, 1985; Duffy PALEOHYDROGEOLOGY IN THE COLORADO PLATEAU 19 109° | lg— - _ 1 I I I 1 / I 1 0 < |Q Z | x O | a Gallup N ; * ° fA a <| & Z I I I I 1 | I Grants J I 0 1 0 10 20 30 40 50 MILES I l | iu u | | | | I l Td r ‘ l I l 35° |- McKINLEY COUNTY | 0 20 30 40 50 KILOMETERS _ o _ _] ! | SOURCES OF DATA & Adams and Saucier (1981) K Reynolds and others (1986) 4 N.S. Fishman (written commun., 1989) Figure 11. Map showing contours of the interval of total ilmenite-magnetite dissolution in the upper part of the Westwater Canyon Member of the Morrison Formation. The thickness of the altered interval is plotted as a percentage normalized to the total thickness of the Westwater Canyon Member. The normalized thickness indicates the intensity of alteration. Compare with maps of the unnormalized thick- ness of the same interval shown in Adams and Saucier (1981), Turner-Peterson (1985), and Turner-Peterson and Fishman (1986). and Al-Hassan, 1988; Winter and Woo, 1990), and com- puter modeling of Late Jurassic-Early Cretaceous ground- water flow in the San Juan Basin shows that Brushy Basin Member mudstone could have been sufficiently permeable to yield significant discharge, approximately 15+10 mm/yr (Sanford, 1994). Thus, the diagenetic and paragenetic relations indicate that the Jackpile Sandstone Member was altered mostly by shallow, relatively dilute ground water early in its history and, later in its history, by alkaline-saline ground water mov- ing upward from the underlying Brushy Basin Member and (or) by saline water moving upward from deeper units such as the Todilto Limestone Member. Alkaline-saline and (or) saline ground water probably ponded in deeper parts of the basin and formed an interface with fresher water above along which tabular sandstone uranium deposits formed. LOWER PART OF THE CHINLE FORMATION The lower part of the Upper Triassic Chinle Formation consists of continental deposits of fluvial and lacustrine ori- gin (Stewart and others, 1972; Blakey and Gubitosa, 1983; Dubiel, 1987, 1989; Dubiel and others, 1991). Clastic sedi- ments were shed from the Uncompahgre highlands on the northeast and from the Mogollon highlands on the southwest. 20 35° JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS 109° ' : 1 Cuba LQ. _ _ _ _ _ _ _ _ _ _ 900 \~\\‘\ bat ~ | e | 5 1 1 \\ I \ \ 0 \ '- | \ 0 *. : \\ 0 i \ 7 \, A\ 'O. i Analcime and O * heulandite- < 1 OQ \ C clinoptilolite z | E \ - +. « N | Z &. ~ &, %. f re "Oy;> o C. o 6 © I 3 "on o ~ 4 > lef * 2 Sfo G53. \ ® A» 076 a ts treat 1 'c. WGrants I I Y., \ ~. I N o e A i McKINLEY COUNTY | * [I] 110 210 3|0 410 5|U MILES L_L_ 1 4 e \_ mT T T T T [Cf C T- \0 _ 10 _ 20 _ 30 _ 40 __ 50 KILOMETERS - ~ | * A 109° I [L_ a Soy $ Me - - - m mmm at m macho as Analcime O and 1 Q potassium 1 0.6 ALLUVIAL PLAIN MUDFLAT (continuous (discontinuous sandstone facies) sandstone- smectite facies) _- Springs CARBONATE MUDFLAT Mixing zone 0.4 0.23 PLAYA LAKE (zeolite-potassium feldspar facies) (calcite facies) Fossil mixing zone and uranium-humate deposits Figure 14. - Detailed schematic cross section of the Brushy Basin Member of the Morrison Formation and underlying units for Late Jurassic time at the end of Brushy Basin deposition in the Colorado Plateau area along line of section shown in figure 13. Arrows show general di- rection of ground-water flow. Gravity-driven downdip flow from southwest meets updip density-driven local flow from playa and regional flow from deep basin. Discharge occurs where the topographic surface has decrease in slope, as indicated by transition from alluvial plain to mudflat. Facies and average sandstone to mudstone ratios for the Brushy Basin Member are from Bell (1983, 1986). Unit abbreviations: Je, Entrada Sandstone; Jwt, Todilto Limestone Member of the Wanakah Formation; Jwh, Beclabito Member of the Wanakah Formation; Jcs, Cow Springs Sandstone; Jmr, Recapture Member of the Morrison Formation; Jmw, Westwater Canyon Member of the Morrison Formation; Jmb, Brushy Basin Member of the Morrison Formation. Major rivers flowed northwestward from northwestern New Mexico to northwestern Utah, and tributaries flowed north- ward from the Mogollon highlands and westward from the Uncompahgre highlands (figs. 2, 16). The lower part of the Chinle Formation consists of the Shinarump, Monitor Butte, Moss Back, and Petrified Forest Members (Stewart and others, 1972; Blakey and Gubitosa, 1983; Dubiel, 1987, 1989). The basal Shinarump Member, characterized by conglomerate and sandstone deposited in braided stream channels, fills valleys and scours in the underlying marine Kaibab Limestone, the marine-marginal Moenkopi Formation, and the eolian De Chelly and Coconino Sandstones. The overlying Monitor Butte Mem- ber shows the same transport directions as the Shinarump Member but consists of tuffaceous, bentonitic mudstone, siltstone, sandstone, and limestone deposited in lacustrine and fluvial environments. The next higher unit, the Moss Back Member, similar to the Shinarump Member, consists mainly of fluvial sandstone and conglomerate. Locally, the Moss Back Member is the basal unit and rests on the Per- mian Cutler Formation (Wood, 1968). At the top of the Chinle, the Petrified Forest Member consists of variegated mudstone and lenticular sandstone deposited on a low-gra- dient alluvial plain having high- and low-sinuosity fluvial channel systems, overbank and floodplain environments, and scattered lakes and marshes. Lithofacies and fossils indicate that precipitation and surface water were abundant, streams and lakes were fresh and perennial, and the water PALEOHYDROGEOLOGY IN THE COLORADO PLATEAU 23 table was typically high; however, paleosols and ichnofos- sils indicate that water tables and lake levels fluctuated epi- sodically owing to a tropical monsoonal climate (Dubiel and others, 1989, 1991). TOPOGRAPHIC CONTROLS ON GROUND-WATER FLOW Based on lithofacies data (Stewart and others, 1972; Blakey and Gubitosa, 1983; Dubiel, 1987, 1989), topo- graphic control caused ground water to flow dominantly northwestward following the main channel systems of the Shinarump and Moss Back Members (fig. 16). Flow in the margins of the depositional area was along the tributary streams toward the main northwest-trending fluvial axis. Active structural features affected ground-water flow at regional and local scales. A large, actively rising highland apparently controlled the distribution of lithofacies in the White Canyon and Monument Valley areas (Young, 1964; Malan, 1968; Fisher, 1972). In the Paradox Basin, particu- larly in the Lisbon Valley area, active anticlines were asso- ciated with salt diapirism (Wood, 1968; Fisher, 1972). The topographic gradient and probably the rates of ground-water flow probably changed systematically with time, as shown by the variations in mean grain size of the dif- ferent members. The coarse-grained, fluvial channel sand- stone of the Shinarump Member suggests a relatively steep gradient. The change to fine-grained rocks of the Monitor Butte Member that were deposited in lacustrine, deltaic, and marsh environments implies a shallower gradient and slower ground-water flow. Low-gradient conditions were locally interrupted during deposition of the dominantly coarser grained Moss Back Member but returned during deposition of the fine-grained Petrified Forest Member. GEOLOGIC CONTROLS ON GROUND-WATER FLOW Geologic control on ground water was dominated by complex variations in thickness and facies of the upper Pale- ozoic aquifers (fig. 16). Southwestward along the flow path from the ancient Uncompahgre highlands, the upper Paleo- zoic aquifer system thickens abruptly into the Paradox Basin and from there it thins (Baars, 1962; Jobin, 1962; Blakey and others, 1988; Geldon, in press and unpublished data). Also in a southwest direction, arkosic rocks decrease in thickness, whereas eolian sandstone, particularly the Cedar Mesa Sand- stone and White Rim Sandstone Members of the Cutler For- mation, increases in thickness. Regional thinning of the Paleozoic section westward from the axis of the Paradox Basin suggests decreasing transmissivity and therefore dis- charge, but more eolian sandstone relative to arkose favors recharge because of increased transmissivity. Farther from the Paradox Basin, depositional and tectonic variations are important locally. The net effect of the various influences is unclear, at least locally; however, in the area between the Uncompahgre highlands and the main lower Chinle fluvial 107°30' 10715 35° | 22 i 5 10 MILES 10 KILOMETERS Figure 15. Isopach map of the thickness (in feet) of the Jackpile Sandstone Member of the Morrison Formation showing facies of underlying Brushy Basin Member, location of uranium deposits (solid black), and directions of paleoflow. Area of calcareous facies of Brushy Basin Member is unpatterned; area of zeolite facies of Brushy Basin Member is diagonal-lined. Modified from Moench and Schlee (1967), Adams and others (1978), and Bell (1983). channel, there may have been three general ground-water zones. Next to the uplift, where the topographic slope was greatest and the upper Paleozoic sediments were dominantly arkosic, recharge probably was dominant. Southwestward from this zone, where the slope was more gradual and thinner sediments (favoring discharge) competed with thicker eolian sandstone (favoring recharge), recharge gradually gave way to discharge. Finally, as ground water neared the main Chinle channel axis, where the topographic slope was minimal, ground water probably discharged. A similar zonation may have been present northward from the Mogollon highlands, although thickness and facies changes were less extreme. From the Mogollon highlands northward almost to the Utah State line, the upper Paleozoic section gradually thins, whereas eolian sandstone units, par- ticularly the De Chelly (Blakey and others, 1988), gradually thicken. Based on the isopachs of total sandstone thickness, the net effect was probably a slight increase in transmissiv- ity. In the vicinity of the Arizona-Utah State line, the thick- ness of Permian sandstone abruptly decreases mainly due to thinning of the De Chelly Sandstone, but, farther to the north, Permian sandstone units thicken again owing to the Cedar Mesa Sandstone and White Rim Sandstone Members (Blakey and others, 1988). 24 JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS 37° Edge of Colorado Plateau Simplified Shinarump channels ARIZONA ) 0 50 100 MILES NEW MEXICO 50 100 KILOMETERS EXPLANATION I: Area of topographic lows in deltaic, lacustrine, and marsh environments (topographically controlled discharge) --------- Contour of combined thickness of upper Paleozoic aquifers (in feet) ( } Uranium deposit cluster Figure 16. Map showing relationship between uranium deposit clusters and ground-water discharge in the Colorado Plateau during deposition of the lower part of the Triassic Chinle Formation. Channels in Shinarump and Moss Back Members indicate north-, west-, and northwestward surface- and ground-water flow (indicated by arrows). Discharge of local and regional ground water was in topographically low areas marked by deltaic, lacustrine, and marsh environments of the Monitor Butte Member. Paleogeographic data simplified from Dubiel (1989); thickness of upper Paleozoic aquifers from A.L. Gelden (unpublished data, 1992); uranium deposit clusters from Finch (1991). PALEOHYDROGEOLOGY IN THE COLORADO PLATEAU 25 Thus, the overall pattern of geologic control on ground- water flow suggests a transition from dominantly recharge to dominantly discharge from basin margin to central fluvial channels. Unlike the Morrison Formation, geologic and topographic controls in the Chinle Formation did not clearly combine to favor discharge, but, instead, topographic control of discharge was partly counteracted by geologic controls that favored recharge. DIAGENESIS AND URANIUM DEPOSITION The distribution of oxidized and reduced rocks aids in identifying recharge and discharge zones and accumulations of organic matter. Green and gray mudstone and sandstone coincide with Shinarump and Moss Back channels (Wood, 1968; Dubiel, 1989), supporting the inference that the Chinle fluvial channels were zones of persistent ground-water dis- charge throughout deposition of the lower part of the Chinle Formation. The green mudstone facies of the Monitor Butte Member (Dubiel, 1989) also suggests reducing conditions, a constant high water table, and perennial ground-water dis- charge. Preserved organic carbon and reduced sandstone and mudstone elsewhere in the Chinle Formation (Dubiel, 1989) further indicate areas of persistent ground-water discharge. The inferred ground-water-flow pattern would have caused zones of mixing between relatively fresh, local ground water and more saline, regional ground water in areas of discharge (fig. 16). The gypsiferous Moenkopi Formation (Blakey, 1974) and evaporites in Pennsylvanian and Permian rocks probably contributed to the salinity of deeper ground water that tended to discharge upward into the Chinle For- mation in areas marginal to and within floodplain, lacustrine, and marsh environments. Tabular sandstone uranium-vanadium and uranium- copper deposits in the Chinle Formation are associated with transitional lithofacies, channel sandstone, syndepositional synclines, and reduced lacustrine mudstone in the Shi- narump and Moss Back Members (Johnson, 1957; Finch, 1959; Witkind and Thaden, 1963; Thaden and others, 1964; Malan, 1968; Wood, 1968; Lupe, 1977; Huber, 1980; Dubiel, 1983). Almost all of the deposits are in areas of inferred topographically controlled discharge, as shown by deltaic, marsh, and lacustrine deposits of the Monitor Butter Member (fig. 16). Ore deposits are further localized at tran- sitional lithofacies in the Shinarump and Moss Back Mem- bers where fluvial channel sandstone pinches out and where distal braided stream deposits grade into floodplain deposits (Finch, 1959; Lupe, 1977). Many of the Chinle-hosted deposits are in a belt 4.8-19 km (3-12 mi) wide and 209 km (130 mi) long that corresponds to a major facies change from dominantly sandstone to dominantly mudstone on the flanks of a large uplift approximately in the area of the present Monument uplift (Young, 1964; Malan, 1968; Fisher, 1972). 110°30' 110°00' @ I but LC®] 0 5 10 MILES Tfi @ 0 _ 50 __ 10 kilomete®s @ # ~ \ O - &A +$ D $6966 Figure 17. Map showing relationship between uranium deposits (solid circles), syndepositional synclines, and organic-rich lacus- trine mudstone (shaded areas) in part of the Chinle Formation in southeast Utah. Modified from Dubiel (1983). 37 30° In the Lisbon Valley area, on a smaller scale, uranium deposits are on the flanks of a rising salt-cored anticline (Wood, 1968; Fisher, 1972). In addition to transitional litho- facies, uranium deposits are closely associated with black, organic-rich, lacustrine-marsh mudstone that was deposited where the water table was high (fig. 17) (Dubiel, 1983). Generally, the uranium deposits are below or, less com- monly, beside the lacustrine mudstone (R.F. Dubiel, oral commun., 1992). In addition, approximately 90 percent of uranium deposits in this area are within 3.2 km (2 mi) of the axis of a syndepositional syncline (fig. 17). In the Lisbon Valley area, a density-stratified interface was proposed as a mechanism for ore-deposit formation, based on the occur- rence of the deposits at a particular horizon within the Chinle Formation (Wood, 1968; Fischer, 1974; Huber, 1980). Minor uranium deposits and concentrations are in the Petri- fied Forest Member of the Chinle Formation. These are associated with thicker than normal sediments that are inter- preted to indicate greater subsidence and a high water table that helped preserve organic matter (Spirakis, 1980). 26 JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS ORIGIN OF TABULAR SANDSTONE URANIUM DEPOSITS In summary, the tabular shape, transitional sedimentary lithofacies, syndepositional synclines, and reduced channel sandstone and mudstone associated with tabular sandstone uranium deposits in the Morrison and Chinle Formations on the Colorado Plateau suggest that deposits formed where decreasing topographic slope and topographic depressions caused shallow, relatively fresh ground water and deeper, more saline ground water to discharge and mix at a density- stratified interface. The next step is to integrate these conclu- sions with the other types of evidence for this ore-deposit type and to reconcile divergent opinions on diagenesis and ore formation. ROLE OF HUMATE The location of tabular sandstone uranium deposits in the San Juan Basin likely is closely controlled by the prior deposition of pore-filling amorphous organic matter that is generally thought to be humate precipitated as a gel from a humic-acid-bearing solution (Granger and others, 1961; Moench and Schlee, 1967; Granger, 1968; Nash, 1968; Schmidt-Collerus, 1969; Squyres, 1970, 1980; Haji-Vassil- iou and Kerr, 1973; Hatcher and others, 1986; Turner-Peter- son, Fishman, and others, 1986). Uranium-vanadium deposits elsewhere in the Morrison Formation are poor in pore-filling organic matter, but humate may have been essential in uranium ore-formation and subsequently destroyed (Hansley and Spirakis, 1992). Humate in uranium deposits in the Chinle Formation has been postulated to exist (Huber, 1980), but studies necessary to determine its pres- ence have not been performed. Thus, the role of humate is critical to the origin of the uranium deposits in the San Juan Basin and possibly elsewhere. Previous workers have suggested that humate came (1) from the overlying Upper Cretaceous Dakota Sandstone and moved downdip in the Westwater Canyon Member ("Dakota source hypothesis") (Granger and others, 1961), (2) from the host sandstone and moved laterally downdip ("internal source hypothesis") (Granger and others, 1961; Squyres, 1970, 1980), (3) from the mudstone beds overlying and interbedded with sandstone and moved downward and upward perpendicular to bedding or laterally outward ("lacustrine-humate model") (Turner-Peterson, 1985; Fish- man and Turner-Peterson, 1986; Turner-Peterson and Fish- man, 1986), or (4) from surface water flowing downdip during deposition of the host sands ("syngenetic source hypothesis") (Granger and others, 1961). None of these interpretations can be ruled out based on availability of organic material. Organic "trash" is available in the host sandstone, and the former presence of organic matter in mudstone is suggested by the fact that the mudstone is reduced today. Alteration patterns of iron-titanium oxide mineral and feldspar dissolution are interpreted as indicating downward flow from the mudflat of the Brushy Basin Mem- ber (Turner-Peterson, 1985; Fishman and Turner-Peterson, 1986; Turner-Peterson and Fishman, 1986); however, replot- ting of the iron-titanium oxide data shows that downdip flow was significant (fig. 11). The downdip flow only has to remain in the upper part of the Westwater Canyon Member, which is normal for a dilute ground water overlying a slightly more saline ground water. The lacustrine-humate model would predict symmetric humate layers above and below mudstone layers; however, lacustrine mudstone beds generally are above the host sandstone, and humate layers do not wrap around the ends of mudstone beds. The Dakota source hypothesis would require that humate move from shoreline marshes downdip and seaward, which would only allow humate to accumulate within a short distance downdip from the Late Cretaceous outcrop of the Westwater Canyon Member. Deposits far downdip in the Westwater Canyon Member are difficult to explain by this mechanism. Both the Dakota source hypothesis and lacustrine- humate model imply that humate was introduced signifi- cantly later than sedimentation. Paragenetic relations sug- gest that humate precipitation and associated iron-titanium oxide dissolution were early, but a variety of interpretations are possible (Adams and others, 1974; Adams and Saucier, 1981; Hansley, 1986). Channels that scoured into slightly older humate- and uranium-impregnated channels (Fitch, 1980; Squyres, 1980) suggest humate precipitation almost immediately after sedimentation; however, humate impreg- nation may have been later, controlled by permeability vari- ations. These types of evidence for a very early age favor the syngenetic source hypothesis. Spatial and chemical evidence tends to favor a synge- netic source in muds deposited very shortly after the host sand. Perhaps the strongest geologic evidence for a specific source is the close spatial association of organic-rich lacus- trine mudstone typically above the uranium deposits (F. Peterson, 1980; Peterson and Turner-Peterson, 1980; Dubiel, 1983; Turner-Peterson, 1985). In addition, analyses of modern dissolved organic carbon suggest that lakes and wetlands were a more likely source of humic acid than ground water. Natural ground water has low dissolved organic carbon (median concentration, 0.7 mg/L), almost equal to that of sea water (mean, 0.5 mg/L); in contrast, oli- gotrophic and eutrophic lakes have mean dissolved organic carbon of 2.2 and 12 mg/L, respectively, and bogs have a mean dissolved organic carbon of 33 mg/L (Thurman, 1985, p. 8ff). The low dissolved organic carbon values in ground water suggest that dissolved organic carbon cannot be trans- ported in large amounts for great distances and that nearby sources of dissolved organic carbon, especially lakes and bogs, are more likely. ORIGIN OF TABULAR SANDSTONE URANIUM DEPOSITS 27 ONE OR TWO SOLUTIONS? Some workers have suggested that tabular humate lay- ers formed by the action of one solution (Squyres, 1980; Turner-Peterson, 1985; Fishman and Turner-Peterson, 1986; Turner-Peterson and Fishman, 1986). According to one ver- sion of the lacustrine-humate model (Turner-Peterson, 1985; Fishman and Turner-Peterson, 1986; Turner-Peterson and Fishman, 1986), humic acid from pore water in mudstone is expelled vertically upward or downward into the adjacent sandstone. The expelled fluid dissolves iron, vanadium, and aluminum from detrital grains and clays. As these cations increase in concentration along the flow path, they cause humate to precipitate. Once precipitation begins, floccula- tion, van der Waals forces, "herding instinct," or "the affinity of organics for each other" (Turner-Peterson, 1985; Fishman and Turner-Peterson, 1986) is supposed to account for the tabular shape. This mechanism is unrealistic on spatial, chemical, and hydrologic grounds and is not supported by a plausible mechanism, modern analogs, or experimental evidence. Flocculation and van der Waals forces act on a very small scale, not over the thousands of meters typical of tabular sandstone uranium deposits. They result in small clumps because the forces act radially from numerous centers. How flocculation or van der Waals forces would yield an exten- sive, subhorizontal layer is not explained. Chemically, the proposed mechanism is implausible because the amount of water from compaction is too small to dissolve and transport the observed amounts of humate precipitated or cations leached. For example, humic acid in ground water is typi- cally 0.5-1.0 mg/L and locally as much as 15 mg/L (Thur- man, 1985); however, humate can constitute more than 5 percent of the rock in a uranium deposit (Levanthal, 1980). To achieve such high concentrations, a much higher water to rock ratio is required than can be provided by pore water from compaction (Hilpert, 1969). In addition, it is improba- ble that the pore water in the mudstone would be so different from that in the sandstone that humic acid would dissolve from the first and precipitate in the second. Hydrologically, the model neglects gravity-driven flow, which would have been orders of magnitude greater than compaction-driven flow. Flow in transmissive fluvial sandstone is normally par- allel with stratification, not perpendicular as depicted by the lacustrine-humate model. Downdip flow of one fluid would not result in the observed tabular layers but would probably create podlike or roll-shaped deposits as, for example, in the roll sandstone uranium deposits of Wyoming (Harshman, 1972). The proposed mechanism appears to be loosely based on the process of podzolization; however, podzolization occurs today at the ground surface under boreal forests in areas of high rainfall and ground-water recharge (Petersen, 1976; Mokma and Buurman, 1982; Birkeland, 1984, p. 120ff), an environment very different from the arid to semi- arid conditions during deposition of the Morrison Formation. Further, "M. Thurman, 1981, oral communication," is the only reference to experiments showing the formation of tab- ular humate layers from one solution. Thurman today states that there are no experiments showing the formation of tab- ular humate layers from one solution (Michael Thurman, U.S. Geological Survey, oral commun., 1992). On the other hand, field, theoretical, and laboratory evi- dence for formation of tabular humate layers at a saline water interface between two solutions is abundant. The subhori- zontal attitude, gently crosscutting relationships, and lack of small-scale lithologic control are strong evidence for a den- sity-stratified interface (Fischer, 1947, 1974; Shawe, 1955, 1962, 1966; Granger, 1968; Wood, 1968; Melvin, 1976; Sanford, 1982, 1990a, b; Granger and Santos, 1986; Northrop and Goldhaber, 1990). On theoretical grounds, a density-stratified interface provides a laterally extensive, subhorizontal site for humate deposition and other reactions. The solutes in the lower fluid also provide a mechanism for humate precipitation, as demonstrated by experimental and field studies (Swanson and Palacas, 1965; Hair and Bassett, 1973; Sholkovitz, 1976; Ortiz and others, 1980; Fox, 1983; Thurman, 1985, p. 26ff and 394ff). The maximum amount of humate removal is at salinities between 15,000 and 20,000 mg/L. Therefore, although models of uranium and humate precipitation based on a density-stratified interface are com- monly called "brine-interface" models, the term "brine" is misleading because the denser ground water need only be saline. As discussed above, topographic depressions associ- ated with the uranium deposits are favorable for discharge of saline ground water, and the abundant evaporitic rocks in the subsurface are likely sources for saline water. Modern ana- logs of mixing between discharging, saline ground water overlain by relatively fresh water are common (Counts, 1957; Toth, 1963; Freeze and Witherspoon, 1966; Gallaher and Price, 1966; Boswell and others, 1968; van Everdingen, 1971; Foreman and Sharp, 1981; Mono Basin Ecosystem Study Committee, 1987; Sharp, 1988; Swanson and others, 1988; Banner and others, 1989; Dutton and others, 1989; Huff, 1990; Fee and others, 1992; Herezeg and others, 1992; Hines and others, 1992; Long and others, 1992; Macumber, 1992; Strobel, 1992). Thus, the saline water interface or two- solution model is the only plausible explanation for the pre- cipitation of tabular humate layers. SEEPAGE, COMPACTION, AND DENSITY Seepage, compaction, and density have been proposed to account for the movement of humic-acid-bearing ground water from the presumed source (lacustrine mudstone) to the presumed site of humate deposition (adjacent sandstone). An early version of the lacustrine-humate model calls on down- ward seepage or compaction to transport humic acid from lacustrine mudstone to sites of deposition in adjacent sand- stone (Peterson and Turner-Peterson, 1980). Later versions 28 JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS of the model rely on compaction and density for moving humic acid (Turner-Peterson, 1985; Fishman and Turner- Peterson, 1986; Turner-Peterson and Fishman, 1986). All the proposed mechanisms have serious difficulties. Downward seepage is proposed as a mechanism; how- ever, no hydrologic reason is given for downward seepage as opposed to upward seepage. As discussed above, each of the regions of uranium ore formation in the Colorado Plateau is dominated by evidence for upward flow of ground water. Syndepositional synclines, decreases in slope, topographic depressions, lacustrine sediments, and reduced rocks suggest a high water table, preservation of organic matter in the sat- urated zone, and the discharge of ground water from local and regional flow systems. Under these conditions, seepage is normally upward into lakes and wetlands (Meyboom, 1967; Lissey, 1971; Stephenson, 1971; Winter, 1976, 1978, 1986, 1988; Friedman and others, 1982; Allison and Barnes, 1985; Spencer and others, 1985; Duffy and Al-Hassan, 1988; Winter and Woo, 1990). Transient, depression-focused recharge has not previously been mentioned as having a role in uranium ore formation. Compaction has been suggested as a significant mech- anism for downward or lateral flow of ground water in the Morrison Formation (Shawe, 1976; Wood, 1980; Adams and Saucier, 1981; Turner-Peterson, 1985; Adams, 1986; Hans- ley, 1986; Turner-Peterson and Fishman, 1986; Northrop and Goldhaber, 1990; Wanty and others, 1990). Compaction is unlikely, however, to cause flow in the proper direction, is a minor ground-water control compared to gravity-driven flow, and cannot account for significant humate transport or diagenetic alteration (Hilpert, 1969; Sanford, 19902, 1994). Compaction-driven flow is generally directed upward, espe- cially near the top of a thick section of sediments (Bethke, 1986), as was present below the Morrison Formation. In the center of a large alkaline-saline playa, downward flow is possible, but more as a result of density than compaction. The marginal mudflat of a large playa lake, which is inferred to have downward ground-water flow (Turner-Peterson, 1985; Fishman and Turner-Peterson, 1986; Turner-Peterson and Fishman, 1986), is the least likely area for downward flow. The mudflat is at a lower elevation than the surround- ing alluvial plain, so gravity-driven flow is directed upward and discharges in the mudflat. Ground water in the mudflat is less dense than that in the central part of the playa, so den- sity drives ground water from the central playa outward toward the mudflat, where it discharges, as shown for mod- ern examples (Allison and Barnes, 1985; Spencer and others, 1985; Duffy and Al-Hassan, 1988; Winter and Woo, 1990). The mudflat of a large playa lake can therefore be expected typically to have ground-water discharge. Second, simple calculations show that compaction is a minor influence as compared to gravity-driven flow (San- ford, 1990a, 1994). The Brushy Basin Member would con- tribute an average of 0.0035 mm/yr of ground-water flow from expulsion of pore water in 5 m.y., whereas gravity- driven flow would contribute approximately 15 mm/yr. The dominance of gravity-driven flow over compaction-driven flow in this type of basin is further demonstrated by more elaborate calculations (Bethke, 1986; Person and others, 1992; Garven and others, 1993). For example, Bethke (1986) estimated that maximum compaction-driven flow in the Illi- nois Basin is 2 mm/yr, whereas the maximum gravity-driven flow is 11,000 mm/year. Thus, available evidence indicates that gravity-driven flow in these settings exceeds compac- tion-driven flow by some two to four orders of magnitude. Finally, compaction cannot account for the observed humate accumulation and diagenetic alteration of detrital minerals. As discussed above, humate accumulation requires a higher water to rock ratio than can be achieved by expul- sion of pore water. Iron-titanium oxide minerals are locally 100 percent altered from the top to the base of the Westwater Canyon Member. It is doubtful that enough pore water would have been expelled from the Brushy Basin Member to displace all the pore water in the Westwater Canyon Mem- ber, much less cause the intense alteration. Again, a higher water to rock ratio would be required than was available from compaction of the Brushy Basin Member. Further, there is no petrographic evidence for compaction. In the playa facies of the Brushy Basin, relict shard textures are perfectly preserved by replacement minerals such as chalce- dony and clinoptilolite; in the mudflat facies, original shard textures are destroyed (Turner and Fishman, 1991). Density of ground water has also been suggested as a mechanism for downward movement of ground water (Turner-Peterson, 1985; Fishman and Turner-Peterson, 1986; Turner-Peterson and Fishman, 1986). The references cited by the authors of the lacustrine-humate model for downward flow of ground water pertain to playa lake water in the center of the playa where solutes are highly concen- trated and abundant evidence shows that dense brines can move downward. The lacustrine-humate model depends, however, on ground water descending in the mudflat, not in the central playa. An argument that demonstrates that ground water can descend in the central part of a playa does not dem- onstrate that it descends in the mudflat. As shown by diage- netic facies in the Brushy Basin Member, the mudflat has relatively fresh water, and the concentration of solutes increases toward the center of the lake. The density is there- fore less in the mudflat than in the playa center. All other fac- tors being equal, density would cause ground water to flow downward in the playa center, then outward, and finally upward in the mudflat. Modeling of modern environments confirms this pattern (Duffy and Al-Hassan, 1988). Another difficulty with the density argument is that it is incompatible with transport of humic acid. Humic acid pre- cipitates in concentrated solutions (Swanson and Palacas, 1965; Hair and Bassett, 1973; Sholkovitz, 1976; Ortiz and others, 1980; Fox, 1983; Thurman, 1985, p. 26ff and 394ff), a fact acknowledged by the authors of the lacustrine-humate ORIGIN OF TABULAR SANDSTONE URANIUM DEPOSITS 29 model. To rely on a dense solution to move humic-acid-bear- ing ground water is to contradict the assumed chemistry of the humic-acid-transporting solution, which is thought to be dilute and mildly alkaline. The solution cannot be both dense and dilute. A third difficulty is that a dense solution cannot explain the upward movement of ground water, which the authors of the lacustrine-humate model consider necessary for ore deposits in sandstone above mudstone, such as in the Jack- pile Sandstone Member and locally in the Westwater Can- yon Member. Much emphasis is placed on an apparent "mirror image" alteration pattern, where the intensity of alteration in sandstone decreases away from the Brushy Basin Member, both down into the Westwater Canyon Member and up into the Jackpile Member. This pattern can- not be explained, however, by density-driven flow. Although density-driven flow probably was a significant mechanism in certain situations, it is incompatible with the lacustrine- humate model. TRANSIENT, DEPRESSION-FOCUSED GROUND-WATER RECHARGE The above discussion raises two critical questions. If there was a saline water interface, why was humate only pre- cipitated in certain places along the interface? If humic acid came from lacustrine muds, how did it move downward, apparently against the prevailing hydrologic gradient? Transient, depression-focused recharge during periods of high runoff is a common phenomenon in modern wetland environments and appears to be the only solution to the dilemma (fig. 18). A stream that normally gains water from ground-water discharge may, during flood stage, lose water to the banks of the channel (Gallaher and Price, 1966; Speer and others, 1966; Fetter, 1988, p. 47; Jacobson and others, 1989). Ground-water flow, which normally is toward the channel, is reversed away from the channel. Similarly, dur- ing periods of high runoff, lakes that normally have ground- water discharge into them may recharge the ground-water system (Meyboom, 1967; Lissey, 1971; Winter, 1976, 1983, 1986; Wood and Petraitis, 1984; Allison and Barnes, 1985; Fetter, 1988, p. 45ff; Logan and Rudolph, 1992). Water- table "mounds" build up temporarily in the ground surround- ing the lake, and the direction of ground-water flow is reversed. More water infiltrates the ground around the tem- porarily deepened lakes than infiltrates the ground over the intervening hills. Thus, the wetland temporarily becomes an area of ground-water recharge. The time period between such events may range from days to decades. For the ancient environment relevant to uranium ore formation, transient, depression-focused recharge in wet- lands can explain the apparent downward transport of humic matter in zones of normally upward discharge (fig. 18). Dis- solved organic matter probably was concentrated in the bottom water of lakes and in the pore water of lacustrine and fluvial sediments that were deposited at topographic depres- sions. A perennially high water table, which favors the pres- ervation of organic matter, accounts for the observed reduced rocks. During baseline or steady-state flow, ground water discharged upward into the channel or lake sediments. Depending on the prevailing climate, discharge consisted of saline ground water (fig. 184, baseline stage-dry) or saline ground water overlain by fresher ground water (fig. 18B, baseline stage-wet). During periods of high water, surface water probably rose and ponded in topographic depressions. The organic-acid-bearing pore water in the sediments may have been flushed downward and outward along the chan- nel. Judging from analyses of modern water (Thurman, 1985), dissolved organic carbon in the descending pore water may have been 2-33 mg/L, in contrast to 0.5-1.0 mg/L in the underlying saline ground water. The relatively fresh water formed a lens that spread out on top of the under- lying saline ground water. The underlying saline fluid was displaced slightly downward owing to the increased pressure of the relatively fresh water on top. The fresh-water lens occupied the upper part of the zone formerly occupied by saline ground water. The less soluble, humic part of the dis- solved organic carbon precipitated in the mixing zone. The localization of humate and uranium only at certain places along the interface can be explained by the localiza- tion of humate sources and depression-focused recharge. The humic-acid-rich water descended from specific loca- tions on the surface where organic-rich sediments preferen- tially accumulated and transient recharge was focused; that is, at topographic depressions. Wetlands may have been the only sites of appreciable accumulation of organic matter because grasses had not yet evolved and the major low- growing plants were ferns and related vegetation that require an abundant and constant water supply. Further, topographic depressions were favorable for accumulation of detrital organic matter, which is widely associated with the uranium deposits. Although an interface between shallow fresher water and deeper more saline water may have been wide- spread, precipitation of humate apparently only occurred at the interface hydrologically downgradient from sources of humic acid at local topographic depressions. Ground-water flow downdip along the channels would account for the commonly observed elongation of deposits. Fluctuations in the position of the interface due to intermittent, seasonal, climatic, and tectonic variations can explain the thickness and multiple positions of uranium- humate bodies. Small fluctuations in the position of the interface (on the scale of meters) can explain thickness variations by means of dispersion. During periods of high water, shallow, relatively fresh ground water displaced saline water below. In the zone of displacement, the humic acids would have been transported in the interconnected pores, whereas saline water would have remained in the less connected pores. Humate precipitation may have taken 30 JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS BASELINE STAGE-DRY Lake or channel BASELINE STAGE-WET Lake or channel B FLOOD STAGE OR WET PERIOD Lake Water-table or mound channel ST, EXPLANATION - Saline water - Organic-rich mud | Fresh water D Unsaturated zone Figure 18. - Schematic cross section across a topographic depression occupied by a channel or lake showing relationships of hydrology and humate. Ground water normally discharges at topographic depressions, but, during high-water conditions, recharge of relatively fresh water can take place. Gravity-driven, transient, downward-flowing ground water carries humic acids from wetlands, and humate precipitates in an interface zone of mixing between relatively fresh water and displaced saline water. A, Possible baseline or steady-state discharge of saline ground water under dry conditions, no fresh-water discharge. B, Possible baseline or steady-state discharge under wetter conditions, rela- tively fresh and saline ground water discharge. Modified from Dutton and others (1989). C, Transient recharge during high-water stage or period. Modified from Winter (1976, 1986) and Fetter (1988, p. 47). Humate layer SUMMARY AND CONCLUSIONS 31 place by dispersive mixing of humic-acid-bearing ground water in the interconnected pores with saline ground water in the less connected pores. Multiple positions of tabular humate bodies are explained by larger scale variations of the interface, which may have been controlled by longer term fluctuations of climate or by tectonic adjustments that affected the steady-state or baseline position of the water table. Because humate can be redissolved and reprecipi- tated by dilute water (Swanson and Palacas, 1965), down- ward movement of the interface could have moved the humate downward, but upward movement of the interface would have left the humate in place. Thus, a single layer of humate could be concentrated by repeated downward excursions of the interface. Whereas compaction is a one-time occurrence involv- ing small amounts of ground water, transient flushing can involve large ground-water fluxes and thus account for intense alteration. The widespread alteration of iron-tita- nium oxide minerals probably can only be explained by many pore volumes passing through the sediments. The pat- tern of dissolution of iron-titanium oxide minerals in the San Juan Basin and the elongation of tabular uranium bodies along the channels throughout the Colorado Plateau are con- sistent with a combination of transient, depression-focused recharge and lateral flow down the channels. Variability in precipitation is necessary for reversals in ground-water flow. In Triassic and Jurassic times, precipita- tion fluctuated in time periods that varied from single-event storms to seasonal and longer term climatic cycles. Paleocli- matic, sedimentologic, and, paleontologic evidence (Dubiel, 1989; Dubiel and others, 1991) indicates a monsoonal cli- mate in the Late Triassic. Generally abundant water supply was punctuated by periods of dry conditions. A generally drier climate prevailed during the Late Jurassic (Peterson and Turner-Peterson, 1987). Lithofacies in the Brushy Basin Member are similar to those of the Lake Eyre region of cen- tral Australia, where major flooding occurs every 5-10 years (Bell, 1986). Conglomeratic layers in the otherwise fine- grained Brushy Basin Member (Phoenix, 1958) are evidence for episodic flooding. During both Late Triassic and Late Jurassic time, variegated red and green mudstone suggests fluctuating water tables. SUMMARY AND CONCLUSIONS Reconstruction of ground-water-flow directions in Jurassic and Triassic fluvial-lacustrine rocks of the Colorado Plateau strengthens the conclusions of previous studies (San- ford, 1982, 1990b, 1992, 1994) that tabular sandstone ura- nium deposits formed in zones of inferred regional ground- water discharge. On the scale of the Colorado Plateau, the deposits are commonly associated with transitional lithofa- cies that indicate topographic controls on ground-water flow. Commonly arcuate zones of transitional lithofacies indicate a decrease in paleotopographic slope and favor discharge and mixing of local and regional ground-water-flow systems in wetland environments. For example, such facies changes in the Salt Wash Sandstone Member of the Jurassic Morrison Formation include the sandstone-mudstone facies, transi- tional between conglomeratic sandstone facies and clay- stone-limestone facies (Craig and others, 1955), and the fluvial sheet-sand facies, transitional between the fluvial sheet gravels and mudflat-lake muds (Peterson and Tyler, 1985). In the Triassic Chinle Formation, decreasing topo- graphic slope is indicated by a transition from distal braided stream to floodplain deposits (Finch, 1959; Young, 1964; Malan, 1968; Lupe, 1977). Tabular sandstone uranium deposits are closely associated with these facies changes, typically where the distal edge of the alluvial plain merges into mudflat. Measurable parameters that indicate a decrease in slope on a regional scale include an increase from low- to high-sinuosity channels, a decrease in thickness of sand- stone, a decrease in sandstone as a percent of total thickness, and an increase in mudstone thickness and percentage (Craig and others, 1955; Mullens and Freeman, 1957; Peterson and Tyler, 1985). On a more local scale, the deposits are com- monly associated with syndepositional synclines, possibly controlled by basement faulting. Such structural control has been inferred for the Henry Basin (Peterson, 1984, 1986), Uravan mineral belt, Grants uranium region (Kirk and Con- don, 1986), Jackpile Sandstone Member of the Morrison (Moench and Schlee, 1967), and Chinle strata (Spirakis, 1980; Dubiel, 1983). The topographic depressions associ- ated with these structural downwarps were favorable for ground-water discharge and wetlands. Sedimentologic evi- dence for structural downwarps includes, in the Henry Basin, thicker sediments, more channel sandstone, higher sinuosity channels, increased upper flow regime horizontal lamina- tions, and lacustrine mudstone (Peterson, 1984, 1986); in the Uravan mineral belt, distributary channels, thick host sand- stone, higher sandstone to mudstone ratios, lenticular rather than flatbedded sandstone, scour-and-fill bedding, and over- lying conglomeratic sandstone (Weir, 1952; Phoenix, 1958; Shawe, 1962; Motica, 1968; Thamm and others, 1981); and, in the Grants uranium region, locally thick sediments, high sandstone to mudstone ratios, and smaller numbers of sand- stone-mudstone interbeds (Kirk and Condon, 1986). Lacus- trine mudstone is associated with uranium deposits in the Henry Basin (Peterson, 1984, 1986), Grants uranium region (Turner-Peterson, 1985; Turner-Peterson and Fishman, 1986), and Chinle strata (Dubiel, 1983). As evidence for closed topographic depressions, lacustrine deposits strongly suggest low hydraulic potential and ground-water discharge during sedimentation. The only generally consistent association between ura- nium deposits and lithofacies is between lithofacies in the host sandstone and uranium deposits. Lithofacies in overly- ing units are not consistently associated with uranium 32 JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS deposits. For example, a proposed association between tabu- lar sandstone uranium deposits and the mudflat facies of the Brushy Basin Member of the Morrison Formation (Turner- Peterson and Fishman, 1986) breaks down for deposits in the Uravan mineral belt and Henry Basin, where uranium deposits are associated with analcime-potassium feldspar and alluvial-plain facies of the Brushy Basin Member, respectively, not the mudflat facies (fig. 13). Other workers noted the relationship between tabular sandstone uranium deposits and lithofacies of the host sandstone but failed to explain the relationship. For example, Craig and others (1955) noted that deposits cluster where the Salt Wash Sandstone Member of the Morrison is more than 73 m (240 ft) thick and stream channel sandstone makes up 40-55 per- cent of the thickness of the member; the deposits are between 27 and 60 m (90 and 200 ft) thick and have a per- centage mean deviation of between 5 and 18 percent. They noted that the permeability of these rocks is intermediate between more sandstone rich facies and more mudstone rich facies, and they hypothesized that intermediate flow rates were optimum for uranium and vanadium supply. I suggest that the intermediate lithologic facies represents paleotopo- graphic conditions that are hydrologically favorable for humate and uranium deposition. The facies changes are con- comitant with hydrologic changes that controlled the trans- port and deposition of organic matter and uranium. The scale of the observation and the influence of local tectonic structures must be accounted for in interpreting the paleotopographic slope. On the scale of the Colorado Pla- teau, sedimentary facies, total thickness, sandstone to mud- stone ratio, and other parameters change systematically in the direction of flow. For example, facies of the Salt Wash Sandstone Member of the Morrison Formation change from conglomeratic sandstone to sandstone and mudstone to clay- stone (Craig and others, 1955) or from fluvial sheet gravels to fluvial sheet sands to mudflat and lake muds (Peterson and Tyler, 1985). In addition, there are regional changes such as decrease in total thickness and sandstone to mudstone ratio. At this regional scale, coarser facies, thicker sandstone, and higher sandstone to mudstone ratio suggest greater topo- graphic slope closer to the source. Local tectonic activity can, however, significantly affect these overall trends (Peter- son, 1984; Kirk and Condon, 1986). On a more local scale, for example in the Henry Basin, local tectonic activity affected stream gradients, and the deposition of more chan- nel sandstone in synclines may have resulted from combing back and forth of streams and winnowing out of finer mate- rial (Peterson, 1984). At this scale, coarser facies, greater thickness, and higher sandstone to mudstone ratio are asso- ciated with streams that were slightly lower in elevation than the surrounding overbank deposits. Geologic controls on ground-water flow favored dis- charge of generally northeast flowing ground water every- where in the Morrison Formation and especially in arcuate zones of abrupt aquifer thinning. Thinning and pinching out of the Lower Jurassic and upper Paleozoic aquifer systems favored geologically controlled discharge. Within these aquifer systems, northward and eastward thinning of the Navajo Sandstone, Wingate Sandstone, Cedar Mesa Sand- stone Member of the Cutler Formation, De Chelly Sand- stone, and Meseta Blanca Sandstone Member of the Yeso Formation most favored deep ground-water discharge. Thin- ning of the section over the Uncompahgre and San Luis Pre- cambrian basement blocks probably contributed to deep ground-water discharge. For deposits in the Chinle Forma- tion, westward and northward flow was in the direction of a thinner Paleozoic section but a thicker eolian aquifer, and the net effect of geologic controls is unclear. Within the major discharge zones where topographic and geologic controls generally favored local and regional ground-water discharge, specific areas of discharge are marked by reduced rocks such as green and gray mudstone. A close association between reduced mudstone and channel sandstone and tabular sandstone uranium deposits has been noted for the Henry Basin (Peterson, 1984, 1986), Paradox Basin (McKay, 1955), San Juan Basin (Turner-Peterson, 1985; Turner-Peterson and Fishman, 1986), and Chinle For- mation (Wood, 1968; Dubiel, 1983). Reduced rocks in dom- inantly redbed sequences suggest perennially high water table conditions (Walker, 1967; Reading, 1978, p. 48-49; Dodson and others, 1980; Dubiel, 1983, 1989; Davis, 1988; Ghiorse and Wilson, 1988; Dubiel and others, 1991) and reducing conditions owing to bacterial degradation of organic matter, both of which are most likely in areas of perennial ground-water discharge. The composition of the discharging ground water may be inferred from the types of rocks through which the ground water passed. Shallow, local ground water was probably dilute meteoric water. As it passed downdip through the aquifer, it reacted with detrital material and any bacterially degraded organic matter. Judging from modern ground water in shallow aquifers in arid environments, the water was prob- ably a dilute Nat-Ca"*-HCO;3;- type (Hanshaw and Hill, 1969; Thackston and others, 1981; Davis, 1988; Sanford, 1990a). Recharging, depression-focused ground water was probably relatively fresh, neutral to slightly acidic, reducing, and humic acid rich. In contrast to the dilute, local ground water, deeper, regional ground water was probably saline owing to the widespread presence of evaporites and evaporitic clastic rocks. Shallow source rocks included the Curtis Formation, Todilto Limestone Member of the Wanakah Formation, and Tidwell Member of the Morrison Formation. Deeper sources included gypsum beds in the Triassic marginal- marine Moenkopi Formation. Still deeper Pennsylvanian and Permian sources of saline water were especially abun- dant and included all or parts of the Paradox Formation, SUMMARY AND CONCLUSIONS 33 Kaibab Limestone, Cutler Formation, and Yeso Formation. Dissolution of salts and migration of saline water has mod- ern-day analogs. For example, Carmel Formation evaporites are dissolved by descending relatively fresh water, which then becomes saline; this saline water moves downdip in the underlying Navajo Sandstone and finally discharges upward from the Navajo (Taylor and Hood, 1988). Modern ground water in the Colorado Plateau today is locally highly saline owing to dissolution of evaporites (Mayhew and Heyman, 1965; Hanshaw and Hill, 1969; Thackston and others, 1981; Sanford, 1990a). Evidence for an interface between regional saline and local fresher ground water is shown by tabular layers of humate, uranium, and dolomite in the Colorado Plateau (Fischer, 1947, 1974; Shawe, 1955, 1962, 1966; Granger, 1968; Wood, 1968; Melvin, 1976; Sanford, 1982, 1990a, b; Granger and Santos, 1986; Northrop and Gold- haber, 1990) and by experimental evidence and modern analogs, as discussed above. Many uranium-humate and uranium-vanadium deposits are elongated along the chan- nels and rise stratigraphically toward the basin (Reinhardt, 1963; Moench and Schlee, 1967; Granger, 1968; Hilpert, 1969; Fischer, 1974; Melvin, 1976; Northrop and Gold- haber, 1990), which would be expected at an interface where an upper dilute ground water flowed over a lower brine and discharged. Thus, abundant evidence exists for an association between tabular sandstone uranium deposits and regional ground-water discharge during ore formation. Further, the discharging ground water included dilute local and saline regional fluids that probably interacted at a density-stabi- lized interface. The fact that a consistent set of characteristics is asso- ciated with uranium deposits in Jurassic and Triassic depos- its throughout the Colorado Plateau indicates that the deposit type can be described by one general genetic model. The recurrent deposit characteristics including geometry, channel sandstone host rock, elongation and stratigraphic rise in the direction of paleoflow, association with reduced rocks, and presence in structurally controlled paleotopographic depressions suggest common physical and chemical controls in which wetlands played a critical role. Humate-rich and vanadium-rich deposits probably are variations on a general theme rather than distinct deposit types (Sanford, 1992). Chemical changes among these vari- ations can be attributed to post-ore diagenesis (Hansley and Spirakis, 1992). The conclusion that tabular sandstone uranium deposits are closely associated with a particular hydrologic environ- ment implies that the ore, or at least the humate, was depos- ited very soon after deposition of the sediments, perhaps after only meters or tens of meters of burial. Many prior esti- mates could narrow the age only to "shortly after deposi- tion" of the host sediments, which could mean anything from days to tens of millions of years. For example, ore deposits in the Westwater Canyon Member of the Morrison Formation are thought to have formed during compaction of the overlying Brushy Basin Member (Turner-Peterson, 1985; Turner-Peterson and Fishman, 1986). The present analysis suggests that ore formed even earlier. In a zone of regional - ground-water discharge, - depression-focused recharge would only displace the uppermost ground water. Humic acid from wetlands would only be transported to shallow depths. Once the humate layer was deposited, ura- nium may have accumulated later as a result of the reducing conditions generated by bacterial degradation of humate. A very early age for humate precipitation is supported by channel scours into humate-impregnated sandstone (Fitch, 1980; Squyres, 1980). The associations among channel sandstone, transi- tional lithofacies, decrease in topographic slope, thinning of aquifers, - syndepositional - synclines, paleotopographic depressions, interbedded sandstone and mudstone, underly- ing marine rocks especially evaporites, reduced mudstone, and tabular sandstone uranium deposits are so significant that any model of uranium deposition must account for them. The only unifying phenomenon yet proposed that is consistent with these observations is the regional discharge of deep gravity-driven ground water that mixed with shal- low dilute ground water during and shortly after deposition of the host sediments in a wetland environment. Depres- sion-focused transient recharge can explain the apparent downward transport of humic matter in areas of normally discharging, saline, regional ground water. The present analysis suggests that exploration for new districts and belts of tabular sandstone uranium deposits should be guided by features that indicate areas of regional ground-water discharge and transient, depression-focused, local recharge during and shortly after sedimentation. Major uranium belts can be expected where coarse- grained, alluvial-plain facies merge into finer grained, lower gradient facies. Areas where aquifers also thin in the direction of flow are still more favorable. Within these broad areas, local structural, sedimentologic, and diage- netic features such as syndepositional synclines, fluvial channels, and reducing conditions that indicate ancient wetlands are most favorable. _ with - Warren - I. Finch, Harry C. Granger, Paula L. Hansley, Fred Peterson, and Charles S. Spirakis were constructive. Thomas E. Bell, Steven M. Condon, Russell F. Dubiel, Christopher J. Duffy, Warren I. Finch, Arthur L. Geldon, Harry C. Granger, Steven E. Ingebritson, and Charles S. Spirakis reviewed the paper. I, of course, take sole responsibility for the interpre- tations expressed. 34 JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS REFERENCES CITED Adams, S.S., Curtis, H.S., and Hafen, P.L., 1974, Alteration of detrital magnetite-ilmenite in continental sandstones of the Morrison Formation, New Mexico, in Formation of uranium deposits: Vienna, International Atomic Energy Agency, p. 219-253. Adams, S.S., Curtis, H.S., Hafen, P.L., and Salek-Nejad, Hossein, 1978, Interpretation of postdepositional processes related to the formation and destruction of the Jackpile-Paguate uranium deposit, northwest New Mexico: Economic Geology, v. 73, p. 1635-1654. Adams, S.S., and Saucier, A.E., 1981, Geology and recognition cri- teria for uraniferous humate deposits, Grants uranium region, New Mexico: U.S. Department of Energy Report GJBX-2(81), 225 p. Allison, G.B., and Barnes, C.J., 1985, Estimation of evaporation from the normally "dry" Lake Frome in South Australia: Jour- nal of Hydrology, v. 78, p. 229-242. Anderson, R.Y., and Kirkland, D.W., 1980, Dissolution of salt deposits by brine density flow: Geology, v. 8, p. 66-69. Baars, D.L., 1962, Permian System of Colorado Plateau: American Association of Petroleum Geologists Bulletin, v. 46, p. 149-218. Baars, D.L., and Stevenson, G.M., 1977, Permian rocks of the San Juan Basin: New Mexico Geological Society Field Confer- ence, 28th, San Juan Basin III, Guidebook, p. 133-138. Banner, J.L., Wasserburg, G.J., Dobson, P.K., Carpenter, A.B., and Moore, C.H., 1989, Isotopic and trace element constraints on the origin and evolution of saline groundwaters from central Missouri: Geochimica et Cosmochimica Acta, v. 53, p. 383-398. Bell, TE., 1983, Deposition and diagenesis of the Brushy Basin and upper Westwater Canyon Members of the Morrison Formation in northwest New Mexico and its relationship to uranium min- eralization: Berkeley, University of California, Ph.D. thesis, 101 p. 1986, Deposition and diagenesis of the Brushy Basin Mem- ber and upper part of the Westwater Canyon Member of the Morrison Formation, San Juan Basin, New Mexico, in Turner- Peterson, C.E., Santos, E.S., and Fishman, N.S., eds., A basin analysis case study-The Morrison Formation, Grants urani- um region, New Mexico: American Association of Petroleum Geologists Studies in Geology 22, p. 77-91. Bethke, C.M., 1986, Hydrologic constraints on the genesis of the upper Mississippi valley mineral district from Illinois Basin brines: Economic Geology, v. 81, p. 233-249. 1989, Modeling subsurface flow in sedimentary basins: Geologische Rundschau, v. 78, p. 129-154. Billodeau, W.L., 1986, The Mesozoic Mogollon highlands, Arizo- na-An early Cretaceous rift shoulder: Journal of Geology, v. 94, p. 724-735. Birkeland, P.W., Soils and geomorphology: New York, Oxford University Press, 372 p. Blakey, R.C., 1974, Stratigraphic and depositional analysis of the Moenkopi Formation, southeastern Utah: Utah Geological and Mineralogical Survey Bulletin, v. 104, 81 p. 1980, Pennsylvanian and Early Permian paleogeography, southern Colorado Plateau and vicinity, in Fouch, TD., and Magathan, E.R., eds., Paleozoic paleogeography of the west- central United States: Rocky Mountain Section, Society of Economic Paleontologists and Mineralogists, p. 239-257. Blakey, R.C., and Gubitosa, Richard, 1983, Late Triassic paleo- geography and depositional history of the Chinle Formation, southern Utah and northern Arizona, in Reynolds, M.W., and Dolly, E.D., eds., Mesozoic paleogeography of the west-cen- tral United States: Society of Economic Paleontologists and Mineralogists, Rocky Mountain Section, Rocky Mountain Paleogeography Symposium, v. 2, p. 57-77. Blakey, R.C., Peterson, Fred, and Kocurek, Gary, 1988, Synthesis of late Paleozoic and Mesozoic eolian deposits of the western interior of the United States: Sedimentary Geology, v. 56, p. 3-125. Boswell, E.H., Cushing, EM., and Hosman, RL., 1968, Quaterna- ry aquifers in the Mississippi embayment: U.S. Geological Survey Professional Paper 448-E, 15 p. Bredehoeft, J.D., Blyth, C.R., White, W.A., and Maxey, G.B., 1963, Possible mechanism for concentration of brines in sub- surface formations: Geological Society of America Bulletin, v. 47, p. 257-269. Breit, G.N., Goldhaber, M.B., Shawe, D.R., and Simmons, E.C., 1990, Authigenic barite as an indicator of fluid movement through sandstones within the Colorado Plateau: Journal of Sedimentary Petrology, v. 60, p. 884-896. Chenoweth, W.L., and Malan, R.C., 1973, The uranium deposits of northeastern Arizona: New Mexico Geological Society Guide- book 24, p. 139-149. Chenoweth, W.L., and McLemore, V.T., 1989, Uranium resources on the Colorado Plateau, in Lorenz, J.C., and Lucas, S.G., eds., Energy frontier in the Rockies: Albuquerque Geological Soci- ety, p. 153-165. Cherkauer, D.S., and Nader, D.C., 1989, Distribution of groundwa- ter seepage to large surface-water bodies-The effect of hydraulic heterogeneities: Journal of Hydrology, v. 109, p. 151-165. Condon, S.M., and Huffman, A.C., 1984, Stratigraphy and deposi- tional environments of Jurassic rocks, San Juan Basin, New Mexico, with emphasis on the south and west sides, in Brew, D.C., ed., Field trip guidebook: Geological Society of Ameri- ca, Rocky Mountain Section, Annual Meeting, 37th, p. 93-107. Condon, S.M., and Peterson, Fred, 1986, Stratigraphy of middle and upper Jurassic rocks of the San Juan Basin-Historical perspective, current ideas, and remaining problems, in Turner- Peterson, C.E., Santos, E.S., and Fishman, N.S., eds., A basin analysis case study-The Morrison Formation, Grants urani- um region, New Mexico: American Association of Petroleum Geologists Studies in Geology 22, p. 7-26. Cooper, H.H., Jr., Kahout, F.A., Henry, H.R., and Glover, RE., 1964, Sea water in coastal aquifers: U.S. Geological Survey Water-Supply Paper 1613-C, 84 p. Coplen, TB., and Hanshaw, B.B., 1973, Ultrafiltration by a com- pacted clay membrane-I, Oxygen and hydrogen isotopic frac- tionation: Geochimica et Cosmochimica Acta, v. 37, p. 2295-2310. Counts, H.B., 1957, Ground-water resources of parts of Lonoke, Prairie, and White Counties, Arkansas: Arkansas Geological and Conservation Commission Water Resources Circular 5, 58 p. REFERENCES CITED 35 Craig, L.C., Holmes, C.N., Cadigan, R.A., Freeman, V.L., Mullens, T.E., and Weir, G.W., 1955, Stratigraphy of the Morrison and related formations, Colorado Plateau region, a preliminary report: U.S. Geological Survey Bulletin 1009-E, p. 125-168. Davis, G.H., 1988, Western alluvial valleys and the high plains, in Back, William, Rosenshein, J.S., and Seaber, PR., eds., Hydrogeology: Geological Society of America, The Geology of North America, v. O-2, p. 283-300. Dodson, Peter, Behrensmeyer, A.K., Bakker, R.T., and Mcintosh, J.S., 1980, Taphonomy and paleoecology of the dinosaur beds of the Jurassic Morrison Formation: Paleobiology, v. 6, p. 208-232. Dubiel, RF., 1983, Sedimentology of the lower part of the Upper Triassic Chinle Formation and its relationship to uranium deposits, White Canyon area, southeastern Utah: U.S. Geolog- ical Survey Open-File Report 83-459, 48 p. 1987, Sedimentology of the Upper Triassic Chinle Forma- tion, southeastern Utah-Paleoclimatic implications: Journal of Arizona-Nevada Academy of Science, v. 22, p. 35-45. 1989, Depositional and climatic setting of the Upper Trias- sic Chinle Formation, Colorado Plateau, in Lucas, $.G., and Hunt, A.P., eds., Dawn of the age of dinosaurs in the American Southwest: Albuquerque, New Mexico Museum of Natural History, p. 171-187. Dubiel, RF., Good, S.C., and Parrish, J.M., 1989, Sedimentology and paleontology of the Upper Triassic Chinle Formation, Bed- rock, Colorado: The Mountain Geologist, v. 26, p. 113-126. Dubiel, RF., Parrish, J.T., Parrish, J.M., and Good, S.C., 1991, The Pangean megamonsoon-Evidence from the Upper Triassic Chinle Formation, Colorado Plateau: Palaios, v. 6, p. 347-370. Duffy, C.J., and Al-Hassan, S., 1988, Groundwater circulation in a closed desert basin-Topographic scaling and climatic fore- ing: Water Resources Research, v. 24, p. 1675-1688. Dutton, A.R., 1987, Origin of brine in the San Andres Formation, evaporite confining system, Texas Panhandle and eastern New Mexico: Geological Society of America Bulletin, v. 99, p. 103-112. Dutton, A.R., Richter, B.C., and Kreitler, CW., 1989, Brine dis- charge and salinization, Concho River watershed, west Texas: Ground Water, v. 27, p. 275-383. Eugster, HP., and Hardie, L.A., 1978, Saline lakes, in Lerman, A., ed., Lakes-Chemistry, geology, physics: New York, Springer-Verlag, p. 237-293. Fee, J.A., Gaudette, HE., Lyons, W.B., and Long, D.T., 1992, Rare-earth distribution in Lake Tyrrell groundwaters, Victoria, Australia: Chemical Geology, v. 96, p. 67-93. Fetter, C.W., 1988, Applied hydrogeology: Columbus, Ohio, Mer- rill Publishing, 592 p. Finch, WL., 1959, Geology of uranium deposits in Triassic rocks of the Colorado Plateau region: U.S. Geological Survey Bulletin 1074-D, p. 125-164. 1967, Geology of epigenetic uranium deposits in sandstone in the United States: U.S. Geological Survey Professional Paper 538, 121 p. 1991, Maps showing the distribution of uranium deposit clusters in the Colorado Plateau uranium province: U.S. Geo- logical Survey Miscellaneous Field Studies Map MF-2080. Finch, WI., and Davis, J.F., eds., 1985, Geological environments of sandstone-type uranium deposits: Vienna, International Atomic Energy Agency, IAEA-TECDOC-328, 408 p. Finch, W.L, and McLemore, V.T., 1989, Uranium geology and resources of the San Juan Basin, in Finch, WI., Huffman, A.C., Jr., and Fassett, J.E., eds., Coal, uranium, and oil and gas in Mesozoic rocks of the San Juan Basin-Anatomy of a giant energy-rich basin: American Geophysical Union, Field Trip Guidebook T120, p. 27-32. Fischer, RP., 1947, Deposits of vanadium-bearing sandstone, in Vanderwilt, J. W., ed., Mineral resources of Colorado: Denver, State of Colorado Mineral Resources Board, p. 451-456. 1974, Exploration guides to new uranium districts and belts: Economic Geology, v. 69, p. 362-376. Fisher, R.S., and Kreitler, C.W., 1987, Origin and evolution of deep-basin brines, Palo Duro Basin, Texas: Texas Bureau of Economic Geology, Report of Investigations 166, 33 p. Fishman, N.S., and Turner-Peterson, C.E., 1986, Cation scaveng- ing-An alternative to a brine for humic acid precipitation in tabular uranium ore, in Dean, W.E., ed., Organics and ore deposits: Denver Regional Exploration Geologists Society Symposium, 1985, Proceedings, p. 197-204. Fitch, D.C., 1980, Exploration for uranium deposits, Grants mineral belt, in Rautman, C. A., ed., Geology and mineral technology of the Grants uranium region 1979: New Mexico Bureau of Mines and Mineral Resources Memoir 38, p. 40-51. Foreman, TL., and Sharp, J.M., Jr., 1981, Hydraulic properties of a major alluvial aquifer-An isotropic, inhomogeneous sys- tem: Journal of Hydrology, v. 53, p. 247-268. Fox, LE., 1983, The removal of dissolved humic acid during estu- arine mixing: Estuarine, Coastal and Shelf Science, v. 16, p. 431-440. Freethey, G. W., and Cordy, G.E., 1991, Geohydrology of Mesozoic rocks in the Upper Colorado River Basin in Arizona, Colorado, New Mexico, Utah, and Wyoming, excluding the San Juan Basin: U.S. Geological Survey Professional Paper 1411-C, 118 p. Freeze, R.A., and Witherspoon, P.A., 1966, Theoretical analysis of regional groundwater flow-1, Analytical and numerical solu- tions to the mathematical model: Water Resources Research, v. 2, p. 641-656. 1967, Theoretical analysis of regional groundwater flow-2, Effect of water-table configuration and subsurface permeability variation: Water Resources Research, v. 3, p. 623-634. Friedman, Irving, Smith, G.L., and Matsuo, Sadao, 1982, Economic implication of the deuterium anomaly in the brine and salts in Searles Lake, California: Economic Geology, v. 77, p. 694-702. Gallaher, J.T., and Price, W.E., Jr., 1966, Hydrology of the alluvial deposits in the Ohio River valley in Kentucky: U.S. Geological Survey Water-Supply Paper 1818, 80 p. Galloway, W.E., 1979, Morrison Formation of the Colorado Pla- teau, in Galloway, W.E., Kreitler, C.W., and McGowen, J.H., eds., Depositional and ground-water flow systems in the explo- ration for uranium: Austin, Texas, Bureau of Economic Geol- ogy, p. 214-228. Garven, Grant, and Freeze, R.A., 1984a, Theoretical analysis of the role of groundwater flow in the genesis of stratabound ore deposits-1, Mathematical and numerical model: American Journal of Science, v. 284, p. 1085-1124. 36 JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS 1984b, Theoretical analysis of the role of groundwater flow in the genesis of stratabound ore deposits-2, Quantitative results: American Journal of Science, v. 284, p. 1125-1174. Garven, Grant, Ge, S., Person, M.A., and Sverjensky, D.A., 1993, Genesis of stratabound ore deposits in the midcontinent basins of North America-1, The role of regional groundwater flow: American Journal of Science, v. 293, p. 497-568. Geldon, A.L., in press, Hydrologic properties and flow systems of the Paleozoic rocks in the Upper Colorado River Basin, excluding the San Juan Basin, in Arizona, Colorado, Utah, and Wyoming: U.S. Geological Survey Professional Paper 1411-B. Ghiorse, W.C., and Wilson, J.T., 1988, Microbial ecology of the terrestrial subsurface, in Laskin, A.I., ed., Advances in applied microbiology: San Diego, Academic Press, p. 107-172. Graf, D.L., 1982, Chemical osmosis, reverse chemical osmosis, and the origin of subsurface brines: Geochimica et Cosmochimica Acta, v. 46, p. 1431-1448. Granger, H.C., 1968, Localization and control of uranium deposits in the southern San Juan Basin mineral belt, New Mexico-An hypothesis: U.S. Geological Survey Professional Paper 600-B, p. B60-B70. Granger, H.C., and Finch, W.I., 1988, The Colorado Plateau urani- um province, USA, in International Atomic Energy Agency, ed., Recognition of uranium provinces: Vienna, International Atomic Energy Agency, p. 157-193. Granger, H.C., and Santos, E.S., 1986, Geology and ore deposits of the Section 23 Mine, Ambrosia Lake District, New Mexico, in Turner-Peterson, C.E., Santos, E.S., and Fishman, N.S., eds., A basin analysis case study-The Morrison Formation, Grants uranium region, New Mexico: American Association of Petro- leum Geologists Studies in Geology 22, p. 185-210. Granger, H.C., Santos, E.S., Dean, B.G., and Moore, F.B, 1961, Sandstone-type uranium deposits at Ambrosia lake, New Mex- ico-An interim report: - Economic Geology, v. 56, p. 1179-1210. Habermeh1, M.A., 1980, The Great Artesian Basin, Australia: BMR Journal of Australian Geology and Geophysics, v. 5, p. 9-38. Hair, ME., and Bassett, C.R., 1973, Dissolved and particulate humic acids in an east coast estuary: Estuarine and Coastal Marine Science, v. 1, p. 107-111. Haji-Vassiliou, A., and Kerr, P.F., 1973, Analytic data on nature of urano-organic deposits: American Association of Petroleum Geologists Bulletin, v. 57, p. 1291-1296. Hammer, U.T., 1986, Saline lake ecosystems of the world: Dor- drecht, Dr W. Junk Publishers, 616 p. Hanor, J.S., 1983, Fifty years of thought on the origin and evolution of subsurface sedimentary brines, in Boardman, J.S., ed., Rev- olution in the earth sciences-Advances in the past half centu- ry: Dubuque, Iowa, Kendall/Hunt, p. 99-111. Hanshaw, B.B., and Hill, G.A., 1969, Geochemistry and hydrody- namics of the Paradox Basin region, Utah, Colorado and New Mexico: Chemical Geology, v. 4, p. 263-294. Hansley, P.L., 1986, Regional diagenetic trends and uranium min- eralization in the Morrison Formation across the Grants urani- um region, in Turner-Peterson, C.E., Santos, E.S., and Fishman, N.S., eds., A basin analysis case study-The Morri- son Formation, Grants uranium region, New Mexico: Ameri- can Association of Petroleum Geologists Studies in Geology 22, p. 277-301. 1990, Regional diagenesis of sandstones in the Upper Juras- sic Morrison Formation, San Juan Basin, New Mexico and Colorado-Geologic, chemical, and kinetic constraints: U.S. Geological Survey Bulletin 1808-H, 35 p. Hansley, P.L. and Spirakis, C.S., 1992, Organic matter diagenesis as the key to a unifying theory for the genesis of tabular urani- um-vanadium deposits in the Morrison Formation, Colorado Plateau: Economic Geology, v. 87, p. 352-365. Harshman, E.N., 1972, Geology and uranium deposits, Shirley Basin area, Wyoming: U.S. Geological Survey Professional Paper 745, p. 82. Hatcher, P.G., Spiker, E.C., Orem, W.H., Romankiw, L.A., Szever- enyi, N.M., and Maciel, G.E., 1986, Organic geochemical stud- ies of uranium-associated organic matter from the San Juan Basin-A new approach using solid-state 3C magnetic reso- nance, in Turner-Peterson, C.E., Santos, E.S., and Fishman, N.S., eds., A basin analysis case study-The Morrison Forma- tion, Grants uranium region, New Mexico: American Associa- tion of Petroleum Geologists Studies in Geology 22, p. 171-184. Herczeg, A.L., Barnes, C.J., Macumber, P.G., and Olley, J.M., 1992, A stable isotope investigation of groundwater-surface water interactions at Lake Tyrrell, Victoria, Australia: Chemi- cal Geology, v. 96, p. 19-32. Hilpert, L.S., 1969, Uranium resources of northwestern New Mex- ico: U.S. Geological Survey Professional Paper 603, 166 p. Hilpert, L.S., and Moench, R.H., 1960, Uranium deposits of the southern part of the San Juan Basin, New Mexico: Economic Geology, v. 55, p. 429-464. Hines, M.E., Lyons, W. B., Lent, R.M., and Long, D.T., 1992, Sed- imentary biogeochemistry of an acidic, saline groundwater dis- charge zone in Lake Tyrrell, Victoria, Australia: Chemical Geology, v. 96, p. 53-65. Hitchon, Brian, 1969, Fluid flow in the western Canada sedimenta- ry basin; 1, Effect of topography: Water Resources Research, v. 5, p. 186-195. Hubbert, M.K., 1940, The theory of ground-water motion: Journal of Geology, v. 48, p. 785-944. 1953, Entrapment of petroleum under hydrodynamic condi- tions: American Association of Petroleum Geologists, Bulle- tin, v. 37, p. 1954-2026. 1956, Darcy's law and the field equations of the flow of underground fluids: Petroleum Transaction, American Institute of Mining Engineers, v. 207, p. 222-239. Huber, G.C., 1980, Stratigraphy and uranium deposits, Lisbon Val- ley district, San Juan County, Utah: Quarterly of the Colorado School of Mines, v. 75, no. 2, 45 p. Huff, G.F., 1990, Extent and sources of saline water in a part of the Mississippi River alluvial aquifer in northeastern Louisiana and southeastern Arkansas: Geological Society of America Abstracts with Programs, v. 22, p. A65. Hunter, R.E., Gelfenbaum, Guy, and Rubin, D.M., 1992, Clastic pipes of probable solution-collapse origin in Jurassic rocks of the southern San Juan Basin, New Mexico: U.S. Geological Survey Bulletin 1808-L, 19 p. Ingebritsen, S.E., Mariner, RH., and Sherrod, D.R., in press, Hydrothermal systems of the Cascade Range, north-central Oregon: U.S. Geological Survey Professional Paper 1044-L. REFERENCES CITED 37 Jackson, RE., and Patterson, R.J., 1982, Interpretation of pH and Eh trends in a fluvial-sand aquifer system: Water Resources Research, v. 18, p. 1255-1268. Jacobson, G., Calf, G.E., Jankowski, J., and McDonald, P.S., 1989, Groundwater chemistry and paleorecharge in the Amadeus Basin, central Australia: Journal of Hydrology, v. 109, p. 237-266. Jobin, D.A., 1962, Relation of the transmissive character of the sedimentary rocks of the Colorado Plateau to the distribution of uranium deposits: U.S. Geological Survey Bulletin 1124, 151 p. Johnson, H.S., Jr., 1957, Uranium resources of the San Rafael dis- trict, Emery County, Utah-A regional synthesis: U.S. Geo- logical Survey Bulletin 1046-D, 54 p. Kelley, V.C., compiler, 1963, Geology and technology of the Grants uranium region: New Mexico Bureau of Mines and Mineral Resources Memoir 15, 277 p. Kirk, A.R., and Condon, S.M., 1986, Structural control of sedimen- tation patterns and the distribution of uranium deposits in the Westwater Canyon Member of the Morrison Formation, north- western New Mexico-A subsurface study, in Turner-Peter- son, C.E., Santos, E.S., and Fishman, N.S., eds., A basin analysis case study-The Morrison Formation, Grants urani- um region, New Mexico: American Association of Petroleum Geologists Studies in Geology 22, p. 105-143. Kreitler, C.W., 1989, Hydrogeology of sedimentary basins: Journal of Hydrology, v. 106, p. 29-53. Levanthal, J.S., 1980, Organic geochemistry and uranium in Grants mineral belt, in Rautman, C.A., ed., Geology and mineral tech- nology of the Grants uranium region 1979; New Mexico Bureau of Mines and Mineral Resources Memoir 38, p. 75-85. Lissey, A., 1971, Depression-focused transient groundwater flow patterns in Manitoba: Geological Association of Canada Spe- cial Paper 9, p. 333-341. Lockley, M.G., Houck, K.J., and Prince, N.K., 1986, North Ameri- ca's largest dinosaur trackway site-Implications for Morrison Formation paleoecology: Geological Society of America Bul- letin, v. 97, p. 1163-1176. Logan, W.S., and Rudolph, D.L., 1992, Microdepression-focused, fresh-water recharge into the saline groundwater of the Rio de la Plata coastal plain, Argentina: Geological Society of Amer- ica Abstracts with Programs, v. 24, p. A253. Long, D.T., Fegan, N.E., Lyons, W.B., Hines, M.E., Macumber, P.G., and Giblin, A.M., 1992, Geochemistry of acid brines: Lake Tyrrell, Victoria, Australia: Chemical Geology, v. 96, p. 33-52. Lupe, R., 1977, Depositional environments as a guide to uranium mineralization in the Chinle Formation, San Rafael Swell, Utah: U.S. Geological Survey Journal of Research, v. 5, p. 365-372. Macumber, P.G., 1992, Hydrological processes in the Tyrrell Basin, southeastern Australia: Chemical Geology, v. 96, p. 1-18. Magaritz, Mordeckai, and Luzier, J.E., 1985, Water-rock interac- tions and seawater-freshwater mixing effects in the coastal dunes aquifer, Coos Bay, Oregon: Geochimica et Cosmochim- ica Acta, v. 49, p. 2515-2525. Malan, R.C., 1968, The uranium mining industry and geology of the Monument Valley and White Canyon districts, Arizona and Utah, in Ridge, J.D., ed., Ore deposits of the United States, 1933-1967: New York, American Institute of Mining, Metal- lurgical, and Petroleum Engineers. Mallory, W.W., ed., 1972, Geologic atlas of the Rocky Mountain region: Denver, Rocky Mountain Association of Geologists, 331 p. Mayhew, E.J., and Heyman, E.B., 1965, Concentrated subsurface brines in the Moab region, Utah: Utah Geological and Miner- alogical Survey Special Studies 13, 28 p. McKay, E.J., 1955, Criteria for outlining areas favorable for urani- um deposits in parts of Colorado and Utah: U.S. Geological Survey Bulletin 1009-J, p. 261-282. McLean, J.S., 1970, Saline ground-water resources of the Tularosa Basin, New Mexico: U.S. Department of the Interior Research and Development Progress Report 561, 128 p. Melvin, J.W., 1976, Systematic distribution of large uranium deposits in the Grants uranium region, New Mexico: New Mexico Geological Society Special Publication 6, p. 144-150. Meyboom, P., 1967, Mass-transfer studies to determine the ground- water regime of permanent lakes in hummocky moraine of western Canada: Journal of Hydrology, v. 5, p. 117-142. Moench, R.H., and Schlee, J.S., 1967, Geology and uranium of the Laguna district, New Mexico: U.S. Geological Survey Profes- sional Paper 519, 117 p. Mokma, D.L., and Buurman, P., 1982, Podzols and podzolization in temperate regions: Wageningen, International Soil Museum Monograph 1, 126 p. Mono Basin Ecosystem Study Committee, 1987, The Mono Basin ecosystem: Washington, D.C., National Academy Press, 272 p. Motica, J.E., 1968, Geology and uranium-vanadium deposits in the Uravan mineral belt, southwestern Colorado, in Ridge, J.D., ed., Ore deposits of the United States, 1937-1967: New York, American Institute of Mining, Metallurgical, and Petroleum Engineers, p. 805-813. Mullens, TE., and Freeman, V.L, 1957, Lithofacies of the Salt Wash Member of the Morrison Formation, Colorado Plateau: Geological Society of America Bulletin, v. 68, p. 505-526. Munn, M.J., 1909, The anticlinal and hydraulic theories of oil and gas accumulation: Economic Geology, v. 4, p. 509-529. Nash, J.T., 1968, Uranium deposits in the Jackpile Sandstone, New Mexico: Economic Geology, v. 63, p. 737-750. Nash, J.T., Granger, H.C., and Adams, S$.S., 1981, Geology and concepts of genesis of important types of uranium deposits: Economic Geology, 75th Anniversary Volume, p. 63-116. Northrop, H.R., and Goldhaber, M.B., eds., 1990, Genesis of the tabular-type vanadium-uranium deposits of the Henry Basin: Economic Geology, v. 85, p. 215-269. Ortiz, N.V., Ferentchak, J.A., Ethridge, F.G., Granger, H.C., and Sunada, D.K., 1980, Ground-water flow and uranium in Colo- rado Plateau: Ground Water, v. 18, p. 596-606. Owen, D.E., Turner-Peterson, C.E., and Fishman, N.S., 1989, X- ray diffraction studies of the <0.5-um fraction from the Brushy Basin Member of the Upper Jurassic Morrison Formation, Col- orado Plateau: U.S. Geological Survey Bulletin 1808-G, 25 p. Owen, D.E., Walter, L.J., Jr., and Beck, R.G., 1984, The Jackpile Sandstone Member of the Morrison Formation in west-central New Mexico-A formal definition: New Mexico Geology, v. 6, p. 45-52. Person, Mark, Toupin, Denah, Wieck, Jim, and Eadington, Peter, 1992, Hydrologic constraints on the thermal evolution of the Cooper and Eromanga Basins, Australia-Implications for 38 JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS petroleum generation: American Geophysical Union, 1992 Fall Meeting Abstract Supplement, EOS, Transactions, v. 73, no. 43, p. 193. Petersen, Leif, 1976, Podzols and podzolization: Copenhagen, DSR Forlag, 293 p. Peterson, Fred, 1980, Sedimentology as a strategy for uranium exploration-Concepts gained from analysis of a uranium- bearing depositional sequence in the Morrison Formation of south-central Utah, in Turner-Peterson, C.E., ed., Uranium in sedimentary rocks-Application of the facies concept to explo- ration: Denver, Society of Economic Paleontologists and Min- eralogists, Rocky Mountain Section, p. 65-126. 1984, Fluvial sedimentation on a quivering craton-Influ- ence of slight crustal movements on fluvial processes, Upper Jurassic Morrison Formation, western Colorado Plateau: Sed- imentary Geology, v. 38, p. 21-49. 1986, Jurassic paleotectonics in the west-central part of the Colorado Plateau, Utah and Arizona, in Peterson, J.H., ed., Paleotectonics and sedimentation in the Rocky Mountain region, United States: American Association of Petroleum Geologists Memoir 41, p. 563-596. 1988b, Pennsylvanian to Jurassic eolian transportation sys- tems in the western United States: Sedimentary Geology, v. 56, p. 207-260. 1988a, Stratigraphy and nomenclature of Middle and Upper Jurassic Rocks, western Colorado Plateau, Utah and Arizona: U.S. Geological Survey Bulletin 1633-B, p. 13-56. in press, Paleogeography of a widespread continental sequence-The Morrison Formation and related beds of the Colorado Plateau and vicinity, in Gillette, D.D., ed., Chemis- try and preservation of dinosaur bones: New York, Springer- Verlag, 76 p. Peterson, Fred, and Turner-Peterson, C.E., 1980, Lacustrine- humate model-Sedimentologic and geochemical model for tabular sandstone uranium deposits in the Morrison Formation, Utah, and application to uranium exploration: U.S. Geological Survey Open-File Report 80-319, 48 p. 1987, The Morrison Formation of the Colorado Pla- teau-Recent advances in sedimentology, stratigraphy, and paleotectonics: Hunteria, v. 2, p. 1-18. Peterson, Fred, and Tyler, Noel, 1985, Field guide to the upper Salt Wash alluvial complex, in Flores, R.M., and Harvey, M.D., eds., Field guidebook to modern and ancient fluvial systems in the United States: International Fluvial Sedimentology Confer- ence, 3rd, Fort Collins, Colorado, Proceedings, p. 45-64. Peterson, J.A., 1980, Permian paleogeography and sedimentary provinces, west-central United States, in Fouch, TD., and Magathan, E.R., Paleozoic paleogeography of the west-central United States: Rocky Mountain Section, Society of Economic Paleontologists and Mineralogists, p. 271-292. Peterson, J. A., Loleit, A.J., Spencer, C.W., and Ullrich, R.A., 1965, Sedimentary history and economic geology of San Juan Basin: American Association of Petroleum Geologists Bulletin, v. 49, p. 2076-2119. Pfannkuch, H.O., and Winter, T.C., 1985, Effects of anisotropy and groundwater system geometry on seepage through lakebeds; 1, Analog and dimensional analysis: Journal of Hydrology, v. 75, p. 213-237. Phoenix, D.A., 1958, Uranium deposits under conglomeratic sand- stone of the Morrison Formation, Colorado and Utah: Geolog- ical Society of America Bulletin, v. 69, p. 403-418. Plummer, L.N., 1975, Mixing of sea water with calcium carbonate ground water, in Whitten, E.H.T., ed., Quantitative studies in geological sciences: Geological Society of America Memoir 142, p. 219-236. Randazzo, A.F., and Bloom, J.I., 1985, Mineralogical changes along the freshwater/saltwater interface of a modern aquifer: Sedimentary Geology, v. 43, p. 219-239. Rautman, C.A., compiler, 1980, Geology and mineral technology of the Grants uranium region 1979; New Mexico Bureau of Mines and Mineral Resources Memoir 38, 400 p. Reading, H.G., 1978, Sedimentary environments and facies: New York, Elsevier, 557 p. Reinhardt, E.V., 1963, The Uravan mineral belt-A review, and a few new theories on its origins: Engineering and Mining Jour- nal, v. 164, p. 107-110. Reynolds, RL., Fishman, N.S., Scott, J.H., and Hudson, MR., 1986, Iron-titanium oxide minerals and magnetic susceptibility anomalies in the Mariano Lake-Lake Valley cores-Con- straints on conditions of uranium mineralization in the Morri- son Formation, San Juan Basin, New Mexico, in Turner- Peterson, C.E., Santos, E.S., and Fishman, N.S., eds., A basin analysis case study-The Morrison Formation, Grants urani- um region, New Mexico: American Association of Petroleum Geologists Studies in Geology 22, p. 303-313. Ridgley, J.L., 1989, Trace fossils and mollusks from the upper member of the Wanakah Formation, Chama Basin, New Mex- ico-Evidence for a lacustrine origin: U.S. Geological Survey Bulletin 1808-C, 16 p. Robinove, C.J., Langford, R.H., and Brookhart, J. W., 1958, Saline- water resources of North Dakota: U.S. Geological Survey Water-Supply Paper 1428, 72 p. Runnels, D.D., 1969, Diagenesis, chemical sediments, and the mix- ing of natural waters: Journal of Sedimentary Petrology, v. 39, p. 1188-1201. Sanford, RF., 1982, Preliminary model of regional Mesozoic groundwater flow and uranium deposition in the Colorado Pla- teau: Geology, v. 10, p. 348-352. 1988, Regional paleohydrology, in Granger, H.C., and Finch, WI., eds., The Colorado Plateau uranium province, USA-Recognition of uranium provinces: Vienna, Interna- tional Atomic Energy Agency, p. 175-178. 1990a, Paleohydrogeology of the Colorado Plateau-Back- ground and conceptual models, in Spencer, R.J., and Chou, I., eds., Fluid-mineral interactions-A tribute to H.P. Eugster: Geochemical Society Special Publication 2, p. 285-311. 1990b, Hydrogeology of an ancient arid closed basin-Implications for tabular sandstone-hosted uranium deposits: Geology, v. 18, p. 1099-1102. 1992, A new model for tabular-type uranium deposits: Eco- nomic Geology, v. 87, p. 2041-2055. 1994, A quantitative model of ground-water flow during formation of tabular sandstone uranium deposits: Economic Geology, v. 89, p. 341-360. Schmidt-Collerus, J.J., 1969, Investigations of the relationship between organic matter and uranium deposits-Part II, Exper- imental investigations: U.S. Atomic Energy Commission Open-File Report GJO-9332, 192 p. REFERENCES CITED 39 Sharp, J.M., Jr., 1988, Alluvial aquifers along major rivers, in Back, William, Rosenshein, J.S., and Seaber, P.R., eds., Hydrogeolo- gy: Geological Society of America, The Geology of North America, v. O-2, p. 273-282. Shawe, D.R., 1955, Significance of roll ore bodies in genesis of ura- nium-vanadium deposits on the Colorado Plateau, in United Nations, ed., Geology of uranium and thorium: International Conference on Peaceful Uses of Atomic Energy, Geneva, Pro- ceedings, v. 6, p. 335-337. 1962, Localization of the Uravan mineral belt by sedimen- tation: U.S. Geological Survey Professional Paper 450-C, p. 1966, Zonal distribution of elements in some uranium-vana- dium toll and tabular ore bodies on the Colorado Plateau: U.S. Geological Survey Professional Paper 550-B, p. B169-B175. 1976, Sedimentary rock alteration in the Slick Rock district, San Miguel and Dolores Counties, Colorado: U.S. Geological Survey Professional Paper 576-D, 51 p. Shawe, D.R., Nash, J.T., and Chenoweth, W.L., 1991, Uranium and vanadium deposits, in Gluskoter, H.J., Rice, D.D., and Taylor, R.B., eds., Economic geology, U.S.: Geological Society of America, The Geology of North America, v. P-2, p. 103-124. Sholkovitz, E.R., 1976, Flocculation of dissolved organic and inor- ganic matter during the mixing of river water and seawater: Geochimica et Cosmochimica Acta, v. 40, p. 831-845. Smith, G.I., 1979, Subsurface stratigraphy and geochemistry of late Quaternary evaporites, Searles Lake, California: U.S. Geolog- ical Survey Professional Paper 1043, 130 p. Speer, P.R., Hines, M.S., Hanson, ME., and others, 1966, Low- flow characteristics of streams in the Mississippi embayment on northern Arkansas and in Missouri: U.S. Geological Survey Professional Paper 448-F, 25 p. Spencer, R.J., Eugster, HP., Jones, B.F., and Rettig, S.L., 1985, Geochemistry of Great Salt Lake, Utah; I, Hydrochemistry since 1850: Geochimica et Cosmochimica Acta, v. 49, p. 727-1737. Spirakis, C.S., 1980, A possible relationship between subsidence and uranium mineralization in the Petrified Forest Member of the Chinle Formation in the Cameron and Holbrook-St. Johns areas of Arizona: U.S. Geological Survey Open-File Report 90-808, 7 p. 1991, Genesis of the tabular-type vanadium-uranium deposits of the Henry Basin, Utah-A discussion: Economic Geology, v. 86, p. 1350-1353. Squyres, J.B., 1970, Origin and depositional environment of urani- um deposits, Grants region, New Mexico: Palo Alto, Stanford University, Ph.D. thesis, 228 p. 1980, Origin and significance of organic matter in uranium deposits of Morrison Formation, San Juan Basin, New Mexico, in Rautman, C.A., ed., Geology and mineral technology of the Grants uranium region 1979; New Mexico Bureau of Mines and Mineral Resources Memoir 38, p. 86-97. Stephenson, D.A., 1971, Groundwater flow system analysis in lake environments, with management and planning implications: Water Resources Bulletin, v. 7, p. 1038-1047. Stewart, J.H., Poole, F.G., and Wilson, RF., 1972, Stratigraphy and origin of the Chinle Formation and related Upper Triassic stra- ta in the Colorado Plateau region: U.S. Geological Survey Pro- fessional Paper 690, 336 p. Stone, W.J., Lyford, F.P., Frenzel, P.F., Mizell, N.H., and Padgett, ET., 1983, Hydrogeology and water resources of San Juan Basin, New Mexico: New Mexico Bureau of Mines and Min- eral Resources Hydrologic Report 6, 70 p. Straw, W.T., Kehew, A.E., Barrese, P.G., Steinmann, W.K., and Kasenow, MC., 1990, Flow systems in the Schoolcraft aqui- fer-A study of regional flow interrupted by local flow sys- tems: Geological Society of America Abstracts with Programs, v. 22, p. A64. Strobel, ML., 1992, Hydrogeological restrictions to saline ground- water discharge in the Red River of the north drainage basin, North Dakota: Geological Society of America Abstracts with Programs, v. 24, p. A337. Swanson, G.A., Winter, TC., Adomaitis, V.A., and LaBaugh, J. W., 1988, Chemical characteristics of prairie lakes in south-central North Dakota-Their potential for influencing use by fish and wildlife: U.S. Fish and Wildlife Service Fish and Wildlife Technical Report 18, 44 p. Swanson, V.E., and Palacas, J.G., 1965, Humate in coastal sands of northwest Florida: U.S. Geological Survey Bulletin 1214-B, 29 p. Taylor, O.J., and Hood, J.W., 1988, Region 3, Colorado Plateau and Wyoming Basin, in Back, William, Rosenshein, J.S., and Seaber, PR., eds., Hydrogeology: Geological Society of America, The Geology of North America, v. O-2, p. 51-58. Taylor, O.J., Hood, J.W., and Zimmerman, E.A., 1986, Hydrogeo- logic framework of the upper Colorado River basin-exclud- ing the San Juan basin-Colorado, Utah, Wyoming, and Arizona: U.S. Geological Survey Hydrogeologic Atlas HA-687. Thackston, J. W., McCulley, B.L., and Preslo, L.M., 1981, Ground- water circulation in the western Paradox Basin, Utah: Rocky Mountain Association of Geologists, Field Conference, p. 201-225. Thaden, RE., Trites, A.F., Jr., and Finnell, TL., 1964, Geology and ore deposits of the White Canyon area, San Juan and Garfield Counties, Utah: U.S. Geological Survey Bulletin 1125, 166 p. Thamm, J.K., Kovschak, A.A., Jr., Adams, S.S., 1981, Geology and recognition criteria for sandstone uranium deposits of the Salt Wash-type, Colorado Plateau province, final report: U.S. Department of Energy Report GJBX-6(81), 136 p. Thorman, C.H., Ketner, K.B., Brooks, W.E., Snee, L. W., and Zim- merman, R.A., 1991, Late Mesozoic-Cenozoic tectonics in northeastern Nevada, in Raines, G.L., Lisle, RE., Schafer, R.W., and Wilkinson, W.H., eds., Geology and ore deposits of the Great Basin: Reno, Geological Society of Nevada, p. 25-45. Thorman, C.H., Ketner, K.B., and Peterson, Fred, 1990, The Elko orogeny-Late Jurassic orogenesis in the Cordilleran miogeo- cline: Geological Society of America Abstracts with Programs, Cordilleran Section, v. 22, p. 88. Thurman, EM., 1985, Organic geochemistry of natural waters: Boston, Martinus Nijhof/Dr. Junk Publishers, 497 p. Toth, J., 1962, A theory of groundwater motion in small drainage basins in central Alberta, Canada: Journal of Geophysical Research, v. 67, p. 4375-4387. 1963, A theoretical analysis of groundwater flow in small drainage basins: Journal of Geophysical Research, v. 68, p. 4795-4812. 40 JURASSIC AND TRIASSIC WETLANDS AND ORIGIN OF URANIUM DEPOSITS 1978, Gravity-induced cross-formational flow of formation fluids, Red Earth region, Alberta, Canada: Water Resources Research, v. 14, p. 805-843. Turner, C.E., and Fishman, N.S., 1991, Jurassic Lake T'oo'dichi'-A large alkaline, saline lake, Morrison Forma- tion, eastern Colorado Plateau: Geological Society of America Bulletin, v. 103, p. 538-558. Turner-Peterson, C.E., 1985, Lacustrine-humate model for primary uranium ore deposits, Grants uranium region, New Mexico: American Association of Petroleum Geologists Bulletin, v. 69, p. 1999-2020. 1986, Fluvial sedimentology of a major uranium-bearing sandstone-A study of the Westwater Canyon Member of the Morrison Formation, San Juan Basin, New Mexico, in Turner- Peterson, C.E., Santos, E.S., and Fishman, N.S., eds., A basin analysis case study-The Morrison Formation, Grants urani- um region, New Mexico: American Association of Petroleum Geologists Studies in Geology 22, p. 47-75. 1987, Sedimentology of the Westwater Canyon and Brushy Basin Members, Upper Jurassic Morrison Formation, Colo- rado Plateau, and relationship to uranium mineralization: Boul- der, University of Colorado, Ph.D. thesis, 169 p. Turner-Peterson, C.E., and Fishman, N.S., 1986, Geologic synthe- sis and genetic models for uranium mineralization in the Mor- . rison Formation, Grants uranium region, New Mexico, in Turner-Peterson, C.E., Santos, E.S., and Fishman, N.S., eds., A basin analysis case study-The Morrison Formation, Grants uranium region, New Mexico: American Association of Petro- leum Geologists Studies in Geology 22, p. 357-388. Turner-Peterson, C.E., Fishman, N.S., Hatcher, P.G., and Spiker, E.C., 1986, Origin of organic matter in sandstone uranium deposits of the Morrison Formation, New Mexico-Geologic and chemical constraints, in Dean, W.E., ed., Organics and ore deposits: Denver Regional Exploration Geologists Society Symposium, 1985, Proceedings, p. 185-195. Turner-Peterson, C.E., Santos, E.S., and Fishman, N.S., eds., 1986, A basin analysis case study-The Morrison Formation, Grants uranium region, New Mexico: American Association of Petro- leum Geologists Studies in Geology 22, 391 p. Tyler, Noel, and Ethridge, F.G., 1983, Depositional setting of the Salt Wash Member of the Morrison Formation, southwest Col- orado: Journal of Sedimentary Petrology, v. 53, p. 67-82. van Everdingen, R.O., 1971, Surface-water composition in south- ern Manitoba reflecting discharge of saline subsurface waters and their subsurface solution of evaporites: Geological Associ- ation of Canada Special Paper 9, p. 343-352. Walker, TR., 1967, Formation of red beds in modern and ancient deserts: Geological Society of America Bulletin, v. 78, p. 353-368. Wanty, R.B., Goldhaber, M.B., and Northrop, H.R., 1990, Geochemistry of vanadium in an epigenetic, sandstone-hosted vanadium-uranium deposit, Henry Basin, Utah: Economic Geology, v. 85, p. 270-284. Ward, W.C., and Halley, R.B., 1985, Dolomitization in a mixing zone of near-seawater composition, late Pleistocene, northeast- ern Yucatan Peninsula: Journal of Sedimentary Petrology, v. 55, p. 407-420. Warner, J. W., Heimes, F.J., and Middelburg, RF., 1985, Ground- water contribution to the salinity of the upper Colorado River # U.S. GOVERNMENT PRINTING OFFICE: 1994-673-046/86050 basin: U.S. Geological Survey Water-Resources Investigations Report 84-4198, 113 p. Weir, D.B., 1952, Geologic guides to the prospecting for carnotite deposits on Colorado Plateau: U.S. Geological Survey Bulle- tin 988-B, p. 15-27. Whitney, Gene, and Northrop, HR., 1987, Diagenesis and fluid flow in the San Juan Basin, New Mexico-Regional zonation in the mineralogy and stable isotope composition of clay min- erals in sandstone: American Journal of Science, v. 287, p. 353-382. Whittemore, D.0., McGregor, K.M., and Marotz, G.A., 1989, Effects of variations in recharge on groundwater quality: Jour- nal of Hydrology, v. 106, p. 131-145. Williams, A.E., and McKibben, M.A., 1989, A brine interface in the Salton Sea geothermal system, California-Fluid geochemical and isotopic characteristics: Geochimica et Cos- mochimica Acta, v. 53, p. 1905-1920. Winter, T.C., 1976, Numerical simulation analysis of the interac- tion of lakes and ground water: U.S. Geological Survey Pro- fessional Paper 1001, 45 p. 1978, Numerical simulation of steady state three-dimen- sional groundwater flow near lakes: - Water Resources Research, v. 14, p. 245-254. 1983, The interaction of lakes with variably saturated porous media: - Water Resources Research, v. 19, p. 1203-1218. 1986, Effect of ground-water recharge on configuration of the water table beneath sand dunes and on seepage lakes in the sandhills of Nebraska, U.S.A.: Journal of Hydrology, v. 86, p. 221-237. 1988, A conceptual framework for assessing cumulative impacts on the hydrology of nontidal wetlands: Environmental Management, v. 12, p. 605-620. Winter, T.C., and Woo, M.-K., 1990, Hydrogeology of lakes and wetlands, in Moss, M.E., Wollman, M.G., and Riggs, H.C., eds., Surface water hydrology: Geological Society of America, The Geology of North America, v. O-1, p. 159-187. Witkind, LJ., and Thaden, R.E., 1963, Geology and uranium-vana- dium deposits of the Monument Valley area Apache and Nava- jo Counties, Arizona: U.S. Geological Survey Bulletin 1103, 171 p. Wood, H.B., 1968, Geology and exploitation of uranium deposits in the Lisbon Valley area, Utah, in Ridge, J.D., ed., Ore deposits of the United States, 1933-1967: New York, American Insti- tute of Mining, Metallurgical, and Petroleum Engineers, p. 771-789. Wood, W.W., and Petraitis, M.J., 1984, Origin and distribution of carbon dioxide in the unsaturated zone of the Southern High Plains of Texas: Water Resources Research, v. 20, p. 1193-1208. Young, R.G., 1964, Distribution of uranium deposits in the White Canyon-Monument Valley district, Utah-Arizona: Economic Geology, v. 59, p. 850-873. 1978, Depositional systems and dispersal patterns in urani- ferous sandstones of the Colorado Plateau: Utah Geology, v. 5, p. 85-102. Cover. View south-southeast across Lhasa He (Lhasa River) flood plain from roof of Potala Pal- ace, Lhasa, Xizang Autonomous Region, China. The Potala (see frontispiece), characteristic sym- bol of Tibet, rises 308 m above the valley floor on a bedrock hill and provides an excellent view of Mt. Guokalariju, 5,603 m elevation, and adjacent mountains 15 km to the southeast. These mountains of flysch-like Triassic clastic and volcanic rocks and some Mesozoic granite character- ize the southernmost part of Northern Xizang Structural Region (Gangdese-Nyaingentanglha Tec- tonic Zone), which lies just north of the Yarlung Zangbo east-west tectonic suture 50 km to the south (see figs. 2, 3). Mountains are part of the Gangdese Island Arc at south margin of Lhasa continental block. Light-tan areas on flanks of mountains adjacent to almost vegetation-free flood plain are modern and ancient climbing sand dunes that exhibit evidence of strong winds. From flood plain of Lhasa He, and from flood plain of much larger Yarlung Zangbo to the south (see figs. 2, 3, 13), large dust storms and sand storms originate today and are common in capitol city of Lhasa. Blowing silt from larger braided flood plains in Pleistocene time was source of much loesslike silt described in this report. Photograph PK 23,763 by Troy L. Péwé, June 4, 1980. ORIGIN AND CHARACTER OF LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA Frontispiece. On a 130-meter-high bedrock hill of Mesozoic lime- stone and schist, the splendid 13-story Potala Palace rises 308 meters above flood plain of the Lhasa He (Lhasa River) in Lhasa, capitol city of Xizang Autonomous Region, China. Construction of this well-known symbol of Tibet started in the 7th century, and most work took place under the direction of the fifth Dalai Lama from 1645 to 1693. The Potala served as home for the fifth and later Dalai Lamas and about 1,000 monks. The striking 1,000-room structure today is a state historical museum with about 35 caretaker monks. Roof provides a spectacular view of terrain of the southern Qinghai- Xizang (Tibet) Plateau (cover photograph). North Mountain, in the distance, consists of coarse-grained biotite granite of the Gangdese batholith of late Cretaceous and Palaeogene age; view to northwest. Photograph PK 23,739 by Troy L. Péwé, June 3, 1980. Origin and Character of Loesslike Silt in the Southern Qinghai-Xizang (Tibet) Plateau, China By T.L. Péwé, Lin Tungsheng, Roger M. Slatt, and Li Bingyuan U.S. GEOLOGICAL SURVEY PROFESSIONAL PAPER 1549 Retransported, tan, loesslike silt is widespread in the southern Qinghai-Xizang (Tibet) Plateau. On low hilltops the silt is primary loess. On the lower slopes and in valley bottoms of the major river valleys, however, it is poorly to well-stratified retransported loesslike silt. The silt probably was originally deposited by winds blowing across broad vegetation-free flood plains. UNITED STATES GOVERNMENT PRINTING OFFICE, WASHINGTON : 1995 U.S. DEPARTMENT OF THE INTERIOR BRUCE BABBITT, Secretary U.S. GEOLOGICAL SURVEY Gordon P. Eaton, Director For sale by U.S. Geological Survey, Information Services Box 25286, Federal Center Denver, CO 80225 Any use of trade, product, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government. Text and illustrations edited by Mary Lou Callas Line drawings by Susan M. Selkirk unless otherwise noted Other graphics production by Susan M. Mayfield Layout by Linnea Larsen Library of Congress Cataloging-in-Publication Data Origin and character of loesslike silt in the southern Qinghai-Xizang (Tibet) Plateau, China / by Troy L. Péwé ... [et al.]. p. cm. - (U.S. Geological Survey professional paper ; 1549) Includes bibliographical references. 1. Loess-China-Tibet, Plateau of. I. Péwé, Troy Lewis, 1918-. II. Series. QE579.075 1995 552'.5-dc20 94-26508 CIP CONTENTS ADSHTACL. .. 2200020200222 verre revere reverse seer er rer ener neenee rere errr errr eee reer rene renee ne nene nene rere rre nene s 1 narnia 2 .css seee aree verre neer renner er verre ren rrr earner renner renee rere enn renner een ees 4 PRYSiCAl SEHNMG ..... cess eevee reer eer enne rrr rene nn rn rrr errr ene errr neenee neer rrr renee rrr nene ree 5 Physiography @Nd G@OIOGY ...... ..... err err errr enne ree nner ees 5 CIMAC verse eave ever er err renee rere rere eer errr rere reer ene rene nen ene ne nne nene rennes 8 sees ss seve rer er ere rere rre rere nere reer er err er errr nen ener e nere ner rere enne 11 GEOMOTDROIOGY ...... ...s eserves ee ner enne rere rre rrr rr renner nre renee nene nein 12 LOGSSIiKG Silt Q@DOSitS seres evere reer reer ener reer ner errr err ner neenee neenee rennes 15 DiStriDUtiON @NG thiCKN@SS seve vere rere reaver errr eer rere ener er err renee rere renee ees 15 io on e 19 MiNETAl COMPOSitON seee verre rear err rer errr rere rrr errr nner errr rn er renee ene neenee 24 Figld erve vere rense revere reese reer revere rarer enne errr ener enne nene nene ener ene nees 32 FOSSil$ in 10@SSIKG Silt verses errs serre erver er rere ener enne nere renee renee ere 32 POIIEN vere reverse reese erve sever er er ere iene nene renee nere ne rene nere nere renner ener ene 36 MEthOGOIOGY ever serre arre rer errr neenee enne nere nner nere rennes 36 RESUIES sessa verse rere revere renner errr rrr nene neer ee ever er ener renee renee nere en ene ne nes 36 erp 37 eevee rer vere revere eres rer vere rere er ern errr ere ener nene nner renner enne ener es 39 OFigint Of 10@SSIK@ seres errr eer rre rrr errr errr ren nner rn err renner renner enne ees 39 WEAtRETINZ RYPOTRESIS seee eee aree err reer rere rere re rr renee nere rrr evere renner rre nene 40 LACUStTiNE RYPOTAESIS ...... ...s sees severe vere rere rere errr rr renner reer nner errr errr ener rere rere nere 40 HYPOTHESIS ...... scenes seres eave eer rrr reer rre rrr rrr nerve rrr re rrr rrr rere rn errr nre nene 41 Figld es erve revere rere errs rse errr rer er ere errr errr rere neer nen enne neer nene es 41 GiStMDUtON eee ere eevee errr reer renner errr ener ees 41 Independent lithOIOGiC CRATACtET eee verve ever rere rrr ern rene eee} 41 Absence Of GiStiNCt StTAtifCAUNION ...... seee rere esen rere aree renee ere eee 41 COIOF ANG eee verve rer erver errr ever er ener ener errr ener enne nees 41 Association with other evidence of wind 41 FoSSil$s Of @ir-Dre@thiNng AMIM@IS eee e rere rere reer errr renee 43 MECHANIC] COMPOSitONM .......... cesses seee reese verre rrr never rer er neer errr renee nner ene rre nes 43 PEtTOQTAPRIC cscs eee ere renee reer nner even renner errr enne nner nees 43 SUIMIMATY neem aree earner nevr ern nevr earner erver rrr renee nen enne neenee nene rre renee nene 43 REtFANSPOTY@U 1O@SS ......... cscs seres errr rer rare renee errr errr rrr never rere enne reer ener renner enne r enne e 43 Age and COITEIAtiION Of Silt Q@DOSiIHS e erve eer ere nner err reer rere nner rene rrr eerie es 45 SUINIMATY ...se eee eevee eevee vre sve rev nerve reer rr errr errr even rene reer enne rere veneer errr renner renner rere ren enn 46 Cit@U erver er errr revere ree verre rer errr er ener ere rre renee ener renee neer nene nene nene nes 51 FRONTISPIECE Frontispiece. - Potala Palace and bedrock hills in Lhasa, Xizang Autonomous Region, China FIGURES 1. _ Map showing distribution of loess in China and location of Qinghai-Xizang (Tibet) Plateau ........................ VI 10. 11-20. 21. 22-24. 25. 26. 27. 28. 29. 30-33. 34. 35. 36. 37. 38. CONTENTS iNGCX MAPS Of WESt@PN CRIMQ errr er Index map of South-Central QinGRAi-XiZ@NG (TiDEt) errr ss. Photograph of steam rising from wells in post-middle Pleistocene Yangbajain thermal field, 90 km Of reer errr Map showing distribution of existing glaciers and SNOW-Iline @I@VAtiONS iN eens. Generalized precipitation and vegetation map Of AUtONOMOUS eee eee nnn. Climatic data for: 7. _ LR@S@ (1951-1980) 11222220000 errr ree 8. _ XIgAZ@ (1955-1 reer eres rere rere eer ees 9. _ TiMgIi (1970-1980) 1122020000 errr reer rere errr Map showing distribution of permafrost and seasonally frOZEN iM CHINA eens. Photographs showing: 11. Eroded remnant of frost blister PATt Of AM iC@ COTE ss. 12. Broad flood plain of a north-flowing tributary of the Yarlung Zangbo about 40 km west of Xigaze...... 13. Broad flood plain of Yarlung Zangbo about 15 km dOWNStre@M frOM 14. View southeast (upstream) Of west Side Of NYANZ QU VAII@Y 15. Active sand dunes on flood plain of Nyang Qu about 7 km SOUtheaSt Of XigAZE 16. Climbing sand dunes on lower flanks of hills @lONg braided LRAS@ HG ekle es. 17. - Sharply dissected fans of retransported loess at base of low west-facing mountains east of Tingri, XiZANG AULONOMOUS REGION errr eer ress eer rere ree rere veces 18. Large dissected fan of retransported loess below site of Tibetan monastery and Xigar-Zhong at Tingri, south-central Xizang Autonomous Region, 1980 AMG 1922 19. Newly made bricks of loess and loess-brick walls 45 km SOUtheaSt Of 20. - Remnants of loess-brick walls of 13th century monastery about 10 km west of Xigaze......................... Diagrammatic sketch of typical dissected loesslike silt with lenses and layers of fluvial sand and eron ci Lm Photographs of: 22. Loess overlying bedrock on knob 15 M @bOVE NYANZ QU @t 23. Deep, sharp-walled gully dissected into retransported loess fan at base of limestone knob east Of XI&AZE ss.. seee severe seee errs erence. 24. - Alluvial fan of retransported loess sharply dissected by steep-walled gullies east of Xigaze.................. Physiographic diagram of Xigaze area at junction of Nyang Qu and Yarlung leek}. Photograph of flat-topped, steep-walled remnant of bedded, vertically jointed retransported loess in QU V@ll@Y NOTth Of Ki§AZE rere reer eevee eee. Sketch map of flood plain and stratigraphic sections of retransported sandy loess on terrace on south side ip er P Photograph of cliff-forming retransported tan loess exposed in road cut on south side of Xigaze .................. Triangular diagram showing proportions of sand, silt, and clay in sediment samples from south-central Qinghai-Xizang (Tibet) Plateau and @ 10€SS S@MpPIG fFOM GEMMANY sree rere rere veers reer Graphs showing cumulative-frequency grain-size curves for: 30. _ Retransported loesslike silt in southern part of Qinghai-Xizang (Tibet) 31. Loess flanking or on top of bedrock knobs near Yarlung Zangbo and loesslike silt from Xizang AULONOMOUS REGIOM eres seee seres eres r eres reese eserves 32. - Clay-rich retransported loess, south-central Qingh@i-Xizang (Tibet) PIAtEAU eee ee 33. - Eolian and fluvial sand, south-central Qingh@i-Xizang (TibEet) PIAtEAU veers reer rere ekke. Generalized geologic map of valley of Yarlung Zangbo and vicinity and weighted peak-percentage diagrams Of CI@Y MiNETAIS @NG NMAJOT IMIN@TAIS seee e vee reeves ress errr rer reese reer reer ree ree. X-ray diffractograms of samples of loess and underlying volcaniclastic sandstone and lithic tuff ................... Diagram showing pollen composition of sediment samples from stratigraphic section on scarp of terrace of NYAMG QU ere reer veer seee essere errr reer errr eer errr ea Photograph of collecting fossils in loesslike silt at river-terrace exposure on south side of Xigaze ................. Photograph of fresh-water gastropods Radix auricularia (L.) from loesslike silt south of XIgAZC 6 8 9 10 11 12 12 13 14 14 15 16 16 17 18 19 20 21 23 23 24 26 28 29 29 30 31 32 33 33 34 34 35 36 37 38 39 J w poro o un 39. 40. 41. 42. 43. 45. 46. CONTENTS Cumulative-frequency grain-size curves for loess and retransported silt from near Xigaze and for volcanic ash and modern wind-blown dust from Other PaIt$ Of the WOTIQ errr ener renee Cumulative-frequency grain-size curves for loess and retransported silt from near Xigaze, from Beijing, ANd frOM Other Of the WOPIG ree renner errr renee neenee nerve rrr nene never verre Photograph of ruins of ancient city of J iaohe near Turpan, Xinjiang Uygur Autonomous Region ................... Block diagram showing retransported loess in valley beSide FUNS Of revere rere errr rere reese reer rer ers Photograph of narrow steep-walled valley entrenched in retransported loess on west side of Jiaohe............... Low oblique aerial photograph of mesa-like remnants of retransported loess at ruins of Jiache NEAF TUIPAN ...... enne enne eee nene renee ern rer Cumulative-frequency grain-size curves for loess and retransported loess from Kuga (Kucha), Jiaohe, and Nilka, south and north flanks of Tian Shan, ANG fTOM XKigAZE ATCA ..........cccce eee ver verre reer rere vre nevr errr er nere ener nne es Schematic block diagram of loess and retransported loess north of Nilka, in Ili He valley on north slope Of TiAN SHAN sree vrea vrea rr rear eer ener reer reer ener ener nner rer errr reer renner renner erver nere ber enne rn rrr nere nnn rene TABLES Data on Quaternary sediment samples collected in southern Qinghai-Xizang (Tibet) PI@te@U ...... Mechanical properties of sand and loesslike silt from southern Qinghai-Xizang (Tibet) Plateau .............c.cc..... Percentage of sand, silt, and clay in sand and loesslike silt, southern Qinghai-Xizang (Tibet) Plateau ................ Mineral compositions of >0.062-mm fraction of sand and loesslike silt, southern Qinghai-Xizang (Tibet) PIANCAU erver erase rere ere rere reer errr erea sever ever errr neer nere reer neer ener errr nene nere rene neer nen en enne rrr renner eee eer rere rrr errr rre ene Mineral composition of bulk samples of sand and loesslike silt, southern Qinghai-Xizang (Tibet) Plateau ......... Mineral composition by weighted peak-area percentages of <0.005-mm fraction of sand and loesslike silt, southern Qingh@i-Xiz@ng (Tibet) eee errr errr rr nner rene enne nner ne VII 44 45 46 47 48 50 50 22 25 27 27 28 28 roms mers Origin and Character of Loesslike Silt in the Southern Qinghai-Xizang (Tibet) Plateau, China By Troy L. Péwé,! Liu Tungsheng,2 Roger M. Slatt," and Li Bingyuan4 ABSTRACT Tan loesslike silt of probable late Quaternary age is widespread in the southern Qinghai-Xizang (Tibet) Pla- teau" of southwest China. Most of the silt has been re- transported and occurs mainly in the lowlands and lower slopes; it is absent from steep slopes and active flood plains. This loesslike silt covers most alluvial fans and is interbedded with the sand and gravel of the fans. It is well exposed in the agricultural fields on low terraces in the valleys and in the steep-walled scarps of dissected valley fill. On low hilltops the silt is primary loess, probably de- posited by winds that picked it up from broad vegetation- free flood plains and perhaps from dry lake basins to the north. In the lowlands, most of it is retransported loess that has been carried downslope and redeposited by water. The primary loess is as thick as 6 m on hilltops or high flanks of hills and as thick as 10 to 15 m on the lower slopes. The thickest deposits are near the rivers in valley bottoms, where they develop vertical cliffs as high as 15 m. The texture and mineral composition of the primary loess is relatively uniform; however, the retransported loess varies widely in texture. The silt is clay rich to sand rich, and some is even gravel rich, depending on how near the original loess source was, on the distance of retrans- portation, and on the amount of mixing with other sedi- ments by fluvial action. The silt is massive, and erosion \ Present address: Arizona State University, Box 871404, Tempe, AZ 85287-1404, U.S.A. 2 Institute of Geology, Academia Sinica, P.O. Box 634, Beijing, 100029, People's Republic of China. 3 Colorado School of Mines, Golden, CO 80401, U.S.A. * Institute of Geography, Academia Sinica, P.O. Box 771, Beijing, 100101, People's Republic of China. 5 The term for the Tibet Plateau that is approved by the U.S. Board on Geographic Names is "Qing Zang Gaoyuan". However, the term "Qinghai-Xizang (Tibet) Plateau" is most widely used in international scientific English-language publications and is used in this report. Manuscript approved for publication May 18, 1994. forms vertical cliffs and deep gullies. It has little or no stratification or jointing, except where it has been retrans- ported to lower slopes and valley bottoms where it is well to poorly stratified. The few fossils that have been found in the loesslike silt on the plateau include Quaternary pollen, ostracodes, gastropods, and vertebrate teeth and bones. They reflect a dry cold area with a few ponds or marshes. Through the years, the origin of the loesslike silt on the plateau has been ascribed to in situ weathering of re- sidual country rock as well as to lacustrine, fluvial, or eoli- an processes. The weathering or residual hypothesis has little support. The lacustrine hypothesis, either separately or in conjunction with fluvial processes, has been consid- ered since the late 1970's and the 1980's. The eolian proc- ess was first proposed in the 1980's. Since the early 1930's the loesslike silt has been de- scribed as a residual deposit formed by the breakdown by freezing and thawing of the underlying rocks. However, the silt bears no chemical, mineralogical, or textural rela- tion to the underlying strata, and it is too thick to represent only a breakdown of rocks in place. The most common explanation in the literature for the origin of these thick deposits of loesslike sediments, especially in valley bottoms, is by lacustrine processes. This origin is unlikely, however, because no shorelines, wave-cut beaches, deltas, mud cracks, or ripple marks have been found in the silt deposits. Neither lacustrine stratifica- tion nor an appreciable amount of clay exists in the silt. Moreover, the deposits have no definite upper boundary, as would be expected if they were lacustrine in origin. The mantle of uniform loesslike silt on top of the bedrock hills is here considered to be primary loess, wind- blown and derived from the braided streams and broad plains, because (1) it occurs as a surficial mantle; (2) it is lithologically independent of the underlying material; (3) the fine silt is stratified indistinctly or not at all, except in retransported materials; (4) it is associated with sand dunes in a broad wind-blown area; (5) it contains fossils of land animals; (6) its sorting and texture on hilltops is similar to that of loess and windblown dust in many other 2 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA places in the world; and (7) its grains are angular and rela- tively unweathered. The thick mantle of poorly to well- stratified loesslike silt on the lower slopes and along the bottoms of major river valleys, however, is retransported loess. It has been transported from a few meters to hun- dreds if not thousands of meters down slopes, primarily by processes involving water; consequently, it has become in- terlayered with sand or with gravel. Almost all of the loesslike silt in the southern part of the Qinghai-Xizang (Tibet) Plateau is retransported loess. t4 £ 444 XLMATHRAREm®kt. ki, #Af iw. k BRVUAk$KXRAL #4%#A6HD%6%Lk. %*% iki t. Tt} MK. Br X to # © A k in $1 66 14 & 1. th ih be # k th & Th th. 51K. LkiTM#, , AHA Tit ® I LMM A & #) ® t. A K 144 46 6 +T "A if . RAX +ih AA {AK L £6h Thi tok Ligik 6m, f#KMpik 10m. # ik15m, MAnKR k 16) RZ. nk 4% +6) #K £1t Hk, MBLASNDMAEZEARKAE , th BR 6) ik . Kkikigs F E Ad &txkihk $. it*#% i EHA, i24##. T AWA 4 bk # i) 6 Kik *% +90 LTP w WA 6) £ . AUB %. MA AAR KUT ¥ A#. i4&k, , , LkGAM»J4GHAR. #RARAOAEAKEk. PMABHET-EZAFTIOGHRLERE®, H1804RL.AXkXHk k T Aki. Zk3044, MK AmKK . {2XRMAAIAzi,. AXEMAAkKkk, AG THk® # L6) K ®. XL&kA## T %. L, AX*AiNnkKikXA+% #M ik , ik +t XkRLXZ. KARAGMME4&MALZR. . »#t3b, && 4A ARF Wik Ataxt t M AA th kW LF. KAN», UMKEBAXLE TAE Whi Kk ik iA RARKKi A»%: (1) TIIUKEGEALRLKL:; (Q) (0) RFA XH4LARMAARIKY;, U) (5 AMLLEABRARHHILL; (6) % + #11; M &*P#HHHZEHkRAk, AMIUIREH. T kK % i, Wig UKTA&*TP4AV%k. , ML4UXAK HK +ki+AAAkiEg H) % i. INTRODUCTION China is well known for some of the most extensive accumulations of loess in the world. Loess is one of the most extensive Quaternary deposits of the country, covering about 1,000,000 kmz, or about 10 percent of China. Al- though most of the loess is in north-central China, it is also widespread in the northwestern and northeastern parts of the country (fig. 1). The earliest known description of Chinese loess appears to be more than 2,300 years old (Liu and others, 1964, p. 10); the earliest known record of the ubiq- uitous dust storms from the deserts was a so-called "dust rain" that occurred in 1150 B.C. and is documented in the historical book, Zhu Shu, Ji Nian (Chronicles Recorded on Bamboo Slips) (see Liu, 1981). Characteristics of Chinese loess and diverse ideas of its origin have been known out- side of China since the explorations of Raphael Pumpelly (1867, 1879), Baron Ferdinand Paul Wilhelm freiherrn von Richthofen (1877, 1882, 1886), and Bailey Willis, Eliot Blackwelder and R.H. Sargent (1907). However, even on recent maps showing loess distri- bution in China (Wang and Zhang, 1980; Liu and Yuan, 1982; Wang and Song, 1983; Liu and Ding, 1984; Liu and others, 1985, 1988; Liu, 1981, 1991), no loess is shown on the Qinghai-Xizang (Tibet) Plateau on a recent map of China (fig. 1). Lack of knowledge of detailed geo- logic reports of the plateau can be attributed to the geo- graphical and political remoteness of this region, often called the "Roof of the World," that is rimmed by even higher mountains. This unique environment has an isolat- ed civilization that was relatively unchanged until the 1951 occupation by Chinese from the east. Since the 1950's, China has organized seven multidisciplinary inte- grated surveys to the Qinghai-Xizang (Tibet) Plateau. More than 50 disciplines, represented by more than 1,600 scientists and technicians from Academia Sinica have made very informative studies, and 32 volumes of reports have already been published. Yet much, if not most, of these efforts were unknown outside of China until the landmark international symposium on the geological and ecological studies of the Qinghai-Xizang (Tibet) Plateau was held in China in 1980 (Liu, 1981; Péwé, 1980a, b, 1981; Reiter and Reiter, 1981; and Sengor, 1981). Aca- demia Sinica arranged this historic symposium that in- cluded a field study trip to the Qinghai-Xizang (Tibet) Plateau. Of 79 scientists from 18 countries who initially met with 240 Chinese scientists in Beijing, 66 participat- ed in a stimulating 2-week trip through the southern part of the plateau and traveled 1,000 km to Nepal. This field trip followed and crossed broad, vegeta- tion-free flood plains and sand dunes on the valley bot- toms and low hillsides that are quite striking in the southern part of the Qinghai-Xizang (Tibet) Plateau. It soon became apparent to Liu and Péwé that the wide- spread yellowish loesslike sediment was undoubtedly re- transported loess because (1) it covers some alluvial fans and is interbedded with the sand and gravel of the fans, (2) it is well exposed in agricultural fields in the valleys, and (3) it forms thick steep-walled terrace scarps in dis- sected valley fill. Winds probably blew it there originally from the sand and silt bars of the broad rivers and perhaps from dry lake basins to the north. The geologic guidebook of the area (Academia Sinica, 1980) does not mention loess but does mention thick lacustrine deposits. INTRODUCTION 3 We saw small valleys as wide as 1 km that had slowly filled to a thickness of 30 m or more with retrans- ported loess. These valley fills had subsequently been dis- sected by streams rejuvenated by uplift, probably in Holocene time. Slow aggradation and interlayering of re- transported loess with some fluvial sand and locally fine gravel has created nearly flat surfaces on the old valley floors that now stand above valley flood plains as buttes or mesas, 15 to 30 m high, with almost vertical walls. These landforms are common at XKigaze and also at Turpan, 1,500 km to the north, for example. As silt accumulated locally in valley bottoms, remains of the extant flora and fauna were incorporated. Sandy prima- ry loess accumulated on low hills next to the rivers and is interfingered downslope with coarse sediments. Péwé col- lected sediment samples and associated fossils in Yangbajain Basin north of Lhasa, in the vicinity of Lhasa, and southwest- ward through major valleys, high passes, and mountain rang- es to Gyangze on the middle Nyang Qu and to Xigaze near the junction of the Nyang Qu and the Yarlung Zangbo (fig. 3, table 1). Detailed collections and observations were made near Xigaze. Sediments, topography, and geomorphology 7 U I 70°E 80°E 90°E [~50°N Turpan fifi \ , XINJIANG UVYGUR f AUTONOMOUS REGIONf QINGHAI od, ix \\PRO VINCE QINGHAI XIZANG """ (% (TIBET) PLATEAU A \\\\\\\\\\\\é\\ t_, _ \.\ [~30°N °E 100° 90LF 0l E | I 100°E Lanzho AUTONOMOUS \ SICHUAN - vi en - PROVINCE' $ ~~ nd o O Shanghai - ~" N | "4-15 \ 130°E - 80°E Guangzhou EXPLANATION PROVINCE -" o [~20°N - Loess 20°N - E Retransported loess O anghal Xizang [~ \\ (Tibet) Plateau 250 0 500 1000 Km - * Boundary of Plateau p SCALE ;~ Boundary of adjacent Province or Region 120°E T I I I | I \ 130°E 110°E 120°E 50°N 7 k ' 2" » All 3 gmmsmrm‘ibdba y (anges im ae us n -- C clk Figure 2. Landform map (A) and index map (B) of western China showing location of Qinghai-Xizang (Tibet) Plateau and features of Xinjiang Uygur Autonomous Region, Xizang Autono- mous Region, and parts of the Qinghai, Sichuan, and Gansu Provinces. Landform map is part of "Landforms of China," (copyright by Erwin Raisz, 1955, and reprinted with permission of Raisz Landform Maps, P.O. Box 773, Melrose, MA 02176). ge Nosy. =) Car ge \ C ’%’v Bo 0 QK mo = Yumeh€§s’//7 soe Cm p z x/, np IEC) Index map by S.M. Mayfield. Names on landform map are Eng- lish equivalents of those in use when map was prepared. Names on index map include many used in this report. (U.S. Board on Geographic Names approved "Qing Zang Gaoyuan" as the name for the Tibet Plateau; however, the term "Qinghai-Xizang (Tibet) Plateau" is most widely used in international scientific English- language publications and is used in this report.) 7 75° 80° 85° 90° 95° 100° I T 45°L Q > Lay" . *-. Hantgggg 'tIARN SHAN /Ururr3m o 40° w 908 Sha, - PCC Turpan® **~% Muztagate 35° 30° PHYSICAL SETTING A Gongga <1 Feng Chengdu, C go‘l’r- 25°~ ¢$ii";- & $41», 0 500 MILES I 1 1 1 - 1 ] 0 | 500 KILOMETERS | B Xinjiang Uygur Autonomous Regi AREA OF MAP C HIN A LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA I I 90° 95° A ) B Ova X) HA N >- l & | « v2 NY AlN Q & 5&2} & Nying 4? Hatchured area covered by geologic map (fig. 34 A) o Zhou ,—’/ -30° Yangbajain-__° “"‘\ y Sa + e NQ S-. ~ ~~=16 Maizhokunggar ___ __ ~~=~-- NyingehiC - ( p Namling Lhasa_Q‘/' 7 mostnge gymgfl,’ / TN... pt--=- . 7 i Egga n e ® Xxgaze een, quu’g’ Sggrl Nangxian /,' s fen y yamringo merino otto t- v8. uu. A seid 25 2s p Yarlung Za ~~*~-6, ./ Mainlin Yarlung Zanzba Lhaz 69 e Bamang Yamzho arlung Zangbo &.. g,4 ,/ oGyirong 11-17 fig») oGyangze Lake 3’7 I H C;, & \ q f ® Xoo qTingri & ( Pumba Loke Comai Lhunze veg Pam Ou ~ Pal oll 22. - 2 ( Mt. Qomolangma f Gamba ~‘\ j --" 1 \, \ ZhamL\ [C “343 g 2 ~~ } + o co \ / ~T R L) /A ,Y A s w Cona $o, 1/ vol. ( 7 /: e $l £ Yadong t ’\./ I nest SIKKIM \ to / BHUTAN \_. / I} NEPAL a t&-.-A~y~ y \ // 35" ~ \ \ ¢ toZ 4 $9 * a *-- \ ~" /— », /" N e - / INDIA ‘-—\-.——\/--Q~ l -" __/e_~/./ I zz - U 0 60 120 180 Kilometers Figure 3. Index map of south-central Qinghai-Xizang (Tibet) Plateau, China. Circled numbers are locations of sediment samples (see table 1). Rectangle in upper left corner outlines the four major tectonic zones of the plateau (from Academia 1980; Shi and Li, 1980, 1981; Shi, 1988). About 83 per- cent of the area covered by modern glaciers in China is in this region (fig. 5). Most are small cirque glaciers and winding alpine glaciers, although a few are nearly 30 km long. The high elevation, topography, and particularly the climate of the plateau influence the development and dis- tribution of these glaciers (Kuhle, 1986). Two main types of glaciers are present on the plateau: Continental glaciers in the cold, dry parts of the plateau, and maritime glaciers in the southeastern part of the plateau, where monsoon precipitation is high. The snow line ranges from about 4,400 m in the southeast to more than 6,200 m in the western part of the plateau (fig. 5). Snow-line variation is mainly due to the rapid decrease of precipitation in the higher mountains toward the inner parts of the plateau. The west-central part of the plateau receives the highest total solar radiation in China. Numerous mass-balance studies of glaciers have been made by Chinese scientists who found that solar radiation accounts for 80 to 90 per- cent of the heat sources in the majority of glaciers in the arid part of the plateau and decreases to about 60 percent in the humid southeastern plateau. Natural vegetation of the southern part of the plateau is that of a montane shrubby steppe or locally an alpine Sinica, 1980) superposed on an outline map of Xizang Auton- omous Region, China. Center rectangle is area covered by geologic map (fig. 344) and mineral distribution maps (figs. 34B, C). steppe (Zheng and others, 1981) of the semiarid region. Most common species are Sopora moorcroftiana, Trikeraia haekeri, Aristida triseta, and Artemisia. Common associ- ates are Astragalus strictus, Poa patens, and Androsace graminofolia (Yang and others, 1983; Li and others 1983). This region is essentially unforested except for species lo- cally growing near streams. Agriculture has been intro- duced in the broad valley lowland near the major villages. Conditions are suitable for growing barley, wheat, pota- toes, peas, turnips, and apples. CLIMATE The climate of the Qinghai-Xizang (Tibet) Plateau is mainly characterized by strong winds, cold temperatures, and aridity, especially in the high western and central parts (fig. 6). Except for the southern major valleys (figs. 2, 3), the mean annual air temperature is colder than 0°C. On the higher plateau, the mean annual air temperature ranges from -6 to -10°C (Academia Sinica, 1980). Some of the earliest temperature records reported in English from the high plateau are from Hedin (1903), who recorded -32.5°C in January 1901. Climatic data have been collected since PHYSICAL SETTING 'saun030,, «e yo uotsstu -13d wim poystrqnd pue (1g61) mag wou; pajunday 'qg6] '¢ dung 'qmgd '7 four, 4q g6L'ET Nd udesSo0ug 'surequunow jo sadofs jsamor uo pa sey ajqreu pug C istyos © ores UerunHag: sno1aJtuoq -I€;) Jo urejunopy Sue, Sunlop1oq 03 UIseq ssoloe jseanos moalA uoiSay snowouojny SuezTX Jo Wy Q6 'UlIseq jng] urefeqguex ay) Jo pjoy feuuay} weajs-jam Jo Jojem-j0y armeJodua}-y31y urefeqguex auaoojstog afppiuw-isod ur som wor}; Susu weary *p ansi © 10 1957 along a road that runs northward from Lhasa to Qing- hai Province (Xie, 1982). At Fenghueo Shan Pass (eleva- tion 4,800 m), about 500 km north of Lhasa (fig. 2), for example, the mean annual air temperature is -6.6°C (Wang LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA and French, 1994). 75° In the southern valleys, the mean air temperature dur- ing the warmest month ranges from 10 to 16°C; during the coldest month (January) it is -8°C. Mean annual air temper- ature ranges from 3 to 8°C. In Lhasa (elevation 3,658 m) the mean annual air temperature is 7.5°C; the average for June 95° 100° 4a5°|_ 40°~ a n :_\ __ h Feng o 35 30° 25° o--o mg? ~.-5400 ~~ z. [CONTOUR INTERVAL 200 METER T N I S500 KILOMETERS | e "#$ 4 Xigaze I I C H I AREA OF MAP, EXPLANATION + «an Glacier - 4000- Snowline elevation in meters above \ mean sea level "\ Approximate boundary of . Peoples Republic of China ._ / " comes hep ~. 4000 t- ent, Lanzhou , Kunming 1 Figure 5. Distribution of existing glaciers and snow-line elevations in China (from Shi, 1980). PHYSICAL is 15.5°C and for January is -2.2°C (Academia Sinica, 1980, p. 5). The frost-free season is 140 days. As one goes west, the climate becomes colder and drier. Mean annual air tem- peratures of Lhasa, Xigaze, and Tingri are 6.3°C, 4.8°C, and 2.17°C, respectively (figs. 7, 8, 9). The western part of the plateau is arid, and most of it receives little more than 100 mm of precipitation annually (fig. 6). The southern valleys are in the semiarid zone (250- 500 mm precipitation), and the far east is humid. The dry season extends from October to May and the wet season from June to September. Mean annual precipitation in Lhasa, Xigaze, Gyangze, and Tingri is 445, 431, 288, and 318 mm respectively (Academia Sinica, 1980) (figs. 7, 8, 9). During winter, westerly winds prevail over the pla- teau, and from October through May there are clear skies and strong winds. PERMAFROST The Qinghai-Xizang (Tibet) Plateau lies in an area of widespread alpine permafrost. Permafrost, or perennially frozen ground, is a naturally occurring material that has been at a temperature of 0°C or colder continuously for 2 or more years. Permafrost is defined on the basis of tempera- ture alone, notwithstanding the type of sediment or rock or the ice content. Most permafrost is consolidated by ice. On the plateau, permafrost extends over an area of about 1,500,000 km and represents about 70 percent of SETTING 11 all the permafrost in China (fig. 10; Tong, 1981; Cheng, 1983; Shi, 1988). Permafrost in the plateau exists at eleva- tions higher than 4,100 to 4,200 m in the far north, where its minimum elevation rises by about 100 m per 1° lati- tude (Cheng and Wang, 1982); to the south it starts at about 5,200 to 5,400 m elevation on the northern slopes of the Himalayas. Tong (1981) estimates that the ground tem- perature may be as cold as -12°C at 6,000 m elevation in the northern part of the plateau and about -7.4°C underly- ing the north slope of the Himalayas. The temperature of permafrost underlying most of the plateau is considerably warmer, and much of it is near 0°C near the topographical lower limit of permafrost. Permafrost thickness is estimated to be about 400 m at an elevation of 6,000 m and not more than 130 m in the valleys or basins at lower elevations. Generally there is no permafrost beneath large rivers and lakes on the plateau. Ice wedges have not been recorded in the Qinghai-Xizang (Tibet) Plateau, and only one or two are known in northeast China (Péwé, 1986a, b). However, they may have been widespread in the perennially frozen ground of China in late Pleistocene time, as indicated by the presence of ice-wedge casts (Lanzhou Institute of Glaciology and Cryopedology, 1983; Sun and Li, 1986; Cheng and Liang, 1987). A few pingos (large, ice-cored frost mounds in per- mafrost; Wang and Yao, 1981; Zheng and Jiao, 1991) and a multitude of high-altitude periglacial phenomena (Kuhle, 1985) exist on the plateau. Palsas (elliptical frost mounds in peat in permafrost) reported in 1980 at an elevation of 80°E 85°E 90°E 95°F f I C- TS, I _- I EXPLANATION beau Vegetation Conifers (with deciduous and broadleaf , evergreens at lower elevations) o Alpine grass (some conifers and shrubs in 35°N K- (I I: s elte§ed locations) { ~> |___l Alpine and drought-resistant plants &_ Approx??? boundary Precipitation -35°N i Autox? omtlyzuasnlgegi on - 25(- - Annual precipitation in millimeters (% i ( I \ / f < 30°N - \\ I, NPSL, a?“ % , o 5% n ame a *" Lh.asa 30°N Approximate boundary \ 14 "* maw mw moe mae 604 e Xi of People's Republic Ni igaze of China _ Tie . . L - Tingri ® ® Gyangze z --\~-~'~\ -" ",\*'~-/~( ll‘ p “ 100 __ 0 ___ 100 __200 _ 300 Kilometers "~*~ {_/ -~ ¥ w 1 | | | | \ 80°E 85°E 90°E 95°E Figure 6. Generalized precipitation and natural vegetation map of Xizang Autonomous Region, China. Modified from Pradyumna (1976) and U.S. Government Printing Office (1971). 12 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA about 5,000 m above sea level southeast of Gyaco La Pass GEOMORPHOLOGY in the southern part of Xizang Autonomous Region (Péwé, 1980a, b, 1981) are now thought to be frost blisters (low The geomorphology of the broad, flat-floored valleys frost mounds in seasonal frost) (fig. 11). Because of the _ (fig. 12) and the adjacent slopes (Lhalungpa, 1983, p. 128- rigorous climate, perennially frozen ground and seasonally 129) is of interest to this study. These areas exhibit sand frozen ground are widespread, and a high number of _ dunes, sharply trenched terraces, and vertical silt cliffs. All freeze/thaw cycles per year favors mechanical breakdown _ the major rivers and many of the minor rivers in the of the rocks (Huang and others, 1981). southern part of the plateau seem to be braided and have JAN FEB MAR APR MAY JUNE JULY AUG SEPT OCT NOV DEC JAN JAN FEB MAR APR MAY JUNE JULY AUG SEPT OCT NOV DEC JAN I I I I I I I I I I I I I I I | I I I I I 20 |- 15.3°C annual mean maximum temperature 14.5°C annual mean 10 ~6.3°C annual mean lemperamre 0° (freezing point) 0.8°C annual mean minimum temperature O° (freezing point) TEMPERATURE, IN DEGREES CELSIUS TEMPERATURE, IN DEGREES CELSIUS -20 1 1 1 1 1 1 1 1 1 L L -20 |_ | | L 1 L 1 N 1 I 1 MEAN TEMPERATURES MEAN TEMPERATURES E 100 TOT T T T E 100 r T T T T T & Above Freezing E : Above Freezing E“) 50 52) 50 E 0 eles . 1 1 1 1 E 0 CK to. 1 1 1 1 MEAN PERCENT OF DAYS WITH TEMPERATURE BELOW FREEZING MEAN PERCENT OF DAYS WITH TEMPERATURE BELOW FREEZING 201.8 mean annual number of days with temperature below freezing 217.7 mean annual number of days with temperature below freezing Z E T T T T I P I I I I Z 2 T T T T T -T T T T T g F 44.48 cm mean annual o E 43.12 cm mean annual TB 10|- __ precipitation - TH 10|- _ precipitation - <2 <2 m m g U 1 1 1 1 1 L L T _| g O 1 1 1 1 1 1 1 1 T a Z a. zZ MEAN PRECIPITATION a MEAN PRECIPITATION E E 100 @ a 2 3 " € k MEAN PERCENT OF CLEAR, PARTLY CLOUDY, AND CLOUDY DAYS MEAN PERCENT OF CLEAR, PARTLY CLOUDY, AND CLOUDY DAYS 87.8 mean annual 178.9 mean annual 98.3 mean annual 113.6 mean annual 172.7 mean annual 78.7 mean annual clear days partly cloudy days cloudy days clear days partly cloudy days cloudy days 4a T T -T T T T T T T T 4 53] 5 é 3}- | 2.2 meters per second mean monthly wind velocity = E a é 3}- | I ' I I I I ll.8 melelrs per gecond I. _ a E a 2} % E 8 s /f mean monthly wind velocity _| 852 i} | > §: af ] E Z E 0 1 1 pu 1 i 1 1 1 1 1 E Z é 0 4 A p | y | R F 4 4 4 MEAN MONTHLY WIND VELOCITY MEAN MONTHLY WIND VELOCITY 0 _ 5% L_ calm 46% 0 5% L_ calm 25% Lhasa (1951-1980) Xigaze (1955-1980) MEAN ANNUAL FREQUENCY OF WIND DIRECTION MEAN ANNUAL FREQUENCY OF WIND DIRECTION Figure 7. Climatic data for Lhasa, Xizang Autonomous Region, Figure 8. Climatic data for Xigaze, Xizang Autonomous Region, China (1951-1980). From Meteorological Bureau, Xizang Au- China (1955-1980). From Meteorological Bureau, Xizang Au- tonomous Region (1983). tonomous Region (1983). PHYSICAL SETTING 13 almost vegetation-free flood plains (cover, fig. 13). The Yarlung Zangbo in the vicinity of Xigaze is a heavily braided stream 8 to 10 km wide (Academia Sinica, 1980, p. 11). Large, low-angle alluvial fans form where major tributaries enter the wide valleys (fig. 14). Broad alluvial basins exist at Xigaze, Gyangze, Lhaze, Lhasa, Tingri, and many other places. JAN FEB MAR APR MAY JUNE JULY AUG SEPT OCT NOV DEC JAN I I I I I I I I I I I [se O T | 11.6°C annual mean maximum temperature 0° (freezing point) {ya 7 -- $ -- 53°C annual mean | $9 minimum temperature $& & 10 -10 TEMPERATURE, IN DEGREES CELSIUS -20 1 N 1 | 1 1 1 | 1 1 1 MEAN TEMPERATURES 100 50 PERCENT 0 > 3 MEAN PERCENT OF DAYS WITH TEMPERATURE BELOW FREEZING 255.4 mean annual number of days with temperature below freezing T I T I I I I I I I I 31.8 cm mean annual 10 - precipitation ~I L 1 L T L L MEAN PRECIPITATION PRECIPITATION, IN CENTIMETERS E 100 a O 50 pL 8 o E MEAN PERCENT OF CLEAR, PARTLY CLOUDY, AND CLOUDY DAYS 143.6 mean annual 164.4 mean annual 57 mean annual clear days partly cloudy days cloudy days 6z2 4 T T T T T T T T T E; é 3 / ZMW Z 28% ° 1 22 a « E Z é 0 1 1 1 1 1 1 1 1 1 1 1 MEAN MONTHLY WIND VELOCITY 0 5% L_] calm 40% Tingri (1970-1980) MEAN ANNUAL FREQUENCY OF WIND DIRECTION Figure 9. Climatic data for Tingri, Xizang Autonomous Region, China (1970-1980). From Meteorological Bureau, Xizang Au- tonomous Region (1983). Striking, active sand dunes and sand stringers have long been recorded in the broad valleys of the southern and central plateau (Rawling, 1905, p. 238). They are common on flood plains and on lower slopes of the bounding moun- tains (fig. 15) (Reiter and Reiter, 1981; Yang and others, 1981, p. 1756; Zhao and others, 1976, p. 6; Wang and Li, 1983, p. 4-5; Kreig and others, 1986; Wang and Fan, 1987). Most of them are conventional blow-out dunes or climbing dunes (cover, fig. 16), although an area of excellent barchan dune development was noted in the upper reaches of the Pum Qu valley near Tingri. Lower slopes of foothills are characterized by fans of silt or fans of silt layers interbedded with sand or gravel layers that include silt. Most of the silt is well sorted, some of it is mixed with sand and gravel, and most is mineralogically different from the upslope bedrock. These fans are ubiquitously trenched by steep-walled V-shaped gullies (fig. 17), that radiate outward from the top of the fan and extend to the toe. These gullies were cut by streams in the easily erodible silt. At the village of Tingri (fig. 3), a gullied, low-angle fan of retransported loess sits at the base of a sharp-peaked hill of tilted sediments (fig. 18A). Because this site is along the road used by per- sons traveling south to Mt. Everest (Qomolangma Feng), numerous photographs show changing construction on the silt fan and at the adjoining Tibetan monastery (for exam- ple, fig. 18B) (Harvard, 1984, p. 77). Liu and Prof. Zhang Ronsu examined the sediments of the fan during the Aca- demia Sinica scientific expedition to the Qinghai-Xizang (Tibet) Plateau of 1966-67, of which Liu was the leader. Zhang (now senior staff member at International Center for Integrated Mountain Development, Kathmandu, Nepal) observed then that the fan was a loesslike deposit. In 1980, it was noted that small storage caves had been exca- vated in the loess, and, because of the ease in excavating loess, part of the fan was planed off in steps to make flat construction sites for more buildings (fig. 18A). Locally, the distal parts of the silt fans have been eroded to form vertical cliffs as much as 15 m high. Like loess cliffs in central China, the Mississippi River valley and central Alaska, U.S.A., France, and elsewhere, these vertical silt cliffs in the southern plateau apparently stand for many years. The stability of these cliffs may be attrib- uted to the angularity of the silt grains and to the strength- ening effects of accretionary rods and tubes from earlier vegetation roots. Such cliffs are common south of Xigaze where they form the terrace scarps along the Nyang Qu (fig. 28). The loesslike silt that water has transported from slopes to valley bottoms forms excellent agricultural land that locally supports large farms. The loesslike silt has been used to make bricks (fig. 19) for thousands of years, and some brick walls are still preserved in interesting ar- cheological remnants (fig. 20) that remain along the busy trade routes of the past. 14 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA Figure 10. Distribution of permafrost 70°E 80°E 90°E, 100°E 110°E 120°E 130°E 140°E and seasonally frozen ground in Chi- a ’ / Fo G L 1 C 3G \ na. Map from Lanzhou Institute of Seasonal frost zone Glaciology and Cryopedology (1983) and published with their permission. 50°N Yitulihe 40°N 40°N Qinghai-Xizang (Tibet) Plateau 30°N K 30°N Figure 11. Eroded remnant of frost blister exposing part of an ice core |- in shallow drainage way, elevation 4,900 m, near Gyaco La Pass, near south border of Xizang Autonomous 20°N - Region, China. Meter bar shows scale. Photograph 4583 by Troy L. Péwé, June 11, 1980; reprinted from Péwé (1981) and published with per- IPC TOVE IGE 120°E 130°E mission of "Geotimes." 20°N LOESSLIKE SILT DEPOSITS 15 LOESSLIKE SILT DEPOSITS DISTRIBUTION AND THICKNESS In the mountainous Qinghai-Xizang (Tibet) Plateau, loesslike silt occurs in the lowlands and on gentle adjacent slopes; it is absent on the steep slopes and active flood plains (fig. 21). The silt is retransported loess that is thick- est near major rivers and on parts of low fans, pediments, and river terraces. It lies in a belt between the major mountain ranges and the active flood plains, which are lo- cally spotted with sand dunes or eolian sand blankets. This distribution of retransported loess is similar to that 1,400 Figure 12. Broad flood plain of a north-flowing tributary of the Yarlung Zangbo about 40 km west of Xigaze, Xizang Autono- mous Region, China. Area has barely enough vegetation to sup- port a few herds of yak and dzo. Retransported loess occurs in km to the north in western Xinjiang Uygur Autonomous Region on the south and north sides of the Tian Shan (fig. 5; Péwé, 1987; Wang and Song, 1983). Most of the loesslike silt has been slightly to considerably reworked and transported downslope by rill wash, sheet wash, stream action, and solifluction. As much as 1 to 2 m of silt still lies on a bedrock knob (fig. 22) 15 m above the Nyang Qu at Xigaze, and as much as 6 m of loesslike silt blankets the high flanks of a limestone knob 100 m above the south side of the Yarlung Zangbo 60 km downstream from Xigaze. The flanking silt is deeply dissected by deep, vertical-walled gullies (figs. 23, 24). lower valleys at base of folded Cretaceous rocks which form the east-facing mountains in background. Photograph 4581 by Troy L. Péwé, June 1980; reprinted from Péwé (1981) and published with permission of "Geotimes." 16 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA Figure 13. Broad, almost vegetation-free flood plain of braided Yarlung Zangbo is a ready source of wind-blown silt about 15 km east of Xigaze, Xizang Autonomous Region, China. Late Mesozoic granodiorite forms south-facing mountains in background. Photograph PK 23,902 by Troy L. Péwé, June 9, 1980. Figure 14. View southeast (upstream) of Nyang Qu valley from fortress thousands of years, a braided stream wandered across this broad, almost on rock knob in valley at Gyangse, Xizang Autonomous Region, China. vegetation-free flood plain, now spotted with sand dunes. Winds picked up Fortress originally built by Tibetans as defense against Mongols. For silt and blew dust onto bounding hills. Broad valley is now irrigated and LOESSLIKE SILT DEPOSITS 17 Figure 15. Active sand dunes on flood plain of Nyang Qu about 7 km southeast of Xigaze, Xizang Autonomous Region, China. View south. Photograph PK 23,871 by Troy L. Péwé, June 8, 1980. supports crops for city of Gyangse, foreground, in center of valley. Note steep-sided gullies cut into fans of retransported loess along left edge of river valley. Surrounding hills composed of Jurassic and Cretaceous sandstone, limestone, and shale. Photographs 4572-4576 by Troy L. Péwé, June 8, 1980. 18 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA About 10 to 20 m of loesslike silt covers the lower northwestern slope of a hill at Yangbajain at the southern foot of the Nyaingentanglha Shan (fig. 3; Academia Sini- ca, 1980, fig. II-2; Wang and Li, 1983, fig. 2-25, p. 52). Loesslike silt is most widely distributed where it spreads out as fans 1 to 2 m thick on or near the edges of flat, gently sloping valley floors. The fans are rarely all silt; they include layers of sand and fine gravel that permit higher angle fans than would be preserved if they were composed only of silt. Narrow, steep-walled gullies in the loesslike silt are similar, but on a much smaller scale, to the spectacular results of loess erosion in north-central China (Wang and Zhang, 1980; Wang and Song, 1983). Except along terrac- es and valley-bottom deposits, thickness is difficult to measure; however, studies of gullies in the fans indicate that the loesslike silt is about 1 to 6 m thick. Reports of loesslike silt on the plateau outside of the route of our 1980 investigation are not common. Zhao and others (1976, p. 20), on the basis of early expeditions to the Himalayas in 1966-68, report a thin (1-m) loesslike silt overlying the first, second, and outwash terraces of Gyirong valley in the southwest part of the plateau (fig. 3). They also describe as much as 10 m of eolian loesslike silt and colluvium overlying the terrain near the bridge over the Pum Qu, 5 km south of Tingri (fig. 3). Also from the Hima- layan expeditions in the 1970's, Wang and Li (1983) de- scribe loesslike deposits on slopes in the southern Xizang Autonomous Region. Xu (1980, 1981) reports what he called extensive "periglacial" loess deposited in late Pleis- tocene time on the slopes of the Tanggula Shan (figs. 2, 5) in the central part of the Qinghai-Xizang (Tibet) Plateau. Derbyshire (1985) records local thin loess deposits in Qing- hai Province northeast of the plateau (fig. 1). Huang (1980) reports yellowish sandy soil at Tingri and as much as 9 m of loessic mud and sandy soil at Xie- gal. At Yangbajain (fig. 3) he reports 19.8 m of yellowish clay over yellow silty sand with some gravel and mamma- lian fossils (Huang, 1980, p. 50). He believes both the silt here and the loesslike silt at Lingtse County (area to the Figure 16. Climbing sand dunes on lower flanks of west-facing hills of Tertiary granite along braided Lhasa He 20 km southwest of Lhasa, Xizang Autonomous Region, China. Photograph 4547 by Troy L. Péwé, June 3, 1980; reprinted from Péwé (1981) and pub- lished with permission of "Geotimes." LOESSLIKE SILT DEPOSITS 19 east around Nyingchi) are retransported loess (Huang Wanpo, oral commun., 1986). Yellowish silt is used for making brick and tile at Nyingchi. The thickest deposits of loesslike silt are in valley bottoms and are commonly exposed in terrace scarps cut by streams (fig. 21). The deposits vary from relatively pure yellowish loess in vertical cliffs 10 m high to yellow- ish, fluvial, silty sand and silty clays that vary from less than a meter to 5 m thick. The most striking and most extensive deposits of thick loesslike silt occur in the bottom of the Nyang Qu valley near its junction with the Yarlung Zangbo at Xigaze (fig. 25). There the Nyang Qu has cut down into a valley- bottom deposit of retransported silt as thick as 15 m. The eroded edges of the terraces on each side of the valley are vertical walls 15 m high of bedded silt with some sandy layers. The most striking land forms are steep-sided flat- topped buttes and mesas, the latter as much as 1 km in diameter, that remain in the valley bottom (fig. 26). These erosional remnants clearly exhibit the vertically jointed re- transported silt. The tops of the erosional remnants are bare of vegetation, perhaps because they are well drained and very dry. Similar land forms in retransported loess have developed on the south side of the Tian Shan in the Xinjiang Uygur Autonomous Region (Péwé, 1987). Probably the most thoroughly studied section of re- transported loess in the Qinghai-Xizang (Tibet) Plateau is along the main highway at the south edge of Xigaze (fig. 27). About 15 m of alternating beds of retransported loess, sandy loess, and sand are exposed, and the upper- most 6 m are easily accessible for collecting samples and fossils. Beds of relatively pure, cliff-forming, yellowish silt as thick as 3 m are quite impressive (fig. 28). COLOR AND TEXTURE The loesslike silt is commonly yellowish tan (fig. 28), although grayish-tan is not rare; when wet it is brown. In many localities thin, dark, carbonaceous and iron-stained bands or mottling are present. Mechanical composition of the loesslike silt and asso- ciated sandy silt and sand was measured on 17 samples from the plateau and 3 samples of loess and retransported loess Figure 17. Sharply dissected fans of retransported loess at base of low west-facing mountains east of Tingri, Xizang Autonomous Region, China. Photograph 4584 by Troy L. Péwé, June 11, 1980. 20 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA Figure 18. Large dissected fan of retransported loess below site of Tibetan monastery and forti- fications of Xigar-Zhong, seat of district administration during Ching Dynasty, at Tingri in south-central Xizang Autonomous Region, China. Mountain of tilted shale and sandstone of Jurassic age. Note difference in use of fan for building sites in each picture. A, Early morning sun strikes remnants of once-splendid fortifications high on mountain. Loess is easily excavated, and small entrances to storage rooms can be seen in scarp at base of fan. Fan has been planed off in steps to accommodate modern buildings as city of Tingri expands. Photograph PK 23,950 by Troy L. Péwé, June 12, 1980. from elsewhere in China (table 1). Grain-size analyses of samples 1 through 10, and a sample of loess from near Heidelberg, Germany, were made by wet sieving to separate the >0.062 mm fraction, then pipetting that fraction at whole phi-size intervals according to the method of Folk (1974). Samples 11 through 17 were wet sieved, and calcium car- bonate was removed before grain-size analyses were made by hydrometer. Samples 18 and 19 were wet sieved and further analyzed by the hydrometer method. Loess sample 20 from near Beijing was analyzed by sedimentation in water. The powdery loesslike silty sediment is widespread in fans of different localities, shapes, and sizes and on LOESSLIKE SILT DEPOSITS 21 Figure 18. B, View southeast toward intact monastery and fortifications on same hillside in 1922. Copyright photograph 092207-D525 by T.G. Longstaff, MEE, while on reconnaissance expedition to Mount Everest. Published with permission of Royal Geographic Society, London. Figure 19. Newly made bricks of loess in foreground, loess-brick walls around house in center. View west across valley of Nyang Qu near Bainang, about 45 km southeast of Xigaze, Xizang Autonomous Region, China. Photograph PK 23,878 by Troy L. Péwé, June 8, 1980. 22 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA Table 1. Data on Quaternary sediment samples collected from southern Qinghai-Xizang (Tibet) Plateau, China, and from other parts of China. [Samples collected by TL. Péwé except as noted. Mineralogy analysis includes clay minerals and rock fragments.] Analyses performed Sample No. Material Date Sample site Position and stratigraphy at site Age (figure) collected Granulo- Mineral- Pollen Ostra- Mol- Verte- metric ogy codes lusca brates 1 Retransported 6/8/80 - Bainang Agricultural Station, _ Surface sample in field at base - Holocene.... X X (figs. 3, 34) - sandy loess. elevation 3,900 m, in Nyang of alluvial fan of silt with Qu Valley, 40 km NW of some gravel. Gyangze. 2 Silty sand.......... 6/8/80 _ In Nyang Qu Valley, 22 km Surface sample in silty field at - Holocene....... X X (figs. 3, 34) upstream from Xigaze. base of rock cliff on low terrace. 3 Sandy loess....... 6/9/80 _ On bedrock hill 60 m above At depth of 2.4 m in loess Late X X (figs. 3, 22, river at junction of Nyang Qu __ plastered on hill of volcani- Pleistocene. 25, 34) and Yarlung Zangbo, north clastic sandstone. side of Xigaze. 4 Sandy loess ...... 6/9/80 _ On flank of bedrock hill 15 m _ At depth of 2 m in gully of Late X X (figs. 3, 34) above and adjacent to loess blanketing a bedrock Pleistocene. Yarlung Zangbo, 50 km hill of limestone. downstream from Zigaze on south side of river. 5 Retransported 6/9/80 _ On face of terraces of silt and _ Unit III, 3.6 m below surface in Late X X X X (figs. 3, 27, - loess (same sand (distal end of fans from bank of retransported fossilif- _ Pleistocene. 34) as sample No. hills), west side of Nyang Qu erous loess and alluvial sand. 15). at south edge of Xigaze. 6 Retransported 6/10/80 Agricultural field on Near-surface sample on very Holocene.... X X (figs. 3, 34) - loess. retransported loess at village gently sloping fan. of Netang, 14 km west of Xigaze. 7 Retransported 6/10/80 1 km west of village of Jaling, _ At depth of 3.0 m in broad fan _ Holocene.... X X (figs. 3, 34) - loess. 75 km west of Xigaze. of black gravel of angular clasts, 2-4 cm diam.; gravel 2.5 m thick over 2.5-m layer of loess with 0.3-cm zone of red soil at base. 8 Retransported 6/11/80 Dissected alluvial fan of silt 15 - Surface silt on fan.................... Holocene.... (figs. 3, 34) - loess. km east of Lhaze. 9 Silty pebbly 6/12/80 70 km west of Tingri on south _ Surface-sediment sample...... Holocene.... X (figs. 3, 34) - retransported side of Pum Qu valley. loess. 10 Eolian sand '...... 6/17/80 45 km southwest of Lhasa Surface-sediment sample...... Holocene.... X X (figs. 3, 34) along west side of Lhasa He. 11 Retransported 6/18/80 Terrace on west side of Nyang - Unit VI, 0.9 m below surface.. - Late X X X (figs. 3, 27) - loess". Qu, south side of Xigaze. Pleistocene. 12 Silty sand bolls. 6/18/80 Terrace on west side of Nyang - Unit V, 1.4 m below surface.. - Late X X X (figs. 3, 27) Qu, south side of Xigaze. Pleistocene. 13 Retransported 6/18/80 Terrace on west side of Nyang - Unit IV, 1.8 m below surface.. Late X X X (figs. 3, 27) - loess'. Qu, south side of Xigaze. Pleistocene. 14 Retransported 6/18/80 Terrace on west side of Nyang - Unit IV, 2-3 m below surface. - Late X X X (figs. 3, 27) - loess!. Qu, south side of Xigaze. Pleistocene. 15 Retransported 6/18/80 Terrace on west side of Nyang _ Unit III, 3.6 m below surface.... Late X X X (figs. 3, 27) - loess (same Qu, south side of Xigaze. Pleistocene. as lsample No. 16 Retransported 6/18/80 Terrace on west side of Nyang _ Unit II, 4.8 m below surface.... Late X X X (figs. 3, 27) - loess and Qu, south side of Xigaze. Pleistocene. sandy loess 1, 17 Crossbedded 6/18/80 Terrace on west side of Nyang _ Unit I, 5.5 m below surface...... Late X X X (figs. 3, 27) - sand '. Qu, south side of Xigaze. Pleistocene. 18 Retransported 8/6/86 _ Ancient city of Jiaohe, near 30-m vertical loess wall of Late X (figs. 5, 42, - loess. Turpan, Xinjiang Uygar small creek valley. Sample Pleistocene. 45 Autonomous Region. from 2.0 m below surface. 19 Retransported 8/10/86 West bank of Sigh He at Kuga, _ Alluvial loess and gravel Late X (figs. 5, 45 _ gravely loess. north of Tarim Basin, layers, 1-3 m thick. Sample Pleistocene. Xinjiang Uygar Autonomous from 3 m below surface. Region. 20 LOSS 1973 Near BEIJING... Surface sample.......................... Late X x3 (fig. 1) Pleistocene. ! Collected by Liu Tungsheng and Li Bingyuan. Collected by Roger J.E. Brown. *See PEwé and Journaux (1983) for mineral analyses. LOESSLIKE SILT DEPOSITS 23 hills of different bedrock types. The silt varies widely in texture and sorting from clay-rich to sand-rich and even to gravel-rich, depending upon nearness to the original silt source and the various distances of retransportation and mixing with other sediments by fluvial and colluvial slope movement. Silt near the hills or mountain slopes is coarser and interbedded with more coarse beds than that at the toe of low-angle silt fans near the center of the valleys. In great contrast to the uniform mechanical composition and texture of loess throughout the world (table 2; Péwé and Journaux, 1983, their figs. 28, 29), this retransported loess is less well sorted and generally has a lower percentage of silt-size particles (table 3, figs. 29, 30) than in situ loess. Percentages of grain sizes in silt samples vary: 8 to 55 Figure 20. Remnants of loess-brick walls of 13th century monastery about 10 km west of Xigaze, Xizang Autonomous Region, China. View west. Photograph PK 23,918 by Troy L. Péwé, June 10, 1980. 2 kilometers and grave Figure 21. Diagrammatic sketch of typical dissected loesslike silt with lenses and layers of fluvial sand and gravel in valley bottoms along major streams in southern Qinghai-Xizang (Tibet) Plateau, China. Vertical exaggeration 33%. Diagram by S.M. Selkirk. 24 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA percent sand; 26 to 75 percent silt; and 2 to 67 percent clay (table 3). However, loess on or flanking bedrock knobs or hill- tops near the Yarlung Zangbo was retransported little or not at all; it is well sorted and has a relatively high sand content and a low clay content (fig. 31). The silt has the same texture throughout the section from the surface to the bedrock. Tex- ture and cumulative-frequency grain-size curves of the sandy loess from near rivers (samples 3 and 4) are similar to those of sandy loess adjacent to braided streams in Alaska, Siberia, and elsewhere (Péwé and Holmes, 1964; Trainer, 1961; Péwé and Journaux, 1983). Samples 5, 7, 13, 14, 15, and 16 have more than 40 percent clay-size particles (figs. 30, 32) and are typical of silt deposits on fans. Huang (1980) refers to loessic mud at Tingri and yellowish clay at Yangbajain. Sample 6, which has the highest silt content (fig. 30) and is best sorted, is from an agricultural field on the toe of the slope far from the mountains. Some samples have fluvial silty sand and some have eolian sand (fig. 33). Fluvial sand is interbedded with the loesslike silt in many localities. Silt grains in the loesslike silt are angular, and the sediment has little or no cementation. A few calcareous concretions (loess kindshen) are present locally. The presence of an expandable mixed-layer clay and stringers of calcite and small roots in some samples indicates that these sediments have been subject to soil-forming processes. MINERAL COMPOSITION Mineralogical analyses of 10 sediment samples from southern Qinghai-Xizang (Tibet) Plateau were compared with that of a control sample of loess from Germany. Min- eral composition of the sand-size fraction of samples 1 through 10 (table 4) was determined by impregnating the sand in epoxy resin, which allowed thin sections to be pre- pared, then using a petrographic microscope for point counting 200 grains. Petrographic examination reveals that the silt grains, although slightly iron stained, are angular and fresh. Figure 22. Loess overlying bedrock on knob 15 m above the Nyang Qu at Xigaze, Xizang Autonomous Region, China. Collection site of sample 3 (table 1). Photograph PK 23,896 by Troy L. Péwé, June 9, 1980. LOESSLIKE SILT DEPOSITS Table 2. Mechanical properties of sand and loesslike-silt samples from southern Qinghai-Xizang (Tibet) Plateau, China (Nos. 1-19), and loess and ash from other parts of the world. [Sedimentological terms from Folk (1974). Compiled by Mu Guijin from cumulative-frequency curves (figs. 30-33). See table 1 for sampling information.] Sediment Graphic mean _ Median Inclusive graphic Inclusive graphic - Graphic kurtosis sample Material (Mz) 6 (Md) 6 standard deviation skewness (Kg) No. (sorting) (01) 6 (5) 1 Retransported sandy 5.42 4.84 2.00 0.43 0.96 2 Silty SAM.... ...... cece nner nne 4.82 3.82 2.65 .91 .91 3 SANGY emmm nnn enn 4.58 4.58 1.08 19 1.14 4 SAMGY 10€8S........c cece nne neenee 4.35 4.13 2.15 23 1.50 5 10C$$S.... ..... c.. 7.62 7.90 2.10 -.27 .87 6 Retransported loess.. 5.32 5.35 1.46 -.05 1.77 7 Retransported loess.. 7.84 7.96 1.79 -.16 T7 8 RetranSPOTt@d ...s ccc 7.45 7.48 1.49 -.04 90 9 Silty pebbly retransported loess.................... 6.176 6.72 2.02 -.O1 .84 10 EON SMG... 3.18 3.12 1.80 .26 2.13 11 REtranSPOFt@d 6.34 5.80 2.24 31 T2 12 Silty No analyses available SMU.. seee cece enne nene neenee nes 13 REtraNSPOTt@d ..... 7.03 7.26 1.92 -15 .88 14 REtranSPOTt@d 10@SS............ ccm 6.57 6.65 2.99 -11 .63 15 REtraNSPOFt@d mene 7.31 8.15 2.35 -A5 .84 16 Retransported loess and sandy loess............ 6.2 6.50 3.18 -16 57 17 CrOSSb@dded SAMG.......... ...er No analyses available 18 Retransported loess......... 7.83 7.46 2.07 23 1.1 19 Retransported gravelly loess! 5.82 5.58 2.55 .064 1.51 20 Lo€SS, BGijing, eee 7.94 7.65 2.84 -.1 83 DUSt, ..... cee neer neer 5.42 5.50 0.95 A7 1.02 Volcanic ash, Fairbanks, Alaska, U.S.A.... 6.24 6.15 1.28 .03 1.23 Dust, Kansas, U.S.A.... « 6.84 5.90 2.41 44 1.26 Dust, ArIZO0N&, U.S.A.... ...s 7.8 6.65 2.87 A6 1.47 LO€SS, NEW 8.1 7.80 2.83 J .92 6.24 5.32 1.91 .63 1.78 Loess, former Czechoslovakia... 7.37 5.95 2.96 .56 .98 LOGSS, AIASKA, U.S.A.... sccm 5.84 5.50 1.36 .26 1.12 LOGSS, UZDEKISE@N...... ccc 5.45 5.11 1.48 .52 1.7 Loess, Illinois, U.S.A.. .. 6.07 5.95 1.23 .08 1.64 Loess, Siberia, Russia. . 5.58 5.27 1.87 43 1.86 LOESS, 5.40 5.38 .84 .09 1.59 25 lOriginal sample included about 15 percent gravel larger than 2 mm, which was removed before this analysis. Composition of bulk (not size-fraction) samples was determined by air drying, crushing to a uniform fine powder, then using X-ray diffraction analysis by using a Phillips APD-3500 instrument. Bulk samples consist mainly of quartz, feldspar, calcite, dolomite, micas, and clay minerals. Estimates of percentage abundance of the minerals (table 5) were made by measuring peak areas of the 4.25-A mica plus clay minerals, 3.36-A quartz, 3.25-A potassium feldspar, 3.20-A plagioclase, 3.03-A calcite, and 2.88-A dolomite diffractogram peaks and by using the weighting factors of Schultz (1964). In order to correct for the 3.36-A illite peak that overlaps that of quartz, the 10-A illite peak was calcu- lated and subtracted from that of the 3.36-A peak because 10-A and 3.36-A illite peaks occur in roughly equal propor- tions (ASTM card 9-334). Composition of the clay-size (<0.005-mm) fraction of samples, separated from bulk samples by centrifuging, was also determined by X-ray diffraction analysis of un- treated and glycolated samples using a scanning speed of 1° 20' per minute. To differentiate chlorite and kaolinite, slow scans (0.20°20' per minute) were run over the range 24° to 26°20" (3.5-A peak) (Biscayne, 1965). This fraction consists of smectite, kaolinite, and chlorite with traces of quartz, feldspar, and calcite. X-ray patterns of some sam- ples indicate the presence of a small amount of mixed- layer illite/smectite or smectite/vermiculite. For compara- tive purposes, the relative abundances of the major clay minerals were determined by measuring and applying weighting factors to the following peak areas (peak height multiplied by peak width at half height) on the glycolated 26 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA pattern: The 17-A smectite peak; four times the 10-A illite peak; and twice the 7-A kaolinite plus chlorite peak (Bis- cayne, 1965) (table 6). The kaolinite/chlorite ratio was determined by measuring areas of the 3.54-A chlorite and 3.58-A kaolinite peaks on the slow-scan diffractogram and apportioning the ratio to the 7-A peak area (Biscayne, 1964). The four weighted peak areas were summed, and each is expressed as a percentage of the total, as 100 per- cent clay minerals. In that part of southern Qinghai-Xizang (Tibet) Pla- teau (figs. 2 and 3) where samples 1 through 17 were col- lected, east-trending belts of Mesozoic sedimentary rocks make up the bulk of potential source rocks for sediment samples in the area. An east-trending belt of plutonic rocks occurs to the north, and a belt of ophiolitic rocks occurs close to many of the sampling sites in the center of the mapped area (fig. 344). Composition of the sand-size fraction of the sedi- ments, particularly varieties of lithic fragments that were identified (table 4), can be used as a rough indicator of sediment sources. However, direct comparisons of sand- fraction petrology cannot be made because the data also reflect variations in composition due to grain-size varia- tions among the sand fractions analyzed (see Slatt and Eyles, 1981, for a discussion of variations in composition of glacially derived sands of different grain size; also see Péwé and Journaux, 1983, for a discussion of mineralogi- cal differences that do not reflect size fractions of loess in Siberia). For example, samples 6 and 7 contain an abun- dance of igneous lithic fragments (table 4) that are of coarse-sand size. Most other samples contain only fine- sand-sized particles, which are enriched in quartz, shale, and carbonate lithic fragments. Although these sands may have had a plutonic or volcanic source, such fragments Figure 23. Deep, sharp-walled gully 3 m wide at top, dissected into retransported loess fan at base of limestone knob 50 km downstream from Xigaze on the Yarlung Zangbo, Xizang Autono- mous Region, China. View east of collection site 4 (table 1). Photograph PK 23,905 by Troy L. Péwé, June 9, 1980. Table 4. Mineral composition of sand-size (>0.062-mm) fraction of sand and loesslike silt from southern Qinghai-Xizang (Tibet) Plateau, China LOESSLIKE SILT DEPOSITS Table 3. Percentage of sand, silt, and clay in sand and loesslike- silt samples from southern Qinghai-Xizang (Tibet) Plateau, China, and loess from Beijing, China. [U.S. Department of Agriculture classification: sand, 1.0-0.05 mm; silt 0.05-0.005 mm; clay, 0.005-0.002 mm. See table l for sampling infor- mation.] Sample Material Sand Silt Clay Total No. 1 Retransported sandy 36.5 47.2 16.3 100.0 loess. 2 Silty sand.... 57.5 25.5 17.0 100.0 3 Sandy loess. 220 43.0 54.5 2.5 100.0 4 Sandy loess................ 55.0 34.4 10.6 100.0 5 Retransported loess... 8.5 36.8 54.7 100.0 6 Retransported loess... 19.2 75.3 5.5 100.0 7 Retransported loess... 3.8 40.2 56.0 100.0 8 Retransported loess... 2.5 50.2 47.3 100.0 9 Silty pebbly retrans- 11.5 53.5 35.0 100.0 ported loess. 10 Eolian sand................ 82.5 10.3 7.2 100.0 11 Retransported loess... _ 20.0 50.0 30.0 100.0 12 Silty 80.0 13.5 6.5 100.0 SANO. .css 13 Retransported loess... 9.0 48.5 41.5 100.0 14 Retransported loess... _ 30.5 28.0 40.5 100.0 15 Retransported loess... 16.0 18.5 64.5 100.0 16 Retransported loess 37.0 21.5 41.5 100.0 and sandy loess. 17 Crossbedded sand...... 99.0 1.0 .0 100.0 18 Retransported loess... 3.2 51.6 21.5 100.0 19 Retransported 20.0 58.5 21.5 100.0 gravelly loess! . 20 Loess, Beijing, 25.0 49.5 25.5 100.0 China. 'Original sample included about 15 percent gravel larger than 2 mm, which was removed before this analysis. 27 would not be fine-sand-sized particles. To compare com- positions of the ten sands accurately, only one sand-size fraction should have been studied petrographically: however, in most instances insufficient material was avail- able to make such a fractionation. In addition, only a few sand samples contain any coarse sand, which is the most valuable fraction for recognizing lithic fragments. Nevertheless, the data in table 4 show that a wide variety of rocks contributed to the sediments. Samples 6 and 7 are rich in mafic plutonic and volcanic lithic frag- ments, probably derived from ophiolite terrane. Enrich- ment of sample 10 in felsic plutonic fragments, as well as in feldspar, indicates derivation from the nearby plutonic terrane. Samples 4, 5, and 8 are enriched in carbonate lith- ic fragments, which could have been derived from the same sequences. The lithic fragments in samples 1 to 10 do not suggest a major plutonic source; however, much of the quartz could have such a derivation. Most quartz grains are angular, suggesting that they are first-cycle de- tritus; however, samples enriched in carbonate lithic frag- ments also contain a high proportion of multicycle rounded quartz grains. The high proportion of sand-size chlorite in samples 1, 2, and 7 probably reflects derivation from mafic ophiolitic rocks. Composition of the clay-size fraction of the sedi- ments (table 6) can be used to compare probable sediment sources. Relative proportions of the clay minerals are pre- sented in figure 34B. Samples 1, 2, 7, and 10 contain the highest proportion of smectite, and samples 1 and 7 con- tain the highest proportions of chlorite; these proportions indicate at least partial derivation from mafic ophiolitic rocks. The site of sample 9, which has the lowest chlorite content, is farthest from the ophiolite terrane (fig. 344). [Calculations in weight percent. Heavy minerals include mainly amphiboles, pyroxenes, and opaque minerals. Tr, trace; -, not detected. Petrographic analyses by Cities Service Company, Tulsa, Okla. Samples collected by TL. Péwé unless otherwise noted. See table 1 and figures 3, 27, and 34 for sampling sites.] Lithic fragments Quartz Feldspar Micas Igneous Sedimentary Metamorphic Sample Heavy No. 3 o o 9 C minerals 10 $ 5 10 $o f 2 S 3 o a a = - # 500 $0 B 4 ., 0 ib fb ® § a 0 §§ 3 § pee go go 0,2 & e -or $8 z 8 - o 9 60 g £0 2 p 2 30 3 30 § $o ff € S Sf bs 2 6 - RA - aA {o 2 - % a ma - 0 0 a 6 - a 1 35 7 3 3 - Tr 1 24 - 3 14 5 5 2 43 9 1 2 Tr Tr 4 2 - 30 2 6 3 57 17 I 4 2 - - Tr - - 8 2 9 4 43 7 5 - - h - - 2 1 - 36 - 2 2 2 5 21 8 I I 2 - 5 3 2 53 - 2 2 Tr 6 15 4 - 2 Tr 2 22 37 - 6 - I - 7 Tr 4 7 26 14 Tr 3 Tr 1 4 29 - 3 3 Tr Tr 14 1 2 8 33 20 - 1 - - - 7 Tr 7 - 14 - 3 3 11 9 40 22 2 2 - - - Tr - 4 7 2 4 Tr 6 10 10' 47 7 4 6 3 9 -o- Tr 1 40 Tr 1 3 8 7 'Collected by Liu Tungsheng and Li Bingyuan. 28 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA Table 5. Mineral composition of bulk samples of sand and loesslike Table 6. Mineral composition by weighted peak-area percentages of silt from southern Qinghai-Xizang (Tibet) Plateau, China (Nos. 1- clay-size (<0.005-mm) fraction of sand and loesslike silt from southern 10), and from Heidelberg, Germany. Qinghai-Xizang (Tibet) Plateau, China. [Calculations in weight percent; -, not detected. X-ray diffraction analyses [Calculations in weight percent. Tr, trace; -, not detected. X-ray diffraction by Cities Service Company, Tulsa, Okla. Samples collected by T.L Péwé analyses by Cities Service Company, Tulsa, Okla. Samples collected by T.L u.nless otherwise noted. See table 1 and figures 3, 27, and 34 for sampling Péwé unless otherwise noted. See table 1 and figures 3, 27, and 34 for sites.] sampling sites.] Sample Quartz - Potassium Plagio- Calcite Dolomite - Micas and Sample - Smectite Illite Kaolinite Chlorite Quartz Feldspar Calcite No. feldspar clase clay minerals No. 1 33 - 11 2 3 51 1 9 48 18 25 Tr Tr - 2 44 9 8 5 3 31 2 16 57 15 11 Tr Tr - 3 32 1 9 7 11 40 3 - 77 7 16 Tr Tr - 4 29 6 25 7 1 32 4 - 69 10 21 Tr Tr - 5 31 6 17 1 45 5 - 77 9 14 Tr Tr - 6 35 10 13 6 - 36 6 Tr 81 - 19 Tr Tr - 7 30 - 8 2 - 60 7 5 68 - 27 Tr Tr - 8 34 1 5 4 1 55 8 - 86 - 14 Tr Tr Tr 9 18 2 8 53 - 19 9 - 92 3 5 Tr Tr Tr 10' 33 22 29 s - 16 10' 8 64 7 21 Tr Tr - Germany 29 3 6 27 32 14 'Collected by Liu Tungsheng and Li Bingyuan. 'Collected by Liu Tungsheng and Li Bingyuan. Figure 24. Alluvial fan of retransported loess sharply dissected by steep-walled gullies as much as 6 m deep. Fans flank limestone bedrock knob 50 km downstream from Xigaze on the Yarlung Zangbo, Xizang Autonomous Region, China. Photograph PK 23,904 by Troy L. Péwé, June 9, 1980. LOESSLIKE SILT DEPOSITS 29 0 200 400 _ 600 SCALE 800 1000m Vertical exaggeration 7x Figure 25. Physiographic diagram of Xigaze area at junction of Nyang Qu and Yarlung Zangbo, Xizang Autonomous Region, China. Terrace remnants of loesslike silt are about 15 m high. No. 3 is site of loess sample 3 (table 1; fig. 22), and "Fossil site" is a fossiliferous terrace (figs. 27, 37). Diagram by S.M. Selkirk from map prepared by Li Bingyuan. Appreciable kaolinite is present in samples 1 to 5 and 10 and is absent in samples 6 to 8 to the west. This distri- bution indicates at least two different types of source rocks. It is possible that the kaolinite is derived from weathered rocks of the northern plutonic terrain, but it may also be a second-cycle kaolinite derived from sedimentary rocks. Some information on source terrains can be gained by comparing the compositions of bulk-sediment samples (fig. 34C, table 5), although, like the sand fraction, bulk compositions are also affected by variations in grain-size distribution among the samples (fig. 29). Sample 9 is rich in calcite, which possibly suggests a local limestone source (Jurassic sediments) quite unlike that of other sam- ples. The feldspar content tends to diminish in sediments from east to west, generally in sampling sites that are far- ther from the plutonic terrain (fig. 344). Samples 4 and 10, located closest to the plutonic terrain, contain the most feldspar (plagioclase is greater than orthoclase). Variations in the feldspar content probably reflect a variable supply of plutonic detritus to the sediments. Combining the above compositional data for sand, clay-mineral, and bulk samples allows some generaliza- tions about sediment sources and the extent of sediment transport. Samples 1, 2, 7, and possibly 6 have a strong source component of ophiolitic rocks. Samples 3, 4, 5, and 10 are at least partially derived from plutonic as well as Cretaceous sedimentary rocks (particularly carbonate rocks), and samples 8 and 9 are probably derived mainly from Triassic and Jurassic sedimentary rocks, respectively. Thus, although the sediments were originally of wind- blown origin and have subsequently been reworked, they have not been transported far enough to lose the composi- tional imprint of a fairly local origin. Figure 26. Flat-topped, steep-walled, 15-m-high remnant of bedded, vertically jointed retransported loess in bottom of Nyang Qu valley near its junction with Yarlung Zangbo on north side of Xigaze, Xizang Autonomous Region, China. View to southwest from bedrock knob near site of sample 3 (figs. 22, 25). Photo- graph PK 23,897 by Troy L. Péwé, June 9, 1980. 30 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA E ___A TERRACE f B Area of detailed stratigraphic section (C) Highway 2 1 " F29°10 E Millwlw 2 l|l|“|l“1‘\| w u 194 o da y! UM'HMI'H { |\ Hill)“ Nyang Qu I|| I‘lll ll + t wth - § ll| |{\| l‘, __7— II Lower shell bed 2 & 3 w Q _> 6 T < i1- v 5 o 10 - & I p= 9 - | 0 1 EXPLANATION Fine sand, silt and silty clay (retransported loess and sandy loess) 12 - Sediment-sample site & - Mollusca fossils Silt, and silty clay (retransported loess) II - Stratigraphic unit g Bird and mammal fossils Coarse sand Figure 27. Sketch map of flood plain and terraces _ side of Xigaze, Xizang Autonomous Region, China. (A) and stratigraphic sections (B, C) of retransported - Note collection sites of sediment samples 11 to 17 sandy loess, silty clay, and coarse sand cropping out (table 1) and of vertebrate and invertebrate fossils on terrace scarp on west side of the Nyang Qu, south (see fig. 28). LOESSLIKE SILT DEPOSITS Figure 28. Cliff-forming retransported tan loess exposed in road cut near Army fuel station along Xigase-Lhasa Highway on south side of Xigaze, Xizang Autonomous Region, China. This classic section is site of bird fossils, thin layers of fresh-water mollusk shells near middle and lower part of the exposure, and numerous sediment and pollen samples (see figs. 27, 37). Note wall made of bricks of loess in upper right. Inspecting scientists are mem- bers of International Union of Quaternary Research (INQUA) field excursion. View toward northeast. Hills of clastic rocks of Cretaceous age in distance lie east of the Nyang Qu and south of the Yarlung Zangbo (figs. 25, 344). Photograph PK 29,937 by Troy L. Péwé, August 20, 1991. 32 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA SAND (>0.063mm) CLAY (<0.004mm) SILT (0.063-0.004mm) Figure 29. Proportions of sand, silt, and clay in sediment sam- ples 1 to 10 from south-central Qinghai-Xizang (Tibet) Plateau, China, and a loess sample from Germany (G). (See figs. 3 and 34 for sampling sites.) Diagram by S.M. Selkirk. In order to verify that sediment samples have in- deed been transported and not derived by in situ weath- ering, a bedrock sample was collected from beneath loess sample 3 (figs. 22, 344) on top of a 15-m-high bedrock knob at the junction of Yarlung Zangbo and Nyang Qu at Xigaze (fig. 3). Petrographic analysis indi- cates the rock is a volcaniclastic sandstone or lithic tuff of basaltic or andesitic provenance, which has undergone considerable alteration under either hydrothermal or low- grade regional-metamorphic conditions that resulted in the authigenic mineral assemblage of subgreenschist facies. The rock consists mostly of altered mafic to intermediate volcanic-rock fragments and plagioclase. Volcanic glass fragments are altered to quartz and pum- pellyite and lined with chlorite and sphene. Most of the quartz is authigenic, primary plagioclase is changed to albite, and chlorite and pumpellyite partially replace pla- gioclase. Results of X-ray-diffraction analyses of whole- rock and clay-mineral fractions of sample 3 (figs. 354, B) are compared with those of the underlying bedrock (fig 35B, D). Analysis of sample 3 shows that the loess at that location is not derived from the underlying bed- rock because (1) pumpellyite is present in the bedrock but absent in the sediment and (2) illite is present in the sediment but absent in the bedrock. FIELD RELATIONS Where the loesslike silt has been transported by flu- vial or colluvial processes or is in an original position of deposition on top of a low hill next to a river (as in fig. 22), it overlies various substrates with a sharp contact. Only on local hilltops is the silt massive, showing little or no stratification or jointing. Most of the loesslike silt has been retransported to lower slopes, alluvial fans, or valley bottoms, where it is thickest (figs. 21, 25, 26) and shows poorly to well-developed stratification. The stratifi- cation varies from iron-stained horizons, organic films, and color banding to well-developed sand-and-gravel lay- ers 10 to 200 cm thick in sharp contact with the silt beds. Nowhere does the stratification resemble lacustrine bed- ding. No tephra layers are known in the loesslike silt, and no major soil horizons, such as are widely reported in cen- tral China (Liu and Chang, 1964; Liu and others, 1964, 1966; Liu and Wang, 1965; Liu and others, 1985; Liu, 1991; Rutter and others, 1991), were observed. A conspicuous feature of the loesslike silt is its abil- ity to stand in sheer cliffs (figs. 26, 27, 28). This charac- teristic is probably due to the angularity of the grains and to the strengthening by concretionary rods and tubes. This well-developed vertical cleavage of loess in China has been discussed from the beginning of the 20th century and attributed to vertical permeability by Willis (Willis and others, 1907) and to carbonate deposits around plant roots by Barbour (1925a, b). As in the loess elsewhere in China, vertical walls 10 to 15 m high in the valley-bottom facies of the loesslike silt at Xigaze have stood for many years. Jointing is not common, but vertical jointing typical of loess is well displayed in cliffs at Xigaze (fig. 28). Sharp, steep-walled gullies form in the loesslike silt and are most common on fans (figs. 17, 18, 23, 24). FOSSILS IN LOESSLIKE SILT Loess and retransported loess from many countries have long been known as excellent material for preserving Quaternary - fossils: Vertebrates, mainly - mammals, including frozen carcasses; invertebrates such as fresh- water and pulmonate mollusca, insects; and flora, pollen and mega-specimens. Fossils in the classic loess areas of north-central and eastern China are widespread and in- clude pollen, insects, ostracodes, mammals, birds (espe- cially large ostrich eggs) and hominids. Fossils in the loess of China have been known to the western world since the days of von Richthofen's work in the 1870's and from later studies (von Richthofen, 1877, 1882; Andersson, 1923; Barbour, 1925a, b, 1930; Liu and others, 1964, 1966; Liu, 1985; Chen and others, 1985; Xue, 1984). Reports of Quaternary fossils from the loesslike silt of the plateau are sparse, but observations are known from about the 1950's (Huang, 1980; Huang and Ji, 1981; Péwé, 1980a, b; Wang and Li, 1983). One of the most thoroughly studied fossil-bearing sections of loesslike silt in the plateau is a 6- m section exposed along the highway on the south side of Xigaze (figs. 27, 28). In 1980, Péwé, Liu, and Li, assisted by several scientists, collected sediment samples, fossils of CUMULATIVE FREQUENCY, IN PERCENT Figure 30. Cumulative-frequency grain-size curves (short dashes where projected) for retransported loesslike silt in southern part of Qinghai-Xizang (Tibet) Plateau, China. Clay-size grains have been incorporated during retranspor- tation, so curves are not typical of those for primary loess. (See figs. 3 and 34 for sampling sites.) CUMULATIVE FREQUENCY, IN PERCENT Figure 31. Cumulative-frequency grain-size curves (short dashed where projected) for samples of loess flanking or on top of bedrock knobs near the Yarlung Zangbo (Nos. 3, 4) and loesslike silt (No. 6) in agricul- 100 90 80 70 60 50 40 30 20 10 1.0 FOSSILS IN LOESSLIKE SILT EXPLANATION Sample No. atl RG 0.5 0.1 0.01 0.005 DIAMETER, IN MILLIMETERS 0.001 100 [TTT U U LAA _ so EXPLANATION L4A-<4--4~~~ ALL Sample No. / ~ ao 3 -f ___-_- 4 _____ z= } / 7 70 7 fl 60 ; / 50 j 40 ./ 30 20 # / # PM 10 - 1.0 0.5 0.1 0.01 _ 0.005 0.001 DIAMETER, IN MILLIMETERS tural field near toe of low-angle silt fan, Xizang Auton- omous Region, China. Curves are similar to those of primary loess. Dots are data points. (See figs. 3 and 34 for sampling sites.) 33 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA 100 LTT T U TI s Al _ so EXPLANATION z* J-" Sample No. / P 2-1 5 ,/// //’ _ 80 (18 & 00 || || -- .k, 14 / // i .--. - 45 A & 70 ___ - 16 7 £ a. RY zZ Asi y 60 .3;/ 1/ is A / py {A R 4 t A" & 3 ~ 1 A ; g ® 7A -." fra AG / / [re Pad , w 40 p 17- /- E lA ~ |C / 3 M 7 / / V/ g 30 [ Z7 24" 20 Z/ [= _- /, y/ yA {+ '// // y // /. 10 & »C - I// F 7 I7 ‘l ’// __/_:;'_. 0 1.0 0.5 0.1 0.05 0.01 0.005 0.001 0.0005 0.0001 DIAMETER, IN MILLIMETERS Figure 32. Cumulative-frequency grain-size curves (short dashed where projected) for clay-rich retransported loess, south-central Qinghai-Xizang (Tibet) Plateau, China. Many clay-sized grains have been incorporated during fluvial retransportation. (See figs. 3 and 34 for sampling sites.) 100 TTT TI [-F- --] EXPLANATION EH L- « 7 90 [-+ nz -> PL Sample No. Pag Poa / e 2 *~" « Bo - f--- _ 10 Z E _- -- - 12 w /// 8 4 z CC 70 mB w u « / o_ z a y- 60 lA y (€ / 2 8 1 / f / LW CC LL & w 40 2 © / 3 30 A E / hel G ] / 20 7 / / A1 _/ 10 / #A |" 4/// 0 1.0 0.5 0.1 0.05 0.01 0.005 0.001 0.0005 0.0001 DIAMETER, IN MILLIMETERS Figure 33. Cumulative-frequency grain-size curves for samples of eolian (No. 10) and fluvial (Nos. 2, 12) sand, south-central Qinghai-Xizang (Tibet) Plateau, China. Dots are data points. (See figs. 3 and 34 for sampling sites.) FOSSILS IN LOESSLIKE SILT 35 mollusca, mammals, and birds, and samples for study of pollen and ostracodes from this site. In 1991, Péwé led participants of the Qinghai-Xizang (Tibet) Plateau field trip of the 13th Congress of International Union for Quaternary Research (INQUA) to this site, and many of the scientists collected mollusca fossils. 89° 90° g1° Xizang Mice stl IL "*> - Autonomous |[~/ X +7 - ace x1 Region \T‘ AxvsT \. enina | NZ Ato iits ,_ > A \ LP _ Pe T.~ L121 PL Nk Lhasa Pratl B a Do's P ou 2 7 {cx ,‘ .,\/-«, N EW _ I] (11 W 60 KILOMETERS EXPLANATION - Quaternary fluvial and glacial sediments Triassic sedimentary rocks Contact 'f¢::] Cretaceous sedimentary rocks El Paleozoic(?) sedimentary rocks 04 Sampling site and number shales, and limestones s I K C Bo 0 100 PERCENTAGE CLAY MINERALS -J-] Jurassic sedimentary rocks Ophiolite belt Undivided Triassic and Jurassic sandstones, Granite and granodiorite PERCENTAGE MAJOR MINERALS Figure 34. Valley of Yarlung Zangbo and vicinity, south-central part of Qinghai-Xizang (Tibet) Plateau, China, from Lhasa to Tingri, showing sites of sediment samples 1 to 10 (white cir- cles). (See fig. 3 for geographic details.) A. Generalized geologic map from Academia Sinica (1980). B, Weighted peak-area per- centages of clay minerals in <0.005-mm fraction of sand and , 4 S _i loesslike silt samples. S, smectite; I, illite; K, kaolinite; C, chlo- rite. Quaternary sediments shown in black. C, Weight-percent- age estimates of major minerals in sand and loesslike silt samples. Q, quartz; F, feldspar (orthoclase and plagioclase); C, carbonate minerals (calcite and lesser dolomite). Quaternary sediments shown in black. 36 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA POLLEN Palynological research on loess of China began in the 1950's (Zhou and others, 1960) in the north-central part of the country and is now actively being pursued (Liu and others, 1985 p. 84-92). Pollen studies in the Qinghai- Xizang (Tibet) Plateau are not numerous, however. Huang and Ji (1981) and Wang and Fan (1987) report pollen in Holocene sediments near lakes in south Xizang Autono- mous Region. Holocene peats have also been examined (Academia Sinica, 1980, p. 43-44; Wang and Fan, 1987, p. 50-52). Huang Cixuan identified 25 types of pollen in seven sediment samples from the site above the highway on the south side of Xigaze in 1980. Her methods follow and her count and identifications are shown on figure 36. METHODOLOGY Samples of 100 to 200 grams were boiled for 1 to 3 minutes in a 5-percent solution of sodium hydroxide. After washing to neutral with distilled water, the samples were treated with dilute hydrochloric acid to dissolve the calci- um carbonate. After again washing in distilled water, the samples were centrifuged in three different heavy liquids: Densities 2.3, 2.1, and 2.0. The remaining heavy material was diluted with three to five times its volume of acetic acid, then washed with distilled water. Next, the material was treated with a mixture of one part sulfuric acid and nine parts acetic acid and then washed with distilled water to neutral. RESULTS Analyses revealed few pollen, except in samples 12 and 16 (figs. 27, 36). Woody species make up less than 10 percent of the total pollen, and herbaceous species make up more than 90 percent; the latter are rich in Ranuncu- laceae and Artemisia (fig. 36). These pollen, particularly Ephedra, Artemisia, Che- nopodiaceae and Palygonaceae, suggest a dry-cold steppe- like condition. However, several pollen types including 200; g160- g 160} 2 2 5,120- é 120} = = o Lu > 80} > 80} E E a 5 H ao) H 40} 0 0 200; 160 3160- g > 3 120 E E a 120} a b < W I 80 2 80} > 3 < m - 40 Lu C 40; CC o a 10 20 30 DEGREES 2-THETA BULK SAMPLE Figure 35. X-ray diffractograms of bulk and <0.005-mm fractions of loess sample 3 (A, C) and sample of underlying bedrock of volcaniclastic sandstone and lithic tuff (B, D) from rock knob 15 m high at junction of Nyang Qu and DEGREES 2-THETA <0.005-mm FRACTION OF SAMPLE Yarlung Zangbo, Xizang Autonomous Region, China. Ca, calcite; Ch, chlorite; D, dolomite; I, illite; PI, plagioclase; Pu, pumpellyite; Q, quartz. (See figs. 22, 25 for sampling sites.) FOSSILS IN LOESSLIKE SILT Ranunculaceae, Cyperaceae, and Gentianaceae, indicate meadows and local moist substrate such as small ponds and wet ground. Myriophyllum is aquatic. GASTROPODS Fossil snails are common in Chinese loess and have been described over many years (Chen and others, 1985). However, most if not all of the work has been restricted to the study of land gastropods. Although pulmanate gastro- pods live today in southern Xizang (Chen and others, 1985), we did not see any modern or fossil species in the loesslike silt. Péwé, Liu, Li, and associates collected many specimens of freshwater gastropods from two different zones in a 6-m-high exposure of loesslike silt on the south side of Xigaze (fig. 27, 37) in 1980. In both the upper and lower zones, the specimens were found in more or less horizontal layers a centimeter or two thick in relatively poorly stratified to almost unstratified silt containing some quartz and rock particles as much as 2 mm in diameter. The lower layer could be traced laterally for only about 10 m and the upper layer for 3 or 4 m. The lower of the two horizons has the highest con- centrations of gastropods and is near the level of the road. The most common type and the largest species is Radix auricularia (L.) (fig. 38). The collection contains both ju- venile and adult specimens that measure as large as 25 to 37 30 mm high and 20 to 30 mm wide. This is a white spiral gastropod; the last spiral is so large that it almost appears to form the entire shell. The larger specimens appear to be standard for this species (Zhadin, 1952, p. 118). The second most common species is a brownish or tan-colored, vertically coiled Fossari truncatula (Muller). It is much smaller than R. auricularia-only 5 to 7 mm high and only 3 or 4 mm wide. The least common gastropod is the small white Planorbidae, Gyraulus gredleri (Gredler). It is coiled in a rather flat horizontal plane, and these speci- mens are only about 1 to 3 mm in diameter. The upper zone of gastropods is 3 to 5 m higher in the section (fig. 27), and contains fewer specimens than the lower zone. Only two different species were found. Most common by far was Fossari truncatula (Muller), and there were a few Gyraulus gredleri (Gredler). Living specimens of these species are chiefly Eur- asian in distribution, in areas of cold fresh water, and they have been reported from Quaternary deposits in the former Soviet Union, Alaska, and northwest Canada (Péwé and Journaux, 1983; Clarke and Harington, 1978). F. truncatu- la also lives in Newfoundland and British Columbia. Un- fortunately, the terminology for these species is not always the same in different countries, and there has been a con- stant evolution of terminology over the years. This prob- lem has been carefully pointed out by Zhadin (1952) and by Clarke (1973, p. 483). For example, Radix auricularia (L.) is also known as Lymnaea auricularia (L.). Percentage (hachured) or T Sample number of pollen counted axa 3 o o 2 Depth o 3 2 o s |4|2 “gig SE .o . No. | below | & | § c s s| s lelfélflilGls| E]) € } surface| g. 3 3 |=) =| 5 o § o |Q 81.5] S| &l.«]|€] a 0 9 (m) 9 2 2 |.S|s| S| SSK] §)s| 5 F S 51> 8lel'E| Alt) %I 8| 8 0 3 5 8 |S] 3 3 x 5 = >< = £ o 3 < | O 0 10 20% 20 40 60 80% 0 10% 0 20 40 60 80% 20% 0 10%10%10%10% | 10% T T T T I T T I T d T 11 0.9 1 1 1 |+ + + + 12 1.4 a +| |+| - mee " n 4 13 1.8 14 | 2 26 | +] + + + +1+] + + 14 2.0 17 | 1 9 + + + 15 3.6 17 | 6 4 + + 16 4.8 2 VA + + ~- asma F I B | + + 17 5.5 15 | 35 5 + + +l+] +] + Figure 36. Pollen composition of sediment samples from stratigraphic section (fig. 27) on scarp of terrace on west side of Nyang Qu, south of Xigaze, Xizang Autonomous Region, China. Hachured bar, percentage sum of individual taxa clas- sified as Xerophytic (dry), Herbaceous (moist), or Pterido- phytic (fern). Solid bar, percentage of individual taxa in assemblage. +, species present but not counted. Analyses by Huang Cixuan. 38 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA Radix auricularia (L.) lives in the marginal zones of large and small bodies of water such as rivers and brooks, and has been reported in Siberia (Zhadin, 1952, p. 118). Lozek (1964, p. 177) reports that the species is Paleoarctic and lives in plant-rich standing water. Péwé has collected it from cold-water ponds in central Alaska near Fairbanks and in western Alaska near Galena. Coworkers have found it in Quaternary sediments else- where in Alaska. The retransported loess of late Wisconsin age in central Alaska is rich in organic re- mains, including some fresh-water gastropods. Mertie (1937, p. 192-193), Taber (1943, p. 1491), and Péwé (1975a, p. 88-89) record Stagnicola and Gyraulus gredleri in the retransported loess of central Alaska. Zha- din (1952, p. 125) mentions that Fossari truncatula (Muller) is widespread in the former U.S.S.R., is Paleo- arctic, and has been found in Quaternary deposits in both east and west Siberia. He states that this species lives in silty biotopes and in cold or hot springs, as well as in swamps, pools, and rice fields; it occurs at eleva- tions as high as 2,000 m above sea level. The gastropods at the Xigaze site, which is at 3,836 m, indicate that the ecological conditions under which they lived were main- ly shallow ponds, probably on very gentle slopes, low terraces, and the flood plain of the Nyang Qu in late Quaternary time. As the fine-grained silt was carried by sheet wash and small rivulets across the gentle slopes from the hillsides to the terraces and flood plain, the ponds became filled, and adjacent land surfaces were ag- graded as the sediment accumulated. Figure 37. Collecting fossils in loesslike silt at river-ter- race exposure south of Xigaze, Xizang Autonomous Re- gion, China. Fresh-water mollusks from lower shell bed, right-center foreground; rodent and bird fossils from sedi- ments beside collector at far upper left. Loess-brick wall in center background and a brick yard at upper far right. (Also see fig. 27, 28.) Photograph 4577 by Troy L. Péwé, June 9, 1980. ORIGIN OF LOESSLIKE SILT 39 VERTEBRATES Reports of Quaternary vertebrate fossils from Xizang Autonomous Region are not common. The few specimens collected in the 1970's during expeditions to the Qinghai-Xizang (Tibet) Plateau expeditions are report- ed by Huang (1980) and Huang and Ji (1981). Except for the Ovis sp. and Cervus reported from a cave in bedrock near Maizhokunggar (fig. 3), the remaining four localities reported by Huang (1980) are probably in loesslike silt. An Equus sp. tooth was found in yellowish soil at Tingri (fig. 3), and two lower right jaw bones of Alticola sp. were in or near vertically jointed loesslike mud at Xie- gal. Parts of Equus sp. and Bos sp. were recovered at Yangbajian from a 19-m-thick sediment layer believed by Huang (oral commun, 1986) to be retransported loess. A late Pleistocene metatarsal bone of Cervus sp. was recov- ered from a brick and tile plant at Nyingchi. Although de- scribed as from fluvial and lake deposits, the sediments are probably retransported loess because, as elsewhere in southern Xizang, it is used to make brick and tiles. A unique collection of vertebrate fossils was recov- ered by Péwé, Liu, Li, and associates from loesslike silt in an exposure near the top of a terrace along the Nyang Qu at the south side of Xigaze (figs. 25, 27, 28, 37). In silt, Figure 38. Fresh-water gastropods Radix auricularia (L.) from loesslike silt from stratigraphic section (fig. 28) on scarp of ter- race, west side of Nyang Qu, south of Xigaze, Xizang Autono- mous Region, China. Photograph PK 24,436 by Troy L. Péwé, Sept. 3, 1981. about 3 m below the surface, Bruno Messerli noted small, delicate vertebrate bones of four different birds and two different rodents, identified below, and the remnants of a bird nest. Asio flammeus, a short-eared owl. At least two very young individuals probably less than half grown and obvi- ously still nestlings at the time of death. cf. Alauda arvensis, a skylark. At least three individuals. These specimens agree rather well with Alauda and do not seem to be similar to Eremophila, Galerita, or Lullula. The specimens are larger than available specimens of Alauda gulgula and therefore are most similar to A. arvensis. Passer sp. (either montanus or rutilans), a sparrow. At least two individuals. Definitely too small for P. domesticus. Passeriformes gen. et sp. indet., a song bird. One ulna and possibly two coracoids and two tarsometatarsi are from a passerine intermediate in size between Alau- da and Passer. Microtus (Pitimys) blythi, a Blyth's vole. At least 11 indi- viduals. Cricetulus - kamensis, individual. The accumulation of vertebrate bones appears to be from a nest of the short-eared owl, and both nestlings and prey are represented. The species are all the same as those found in the southern plateau today. The owl and the lark are certain indicators of open country or grassland, and the owl may also indicate the presence of marshes. The area was probably one of a slowly aggrading flood-plain sur- face with small ponds and marshes. The bird and rodent bones in the old nest were covered by silt washed down from the slopes. Another interpretation is that the bones and the nest are much younger than the enclosing sedi- ments and were deposited in a hole in the face of the silt cliff. Although small birds do live in holes in loess, the typical short-eared owl does not. a Tibetan hamster. One ORIGIN OF LOESSLIKE SILT The origin of loess has been a controversial subject worldwide for about 100 years, and most of the discus- sions have centered on the loess in north-central China, central United States, and, later, Alaska and eastern Sibe- ria. Less is known about the probable origin of the wide- spread loesslike silt in the valley bottoms and on flanks of the lower hills in the unglaciated southern part of the Qinghai-Xizang (Tibet) Plateau. Lacustrine, fluvial, weath- ering, and eolian processes have been invoked to explain the loesslike silt. The weathering or residual hypothesis has little support. The lacustrine hypothesis, either sepa- rately or with fluvial processes, has been held since the late 1970's and 1980's (Academia Sinica, 1980). The 40 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA eolian hypothesis has been strongly suggested since 1980 (Péwé, 1980 a, b, 1981; and Péwé and others, 1981, 1985, 1987). Whatever the origin, it is evident that most of the silt has been retransported downslope by sheet wash, rill wash, creep, solifluction, and (or) by small ephemeral streams or rivulets. WEATHERING HYPOTHESIS In the early part of this century, L.S. Berg proposed and expounded in many publications of the former U.S.S.R. that loess forms in place as the underlying rock breaks down (summarized in Berg, 1964 [1960]). His hy- pothesis exerted a profound influence over many Soviet scientists, but now it seems to have little support in the western part of the former Soviet Union. In the 1950's, Soviet scientists carried the concept of the origin of loess by residual weathering to China. For example, Pavlinov (1959) stated that loess formed from destruction of local underlying bedrock. Many scientists in eastern Siberia continued to hold the view that the loesslike silt is a prod- uct of rock material broken down by seasonal freezing and thawing. The eolian origin for the loess in eastern Siberia was demonstrated by Tomirdiaro (1972) and Péwé and Journaux (1983). It has been suggested that the loesslike silt in the southern part of the Qinghai-Xizang (Tibet) Plateau formed in place by disintegration of local country rock. If the loess- like silts were formed in place by disintegration of the local country rock, three conditions would be expected. (1) The mineralogy of the silt and the underlying rock would be similar, allowing for differential resistance to weathering. However, the loesslike silt in the area over- lies many substrates, including alluvium, poorly consoli- dated sands, glacial outwash, limestone, and various other types of bedrock. In the vicinity of Xigaze, loesslike silt overlies a 15-m-high bedrock knob near the junction of the Nyang Qu and the Yarlung Zangbo. Mineralogical analyses show that the silt (sample 3; tables 1, 4, 6) could not have been derived from weathering, either chemical or mechanical, of the underlying rock, which is a volcani- clastic sandstone or lithic tuff that contains pumpellyite that is not present in the silt. Furthermore, the silt contains illite, which is not in the underlying rock. On the south side of the Yarlung Zangbo, about 50 km downstream from Xigaze, a thick deposit of loesslike silt (sample 4, table 1) blankets the upper flanks of the valley almost to the top of a small knob of limestone (figs. 23, 24). Although the loesslike silt includes grains of limestone, most of it is composed of multicycle quartz grains and abundant feldspar grains. The silt could not have resulted from weathering of the underlying bedrock. (2) If silt accumulated on top of a hill or a flat-lying area through the disintegration of underlying bedrock, the material at the surface should be finer and more weathered than that closer to bedrock. However, examination of the section of loesslike silt on the knob of rock near the Nyang Qu at Xigaze did not demonstrate this relation at all. (3) The sediments produced by bedrock disintegra- tion would undoubtedly contain some large bedrock parti- cles, especially of the more resistant materials, and the contact with the bedrock would be gradational. No such particles were noted in the field, especially above the bed- rock knob near Xigaze but also above other bedrock. The loesslike silt overlies the bedrock with a sharp contact, and coarse particles in the lower part of the section are scarce. The contact marks a striking change in textural and miner- alogical composition. In summary, a number of characteristics that would be expected in the loesslike silt if it were formed in place by disintegration of bedrock are not observed. The lack of all these conditions argues convincingly against the weath- ering hypothesis. LACUSTRINE HYPOTHESIS Only two publications suggest that thick deposits of loesslike sediments, especially in the bottoms of the val- leys of Nyang Qu and elsewhere, may have been deposit- ed in a now-drained Pleistocene lake at Xigaze (fig. 3) (Academia Sinica, 1980, p. 11; Zheng and Jiao, 1991). This idea was offered to explain the prominent silt terraces on the sides of the valley and the presence of fresh-water gastropods and ostracodes (figs. 25, 26, 27). The lacustrine hypothesis is improbable for several reasons: (1) Barriers capable of damming water up the Yarlung Zangbo and Nyang Qu have not been proven to exist. Years ago, Ward (1926) examined the gorges for barriers of the Yarlung Zangbo without success. However, it is possible that the Yarlung Zangbo and its tributaries could have been flooded in the past. (2) Evidence of shorelines, wave-cut benches, and deltas would be preserved in a lacustrine environment, but no such features are associated with the silt. (3) Lacustrine deposits have a limited vertical ex- tent, but the loesslike silt is found at various elevations, even on moraines in the Yangbajain valley. (4) Lacustrine silt would be stratified and might even contain varves. This silt is poorly stratified, but some well-stratified sand-and-gravel beds occur locally. In no in- stance does the stratification resemble lacustrine bedding. (5) An appreciable amount of clay could be expected in deposits of any large lake. Most of the sediment sam- ples in the valley-bottom deposits are of silt or sand; how- ever, locally the clay component is 30 to more than 40 percent. (6) Lacustrine deposits would bear little lithologic relation to the underlying rock. This characteristic, howev- ORIGIN OF LOESSLIKE SILT 41 er, is typical of any transported sediment, including the loesslike silt. (7) Some forms of life would have existed in these lakes. So far as we are aware, no lacustrine fossils such as fish have been found in the silt, although gastropods, os- tracodes and pollen occur. Most of the pollen suggests an origin in meadows and small ponds, not in large deep lakes. The ostracodes indicate the same environment. The gastropods indicate mainly shallow ponds and marshes, probably on very gentle slopes, low terraces, and flood plains; they are not characteristic of large lakes. (8) Land fossils would be absent or scarce in lacus- trine deposits. The few mammal remains found in the silt indicate a land environment. In the thick loesslike silt de- posits at Xigaze, the bones of four different birds and two different rodents that were found indicate an environment of small ponds and marshes, not a large lake. (9) Mud cracks and ripple marks would be expected in lacustrine silt, but none have been observed in the loesslike silt. All but one of these features that are characteristic of lacustrine deposits are contrary to the observed nature of the loesslike silt. The one feature (sample 6) that is consistent is typical of deposits transported by any mecha- nism. The observed characteristics of the silt in the south- ern part of the Qinghai-Xizang (Tibet) Plateau cannot be explained by the lacustrine hypothesis. EOLIAN HYPOTHESIS Scientists with the Qinghai-Xizang (Tibet) Plateau expeditions in the 1970's referred to some of the silt on and in the fans flanking the lower parts of the mountains and in valley bottoms in the southern part of the area as loesslike deposits (Zhao and others, 1976; Li and others, 1983). The first published suggestions that the silt was re- transported loess probably were by Péwé (1980 a, b, 1981) and Péwé and others (1981, 1985, 1987). This study of the silt in the southern part of the Qinghai-Xizang (Tibet) Pla- teau, and our comparison with similar deposits elsewhere in China and other parts of the world, supports the identi- fication of the deposits as loess and retransported loess. Features and characteristics of eolian-deposited silt fall into three groups: field relations, mechanical composi- tion, and petrographic character (Smith and Fraser, 1935). FIELD RELATIONS TOPOGRAPHIC DISTRIBUTION As clearly stated by Barbour (1930, p. 468), in refer- ence to the loess of central China, the mantle of loess "conforms closely to the general contouring of the buried preloess topography, filling up the gullies, covering the minor elevations, lying deepest in the depressions and thinning-out up the flanking slopes of the higher ridges." The loesslike silt in the Qinghai-Xizang (Tibet) Plateau occurs as a surficial mantle and fits Barbour's description. It is thin on top of low hills, is thicker on the flanking fans, and reaches maximum thickness in the broad valley bottoms. A thin veneer of the silt remains on the rather flat passes or highlands where it has not been removed by down-slope processes. INDEPENDENT LITHOLOGIC CHARACTER Wind-deposited material usually differs mineralogi- cally and texturally from the underlying material. In the southern part of the plateau, the loesslike silt covers sand and gravel, glacial till, and various types of bedrock. The fine texture of the silt contrasts strongly with the texture of the underlying material. Petrographic study of samples from the silt and bedrock at the knob at Xigaze, and at the limestone hill 50 km downstream from Xigaze, indicates mineralogical differences between the silt and the underly- ing bedrock. ABSENCE OF DISTINCT STRATIFICATION One characteristic of loess is the absence of distinct stratification in silt on top of hills. Within the study area, the loesslike silt on top of hills also lacks stratification. However, transportation of the silt from the slopes to the alluvial fans and to the valley bottoms produces poor to well-developed stratification, especially if interbedded with sand or gravel layers. COLOR AND TEXTURE Wind-deposited silt is generally light colored and stained by moving ground water or carbonaceous inclu- sions; the texture is generally uniform. In loess away from a river or not on slopes or valley bottoms, clay makes up no more than 10 to 15 percent of the sediment. The loess- like silt in the study area is comparable to loess elsewhere in the world, except that much of it is interbedded with sand and gravel. ASSOCIATION WITH OTHER EVIDENCE OF WIND ACTION Smith and Fraser (1935, p. 19) state that sand dunes and ventifacts may be expected near the source of material. The plateau is conventionally referred to as a wind-swept area and a land of blowing sand and dust. Inhabitants have 42 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA long been aware of these conditions, and early western ex- plorers remarked about the "raving wind," "wind roaring over the sandbanks," "perpetual harassing blasts," and "ruth- less winds" (Ward, 1926). Two French Vincentian mission- aries to Lhasa in 1896 bitterly complained of the terrible north wind that lasted 15 days and menaced them with distraction (Huc and Gabet, 1928, v. 2, p. 147-148). A pro- fessor of Oriental language from London, entering Tibet in disguise in 1923, remarked of the effect of the wind on the Yarlung Zangbo "* * * so strong was the wind in the opposite direction [of the flow of the river] that its blasts on the river made it seem as if the water was flowing backward and uphill; in fact, so strong was the illusion that [the natives] thought it to be real and bowed down in worship of a supposed miracle." (McGovern, 1924, p. 285-286). Early western ex- plorers also commented on the presence of immense sand dunes, especially in the Zangbo corridor (Rawling, 1905). Various types of dunes, especially climbing dunes (cover photograph, fig. 16), occur along the broad, gravel-floored valleys of major streams (Zhao and others, 1976, p. 6; Wang and Fan, 1987; Reiter and Reiter, 1981). David-Neel (1927, p. 299-300) commented about invading sand in 1923: Quite near Lhasa, on the left bank of the Brahmaputra [Yarlung Zangbo), one finds a miniature Sahara whose ex- tensive white dunes are invading the whole country* * *. * * * In spite of a ridge which stood in their way, the sands have taken a firm footing in the Kyi Valley, and though still shallow on that side, their fine dust is beginning to accumu- late along the hedges which border Norbu ling, the pleasure garden of the Dalai Lama. It may possibly be that in a few generations Lhasa will be reached. Who knows whether, in a still more distant future, some savant, excavating the en- tirely submerged city, may not discover the Potala and the Jo Khang, just as we now lay bare the palaces and temples which the sands of the great Gobi have overwhelmed* * *. On our way * * * there were a few more villages whose fields are gradually disappearing beneath the mounting sands. These dunes are evidence of eolian action still active on the almost vegetation-free flood plains (figs. 13, 15). Bare sand-and-gravel bars of the braided streams (cover photograph and figs. 12-14) also provide an excel- lent source area for enormous clouds of blowing dust that are common in late winter, spring, and early summer. A rare photograph of clouds of dust beginning to form on the bare flood plain of the Lhasa He near Lhasa in 1954 is illustrated by Sis and Vanis (no date, pl. 50). The duststorms range from a "maze of dancing dust devils" (Ward, 1926, p. 43) to "huge black clouds approaching rapidly" (Harrer, 1953, p. 163). Harrer further states (1953, p. 163) that in Lhasa, "* * * the Potala Palace disappears and at once everyone rushes for home. Street life stops." and (p. 316) that the departure of the Dalai Lama from Lhasa to India on March 19, 1959, occurred during a notorious sandstorm and dust- storm that immobilized Lhasa in the early evening. Commenting further in this vein, Alexandra David- Neel (1927) reported that she became the first foreign woman in Lhasa in 1923 when an immense dust storm permitted her to enter the Forbidden City undetected as she states (p. 256-257): All of a sudden a furious storm arose, lifting clouds of dust high into the sky. I have seen the simoon in the Sahara, but was it worse than this? No doubt it was. Yet, that terrible dry lashing rain of dust gave me the impression of being once more in the great desert. Indistinct forms passed us, men bent in two, hiding their faces in the laps of their dresses, or whatever piece of cloth they might happen to have with them. Who could see us coming? Who could know us? An im- mense yellow curtain of whirling sand was spread before the Potala, blinding its guests, hiding from them Lhasa, the roads leading to it, and those who walked upon them. I in- terpreted it as a symbol promising me complete security, and the future justified my interpretation. For two months, I was to wander freely in the lamaist Rome, with none to suspect that, for the first time in history, a foreign woman was beholding the Forbidden City. "The gods threw a veil over the eyes of his adversaries and they did not recognise him." So went an old Thibet's tale which I had heard long ago. Present day reports support these early views (Han, 1979, p. 166): The wind. A blinding, laceration blizzard of sand and stone, and the whole world destroyed by this maelstrom of small pebbles whirling in the choking air. Lips, hands and face are flayed. The turquoise sky, the white houses, even the Potala, are no longer, all swallowed in a yellow opaqueness which is the wind * * *. For three days the wind blew, and no planes came * * *. Then, relief. With the night the wind abates though an overhang of sand [dust] will remain in the air for another day or more. In the past, the lamas * * * blew long silver trumpets to compel the wind away. In the absence of silver trumpets, similar or larger clouds of dust in late Pleistocene times undoubtedly de- posited silt onto the slopes of the hills and mountains (cover photograph). Many workers have demonstrated that no more wind is necessary to blow that dust than is present today, as winds of 20 m/s are more than strong enough to bear quantities of dust in both periglacial areas and hot desert areas (Bryan, 1927, p. B39-B40; Péwé, 1951; Warn, 1953; Péwé and others, 1976, 1981b). If there were larger source areas than today in late Pleistocene time, such as wider unvegetated flood plains of braided streams, from more extensive glaciated areas (Shi and oth- ers, 1993, fig. 3; Li and Li, 1991), wind-blown dust must have been abundant. In fact, Feng and Thompson (1989) and Zhang and others (1991) state that silt from the north- ern part of the Qinghai-Xizang (Tibet) Plateau may have RETRANSPORTED LOESS 43 blown eastward to the great loess plateau of central China north of Xian (fig. 1) during Pleistocene time. FOSSILS OF AIR-BREATHING ANIMALS Fossils of air-breathing animals should logically be expected in the silt if the material is eolian. Late Pleisto- cene vertebrate fossils recorded in the loesslike silt in the southern part of the Qinghai-Xizang (Tibet) Plateau (Huang, 1980) include remains of horse, bison, and other mammals. Bones of two types of rodents and four types of birds were found in the loesslike silt near Xigaze. The preservation of many of the specimens shows very little evidence of transportation. Although land gastropods are well known from loess in central China, none are reported in the loesslike silt in the southern part of the Qinghai-Xizang (Tibet) Plateau. Nevertheless, Succinea representatives live in the area today. Russell (1944, p. 34) regarded land gastropods as being characteristically present in loess. However, land gastropods have not been found in some U.S.A. loess in the upper Mississippi River valley or in most of the loess in Alaska (Péwé, 1955, p. 720-721). MECHANICAL COMPOSITION In loess, the grain-size and degree of sorting is simi- lar to that of windblown dust or volcanic ash that has been transported a considerable distance (Udden, 1898, p. 31- 60; Swineford and Frye, 1945, p. 252; Warn and Cox, 1951, p. 559; Péwé, 1955, fig. 11; Péwé and others, 1976). Some of the material on the Qinghai-Xizang (Tibet) Pla- teau is similar to silt known to be loess. Cumulative- frequency grain-size curves of mechanical analyses of windblown dust from Arizona, Kansas, and Germany are similar to those of loesslike silt from the top of a bedrock knob (sample 3) and from other localities in the plateau, where the silt is near the river (sample 4) or is slightly retransported (sample 6; fig. 39). Samples 3, 4, and 6 from the southern part of the plateau were compared with samples from deposits of silt of known eolian origin from elsewhere in the world (fig. 40). The great similarity of grain size and degree of sorting between the samples constitutes a strong argu- ment for the eolian hypothesis. Cumulative-frequency grain-size curves of that loesslike silt from near Xigaze are compared with curves of known loess samples col- lected from Beijing, New Zealand, Alaska and Illinois in the U.S.A., France, the former Czechoslovakia, Siberia, and Uzbekistan. The sample of loesslike silt that has been retransported (sample 6) is mixed with different sediment sizes and bears less resemblance to primary loess. PETROGRAPHIC CHARACTER Angular grains have long been considered typical of loess (Barbour, 1925a, fig. 1; Twenhofel, 1932; Char- lesworth, 1957), and scanning-electron-microscope anal- ysis supports that assumption (Cegla and others, 1971; Pye, 1987). Silt grains from the Qinghai-Xizang (Tibet) Plateau are clearly angular, especially light minerals such as quartz. Loess grains probably should also be largely unweathered, and clay minerals among them should be largely unweathered (Smith and Fraser, 1935; Charlesworth, 1957), although the degree of weathering would depend on the age and rate of deposition of the sediment and the effects of diagenetic processes. The loesslike silt from the plateau, like upper Pleistocene loess from other parts of the world, is relatively fresh but shows some weathering under the scanning electron microscope. SUMMARY The characteristics of loesslike silt in the Qinghai- Xizang (Tibet) Plateau and the similarity between it and primary loess support the eolian hypothesis. In the light of this evidence, we conclude that the unstratified loesslike silt on top of bedrock hills of the plateau is primary loess, and that the poorly to well-stratified loesslike silt on the lower slopes and in valley bottoms of the major river val- leys is retransported loess. RETRANSPORTED LOESS Very little of the loesslike silt in the southern part of the Qinghai-Xizang (Tibet) Plateau is primary loess; most of it is called "retransported loess" in this report. The re- transported loess has been transported a few meters to hundreds, if not thousands of meters downslope, primarily by processes involving water. It is poorly to fairly well stratified and is intercalated with sand-and-gravel layers. With increasing distance from the hills and mountains, the sand, gravel, and other detritus became finer, and the num- bers of layers of coarse sediments also decrease. Retransported secondary loess is not unique to this area; it is widespread in the world, especially in areas of hills and mountains. Andersson (1923, p. 129) reported such deposits from central China many years ago and used the term "redeposited" loess: In many valleys within the loess areas there are deposits of loess interstratified with gravel and occurring under such conditions that they must evidently be regarded as repre- senting later depositions derived from the genuine loess. They are valley deposits of very limited extent with rapid and irregular changes of the stratification. Thick beds of 44 gravels suddenly appear, and the gravel component of these deposits is far more abundant than is the case in the genu- ine loess. Everything goes to show that in the formation of the secondary loess-gravel beds water action has been the dominant factor and the wind has played only a very subor- dinate role. I call these deposits Redeposited Loess in order to distinguish them from the genuine Primary Loess. Such secondary loess deposits are known elsewhere in the world, and they have been described in various ways in different languages. Other terms include "alluvial loessial deposits" (Wang and Song, 1983), "loess-like slope deposits," "deluvial loess," "slope loess," a quite ap- propriate term "mountain loess" (Pecsi, 1965, 1967), and "colluvial silt" (Li and others, 1983). In Alaska, the term "retransported silt" has been used, and "muck" is used in the sub-Arctic and Arctic Alaska where the retransported material has a high organic content that has been frozen since the time it was deposited (Péwé, 1968, 1975a, b; Péwé and others, 1989). The frozen material has preserved an abundant number of mammal remains, including peren- nially frozen carcasses. In some regions where thick loess is well studied, the primary loess deposits interfinger and are interleaved with LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA redeposited loess deposits. This interfingering indicates changes in climate and (or) the activity of erosional processes. To more fully understand the characteristics of re- transported loess in other parts of arid western China, and to compare such deposits with loesslike silt in the Qing- hai-Xizang (Tibet) Plateau, Péwé examined retransported loess deposits and land forms on the south side of the Tian Shan in the Xinjiang Uygur Autonomous Region (fig. 1) in 1986 (Péwé, 1987) and discussed in detail with Mu Guijin in 1984 to 1986 the retransported loess deposits and landforms in the middle Ili He valley on the north slope of the Tian Shan. Mu is familiar in detail with the valley fills of loess in the Ili He valley and at Turpan (Mu, written commun., 1987; Yuan and others, 1983; and Wang and others, 1991). These deposits were classified as alluvi- al loessial deposits by Wang and Song (1983). It was im- mediately apparent to Péwé that this material in northwestern China exhibits the same characteristics as the loesslike silt in the southern part of the Qinghai-Xizang (Tibet) Plateau. Turpan (figs. 1, 2, 5) lies in the hottest and lowest (154 m below sea level) part of China and is on the histor- ical Silk Road. With irrigation from the mountains by the 100 TT OT TOI I T TT TOT 14- L -. EXPLANATION sel a | 4 * w-: -A | A -A- {+--- 90 |- Samples AL, nr Te. -' --- 3, loess, east of Xigaze ’;/:, —;’_f" ---- 4, loess, Xigaze / ////Z 80 |- ----- 6, silt, Xigaze {le LAT |4 & Volcanic ash, Alaska A If // o | -- --- - Dust, Arizona // JV :/ / E 70 - - Dust, Kansas 7 77 [I 7 moo | ----- Dust, Germany / 17 | i[7 Z 1 If ~ eo . / 1¢ /‘/ 3 10 M E] f IF / 2 50 {LP L iP 0 1j / ( w A U C F 1i | if1 L_. i! |f; w 40 J_ PLL > / A [/! (<3 / 3 A /I g 30 A7 // 2 / & W 4 [i ° zo L UAL lf " THF * /| 4 /.’ 10 /- -p A 4" ,_/"/ nzflZ—l"’ o -- 1.0 0.5 0.1 0.05 0.01 0.005 0.001 0.0005 0.0001 DIAMETER, IN MILLIMETERS Figure 39. Cumulative-frequency grain-size curves for samples of loess (Nos. 3, 4) and retransported silt (No. 6) from near Xigaze, Xizang Autonomous Region, China, compared with those for volcanic ash (Ester Ash Bed) from near Fairbanks, Alaska (Péwé, 1955, fig. 11; pl. 2, fig. 1), and modern wind-deposited dust from roof of house at 538 E. Fairmont Dr., Tempe, Ariz. (Péwé and others, 1981, fig. 5), the third floor outside window sill, Lakeway Hotel, Meade, Kan. (Swineford and Frye, 1945, p. 252), and from Breslau, Selesia, Germany (now Wroclaw, Po- land) (Zeuner, 1949, p. 27). Samples from Alaska, Arizona, and China collected by T.L. Péwé. Alaskan sample analyzed by U.S. Army Corps of Engineers, Rock Island, III.; Arizona sample analyzed by André Journaux at Centre de Geomorphologie du Centre National de la Recherche Scientifique, Caen, France; Chinese samples ana- lyzed by R.M. Slatt at Cities Service Company, Tulsa, Okla. AGE AND CORRELATION OF SILT DEPOSITS 45 method of underground canals (karez), the areas of ree _ to those at Xigaze (fig. 46). The upper slopes of the valley transported loess have become leading producers of excel- _ walls are blanketed with primary loess of late Pleistocene lent melons and grapes. Only 8 km west of Turpan, at the _ age. A cumulative-frequency grain-size curve of clay-rich ruins of the ancient city of Jiaohe (figs. 41, 42), small _ primary loess 15 km north of Nilka (sample M7, fig. 45; streams are entrenched 30 m or more into the valley fill of _ Mu Guijin, written commun., 1992) on the Ili He is simi- retransported loess to produce vertical walls and mesas _ lar to one for primary loess in the Qinghai-Xizang (Tibet) (figs. 43, 44) that are strikingly similar to the terraces and Plateau (sample 3). Poorly bedded retransported loess of mesas cut into the poorly bedded loesslike silt at Xigaze late Pleistocene and Holocene age overlies sand-and-grav- (figs. 26, 27) in the Qinghai-Xizang (Tibet) Plateau. Gran- __ el alluvium of middle Pleistocene age (fig. 46) in the val- ulometric analysis of the retransported loess at Jiache _ ley bottoms. At least two flat-topped terraces have been (sample 18) yields a cumulative-frequency grain-size _ cut into the retransported loess by the river. curve similar to those of loesslike silt from near Xigaze (table 1, figs. 39, 45). The curves are not like those of primary loess. Retransported loesslike silt at Kuga (figs. 1, AGE AND CORRELATION 2, 5; table 1, sample 19), 600 km west of Turpan, has in- OF SILT DEPOSITS terbedded gravel layers similar to those in fans in the Qinghai-Xizang (Tibet) Plateau. Cumulative-frequency Although loess has been accumulating throughout grain-size curves show that the loesslike silt at Kuga is - Quaternary time in the areas of classic thick loess of coarser than that at Jiaohe and has a lower clay-size con- north-central China (Burbank and Li, 1985; Heller and tent (fig. 45). Liu, 1982, 1984; Heller and others, 1989; Liu and Yuan, Retransported loess along the middle Ili He valley 1982; Liu, 1987; Rutter and others, 1991), loesslike silt of on the north slope of the Tian Shan forms terraces similar the southern part of the Qinghai-Xizang (Tibet) Plateau 100 -t- T TTTTT 7 EXPLANATION {A11 4 f" 90 |- Samples / f/y; < --- 3, loess, east of Xigaze %4;’ PE ---- 4, loess, Xigaze / t 80 | ----- 6, silt, Xigaze T7 TWA LF A7 . (Lg ------ - New Zealand / / 247 /// § 70} - Czechoslovakia Af _L |flls" F m CV] ......-- Siberia, Russia / /," // /’ 4A & - | ----- Alaska, U.S.A. A/? A A" x co | 20, Beijing / L9 -A € ---- illinois, U.S.A. [ /,/} /, [m ----- Uzbekistan I A a] ta / 2 | [ l | aat 6 Q 50 husks seek France 1h -this 7 o i VlP} LL. It M/s ¢ w 40 Lif iL 2 Alk 7 E ii 2 5 [70% 2 so T 5 118 / (6) v / 20 7 f + 7 AIM : / “I, / AHP) 10 £ f L yA ”wk/Ir" p Lhakdscks 1.0 0.5 0.1 0.05 0.01 0.005 0.001 0.0005 0.0001 DIAMETER, IN MILLIMETERS Figure 40. Cumulative-frequency grain-size curves for samples of loess (Nos. 3, 4) and retransported silt (No. 6) from near Xigaze, Xizang Autonomous Region, China, compared with those for loess from around the world: West side of Banks Peninsula, 40 km south of Christchurch, New Zealand; Brno, Czechoslovakia; 60 km northeast of Yakutsk, Yakutia (now Siberian Russia) (Péwé and Journaux, 1983, fig. 29, sample K); Fairbanks, Alaska (Péwé, 1955, fig. 11); Beijing, China (sample 20); Rock Island, Illinois; 50 km south of Samarkand, Uzbekistan; and near Caen, Normandy, France. Sample from France collected by André Journaux; sample 20 collected by R.J.E. Brown; all others collected by TL. Péwé. Illinois and Alaska samples analyzed at U.S. Army Corps of Engineers, Rock Island, III.; samples 3, 4, and 6 analyzed by R.M. Slatt at Cities Service Company, Tulsa, Okla. Other samples analyzed at Centre de Geomorphologie du Centre National de la Research Scientifique, Caen, France, under direc- tion of André Journaux. 46 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA seems to represent only late Quaternary time. Vertebrate remains in the retransported loess are late Quaternary, al- though Huang (1980) states that paleontological studies of loess in the southern Qinghai-Xizang (Tibet) Plateau are only just beginning and that middle Quaternary vertebrates may be recovered during further investigations. Attempts to date gastropod shells from the 15-m- high section of retransported loess at Xigaze (fig. 27) give interesting results. The shells were taken from the upper shell bed near the top of a section, which is thought to be of late Pleistocene age, where considerable subsequent downcutting has left high terraces on both sides of the Nyang Qu (figs. 25, 26, 27). The shells submitted by Anders Martinsson show a radiocarbon age of 2,010+85 yr B.P. (U. Olsson, written commun. to Liu Tungsheng, 1981). Olsson also stated that the "sample was a little too small." Gastropod shells collected by Péwé show a radio- carbon age of 3,645+110 yr B.P. (Robert Stuckenrath, written commun., 1981, laboratory sample SI-4651). Stuckenrath found the date puzzling and concluded from his testing that the snails must have been heavily charged with radon from their local microenvironment. Both labo- ratories, therefore, consider the ages questionable. SUMMARY Retransported loesslike silt occurs as fans and val- ley-bottom deposits in the southernmost part of the Qinghai-Xizang (Tibet) Plateau. The loesslike silt, reach- ing thicknesses of as much as 15 m, is well sorted, poorly stratified, and interbedded locally near the hills and moun- tains with layers and lenses of sand and gravel. It overlies all substrates with a sharp contact and contains minerals not present in the underlying bedrock. The loesslike silt is massive, stands in sheer cliffs, and is strikingly eroded by narrow steep-walled gullies. It contains a few land verte- brate fossils and fossil gastropods that lived in shallow fresh-water ponds. Hypotheses to explain the origin of the loesslike silt have included residual, lacustrine, and eolian processes. The hypothesis that loesslike silt is a residual product of Figure 41. Ruins of ancient city of Jiaohe 8 km west of Turpan, Xinjiang Uygur Autonomous Region, China (figs. 1, 2). City in this arid region existed about 2,000 years ago; walls of loess bricks were built on mesa of retransported loess 30 m high and bounded on three sides by two steep-walled narrow valleys. p ® k ‘A' al View southeast toward flat aggradational surfaces of retransport- ed loess beyond valley that crosses near top of photograph. (Also see figs. 42, 43, and 44 and compare with fig. 26 that shows mesa-like remnants of retransported loess at Xigaze, China.) Photograph 4900 by Troy L. Péwé, Aug. 5, 1986. SUMMARY 47 breakdown in place of the underlying rocks, mainly by frost action, has little support. Evidence against this hy- pothesis includes (1) minerals in the silt are not every- where similar to those in the underlying bedrock, (2) no large particles of more resistant minerals are present, and (3) the silt does not become progressively coarser toward bedrock. The lacustrine hypothesis is strongly supported by scientists in some areas because of the prominent silt ter- races along the sides of valleys and because of the pres- ence of fresh-water gastropods and ostracodes. But several circumstances indicate this origin is unlikely. No shore- lines, wave-cut beaches, or deltas are present; nor are rip- ple marks, mud cracks, and deep-water lake fossils found. Neither lacustrine stratification nor an appreciable amount of clay exists in the silt. No definite upper limit of the deposits, to be expected under a lacustrine hypothesis, is present. Evidence for the eolian origin of the upland silt is abundant. (1) The silt blankets older topography. (2) It is lithologically independent of the underlying material. (3) Stratification is indistinct or absent on small hilltops where z L_ Retransportedfr . the silt has not been retransported. (4) The silt is abundant in areas having sand dunes and active duststorms. (5) It contains fossils of air-breathing animals. (6) Sorting and texture of the loesslike silt on hilltops or near hilltops are similar to those of loess and windblown dust from many different areas in the world. (7) Individual silt grains are angular and relatively fresh. The only glaciation in the southern part of the Qing- hai-Xizang (Tibet) Plateau has been in the high areas of the Himalayas, the Nyaingentanglha Shan, and associated mountains in the upper reaches of the Yarlung Zangbo. Valley glaciers in these mountains were more extensive during glacial maxima, and unvegetated floodplains, char- acteristic of the modern major streams, were more wide- spread then. We believe the upland silt on the plateau is loess that was deposited against the hills and mountains by winds blowing across the unvegetated floodplains and the desert rocky plains to the north during late Quaternary time. Most of the loesslike silt on the plateau is silt re- transported loess that has been moved downslope and re- deposited by rillwash, sheetwash, stream action, and solifluction. Figure 42. Schematic block diagram of late Pleistocene re- transported bedded loess (Wang and Song, 1983) in small val- ley at Jiaohe 8 km west of Turpan, Xinjiang Uygur Autonomous Region, China (fig. 1). Arrows on circled num- bers show direction of view of photographs in figures 41 and 43 (fig. 44 faces opposite direction of fig. 41 and also in- cludes southeast end of mesa). Sketch by Mu Guijin, 1986; vertical exaggeration 7.5.x. 48 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA Figure 43. View upstream (northwest) of narrow, irrigated, steep-walled valley entrenched 30 m into retransported loess on west side of ruins of ancient city of Jiaohe near Turpan, Xinjiang Uygur Autonomous Region, China (figs. 1, 42). Photograph 4898 by Troy L. Péwé, August 5, 1986. SUMMARY ¥" Figure 44. Oblique northwest-facing aerial photograph of mesa-like remnants of retransported loess 8 km west of Turpan, Xinjiang Uygur Autonomous Region, China (figs. 1, 41, 42). Ruins of ancient city of Jiaoche on mesa outline house structures built of loess bricks. Photograph by Georg Gerster, Copyright National Geo- graphic Society; taken September 5, 1987, at 9:30 a.m., and published with permission. 49 50 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA 100 ULTLT T T U TTT MLH 90 EXPLANATION |-- ---] ~ Sample No. +11 x' ----- __ 7, west of Xigaze j-" l- 80 || ------- _ 19, Kuga a e4 & _ _|| ------ 4, east of Xigaze /" Epale / 5 18, Jiaohe M WEEE E 70 ports 9 M7, north of Nilka 7 - a. p ® ./ z . - 60 {a- Z / pa 7. [m) ¢ , 3 50 , C / i in 40 f { w # L / 2 £ / 3 30 /* * 3 A 3 / 0 213 A 20 / iI |L // // eo in 1" / '/’/.: 10 (- + + -4- id 0 71 ® 1.0 0.5 0.1 0.05 0.01 0.005 0.001 0.0005 0.0001 DIAMETER, IN MILLIMETERS Figure 45. Cumulative-frequency grain-size curves for samples of primary loess and retransported loess from XKin- jiang Uygur Autonomous Region in northwestern China (figs. 1, 2) and primary loess and retransported loess from southern Qinghai-Xizang (Tibet) Plateau (fig. 34). Sample 4 is sandy primary loess from east of Xigaze, Xizang Autonomous Region (fig. 34; table 1); M7 is clay-rich primary loess from 15 km north of Nilka in Ili He valley, Xinjiang Uygur Autonomous Region (figs. 1, 46); 18 and 19 are retransported loess from Xinjiang Uygur Autono- mous Region (figs. 1, 2, 5; table 1); and 7 is retransported loess from west of Xigaze (table 1). Samples 4 and 7 collected by T.L. Péwé and analyzed by R.M. Slatt at Cities Service Company, Tulsa, Okla.; samples 18 and 19 collected by T.L. Péwé (Péwé, 1987) and analyzed by Mu Guijin at Arizona State University, Tempe, Ariz.; sample M7 collected and analyzed by Mu Guijin, Institute of Geography, Academia Sinica, Uriimgi, Xinjiang Uygur Au- tonomous Region, China. Dots are data points. \ Jln Sy gm I N No maim Influwllwl} A as > , i, 2 * M7 Loess 27 ITT! Fp U Retrans- j Loess 7 Bedrock ported loess efter i _- fe dee - < 4 *** =-- A SCALE B T1 Sand 23nd TZ -reez- __ Sand and gravel hZ—zsoom gravel alluvium 7, > alluvium Figure 46. Schematic block diagram of loess and retransported loess 15 km north of Nilka, in Ili He valley, north slope of Tian Shan in Xinjiang Uygur Autonomous Region, China (200 km north of Kuga) (fig. 1). Valley has three levels of terraces. Pri- mary loess blankets high terrace and uplands, and retransported stratified loess of late Pleistocene to Holocene age is exposed on two lower levels of terraces in valley, beside and over Middle Pleistocene sand-and-gravel alluvium. Retransported loess and terrace scarps are similar to those at Xigaze and elsewhere in the southern part of the Qinghai-Xizang (Tibet) Plateau (figs. 26, 27). Tl, terrace number; M7, location of loess sample. Sketch by Mu Guijin in 1986; vertical exaggeration 10x. REFERENCES CITED 51 REFERENCES CITED Academia Sinica, 1980, A scientific guidebook to south Xizang (Tibet): Beijing, Organizing Committee, Symposium on Qinghai-Xizang (Tibet) Plateau: Beijing, 1980, 104 p. Andersson, J.G., 1923, Essays on the Cenozoic of northern China: Peking, Geological Survey of China Memoir 3, series A, 152 p. Argand, Emile, 1924, La Tectonique de 1'Asie: Congrés Géologique International Comptes Rendus [13th, Bruxelles, Belgium, 1922], p. 170-372. Barbour, G.B., 1925a, Chinese loess: Association of Chinese and American Engineers Journal, v. 6, no. 4, p. 1-10. 1925b, The loess of China: The China Journal of Science and Arts, v. 3, no. 8, p. 454-463; v. 3, no. 9, p. 509-519. 1930, The loess problem of China: Geological Magazine, v. 67, p. 458-475. Berg, LS., 1960, Less kak produkt vyvetrivaniya i pochvoobra- zovaniya, iz Klimat i zhizn [in Russian]: Moscow, Izdatel'stvo Akademii Nauk SSSR. Translated, 1964, as Loess as a product of weathering and soil formation, from Climate and life, selected works of L.S. Berg, Vol. 3: Jerusa- lem, Israel Program for Scientific Translations, 207 p. Biscayne, PE., 1964, Distinction between kaolinite and chlorite in recent sediments by X-ray diffraction: American Mineral- ogist, v. 49, p. 1281-1289. 1965, Mineralogy and sedimentation of recent deep-sea clay in the Atlantic Ocean and adjacent seas and oceans: Geological Society of America Bulletin, v. 76, p. 803-832. Bryan, Kirk, 1927, The "Palouse soil" problem, Chap. B in Con- tributions to the geography of the United States, 1926; U.S. Geological Survey Bulletin 790, p. 21-46. Burbank, D.W., and Li Jijun, 1985, Age and paleoclimatic signif- icance of the loess of Lanzhou, north China: Nature, v. 316, p. 429-431. Cegla, J., Buckley, T., and Smalley, LJ., 1971, Microtextures of particles from some European loess deposits: Sedimentolo- gy, v. 17, p. 129-134. Chang Cheng-fa and Pan Yu-shen, 1981, A brief discussion on the tectonic evolution of Qinghai-Xizang Plateau, in Liu Dong-sheng, ed., Geological and ecological studies of the Qinghai-Xizang Plateau; Vol. 1, Geology, geological history and origin of Qinghai-Xizang Plateau: Beijing, Science Press, p. 1-18. Charlesworth, J.K., 1957, The Quaternary era, with special refer- ence to its glaciation: London, Edward Arnold, 2 vols., 1700 p. Chen Deniu, Gao Jiaxiang, Gao Fuquing, and Liu Tungsheng, 1985, Fossil snails and their environment, in Liu Tungsheng, ed., Loess and the environment: Beijing, China Ocean Press, sec. 2, chap. 4, p. 72-84. Chen Zhiming, 1986, Lake retreat in Qinghai-Xizang Plateau, with additional reference to its climatic significance [in Chi- nese]: Oceanologia et Limnologia Sinica, v. 17, no. 3, p. 207-216. Cheng Guodong, 1983, Vertical and horizontal zonation of high- altitude permafrost, in Proceedings, Fourth International Conference on Permafrost, 1983, Fairbanks, Alaska: Wash- ington, D.C., National Academy of Science, p. 136-141. Cheng Guodong and Liang Fengzian, 1987, Soil wedge polygons along the Qinghai-Xizang Highway [abs.]: International Union for Quaternary Research, Loess Commission, 12th Congress, Ottawa, Canada, Programme with Abstracts, p. 142. Cheng Guodong and Wang Shaoling, 1982, On the zonation of high-altitude permafrost in China [in Chinese]: Journal of Glaciology and Cryopedology, v. 4, p. 1-17. Chinese Academy of Sciences, 1974, A photographic record of the Mount Jolimo Lungma Scientific Expedition (1966- 1968): Beijing, Science Press, 108 p. Clarke, A.H., 1973, The freshwater mollusks of the Canadian In- terior Basin: Malacologia, v. 13, 509 p. Clarke, A.H., and Harington, C.R., 1978, Asian freshwater mol- lusks from Pleistocene deposits in the Old Crow Basin, Yukon Territory: Canadian Journal of Earth Sciences, v. 15, p. 45-51. David-Neel!, Alexandra, 1927, My journey to Lhasa: New York, Harper Brothers; republished 1986, Boston, Beacon Press, 310 p. Derbyshire, Edward, 1985, Loess and loessic colluvia in Qinghai Province, China [abs.]: International Conference on Geomor- phology, 1st, Manchester, England, 1985, Abstracts of Papers, p. 183. Feng Zhaodong and Thompson, LG., 1989, A discussion on multigenesis of the loesses in China and the United States [abs.]: Geological Society of America Abstracts with Pro- grams, v. 21, no. 6, p. A154. Folk, RL., 1974, Petrology of sedimentary rocks, 2d edit.: Aus- tin, Tex., Hemphill Publishing Co., 182 p. Gao Shiyang and Li Bingxiao, 1981, Borate minerals on the Qinghai-Xizang Plateau, in Liu Dong-sheng, ed., Geological and ecological studies of the Qinghai-Xizang Plateau, Vol. 2, Environment and ecology of the Qinghai-Xizang Plateau: Beijing, Science Press, p. 1725-1731. Han Suyin, 1979, Lhasa, the open city: Frogmore, England, Tri- ad/Panther Books, 192 p. Harrer, Heinrich, 1953, Seven years in Tibet: Los Angeles, Calif., J.P. Tarcher, Inc., 318 p. Harvard, Andrew, 1984, The forgotten face of Everest: Washing- ton, D.C., National Geographic Society, National Geograph- ic, v. 166, no. 1, p. 71-77. Hedin, Sven, 1903, Central Asia and Tibet: London, Hurst and Blackett, Ltd, v. 1, 608 p.; v. 2, 664 p. Heller, Friedrich, and Liu Tungsheng, 1982, Magnetostratigraphi- cal dating of loess deposits in China: Nature, v. 300, p. 431- 433. 1984, Magnetism of Chinese loess deposits: Royal Astro- nomical Society Geophysical Journal, v. 77, p. 125-141. Heller, Friedrich, Ruegg, Felix, Liu Xiuming, Liu Tungsheng, and Sun Jiangzhong, 1989, Magnetic properties of the loess sediments in China [abs.]: International Geological Con- gress, 28th, Washington, D.C., v. 2, p. 2-48-2-49. Huang Wanpo, 1980, The Quaternary mammalian fossil localities from Xizang, in Paleontology of Xizang [in Chinese]: Beijing, Academia Sinica, v. 1, p. 48-53. Huang Wanpo and Ji Hongxiang, 1981, The climate and uplift of the Qinghai-Xiang (Tibet) Plateau in the late Pleistocene and Holocene, in Liu Dong-sheng, ed., Geological and ecologi- cal studies of the Qinghai-Xizang Plateau, Vol. 1, Geology, 52 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA geological history and origin of Qinghai-Xizang Plateau: Beijing, Science Press, p. 225-230. Huang Xiaoming, Yang Hairong, and Liu Tieliang, 1981, Some behaviors of seasonal freezing and thawing of natural ground in the interior of Qinghai-Xizang Plateau, in Liu Dong- sheng, ed., Geological and ecological studies of the Qinghai- Xizang Plateau, Vol. 2, Environment and ecology of Qinghai-Xizang Plateau: Beijing, Science Press, p. 1803- 1816. Huc, ER. and Gabet, Joseph, 1928, Travels in Tartary Tibet and China-1844-1846: London, George Routledge and Sons, Ltd., v. 1, 387 p.; v. 2, 406 p. Kidd, W.S., Pan Yusheng, Chang Chengfa, Coward, M.P., Dew- ey, J.F., Gansser, H., Molnar, P., Shackleton, R.M., and Sun Yiyin, 1988, Geological mapping of the 1985 Chinese-Brit- ish Tibetan (Xizang-Qinghai) Plateau geotraverse route: Royal Society of London Philosophical Transactions, v. 327, p. 287-305, scale 1:500,000. Kreig, R.A., Cheng Guodong, and Brown, Jerry, 1986, Terrain analysis from space shuttle photographs of Tibet: American Society of Civil Engineers, International Conference on Cold Regions Engineering, 4th, Anchorage, Alaska, 1986, Proceedings, p. 400-409. Kuhle, Matthias, 1985, Permafrost and periglacial indicators on the Tibetan Plateau from the Himalaya Mountains in the south to the Quilian Shan in the north (28°-40° N): Zeitschrift fiir Geomorphologie, v. 29, no. 2, p. 183-192. 1986, The upper limit of glaciation in the Himalayas: GeoJournal, v. 13, no. 4, p. 331-346. Lanzhou Institute of Glaciology and Cryopedology, 1983, The tentative map on snow, ice and frozen ground in China: Beijing, Academia Sinica, scale 1:4,000,000. Li Bingyuan and Li Jijun, 1991, Quaternary glacial-distribution map of the Qinghai-Xizang (Tibet) Plateau [in Chinese and English]: Beijing, Science Press, 10 p., scale 1:3,000,000. Li Bingyuan, Wang Fubao, Zhang Qingsong, and others, 1983, Quaternary geology in Xizang, the series of the scientific ex- pedition to Qinghai-Plateau [in Chinese]: Beijing, Science Press, 192 p. Liu Dong-sheng6, ed., 1981, Geological and ecological studies of Qinghai-Xizang Plateau: Beijing, Science Press, v. 1 p. 1- 974, v. 2, p. 975-2138. Liu Tungsheng, ed., 1987, Aspects of loess research: Beijing, China Ocean Press, 477 p. Liu Tungsheng, 1988, Loess in China: Berlin, Heidelberg, Springer-Verlag, 234 p. Liu Tungsheng, ed., 1991, Loess, environment and global change: Beijing, Science Press, 288 p. Liu Dongsheng and Chang Tsunghu, 1964, The "Huangtu" (loess) of China: International Congress on the Quaternary, 6th, Warsaw, 1961, Report, v. 4, p. 503-523. Liu Dongsheng and Ding Menglin, 1984, The characteristics and evolution of the palaeoenvironment of China since the late Tertiary, in Whyte, R.D., ed., The evolution of the east Asian environment; Vol. 1, Geology and palaeoclimatology: Uni- versity of Hong Kong, Centre of Asian Studies, p. 11-40. Previously published variations on Liu's given name include Dong- sheng, Dong-sheng, Tongsheng, Tong-sheng, Tungsheng, and Tung Sheng. Liu Tongsheng and Wang Kelu, 1965, Loess deposits of China [in Chinese]: Beijing, Science Press, 244 p. Liu Tungsheng and Yuan Baoyin, 1982, Quaternary climatic fluctuation-A correlation of records in loess with that of deep sea core V28-238, in Research on geology (I) [in Chinese with English abstract]: Beijing, Academia Sinica, Cultural Relics Publishing House, Institute of Geology, p. 113-121. Liu Dongsheng and others, 1964, Loess of the middle reaches Huang He [Yellow River] valley [in Chinese]: Beijing, Sci- ence Press, 234 p. Liu Tongsheng and others, 1966, The composition and texture of loess [in Chinese]: Beijing, Academia Sinica, Science Press, 132 p. Liu Tung Sheng, Gu Xiong Fei, An Zhi Sheng, and Fan Yong Xiang, 1981, The dust fall in Beijing, China, on April 18, 1980, in Péwé, T.L., ed., Desert dust-origin, characteristics, and effect on man: Geological Society of America Special Paper 186, p. 149-157. Liu Tungsheng and others, 1985, Loess and the environment: Beijing, China Ocean Press, 251 p. Lozek, Vogen, 1964, Quartarmollusken der Tschechoslowakei: Prague, Akademie der Wissenschaften, Verlag der Tschech- slowakischen, v. 31, 374 p., 32 pl. Lhalungpa, L.P., 1983, Tibet, the sacred realm: New York, Aper- ture, Inc. 159 p. McGovern, W.M., 1924, To Lhasa in disguise: New York, Gros- set and Dunlap, 462 p. Mertie, J.B., Jr., 1937, The Yukon-Tanana region, Alaska: U.S. Geological Survey Bulletin 872, 276 p. Meteorological Bureau, Xizang Autonomous Region, Xizang climatic data (1951-1980): 1079 p. Molnar, Peter, 1989, The geologic evolution of the Tibetan Pla- teau: American Scientist, v. 77, p. 350-360. Pavlinov, VIL, 1959, Some data about the genesis of Chinese loess, in Popov, V.V., ed. Loess of northern china; Proceed- ings of the Commission for the Study of the Quaternary Pe- riod [in Russian]: Moscow, Izdatel'stvo Akademii Nauk SSSR, v. 14. Translated, 1964, by Program for Scientific Translations, Jerusalem, Israel, p. 57-84. Pésci, Marton, 1965, Genetic classification of the deposits consti- tuting the loess profiles of Hungary: Budapest, Academiae Scientiarum Hungaricae Acta Geologica, v. 9, p. 65-84. 1967, Horizontal and vertical distribution of the loess in Hungary: Krakow, Studia Geomorphologica Corpatho-Bal- canica, v. 1, p. 13-20. Péwé, TL., 1951, An observation on wind-blown silt: Journal of Geology, v. 59, p. 399-401. 1955, Origin of the upland silt near Fairbanks, Alaska: Geological Society of America Bulletin, v. 66, p. 699-724. 1968, Loess deposits of Alaska: International Geological Congress, 23rd, Prague, 1968, Report, v. 8, p. 297-309. 1975a, Quaternary geology of Alaska: U.S. Geological Survey Professional Paper 835, 145 p. 1975b, Quaternary stratigraphic nomenclature in unglaci- ated central Alaska: U.S. Geological Survey Professional Paper 862, 32 p. 1980a, A geologist in Tibet-National Science Founda- tion trip report: Tempe, Ariz., Arizona State University, manuscript in files of Geology Department, 37 p. 1983, REFERENCES CITED 53 1980b, Report of the Symposium of the Qinghai-Xizang (Tibet) Plateau 1980: Innsbruck, Austria, Zeitschrift fiir Gletscherkunde und Glazialgeologie, v. 16, no. 1, p. 135-144. 1981, Tibetan science updated: Geotimes, v. 26, no. 1, p. 16-20. 1986a, A report on the Third Chinese Conference on Permafrost-China expands research on frozen ground: The Northern Engineer, v. 18, p. 1-8. 1986b, China expands research in frozen ground-a report of the Third Chinese Conference on Permafrost: Innsbruck, Austria, Zeitschrift fiir Gletscherkunde und Glazialgeologie, v. 22, p. 89-95. 1987, Retransported loess on the south side of the Tian Shan, Xinjiang Uygur Autonomous Region, northwestern China [abs.]: Geological Society of America Abstracts with Programs, v. 19, no. 7, p. 804. Péwé, TL., and Holmes, G.W., 1964, Geology of the Mt. Hayes D-4 quadrangle, Alaska: U.S. Geological Survey Miscellaneous Geological Investigations Map 1-394, scale 1:63,360. Péwé, TL., and Journaux, André, 1983, Origin and character of loesslike silt in unglaciated south-central Yakutia, Siberia, U.S.S.R: U.S. Geological Survey Professional Paper 1262, 46 p. Péwé, TL., Liu Tung-Sheng, and Slatt, R.M., 1981a, Loess in southern Tibet [abs.]: Geological Society of America Ab- stracts with Programs, v. 13, no. 7, p. 528. 1985, Loess in southern Tibet [abs.], in Proceedings (ab- stracts) of the International Symposium on Loess Research, Xian, 1985: Beijing, Academia Sinica, p. 29. 1987, Retransported loess in the southern part of the Qinghai-Xizang (Tibet) Plateau, China, in Liu Tungsheng, ed., Aspects of loess research: Beijing, China Ocean Press, p. 70-73. Péwé, TL., Péwé, E.A., Péwé, RH., 1976, Rate of desert dust deposition [abs.]: Arizona Academy of Science Journal, v. 11, Proceedings Supplement, p. 95-96. Péwé, TL., Péwé, E.A., Péwé, R.H., Journaux, André, and Slatt, R.M., 1981b, Desert dust-characteristics and rates of depo- sition in central Arizona, in Péwé, TL., ed., Desert dust- origin, characteristics, and effect on man: Geological Society of America Special Paper 186, p. 169-190. Péwé, TL., Westgate, J.A., and Stemper, B.A., 1989, Refinement of age interpretation of Quaternary events in Fairbanks area, Alaska [abs.]: International Geological Congress, 28th, Washington, D.C., 1989, Abstracts, v. 2, p. 2-602. Pradyumna, PK., 1976, The changing face of Tibet: Lexington, Ky., University of Kentucky Press, 114 p. Pumpelly, Raphael, 1867, Geological researches in China, Mon- golia, and Japan: Washington, D.C., Smithsonian Contribu- tions to Knowledge, no. 202, v. 15, no. 143 p. 1879, Relation of secular rock disintegration to loess, gla- cial drift, and rock basins: American Journal of Science and Arts, v. 17, p. 133-144. Pye, Kenneth, 1987, Aeolian dust and dust deposits: London, Ac- ademic Press, 334 p. Rawling, C.G., 1905, The great plateau: London, Edward Arnold, 319 p. Reiter, R., and Reiter, G.J., 1981, Tibet-the last frontier: Ameri- can Meteorological Society Bulletin, v. 62, no. 1, p. 4-13. Richthofen, EP.W. von, 1877, China: Berlin, D. Reimer, v. 1, 758 p. [In German.] 1882, On the mode of origin of the loess: Hertford, Eng- land, Geological Magazine, decade 2, vol. 9, p. 293-305. 1886, Fuher fiir Forschungsreisende: Hanover, Germany, 734 p. Russell, R.J., 1944, Lower Mississippi Valley loess: Geological Society of America Bulletin, v. 55, p. 1-40. Rutter, Nat, Ding Zhongli, and Liu Tungsheng, 1991, Compari- son of isotope stages 1-61 with the Baoji-type pedostrati- graphic section of north-central China: Canadian Journal of Earth Sciences, v. 28, p. 985-990. Schultz, LG., 1964, Quantitative interpretation of mineralogical composition from X-ray and chemical data for the Pierre Shale: U.S. Geological Survey Professional Paper 391-C, 31 p. Sengor, A.M.C., 1981, Symposium on Qinghai-Xizang (Tibet) Plateau-Beijing (Peking), China: Eos, Transactions, Ameri- can Geophysical Union, v. 62, no. 30, p. 601-603. Shi Yafeng, ed., 1980, Glaciers in China: Shanghai, Shanghai Scientific and Technical Publishers. [~100 p.] 1988, Map of snow, ice, and frozen ground in China: Beijing, China Cartographic Publishing House, Lanzhou In- stitute of Glaciology and Geocryology, scale 1:4,000,000 [explanation, 32 pJ. Shi Yafeng, Hsich Tzechu, Chang Penhsing, and Le Chichum, 1980, Distribution, features and variations of glaciers in China, in World glacier inventory: Gentbrugge, Belgium, In- ternational Association of Scientific Hydrology Publication 126, p. 111-116. Shi Yafeng and Li Jijun, 1980, A glaciological research of the Qinghai-Xizang Plateau, in Organizing Committee, Proceed- ings of Symposium on Qinghai-Xizang (Tibet) Plateau [abs.]: Beijing, Academia Sinica, p. 221-222. 1981, Glaciological research of the Qinghai-Xizang Pla- teau in China, in Liu Dong-sheng, ed., Geological and eco- logical studies of Qinghai-Xizang Plateau; Vol. 2, Environment and ecology of Qinghai-Xizang Plateau: Beijing, Science Press, p. 1589-1597. Shi Yafeng, Zheng Benxing, and Li Shijie, 1992, Last glaciation and maximum glaciation in the Qinghai-Xizang (Tibet) Plateau-a controversy to M. Kuhle's ice sheet hypothesis: Zeitschrift fiir Geomorphologie, Supplement, v. 84, p. 19- 35. Sis, Vladimir, and Vanis, Josef, no date [late 1940's?], On the road through Tibet: London, Spring Books, 51 p., 224 photo pls., printed in Czechoslovakia, translated by Iris Urwin. Slatt, R.M., and Eyles, Nicholas, 1981, Petrology of glacial sand-implications for the origin and mechanical durability of lithic fragments: Sedimentology, v. 28, p. 171-183. Smith, H.T.U., and Fraser, H.J., 1935, Loess in the vicinity of Boston, Massachusetts: American Journal of Science, 5th se- ries, v. 30, p. 16-32. Sorkhabi, R.B., and Stump, Edmund, 1993, Rise of the Hima- laya-a geochronologic approach: Geological Society of America, GSA Today, v. 3, no. 4, p. 85, 88-92. Sun Jianzhong and Li Xingguo, 1986, Environment during the last glaciation, North China [in Chinese]: Xian College of Geology, 18 p. 54 LOESSLIKE SILT IN THE SOUTHERN QINGHAI-XIZANG (TIBET) PLATEAU, CHINA Swineford, Ada, and Frye, J.C., 1945, A mechanical analysis of wind-blown dust compared with analyses of loess: American Journal of Science, v. 243, p. 249-255. - Taber, Stephen, 1943, Perennially frozen ground in Alaska-its origin and history: Geological Society of America Bulletin, v. 54, p. 1433-1548. Tong Boliang, 1981, Some features of permafrost of the Qinghai- Xizang Plateau and the factors influencing them, in Liu Dong-sheng, ed., Geological and ecological studies of the Qinghai-Xizang Plateau; Vol. 2, Environment and ecology of the Qinghai-Xizang Plateau: Beijing, Science Press, p. 1795-1801. Tomirdiaro, S.V., 1972, Permafrost and economic development of uplands and plains [in Russian]: Magadan, Knigoizdat, 174 p. Trainer, FW., 1961, Eolian deposits of the Matanuska Valley ag- ricultural area, Alaska: U.S. Geological Survey Bulletin 1121-C, p. C1-C35. Twenhofel, W.H., 1932, Treatise on sedimentation: Baltimore, Md., Williams and Wilkins, 926 p. Udden, J.A., 1898, The mechanical composition of wind depos- its: Rock Island, III., Augustana Library Publications 1, 69 p. U.S. Government Printing Office, 1971, People's Republic of China, Atlas: Washington, D.C., 82 p. Wang Baolai and French, H.M., 1994, Climate controls and high- altitude permafrost, Qinghai-Xizang (Tibet) Plateau, China: Permafrost and Periglacial Processes, v. 5, No. 2, p. 87-100. Wang Fubao and Fan, C.Y., 1987, Climatic changes in the Qinghai-Xizang (Tibetan) region of China during the Holo- cene: Quaternary Research, v. 28, p. 50-60. Wang Fubao and Li Bingyuan, 1983, Introduction, in Li Bingyuan, Wang Fubao, Zhang Qingsong, and others, Qua- ternary geology in Xizang, the series of the scientific expe- dition to the Qinghai-Plateau [in Chinese]: Beijing, Science Press, p. 1-15. Wang Shaoling and Yao Heqing, 1981, Pingos of the Qingshui River valley in Qinghai-Tibetan Plateau [in Chinese]: Jour- nal of Glaciology and Cryopedology, v. 3, no. 3, p. 58-62. Wang Shuji, Yan Shun, Mu Guijin, and Gao Cunbai, 1991, Qua- ternary deposits and neotectonics in the southern Tian Shan area: International Union for Quaternary Research, 13th Congress, Beijing, 1991, Excursion Guidebook I, 32 p. Wang Yongyan and Song Hanliang, eds., 1983, Rock desert, gravel desert, sand desert and loess: Xian, Shannxi People's Art Publishing House, 186 p. Wang Yongyan, Teng Zhihong, and Lue Leping, 1982, Microtex- tures of loess and the genesis of loess in China, in Liu Tung- sheng, ed., Quaternary geology and environment of China: Beijing, China Ocean Press, Quaternary Research Associa- tion of China, p. 43-48. Wang Yongyan and Zhang Zonghu, eds., 1980, Loess in China: Xian, Shannxi People's Art Publishing House, 169 p. Ward, EK., 1926, The riddle of the Tsangpo Gorges: London, Edward Arnold, 328 p. Warn, G.E., 1953, Drought and dust on the plains: Weatherwise, v. 6, p. 67-71. Warn, G.F., and Cox, W.H., 1951, A sedimentary study of dust storms in the vicinity of Lubbock, Texas: American Journal of Science, v. 249, p. 553-568. Willis, Bailey, Blackwelder, Eliot, and Sargent, R.H., 1907, Re- search in China, Vol. 1, Part 1, Descriptive topography and geology: Carnegie Institution of Washington Publication 54, 353 p. Xie Yingqin, 1982, The temperature condition of permafrost de- velopment in the Qinghai-Xizang Plateau, in Proceedings of the Symposium on Glaciology and Cryopedology, Lanxhou [in Chinese]: Beijing, Science Press, Academia Sinica, p. 1-6. Xu Shuying, 1980, Evolution of paleographical environment of the Tanggula Mountains Pliocene-Quaternary [abs.]: in Organizing Committee, Proceedings of Symposium on Qinghai-Xizang (Tibet) Plateau: Beijing, Academia Sinica, p. 253. 1981, The evolution of the paleogeographic environments in the Tanggula Mountains in the Pliocene-Quaternary, in Liu Dong-sheng, ed., Geological and ecological studies of the Qinghai-Xizang Plateau; Vol. 1, Geology, geological his- tory and origin of Qinghai-Xizang Plateau: Beijing, Science Press, p. 247-255. Xue Xiangxu, 1984, The Quaternary mammalian fossils in the loess area of China, in Sadao Sasajima and Wang Yongyan, eds., The recent research of loess in China: Kyoto, Japan, Kyoto University and Northwest University, p. 112-159. Yang Yichou, Li Bingyuan, and Yin Zeheng, 1983, Geomorphol- ogy of Xizang (Tibet): Beijing, Science Press, 238 p. Yang Yichou, Li Bingyuan, Yin Zeheng, Zhang Qingson, and Wang Fubao, 1981: Basic characteristics of landforms on the Xizang Plateau, in Liu Dong-sheng, ed., Geological and eco- logical studies of Qinghai-Xizang Plateau; Vol. 2, Environ- ment and ecology of Qinghai-Xizang Plateau: Beijing, Science Press, p. 1745-1757. Yuan Fangee, Mu Guijin, and Yan Shan, 1983, Loess of Tian Shan, China [in Chinese]: Uriimgi, Xinjiang Geography, v. 6, no. 4. Zeuner, EE., 1949, Frost soils on Mount Kenya, and the relation of frost soils to aeolian deposits: Oxford, England, Journal of Soil Science, v. 1, p. 20-30. Zhadin, V.L., 1952, Mollusks of fresh and brackish waters of the U.S.S.R.-keys to the fauna of the U.S.S.R.: Academy of Sciences of the U.S.S.R. Zoological Institut, no. 46, 357 p. Translated from Russian by Gourevitch, A., 1965, from Moskva-Leningrad, Izdatel'stvo Akademii Nauk SSSR., by Israel Program for Scientific Translations, Ltd., Jerusalem, 384 p. [Available from U.S. Department of Commerce, Springfield, Va.] Zhang Linyuan, Dai Xuerong, and Shi Zhenguao, 1991, The sources of loess material and the formation of the loess pla- teau in China, in Okuda, Setsuo, Rapp, Anders, and Zhang Linyuan, eds., Loess, geomorphological hazards and pro- cesses: Catena Supplement 20, p. 1-14. Zhao Sichu, Ge Xudong, and Gao Fuging, 1976, Quaternary stratigraphy of the Mt. Himalaya area, in Academia Sinica, The scientific exploration team of Xizang, report of the sci- entific exploration in the Mt. Himalaya area- Quaternary geology [in Chinese]: Beijing, Science Press, p. 2-28. Zheng Benxing and Jiao Keqin, 1991, Quaternary glaciations and periglaciations in the Qinghai-Xizang (Tibetan) Plateau: In- ternational Union for Quaternary Research, 13th Congress, Beijing, 1991, Excursion Guidebook XI, 54 p. REFERENCES CITED 55 Zheng Benxing and Shi Yafeng, 1976, A discussion on the Quaternary glaciation of the Mt. Himalaya area, in Aca- demia Sinica, The scientific exploration team of Xizang, re- port of the scientific exploration in the Mt. Himalaya area- Quaternary geology [in Chinese]: Beijing, Science Press, p. 29-62. Zheng Du, Zhang Yong-zu, and Yang Qin-ye, 1981, Physico-geo- graphical differentiation of the Qinghai-Xizang Plateau, in Liu Dong-sheng, ed., Geological and ecological studies of the Qinghai-Xizang Plateau; Vol. 2, Environment and ecolo- gy of the Qinghai-Xizang Plateau: Beijing, Science Press, p. 1851-1860. Zhou Kunshu, Liang Xiulong, Ye Yongying, and Wang Wenlin, 1960, Sporo-pollen and plant relics of buried soils in older loess at Chenjiaya of Wangjiagou, Lishi County, Shanxi Province: Quaternaria Sinica, v. 1, p. 104-111. # U.S. Government Printing Office: 1995-685-633 sau t aoe Bedding Strike a Slope Directions Slip Directions AVAILABILITY OF BOOKS AND MAPS OF THE U.S. GEOLOGICAL SURVEY Instructions on ordering publications of the U.S. Geological Survey, along with prices of the last offerings, are given in the current-year issues of the monthly catalog "New Publications of the U.S. Geological Survey." Prices of available U.S. Geological Survey publications re- leased prior to the current year are listed in the most recent annual "Price and Availability List." Publications that may be listed in various U.S. Geological Survey catalogs (see back inside cover) but not listed in the most recent annual "Price and Availability List" may no longer be available. Reports released through the NTIS may be obtained by writing to the National Technical Information Service, U.S. Department of Commerce, Springfield, VA 22161; please include NTIS report number with inquiry. Order U.S. Geological Survey publications by mail or over the counter from the offices listed below. BY MAIL Books Professional Papers, Bulletins, Water-Supply Papers, Tech- niques of Water-Resources Investigations, Circulars, publications of general interest (such as leaflets, pamphlets, booklets), single copies of Earthquakes & Volcanoes, Preliminary Determination of Epicen- ters, and some miscellaneous reports, including some of the forego- ing series that have gone out of print at the Superintendent of Documents, are obtainable by mail from U.S. Geological Survey, Information Services Box 25286, Federal Center Denver, CO 80225 Subscriptions to periodicals (Earthquakes & Volcanoes and Preliminary Determination of Epicenters) can be obtained ONLY from the Superintendent of Documents Government Printing Office Washington, DC 20402 (Check or money order must be payable to Superintendent of Documents.) Maps For maps, address mail orders to U.S. Geological Survey, Information Services Box 25286, Federal Center Denver, CO 80225 Residents of Alaska may order maps from U.S. Geological Survey, Earth Science Information OVER THE COUNTER Books and Maps Books and maps of the U.S. Geological Survey are available over the counter at the following U.S. Geological Survey offices, all of which are authorized agents of the Superintendent of Documents. +_ ANCHORAGE, Alaska-Rm. 101, 4230 University Dr. +_ LAKEWOOD, Colorado-Federal Center, Bldg. 810 +_ MENLO PARK, California-Bldg. 3, Rm. 3128, 345 Middlefield Rd. +_ RESTON, Virginia-USGS National Center, Rm. 1C402, 12201 Sunrise Valley Dr. *+ _ SALT LAKE CITY, Utah-Federal Bldg., Rm. 8105, 125 South State St. +_ SPOKANE, Washington-U.S. Post Office Bldg., Rm. 135, West 904 Riverside Ave. +_ WASHINGTON, D.C.-Main Interior Bldg., Rm. 2650, 18th and C Sts., NW. Maps Only Maps may be purchased over the counter at the following U.S. Geological Survey offices: *+ _ FAIRBANKS, Alaska-New Federal Bldg, 101 Twelfth Ave. *+ _ ROLLA, Missouri-1400 Independence Rd. *+ _ STENNIS SPACE CENTER, Mississippi-Bldg. 3101 The Loma Prieta, California, Earthquake of October 17, 1989-Aftershocks and Postseismic Effects U. S. DEPosiTORy PAUL A. REASENBERG, Editor Oct 1 0 1997 EARTHQUAKE OCCURRENCE WILLIAM H. BAKUN and WILLIAM H. PRESCOTT, Coordinators U.S. GEOLOGICAL SURVEY PROFESSIONAL PAPER 1550-D Prepared in cooperation with the National Science Foundation UNITED STATES GOVERNMENT PRINTING OFFICE, WASHINGTON : 1997 DEPARTMENT OF THE INTERIOR BRUCE BABBITT, Secretary U.S. GEOLOGICAL SURVEY Gordon P. Eaton, Director Any use of trade, product, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government Manuscript approved for publication, January 2, 1997 Library of Congress catalog-card No. 92-32287 For sale by the U.S. Geological Survey Information Services Box 25286 Denver Federal Center Denver, CO 80225 CONTENTS Introduction -- By Paul A. Reasenberg Aftershocks of the Loma Prieta earthquake and their tectonic implications ___ By Lynn D. Dietz and William L. Ellsworth Response of regional seismicity to the static stress change produced by the Loma Prieta earthquake --------------- By Paul A. Reasenberg and Robert W. Simpson Spatial variations in stress from the first six weeks of aftershocks of the Loma Prieta earthquake By John W. Gephart Loma Prieta aftershock relocation with S-P travel times from a portable array By Susan Y. Schwartz and Glenn D. Nelson Empirical Green's function study of Loma Prieta aftershocks: determination of stress drop ------------------------ By H. Guo, A. Lerner-Lam, W. Menke, and S.E. Hough U.S. Geological Survey aftershock ground-motion data By Leif Wennerberg Response of U.S. Geological Survey creepmeters to the Loma Prieta earthquake -__- By K. S. Breckenridge and R.W. Simpson Increased surface creep rates on the San Andreas fault southeast of the Loma Prieta main shock ------------------------ By Jeff Behr, Roger Bilham, Paul Bodin, Kate Breckenridge, and Arthur G. Sylvester Effect of the Loma Prieta earthquake on fault creep rates in the San Francisco Bay region ---------------------------- By Jon S. Galehouse Postseismic strain following the Loma Prieta earthquake from repeated GPS measurements ------------------ By Roland Biirgmann, Paul Segall, Mike Lisowski, and Jerry L. Svarc Page D1 49 73 91 105 121 143 179 193 209 Shallow, postseismic slip on the San Andreas fault at the northwestern end of the Loma Prieta earthquake rupture zone By John Langbein Models of postseismic deformation and stress transfer associated with the Loma Prieta earthquake --------------------- By MF. Linker and J. R. Rice A shear strain anomaly following the Loma Prieta earthquake By M.T. Gladwin, RL. Gwyther, and R.H.G. Hart A magnetotelluric survey of the Loma Prieta earthquake area: implications for earthquake processes and lower crustal conductivity ------------------------- By Randall L. Mackie, Theodore R. Madden, and Edward A. Nichols 245 253 277 289 THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989; EARTHQUAKE OCCURRENCE AFTERSHOCKS AND POSTSEISMIC EFFECTS INTRODUCTION By Paul A. Reasenberg, U.S. Geological Survey CONTENTS Page Introduction D1 Seismological studies 1 Postseismic movements on the San Andreas and nearby faults Postseismic deformation Of the CMUSt -------------------------------- Crustal fluids and electrical conductivity --------------------------- w t h INTRODUCTION While the damaging effects of the earthquake represent a significant social setback and economic loss, the geo- physical effects have produced a wealth of data that have provided important insights into the structure and mechan- ics of the San Andreas fault system. Generally, the period after a large earthquake is vitally important to monitor. During this part of the seismic cycle, the primary fault and the surrounding faults, rock bodies, and crustal fluids rapidly readjust in response to the earthquake's sudden movement. Geophysical measurements made at this time can provide unique information about fundamental prop- erties of the fault zone, including its state of stress and the geometry and frictional/rheological properties of the faults within it. Because postseismic readjustments are rapid com- pared with corresponding changes occurring in the preseismic period, the amount and rate of information that is available during the postseismic period is relatively high. From a geophysical viewpoint, the occurrence of the Loma Prieta earthquake in a section of the San Andreas fault zone that is surrounded by multiple and extensive geo- physical monitoring networks has produced nothing less than a scientific bonanza. The reports assembled in this chapter collectively ex- amine available geophysical observations made before and after the earthquake and model the earthquake's prin- cipal postseismic effects. The chapter covers four broad categories of postseismic effect: (1) aftershocks; (2) postseismic fault movements; (3) postseismic surface de- formation; and (4) changes in electrical conductivity and crustal fluids. SEISMOLOGICAL STUDIES The earthquake occurred inside the U.S. Geological Survey's Northern California Seismographic Network (NCSN) and is the largest earthquake within the network to be recorded by it. More than 75 high-gain seismographs recorded the main shock and nearly 11,000 aftershocks during the first 2 years of the earthquake sequence. Many of these instruments had been recording for 20 years be- fore the earthquake. The approximately 5,600 earthquakes that they recorded in the region before the earthquake allow detailed comparisons of the pre- and post-earth- quake activity. In addition, some 170 records were ob- tained from strong-motion instruments within 200 km of the epicenter. After the earthquake, 38 temporary seismo- graphs (mostly strong motion instruments) were deployed to record the aftershock sequence. The first three papers in this chapter analyze the seismological data and focus- their analyses on the structure of the San Andreas fault, its interaction with the major nearby faults, and the con- temporary tectonics of the greater Pacific-North Ameri- can plate boundary in the San Francisco Bay area. Dietz and Ellsworth provide a comprehensive survey of the aftershock locations and magnitudes, the focal mecha- nisms for the main shock and the larger aftershocks, and the time-evolution of the aftershock sequence. From these observations they infer possible geometries of the faulting structures active in the earthquake and their possible ki- nematic and geometric relationship to the San Andreas fault. The aftershock focal mechanisms were highly di- verse in orientation and unlike those of the pre-earthquake seismicity, leading Dietz and Ellsworth to conclude that the stress drop in the main shock may have been nearly complete. At greater distances from the earthquake, Reasenberg and Simpson examine the post-earthquake seismicity changes in central California. They compare the 20 years of seismicity recorded in central California before the earthquake to the activity during the 20-month postseismic period. They find that some regions (for example, the San Francisco peninsula) sustained an increase in seismic ac- tivity after the earthquake, while at least one other area (the Hayward fault) sustained a decrease in activity. These D1 D2 AFTERSHOCKS AND POSTSEISMIC EFFECTS changes are consistent with the stress changes calculated for simple elastic dislocation models of the earthquake when the effective coefficient of friction on the faults is assumed to be low (less than 0.3). A surprisingly small amount of stress change (0.1 bar) was apparently suffi- cient to affect the seismicity on these nearby faults. Another view of the stress changes produced by the earthquake is provided by the aftershock focal mecha- nisms. Gephart uses fault plane solutions for the first 6 weeks of the aftershock sequence to map the spatial varia- tions in stress in the vicinity of the main shock rupture. Near and northwest of the main shock hypocenter, he finds low shear stress remaining on the main shock fault plane after the earthquake, again consistent with the inference of a nearly complete stress drop in the earthquake. South- east of the hypocenter, near Pajaro Gap and the creeping segment of the fault, significant right lateral shear appar- ently remained on the fault. Aftershock observations also were used to hone seis- mological technique. Schwartz and Nelson use synthetic calculations and data from portable (PASSCAL) three- component seismographs in order to compare earthquake hypocenter location methods based on various combina- tions of P, S, and S-P arrival times. They find that the unsynchronized S-P arrival times observed with a sparse, portable array provided remarkably good locations rela- tive to those obtained from the more extensive, permanent NCSN array. The PASSCAL data also were used by Guo and others to estimate source characteristics of the aftershocks. Using an empirical Green's function tech- nique, they estimate Brune-model stress drops from the corner frequencies and moments of the aftershocks. They report an apparent dependence of stress drop on earth- quake moment. Other portable seismic instruments, in addition to the PASSCAL instruments, were deployed after the earthquake to record aftershocks. Altogether, 38 digital seismographs were deployed at 195 sites during the 3-month period after the earthquake. Wennerberg documents these deploy- ments and the massive and disparate data sets they pro- duced. He provides us with detailed maps, instrument descriptions, and data logs for the aftershock deployments. He also provides a table of references to studies utilizing these data and published elsewhere. In so doing, he as- sures that future seismological investigations of the earth- quake will be able to proceed efficiently with a map of these essential, recorded seismological data sets. POSTSEISMIC MOVEMENT ON THE SAN ANDREAS AND NEARBY FAULTS At the time of the earthquake, 27 creepmeters were recording surface movements at points on the San An- dreas and Calaveras faults within about 200 km of the epicenter. These instruments had been recording fault movement for periods up to 25 years before the earth- quake. In addition, small-scale (50-200 m) triangulation networks at about two dozen sites in the San Francisco Bay Area on the San Andreas, Calaveras, Hayward, and Concord-Green Valley faults had been observed regularly to measure surface fault displacements for up to 13 years before the earthquake. These observations form the basis for three papers describing and modeling the surface move- ments of the major faults in the Bay Area after the earth- quake. In a detailed study of surface fault displacements (creep) at the eight nearest creepmeter sites on the San Andreas and Calaveras faults, Breckenridge and Simpson find that both the sense (right-lateral or left-lateral) and magnitude of long-term creep rate changes on nearby faults agree with coseismic changes in horizontal shear stress calcu- lated for elastic dislocation models of the earthquake. In contrast, the amplitudes of coseismic steps in fault dis- placement measured at the same sites apparently bear little similarity to the calculated static stress changes; they pro- pose that dynamic stresses (shaking) likely account for these coseismic steps. The correlation between the regional creep rate changes and the earthquake-induced static stress changes parallels changes observed in the regional seis- micity rates (Reasenberg and Simpson) and suggests that over a period of a few years a dominant effect of the earthquake on the major Bay area faults was the sympa- thetic advancement and retardation of ongoing processes of seismicity and creep, driven by the static stress changes. Behr and others also investigate the increases in creep rate at the five USGS creepmeter sites on the San An- dreas fault studied by Breckenridge and Simpson, and at three additional sites between San Juan Bautista and Pajaro Gap. In contrast to the ephemerally enhanced fault creep observed after some earthquakes, they find a sustained level of elevated creep rate during the first 3 years after the Loma Prieta earthquake. Their two-dimensional elas- tic boundary models of the San Andreas and Calaveras faults can qualitatively explain the observed decrease in postseismic slip with distance from the earthquake. Triangulation measurements over small (50-200 m) ap- erture arrays spanning Bay area faults tell a similar story. Galehouse reports that the creep rate on the San Francisco peninsula section of the San Andreas fault was nil both before and after the earthquake, while creep rates approxi- mately doubled on the San Andreas fault southeast of the rupture. At the same time, creep on the Hayward and southern Calaveras faults decreased and, at some stations along the southern Hayward fault, creep may have stopped or become left lateral. Galehouse's findings are consistent with the seismicity and other creep observations presented in this chapter and with static stress change calculations (Breckenridge and Simpson), suggesting that the East Bay faults were at least temporarily relaxed by the earthquake. INTRODUCTION D3 How long this relaxation will last before being overtaken by continuing tectonic and additional seismogenic load- ing will surely be the subject of continuing measurements and modeling efforts. POSTSEISMIC DEFORMATION OF THE CRUST Prior to the earthquake, horizontal deformation in the epicentral region had been monitored for about 20 years by land-based trilateration measurements, and for several years by Global Positioning System (GPS) measurements. Thus, from a geodetic monitoring point of view, the earthquake's location was indeed advantageous. Resur- veying of these sites and the rapid and repeated survey of new sites by several campaigns launched after the earth- quake has provided an extensive and varied data set for constraining the surface deformation and inferring the de- formation at depth associated with the earthquake. The data and conclusions of these studies compliment the seis- mological and fault slip studies; together they provide one of the most extensive sets of geophysical observations of the crustal response to a large earthquake. Biirgmann and others investigate the deformation of the crust after the earthquake using data from repeated GPS measurements in two fault-crossing networks and com- pare it to the deformation inferred from preseismic obser- vations. They find that horizontal displacement rates northwest of the epicenter did not change significantly at the time of the earthquake, while the fault-parallel station velocities in the epicentral region increased significantly over pre-earthquake rates. They also find evidence for San Andreas fault-normal shortening near the epicenter and within the Foothills thrust. Using linear, elastic dislo- cation models, they find that a two-plane slip model with maximum depth of about 16 km best explains these data. They suggest that the inferred reverse slip on the Foot- hills thrust fault appears to have been triggered by stress changes produced by the earthquake. At the northwest end of the main shock rupture, just northwest of the study region of Biirgmann and others, a 10-km aperture trilateration network was repeatedly mea- sured by Langbein in the 2-year period after the earth- quake. Although no preseismic observations of these lines had been made, comparisons are made to preseismic mea- surements of the Black Mountain geodetic net 15 km to the northwest. Langbein's analysis of these data finds that the rapid temporal decay in the postseismic deformation rate in the network is typical for a postseismic geodetic response. It fails to find evidence for the fault-normal horizontal contraction inferred by Biirgmann and others from the regional GPS data. Linker and Rice use the geodetic observations to explore viscoelastic, time-dependent models of the postseismic deformation. Their models contrast with those of Biirgmann and others in that they allow for transient, postseismic slip in the deep, aseismic region of the fault zone, well below the maximum depth of slip in the elastic dislocation models. While the deep, viscoelastic models are compatible with most of the strike-slip deformation observed in the epicentral region, they do not explain the fault-normal contraction observed there and may overpredict strike-slip deformation near Black Mountain to the northwest of the earthquake. The paper explores a wide range of alternative assumptions about fault zone rheology and slip distribution in the crust. Near the southeast end of the main shock rupture, a borehole strainmeter recorded the tensor strain during and after the earthquake. This instrument had been recording changes in tensor strain for 10 years before the earth- quake. Gladwin and others compare the pre- and post- earthquake strain rates at this site and find an apparent increase in fault-parallel shear strain rate. The timing of this strain anomaly relative to the Loma Prieta earthquake and its regional extent are addressed in detail through comparisons with surface creep and geodetic measure- ments. The occurrence of the Chittenden earthquake se- quence in April 1990, located near the southeast end of the main shock rupture and only 10 km from the bore- hole, apparently increased the shear strain rate at this site. To explain these strain observations, Gladwin and others propose that a shallow asperity southeast of the main shock rupture initially resisted the increased shear load on the San Andreas fault, and then yielded in the April 1990 sequence, thereby unpinning the San Andreas fault and resulting in even higher shear strain rates. CRUSTAL FLUIDS AND ELECTRICAL CONDUCTIVITY Finally, Mackie and others present the results of a magnetotelluric survey of the epicentral area. Working with observations made one year after the earthquake, they find evidence for a vertical tabular zone of anoma- lously high electrical conductance in the lower crust be- low the earthquake. Unfortunately, no corresponding measurements were made before the earthquake, so their interpretation of these data cannot conclusively associate the anomaly with the earthquake; it might instead be a permanent feature of the San Andreas fault. However, they suggest an intriguing scenario in which fluids (wa- ter) that were already present in the crust experienced increased flow due to an increase in porosity and fluid connectivity in the fault zone caused by fracturing during the earthquake. THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989; EARTHQUAKE OCCURRENCE AFTERSHOCKS AND POSTSEISMIC EFFECTS AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE AND THEIR TECTONIC IMPLICATIONS By Lynn D. Dietz and William L. Ellsworth, U.S. Geological Survey CONTENTS Page Abstract DS Introduction 6 Data and methods 6 Hypocenter analysis PrOCedures -----------------------------.-- 6 Crustal model and location accuracy -------------------------- 7 Focal mechanism determination -------------------------------- 10 Display Of CATtHGUAKES nee 12 Overview of sequence 12 Main shock 14 Focal depth 14 Nucleation process 14 Development of the aftershock ZzON@-------------------------..-.-.-.- 15 Initiation 15 Formation 15 Decay 19 Structure of the aftershOCK ZON 21 Main zone 23 San Andreas fault 37 Castro and southern Sargent faults ----------------------------- 37 Faults north of the main zone and east of the San Andreas 39 Hanging Wall ACtivity 39 Relationship of the preshocks to the main shock and aftershocks 40 Discussion 41 The earthquake and the San Andreas fault ------------------- 41 Acknowledgements 42 References cited 43 ABSTRACT The Loma Prieta earthquake and its aftershocks rup- tured faults within a broad zone along the San Andreas fault in the southern Santa Cruz Mountains. The main body of aftershocks forms a southwest-dipping zone, ris- ing from the main shock hypocenter at 15.9 km depth to the San Andreas fault, and extending bilaterally along strike for a distance of 42 km. This zone, however, does not mark a simple fault plane, as it has a width of 1 to 2 km. Aftershocks also occur on faults well removed from the main zone on both sides of the San Andreas fault. The aftershock zone assumed essentially its final form within less than 1 day and only showed apparent growth to the southeast, where aftershock activity occurred on the nor- mally active part of the San Andreas and Castro faults. Aside from these two faults, the aftershocks occurred on structures that were essentially aseismic in the 20 years preceding the earthquake. Focal mechanisms of the after- shocks bear little resemblance to the main shock or pre- ceding seismicity. The mechanisms tend to have fault-normal P-axes, although a very wide range of mecha- nism types is observed. The diversity and orientation of the aftershock mechanisms and the lack of spatial correla- tion with the preshocks argue for a major reorientation of the stress field by the main shock and are consistent with a complete stress drop on the main shock fault. The main shock nucleated with a foreshock 1.6 s before the start of large-amplitude high-frequency seismic radiation. Although the foreshock was too feeble to trig- ger many of the strong-motion stations in the epicentral region (vertical acceleration<0.01 g), it had an amplitude magnitude of 5. Nearfield displacement seismograms indicate significantly larger moment release at long periods, equivalent to My=5 '/2. The foreshock thus ap- pears to mark the beginning of a continuous and rela- tively smooth process of moment release that only began to propagate away from the initial hypocenter after 1.6 s. The aftershock sequence has no distinct beginning, with the earliest locatable events emerging from the main shock coda. The Loma Prieta fault, as mapped by the main zone of aftershocks, exists as a structure distinct from the San Andreas fault along most of its length. Along its southern end, where the sense of displacement changes from re- verse to strike slip, it merges seamlessly with the San Andreas fault near Pajaro Gap. We speculate that the fault represents a former part of the San Andreas that has been abandoned as a strike slip fault and has evolved into a right-reverse fault to accommodate motion around the com- pressional bend in the San Andreas fault through the south- ern Santa Cruz Mountains. DS D6 AFTERSHOCKS AND POSTSEISMIC EFFECTS INTRODUCTION The Loma Prieta earthquake presents an unparalleled opportunity to study the source process of a major earth- quake using multiple and complementary data provided by geology, geodesy, and seismology. Our objectives in this paper are to describe the spatial distribution and tem- poral characteristics of the sequence, the connection be- tween the earthquake and the San Andreas fault, and the relationship between the preshocks, the main shock, and the aftershocks from the vantage point of short-period net- work seismology. The Loma Prieta earthquake occurred within a portion of the San Andreas fault system in one of the most densely instrumented regions for the observation of micro- earthquake activity. Systematic cataloging of accurately located hypocenters, magnitude >1.5, as recorded by the U.S. Geological Survey's telemetered seismic network (CALNET), began in this area over two decades before the 1989 earthquake. The CALNET thus provides a wealth of information on not only the complete after- shock sequence of a major event, but also on its relation- ship to background seismicity over an extended period of time. This study builds on the work of many of our col- leagues who have analyzed various aspects of the CALNET data to determine the crustal structure, hypocentral loca- tions, focal mechanisms, and state of stress as reported in this chapter and other chapters in this volume [also Eberhart-Phillips and others (1990), Michael and others (1990), Foxall and others (1993), Olson (1990), Oppenheimer (1990), Schwartz and others (1990), and Zoback and Beroza (1993)]. In this paper we extend the results of our earlier paper (Dietz and Ellsworth, 1990) to present a comprehensive picture of the earthquake and attendant seismicity through the first two years of the aftershock period, and we compare the patterns of hypo- centers and focal mechanisms with those observed during the preceding 20 years using uniform analysis methods. DATA AND METHODS The primary data sets examined in this study are the locations, magnitudes, and focal mechanisms of earth- quakes that occurred in the vicinity of the Loma Prieta earthquake. These data are derived from the analysis of seismograms recorded by CALNET, and are based on the measured arrival times and polarities of P-wave onsets and the duration of coda waves, recorded as the time from the P-wave onset to the time when the average absolute amplitude (2-second window) falls below a threshold value of 60 millivolts (Eaton, 1992). Earthquakes within the dashed box in figure 1 from January 1, 1969, through October 17, 1991, are consid- ered in this study. About 5,600 events were recorded in this region before the Loma Prieta earthquake and nearly 11,000 were processed in the first 2 years of the sequence. The region completely encompasses the primary aftershock sequence and is approximately twice as long and twice as wide as the extent of main shock rupture determined by geodetic and seismic means. HYPOCENTER ANALYSIS PROCEDURES Earthquake magnitude, denoted My, is determined us- ing the formulation of Eaton (1992). Supplementary mag- nitudes, principally M;, are used for the largest events. During the 1970's, the CALNET's detection capability increased to lower magnitudes as the network was en- larged and densified. Also, a change in network amplifi- cation in 1977 is thought to have introduced an artificial shift in catalog magnitudes (Marks and Lester, 1980; Habermann and Craig, 1988). Since 1980 the detection threshold in the study area has remained relatively con- stant and no network change affecting magnitudes is known to have occurred. Figure 2 shows the frequency-magnitude density distribution, binned by 0.1 magnitude units, of the 20° 37° 40° 20° 122° 40° 20° Figure 1.-Map showing study region (dashed line) and locations of CALNET seismic stations (triangles) within ~60 km of the Loma Prieta main shock (star). Solid triangles are assigned to velocity model 1; open triangles to model 2 (see table 1). Shaded area represents the extent of the aftershock zone. Labeled faults are the San Andreas, SAF; Sargent, S$; Calaveras, CAL; Hayward, HAY; and San Gregorio, SG. AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE AND THEIR TECTONIC IMPLICATIONS pre-Loma Prieta earthquake catalog. Due to the network changes mentioned above, we make no estimate of cata- log completeness or b-value for the early time range (fig. 2A). From 1980 through the end of the study, the catalog appears to be complete to magnitude 1.0 (fig. 2B). It should be understood, however, that a small percentage of the earthquakes with magnitudes above these thresholds are missing because their seismograms are obscured by larger events, such as in the coda of the main shock. The b-value of the preshocks is 0.64+0.02 calculated from M>1.0 events (0.74+0.04 from M>2.0 events). The aftershock b-value, omitting the incomplete record of the first day's activity, is 0.83+0.01 calculated from M>z1.0 events (0.74+0.03 from events). Event locations were derived from a merged set of hand- and machine-picked P-wave arrivals at CALNET stations (fig. 1) using the location program HYPOINVERSE (Klein, 1989), with the station corrections and P-wave velocity model described below. In locating events, the data are weighted as the product of a function of their quality (inversely proportional to the variance of the pick uncertainty), and a function of the epicentral distance of the station. The distance weight is unity out to an epicen- tral distance of 52.5 km, and decreases to zero at a dis- tance of 105 km. When examining the spatial distribution of hypocenters, we restrict our attention to the most accurate locations (fig. 3). The selected hypocenters have root mean square residual (rms) <0.15 s, horizontal standard error <1.0 km, vertical standard error <2.0 km, number of stations >8, and magnitude 21.0. The 3,085 selected preshocks include 78 percent of all My>1.0 and 82 percent of all M,22.0 events. The 4,998 selected aftershocks include 79 percent of the total number of events with M 421.0 and 86 percent of those M >2.0. The mean rms is 0.07 s. D7 CRUSTAL MODEL AND LOCATION ACCURACY In developing a traveltime model for locating the events, our objective was to find a simple model that accounted for the pronounced near-surface variations in velocity known to exist in the region (Wesson and others, 1973a) without introducing biases in the overall distribution due to incorrect modeling of the true three-dimensional earth structure. In the main shock, for example, traveltimes de- part from a laterally homogeneous model by up to 2 s (fig. 4). Fortunately, the fact that traveltime residuals at individual stations change slowly as a function of event location within the sequence implies that near-station structure and path-averaged delays account for most of the observed variance in traveltime. Accordingly, we model the traveltimes using a one-dimensional velocity model with station corrections. Obviously, locations made with these assumptions will contain biases due to unmodeled three-dimensional effects, but these appear to be both predictable and of minor importance in light of similar locations from the three-dimensional studies of Michael and Eberhart-Phillips (1991), and Foxall and others (1993). Using P-wave traveltimes from the main shock and 107 aftershocks distributed evenly throughout the aftershock zone, we calculated station traveltime corrections and one-dimensional velocity models with the joint hypocenter-velocity inversion program VELEST (Roecker, 1981; Kradolfer, 1989). To account for the obvious dif- ferences in crustal geology across the San Andreas fault, we partitioned the stations into two sets corresponding to the northeast (Franciscan basement) and southwest (Salinian basement) sides of the San Andreas fault (fig. 1). Then we simultaneously derived station corrections and a separate velocity model (seven layers over halfspace) - . A 7 . -_ Whtw - 4b ® o F 4 © h J - 8 &e ~ 1969-1979 NUMBER of EARTHQUAKES a00e000000008- 0 _ 1 2 = 38 4 50 6 7 MAGNITUDE 0 B 41000 S\ _ o as 4100 ® a ® BP F f 410 ® 6 - 0 se o a 4 1980-1989 L__1___1___!_ ¢ somim-ecécc006! 0 1 2 = 3 4 50 6 7 MAGNITUDE Figure 2.-Pre-Loma Prieta seismicity catalog. Number of earthquakes in each 0.1 magnitude-unit bin (A) from 1969 through 1979 and (B) from 1980 through October 17, 1989. D8 AFTERSHOCKS AND POSTSEISMIC EFFECTS for each side of the fault. Following the recipe to calcu- late a minimum one-dimensional velocity model (Kissling and others, 1994), we initially considered several candi- date models; our preferred velocity model (table 1) is used to locate all events in this study. Velocities below 18 km are poorly resolved, principally due to inadequate sam- pling. As a consequence, a tradeoff exists between focal depth and velocity for the deepest events in the sequence. In particular, the depth of the main shock could be in error by about 1 km. 20' 10' 37° 50' :..'.luv..l‘v\.v1..-.rl::\;..|..;\.I' '.'....|....l....|....l.-.: 10' 122° 50' 40 30' Figure 3.-Best located events within the study region (dashed line). (A) January 1, 1969, through October 17, 1989, and (B) October 18, 1989, through October 17, 1991. Symbols vary by magnitude: star = M27.0, circle = M25.0, diamond = M24.0, triangle = M23.0, and += M2z1.0. Dots labeled Z and Z" are depth section endpoints. AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE AND THEIR TECTONIC IMPLICATIONS D9 When considering the location results, two types of er- fidence ellipse, of the hypocenter location and averages rors need to be distinguished: relative location error, which _ +0.3 km in epicenter and +0.6 km in depth for our cata- measures the precision of the location in the reference _ log. This means that internal structures with dimensions frame of other locations; and absolute location error. The of about 1 km may be reliably interpreted in most cases. relative error is estimated from the standard error, or con- Figure 5 illustrates the distribution of standard error for 20' 10' 37° 50° 10° Figure 3.-Continued. D10 two portions of the aftershock zone, shown in transverse cross section. In figure 54, the central portion of the af- tershock zone has an apparent width of about 2 km, which is supported by the confidence ellipsoids. In contrast, the San Andreas fault at the southern end of the zone may have zero width (fig. 5B). Absolute location errors are more difficult to assess. It is well known (for example, Reasenberg and Ellsworth, 1982) that many velocity models may be constructed which produce equivalent patterns of relative event locations that differ in absolute location by up to several kilometers. We found this to be true of the Loma Prieta data in deriving our final velocity model. Using the method to calculate a minimum one-dimensional velocity model (Kissling and others, 1994), we constructed several models that fit the data equally well and have the same pattern of relative locations but have significantly different absolute loca- tions. The absolute locations vary in a northeast-southwest direction, normal to the fault and the geologic fabric of the region. To choose among candidate models, we se- lected the one that best reproduces the locations of sur- face explosions in the center of the aftershock zone (fig. 6). This model relocates the explosion nearest the main shock epicenter to within 0.3 km, and the nearby shots to within 1 km or better (table 2). A dense network of portable instruments deployed shortly after the main shock in the center of the after- shock zone recorded several events with sufficient spatial resolution to permit a direct contouring of their traveltimes. The event shown in figure 7 exhibits clear evidence for slower velocities southwest of the San Andreas fault as demonstrated by the more closely spaced isochrons to the southwest of the fault and by the suggestion of a refrac- tion angle at the fault. Locations produced by the three-dimensional model of Michael and Eberhart-Phillips (1991) and by our model agree well with each other, but both are systematically displaced ~1 km southwest from the apparent center of the isochrons. We expect the isoch- AFTERSHOCKS AND POSTSEISMIC EFFECTS ron center to be displaced to the northeast of the true epicenter when higher velocities exist to the northwest of the fault, as appears to be the case here. Thus, this quali- tative evidence supports the absolute accuracy (~0.5 km) of both our location and the three-dimensional location. The performance of the model at the southeastern end of the area is substantially poorer. The southernmost shot, in the San Andreas fault zone near San Juan Bautista, locates almost 3 km west of its actual position (fig. 6, table 2). Earthquake hypocenters in this vicinity also lo- cate about 3 km west of the San Andreas surface trace. The shot's mislocation suggests that our model does not account for velocity variations in the southernmost por- tion of the aftershock zone. This segment of the fault is known to have a very strong velocity contrast across it (Wesson and others, 1973a; Michael and Eberhart-Phillips, 1991), with very low velocity material located to the north- east of the fault. Traveltime contours for the main shock (fig. 8) show a refraction angle in the isochrons along the fault at San Juan Bautista, strong evidence for low veloci- ties to the northeast of the San Andreas in this region. This velocity contrast is of the opposite sense than the smaller contrast observed along the main body of the af- tershock zone (Lees, 1990; Mooney and Colburn, 1985; fig. 7). We infer that the seismicity near San Juan Bautista occurs on the San Andreas fault and that the westward mislocation is an artifact of improperly modeled velocity perturbations in this region. FOCAL MECHANISM DETERMINATION Focal mechanisms were calculated for the best-recorded aftershocks from P-wave first motions using FPFIT (Reasenberg and Oppenheimer, 1985), a grid-search algo- rithm that finds acceptable double-couple mechanisms. A minimum of 15 hand-picked first motions and a misfit value <0.15 were required. For the mechanisms studied 2 00 Reduced Traveltime (t-delta/6.0 s) (0) 1 © 40 60 Epicentral Distance (km) Figure 4.-Main shock traveltimes to CALNET seismic stations versus epicentral distance. Times are reduced by 6 km/s. AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE AND THEIR TECTONIC IMPLICATIONS Table 1.-Seismic velocity models (P-wave) Depth to Top Model 1 Model 2 of Layer SW of San Andreas _ NE of San Andreas km km/s km/s 0.0 3.35 2.99 1.0 4.58 3.99 2.0 5.26 5.57 5.0 5.96 5.98 8.0 6.16 6.26 13.0 6.22 6.44 18.0 6.52 6.58 25.0 8.00 7.98 here, the average uncertainties in strike, dip, and rake are 16°, 21°, and 30°, from an average of 57 first motions per event. In addition to determining the most probable mecha- nism for a particular earthquake, we also use FPFIT to test the hypothesis that a particular mechanism or class of mechanisms will fit the first motion data of an event within the confidence limits of its unconstrained (most probable) solution. For example, we can search for all earthquakes that have mechanisms that resemble the main shock. In almost all parts of the study area, we find that the FPFIT focal mechanisms for our HYPOINVERSE- located events explain the first-motion data as accurately as mechanisms of the same events located by D11 three-dimensional methods. However, minor first-motion discrepancies exist for most mechanisms, whether com- puted using one-dimensional or three-dimensional mod- els. The main shock, for example, has discrepant motions at a number of stations (fig. 9). The three-dimensional models of Michael and Eberhart-Phillips (1991) and Foxall and others (1993) do not account for these discrepancies. Discrepant first-motions at stations to the northwest of the epicenter cannot be explained by a lateral refraction, which would be apparent in the main shock isochrons of figure 8. These discrepancies suggest that significant lateral variations in structure may be present near the hypocenter. The one area where unmodeled effects of lateral struc- ture appear to significantly bias the focal mechanisms is the southeastern part of the study region. For events on the San Andreas fault near San Juan Bautista, the velocity contrast across the San Andreas fault is so strong that first P-arrivals to all stations on the northeast side undergo significant lateral refraction (Spieth, 1981). Because this refraction is not accounted for in the one-dimensional velocity model, the calculated take-off angles are incor- rect, and the resulting focal mechanisms for these San Andreas events have an apparent shallow dip on their northwest-trending fault plane, whereas the northeast- trending auxiliary plane is vertical and strikes normal to the fault. We believe that these events actually have nearly vertical nodal planes and represent horizontal strike slip movement along the San Andreas fault proper. The appar- ent dip of the San Andreas fault as defined by the hypo- B G a' 0 al Y TOT I T OTG T T T T POF P OP OP OP OTT T "I I T T T T T T T - + - I z; f++ é“, - : J of a 6 #1 tL I § } 1 of to o 4 I [ _] L _| E 10 |- { is [ 1 [ | n [ | [ +4 | L - L f - -I - -I - J+ | - - | - 20 1 1 1 1 1 1 1 Loa 1 L L L 1 1 14 1 1 1 1 1 1 1 L 1 oa 1 T L L 14 0 10 0 10 DISTANCE (KM) Figure 5.-HYPOINVERSE error ellipse axes of well-located M,22.0 aftershocks projected in cross section for (A) the region just north of the main shock (section B-B ' of figure 214) and (B) the creeping section of the San Andreas fault (section G-G ' of figure 214). Solid inverted triangle marks the surface trace of the San Andreas fault; open triangle marks the Sargent fault. D12 centers (fig. 5B) may also be an artifact of the unmodeled strong velocity contrast across the fault. DISPLAY OF EARTHQUAKES Many cross sections of hypocenters in this paper por- tray each earthquake as a circular area with its dimension scaled to correspond to the rupture area of a constant stress drop circular crack. By displaying earthquakes as rupture areas, cross sections in the plane of the earth- quake display the approximate faulting area of the events. For our figures, we assume a 30 bar (3 MPa) stress drop (Ag) and use M, to determine seismic moment (Mq), and thereby radius (r), using the formulas 10g 19My = 1.5M4 + 16.0 and 1 (7M0)? r =| --- __ -T 20° h LN C { a C - 10' —_ l C I 37° 50° |- i I Ns - pope oN paese a a 1 aa ala Pally bull t 10° 122° 50° 40' 30' Figure 6.-Map of study region shows the main shock (star), actual shot locations (numbered squares, listed by shot number in table 2) and shot locations using our velocity model (x). Dots show landmarks: SC, Stevens Creek reservoir dam; LR, Lake Ranch reservoir; LX, Lexington dam; EL, Lake Elsman; LP, Loma Prieta peak; GF, Grizzly Flat; HZ, Hazell Dell and Mt Madonna Rd. intersection; PG, Pajaro Gap; SJ, San Juan Bautista. The study area is divided into 7 subregions: SCV, Santa Clara Valley; LPN, Loma Prieta North; LPC, Loma Prieta Central; LPS, Loma Prieta South; CSA, Creeping San Andreas; SAR, Sargent and Castro faults; MON, Monterey Bay area. AFTERSHOCKS AND POSTSEISMIC EFFECTS Note that r scales as AG '>. Because a change in AG of a factor of 3 translates into a change in r of only a factor of 0.7, the plotted symbols only weakly depend on the assumed value of AG. The analysis of small subsets of aftershocks by Hough and others (1991) and Fletcher and Boatwright (1991) suggest that the mean AG value for the sequence may be greater than our assumed value by as much as a factor of 3. Thus our rupture areas may overestimate the area involved in faulting up to a factor of 2. Earthquake focal mechanisms appear in back- hemisphere projection on cross sections and as lower-hemisphere projections in map view, with compres- sional quadrants shaded. Strain axes for the equivalent double couple appear as correctly proportioned horizontal projections of the three-dimensional axes. OVERVIEW OF SEQUENCE Hypocenters of the main shock and its aftershocks acti- vated a 45- to 60-km-long and 10- to 20-km-wide volume of the crust along the San Andreas fault in the southern Santa Cruz Mountains. The events occur over the entire depth range of the seismogenic crust. The main body of aftershocks defines a tabular, southwest-dipping zone, with the main shock near the base and center of the distribu- tion (fig. 10). Aftershocks deeper than 10 km define a plane dipping 65°+5°SW and striking N51°W+2°, consis- tent with the main shock focal mechanism. Rupture in the main shock evidently spread upward and bilaterally along the strike of this tabular zone. Most aftershocks cluster around the perimeter of the distribution and surround a relatively aseismic interior (fig. 10) which generally cor- responds to the extent of main shock rupture, as described in detail in Spudich (1996). The less-seismic interior is somewhat obscured in figures 10C and 10D by off-fault activity projected onto the main aftershock surface; later figures offer a clearer view. At its southeastern end, the aftershock zone warps into a near-vertical surface and merges with the San Andreas fault. The diversity of focal mechanisms in the aftershock series is particularly great, with few events possessing mechanisms similar to the main shock. The larger events have mechanisms that span the full range of possibilities: right-lateral, left-lateral, reverse, and normal. Analysis of the state of stress implied by the aftershock mechanisms suggest near-total shear-stress release in the main shock (Michael and others, 1990; Beroza and Zoback, 1993). In addition to the main dipping zone, aftershocks de- fine numerous other clusters, many of which are well re- moved from the primary rupture surface of the earthquake. Hypocenters within these clusters align with the nodal planes of their focal mechanisms, providing strong sup- port for the argument that they represent coherent faults AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE AND THEIR TECTONIC IMPLICATIONS Table 2.-Actual shot locations and misfits of their relocations using our velocity model D13 Shot Location Misfit (Actual -> Relocation) Shot Ref Time Latitude Longitude At Epicentral Error - Depth yrmoda - hrmn s N W s azm km km 1 sp2* 810612 730 0.00 37° 0.42 121° 41.15 0.22 20° _ 1.46 1.50 2 800 0.02 _ 36° 54.01' 121° 48.02 0.08 267° - 0.79 0.17 3 sp3t 910522 604 0.00 37° 10.37 121° 47.49" 0.09 23° _ 0.26 1.10 4 spst 910524 700 0.00 37° 5.10' 121° 51.68" 0.11 48° _ 0.94 0.99 5 spot 702 0.00 _ 36° 51.43 121° 33.68" 0.49 256° _ 2.62 3.52 6 sp2t 704 0.00 37° 6.34' 121° 50.83" 0.15 46° _ 0.69 1.30 7 spit 804 0.00 37° 1.28" 121° 54.17 -0.01 178° _ 0.31 0.44 *Mooney and Colburn (1985) TMurphy and others (1992) (L. Seeber, written commun., 1992). While some of the secondary faults activated in the aftershock sequence were also active in the two decades preceding the main shock, most were not; thus they represent a triggered response to the stress redistribution of the main shock. The complexity of the sequence, in particular the large component of vertical displacement in the earthquake, Figure 7.-Map showing the epicenter of an aftershock (October 19, 1989, at 0634 UTC) as determined from one-dimensional velocity model (diamond) and three-dimensional model (hexagon); positions of CALNET seismographic stations (triangles) and portable seismic instruments (dots); and contours of relative P-wave traveltimes of the event (dotted lines, contour interval 0.2 s). while surprising at first appearance, can be understood in terms of the local geometry of the San Andreas fault (Dietz and Ellsworth, 1990). Within the southern Santa Cruz Mountains, the San Andreas fault makes a prominent left 0 40 Distance from Main Shock Epicenter (km) -40 Distance from Main Shock Epicenter (km) Figure 8.-Map of epicentral region showing epicenter of the Loma Prieta earthquake (dot), first motions at CALNET seismographic stations (U or D), contours of relative P-wave traveltimes of main shock (dotted lines, contour interval 1 s), and surface projection of main shock focal mechanism nodal lines from figure 9 (heavy lines). Note the field of compressional first-motions (U) in the dilatational quadrant to the north of the epicenter. Discrepant dilatational first-motions (D) may also be seen to the east of the epicenter. SJ, San Juan Bautista. D14 AFTERSHOCKS AND POSTSEISMIC EFFECTS (compressional) bend, connecting straighter subparallel segments to the north and south. Over geologic time, the compression within the bend must be relieved as the plates move either by lateral flow, subduction or mountain- building. This event contributed to the uplift of the moun- tain range (although Loma Prieta peak was downdropped about 10 cm), and the ratio of horizontal to vertical slip on the fault is predicted well by simple kinematic argu- ments (Dietz and Ellsworth, 1990). MAIN SHOCK The Loma Prieta My =7.1 earthquake initiated at 00:04 15.28 UTC on October 18, 1989, at 37° 2.01" N 121° 53.08 W and 15.9 km depth. The 95-percent confi- dence ellipsoid for this hypocenter has semi-major axes with azimuth, plunge, and length (km) of 133°, 0°, 0.29; 43°, 15°, 0.58; and 226°, 74°, 0.89. Our preferred hypo- center is slightly southwest and significantly shallower than our previously published estimate (Dietz and Ellsworth, 1990). It is also slightly shallower than several other recent estimates (table 3). The revised depth results from slight changes in the crustal structure and station corrections. We prefer the present value principally be- cause our revised velocity model fits the explosion travel- time data significantly better than our previous model for stations with epicentral distances of less than the main shock focal depth. The focal mechanism of the earthquake, or more prop- erly, of the initial slip at the hypocenter, corresponds to a fault plane striking NSO°W and dipping 70°SW, with a rake of 140° (Oppenheimer, 1990). This regional P- and T-axis 90% confidence Figure 9.-Lower hemisphere projection of main shock focal mecha- nism showing dilatational (open circle) and compressional (+) first ar- rivals at CALNET stations. first-motion solution agrees well with teleseismic solu- tions (Choy and Boatwright, 1990; Kanamori and Satake, 1990; Romanowicz and Lyon-Caen, 1990), and simple geodetic models of the static displacement field (Arnadottir and others, 1992; Marshall and others, 1991; Lisowski and others, 1990). FOCAL DEPTH A trade-off exists between origin time, earthquake depth, and model parameters for the deepest events in our cata- log due to poor resolution of the deep velocity structure. Therefore, we conducted a series of experiments to test the stability of the main shock hypocenter. In one experi- ment, a range of starting depths between 1 and 22 km was used to test the convergence of the HYPOINVERSE soluton. All runs yielded main shock depths between 15.6 and 15.9 km. When the depth of the shock was held fixed at the trial starting depth (again varying from 1 to 22 km), the hypocenter at 16.5 km depth had the lowest rms (0.09 s; fig. 11). Fixed-depth solutions outside of the range from 15 to 17.5 km had significantly higher rms. Finally, we examined the performance of a suite of veloc- ity models computed using the same 108 events used to develop the final model. Here we fixed the depth of the main shock and simultaneously solved for the aftershock hypocenters and station corrections that minimized the overall data misfit. The lowest overall data variances were obtained when the initial slip of the main shock was fixed between 15 and 17 km; the minimum data variance was obtained with the main shock depth fixed at 16 km (fig. 12). NUCLEATION PROCESS The process by which the earthquake started is of inter- est, as it did not simply propagate uniformly away from the hypocenter beginning at the above origin time. As noted by Wald and others (1991, 1996), the CALNET origin time precedes the initiation of strong motion from the hypocenter by about 2 s, suggesting that our origin time corresponds either to a foreshock or to a more com- plex initiation process. Amplitudes of the high-frequency waves radiated in the first 1.5 s of the event were gener- ally too weak to trigger strong-motion instruments (typi- cally a vertical acceleration of 0.01 g), indicating relatively small slip at the initiation of rupture. Most strong-motion recorders, however, did trigger before the arrival of the shear waves radiated by the initial slip, and thus recorded them. Peak velocities of the shear waves corresponding to the S waves of the initial rupture were compared to peak velocities predicted in the frequency band from 2 to 5 Hz for events of various magnitudes using the relation of Joyner and Boore (1988): AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE AND THEIR TECTONIC IMPLICATIONS D15 logy = a + b(M -6) + c(M -6) + dlogr + kr +s 5.0 < M-7.7 r = (ro2 +h )1/2, where y is the ground motion parameter to be predicted, M is the earthquake magnitude, r, is the shortest distance from the recording site to the vertical projection of the fault rupture on the surface, / is the depth of the earth- quake, and the parameters for predicting peak velocity are: a = 2.17, b= 0.49, c= 0, d=-1.0, k=-0.0026, and s= 0.17. The observed peak velocities of the initial rupture generally fall within one standard deviation of the values predicted for a M=5.0 event at 16 km depth (fig. 13). Large-amplitude, high-frequency motions, correspond- ing to the start of the "main shock" as inferred from strong motion modeling studies (Beroza, 1991, 1996; Wald and others, 1991, 1996; Steidl and others, 1991, 1996) began 1.6 s after the initial event, or at 00:04 16.9 UTC. P- and S-arrival times for both the foreshock and the main shock could be read at a number of stations (table 4). P arrivals were picked from vertical components and $ arrivals were picked from horizontal components of acceleration after integration to velocity. Using the program VELEST to determine the separation between their hypocenters, we find the main shock locates 0.3 km to the southeast, 1.1 km above the initial event, and 1.61 s later. However, the uncertainty in this location due to the uncertainty in the identification and timing of the phases and in the shear velocity structure precludes us from rejecting the hypoth- esis that the two events nucleated at the same point. While it is possible to interpret the initial event as an immediate M=5 foreshock, examination of digital strong-motion records from instruments in the near-field of the event suggest otherwise. Integration of the observed accelerograms to displacement shows a continuous growth in displacement from the moment of initiation (fig. 14). To model the displacement seismograms, we use the com- plete synthetic seismogram for a point dislocation in a half-space (Johnson, 1974) and assume the fault orienta- tion determined from the first-motion data. The seismo- grams are modeled by determining the seismic moment release as a function of time, a well-posed linear inverse problem. The best fit to the observed displacement seis- mograms shows a continuous growth in seismic moment beginning at the high-frequency origin time of 15.28 s (fig. 15A). The total moment release by 1.6 s into the event, just up to the time of the main shock, is equivalent to a M=5"/2 earthquake (fig. 15B), significantly larger than the magnitude determined from 2 to 5 Hz waves (fig. 13). Thus, the initial event appears to mark the beginning of a continuous and relatively smooth process of moment re- lease that only began to propagate away from the hypo- center after 1.6 s. DEVELOPMENT OF THE AFTERSHOCK ZONE INITIATION The divide separating the main shock from its after- shocks is no less ambiguous than the boundary between the foreshock and the main shock. Modeling studies of the main shock, summarized by Spudich (1996) suggest a duration of at least 6 s and not more than 15 s, with most of the slip occurring in the first 10 s of the event. If aftershocks are defined as distinct earthquakes removed in time but not necessarily in space from faulting during the propagation of the main shock, the earliest recogniz- able aftershock, M; = 4.7, occurred 32 s after the start of the main shock (Simila and others, 1990). Earlier after- shocks of similar or larger magnitude, however, might have occurred but might have been obscured by the ar- rival of scattered (coda) waves from the main shock, as this first aftershock just emerges from them (see fig. 2-6 of McNally and others, 1996). During the first 10 minutes of the sequence about 20 aftershocks (M,~4) can be identified in the CALNET seis- mograms. Locations could be determined for eight of these aftershocks. McNally and others (1996) used records from their digital accelerometers to locate these and seven ad- ditional events. After 10 minutes, our catalog appears to be complete to M,;=4.0. Using the frequency-magnitude density distribution (for example, in fig. 16) to estimate completeness at later times, hours 1 to 6 appear complete to My~3.0, and hours 6 to 13 complete to M;~1.5. After hour 13 of the aftershock sequence, our catalog appears complete to My~1.0. A cautionary note is in order, however, as systematic underreporting of earthquakes of all sizes will occur when smaller events are buried in the codas of larger events. At the end of the first day, the aftershock rate was down to one event (M,;21.0) every 2 minutes, with an average event duration of ~10 s. Missing events after the first day should, therefore, be few in number, which agrees with the frequency-magnitude evidence for completeness dis- cussed above. FORMATION The aftershock zone formed extremely rapidly, assum- ing essentially its final shape by the end of the first day (fig. 10). Even the incompletely recovered first 46 min- utes of activity conforms well with the later activity, clearly mapping out the main dipping zone as extending approxi- mately 45 km along strike. Some of the early aftershocks fall outside of the main zone and are associated with AFTERSHOCKS AND POSTSEISMIC EFFECTS D16 ¢ ut se aures ay are ayenbyrreq ©,_ 1 = p/p se aum podesop ares yooyslojye ay; Jt sjuona Jo fenba yoy sfeauogut oum orunyjtresof 01 puodsanoa suonoas ssor ( asioAsuen pue (7-7) pue 'sdepy 'yoo0us ureur ay} oue sXep 01 9°1T€ (@) pue 'sXep 91g 01 0 I (3) 'sAep 0'I 01 7E0'0 (g) 's4ep TE0'0 01 100°0 (¥) uyitm soyenbyrreq 'asuanbas ay; yo juaudojamap resodura; pue am3siq -mm - A & acy aa fag agg aa cta Fag + aa a a a g a afa T aga a ga aaa f a gag aula ® %. ; V y* fe 21V J a aa a la a aaa f a a laa aaa a f a galaa aca a gaga aaa f a gaa aa aul 6 ra ® x +~4 + + 2 4 A A - Z (Wy) 3JONYVLSIG O02 (oes 09 Of Z $9 \ o ._...._..._../.// .,.v_/........ di agg 02 0 10€ 06 ,oo} fod I T C v b v © O [o & qa f aga aa Fig o o °C .......->_.....>L.._...>....._........._..>>....._>>.......-.<......._.>....... - R mtemrmgqemr hel \ R A ~ \ \ Wx OL \ 1 (WM) HLd3GQ \ 108 S .0 . ©+& * «$ © / y; e ~ \o v_ /f s «48 A A - aZ 4 /4/ ,xfi, 10 . ,, t- - SRV Wuk cy 02 aoa aca a Bac fagagaSigad gr- AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE AND THEIR TECTONIC IMPLICATIONS D17 ~T pr Yl C 4 4 F. 4 4 a ] 1 a 1 1 E net .c 1 3 [ _ yiP 1 1 buoy t. ] | * ] >- Licas ald FAP >- [agua ula aff Fewi & z: e P - - a o t e t Foss. t pote . E a L ** F t 3 F. a "” mmrr . * It f Fuad: 3 rary [0 pry ~ thas t 6 ha 43 0st - [ Log et _ a> ' F. [+ e. ,* t C + ai uo a ua f a ala ala a al a al lil aca ua al ala aaa ala a a f a a ala aa aa fag aaa a a aa fag aa g ag aa aao al a f a a a a a aa u f uca a ua ua al f acla a ua ula f ala aa a aa a a f a a a aa aa aa fag ga g aa ga fag aa g aa aa - psk 89 - + fu en cas > p* [~ 2 a a t \, ° F t 20% F . + *, . He F - C N C.. alld a 1 N (+ 1 a > ] aot aa aa agaac aaa da imu yor lvvr..v7Frv--rv—rv.r ~ - a* *, J - tarry: \_\ I F 1] g L L 4 [-l »A E= t 12 b p S aio 7 a {1 9 bes [o of | s FW _43 F 4 -$ EAP F 1 - 8 lirik d - it D18 AFTERSHOCKS AND POSTSEISMIC EFFECTS Table 3.-Main shock hypocentral estimates Source Origin Time Latitude Longitude Depth yrmoda hrmn _s N W km This study 891018 0004 15.28 37° 2.01" 121° 53.08' - 15.94 Dietz and Ellsworth (1990) 15.21 37° 2.37 121° S52.81' - 17.85 U.S. Geological Survey Staff (1990) 15.25 37° 2.19 121° 52.98' - 17.6 Eberhart-Phillips and Stuart (1992) 15.34 37° 2.02 121° 52.85 - 17.16 Pujol (1995) 15.26 37° 1.92 121° 52.98' - 16.3 Roecker and Gupta (this volume) 15.45 37° 2.22 121° 53.46' - 15.60 aftershock activity on secondary faults, principally to the northeast of the San Andreas fault. The aftershock zone apparently lengthens with time to the southeast where it overlaps the seismically active cen- tral segment of the San Andreas fault (figs. 3, 10). While some of this extension reflects the occurrence of earth- quakes normally expected to occur along this part of the San Andreas fault, the analysis of Reasenberg and Simpson (1992, and this chapter) shows that these events occurred at a higher rate after the mainshock and decayed in an aftershock-like manner. Including them in the definition of the aftershock zone would increase its length to about 60 km. 0-5 IIIIIIUTTIIIIIIIIIIIIII o p I | RMS Residual (s) 00 L LO _ l LOO | LOL ._ l L4 | L1 4 0 5 10 15 20 Fixed Depth (km) Figure 11.-Final traveltime rms-residual versus depth for the Loma Prieta main shock location. In these HYPOINVERSE tests the epicenter was free to move, but the depth was fixed. The main shock's depth of 15.9 km from an unconstrained relocation (arrow) falls within the fixed-depth range (15.0-17.5 km) yielding the lowest rms values. In general, the aftershocks are highly clustered and tend to populate the periphery of the main shock rupture as determined from studies of the main shock seismograms (Beroza, 1991, 1996; Wald and others, 1991, 1996). In nearly every case, these clusters contain at least one large aftershock (M4>4) from the first few hours' activity. The persistence of activity in these clusters is particularly strik- ing (figs. 10, 17) and clearly cannot be explained as sec- ondary aftershock activity spawned by a large aftershock from early in the sequence. One secondary aftershock se- quence, however, can be recognized following the April 18, 1990, M; =5.4 Chittenden earthquake (north of PG, fig. 17), which was also the largest event in the entire aftershock sequence. Note the rapid decay of its after- shocks, compared to the overall decline in activity (figs. 17, 18). 0.026 Residual Variance (sec**2) 0.022 0.018 12 14 16 18 20 Fixed Depth of Main Shock (km) Figure 12.-Residual variance of traveltimes for 108 earthquakes used to develop final velocity model for models computed with the main shock depth held fixed. Each data point corresponds to the residual variance obtained by inverting the traveltime observations for optimum earthquake hypocenters and station delays, with the constraint that the main shock depth has the value shown. Results show a clear preference for the unconstrained model solution that places the main shock at a depth of 15.9 km. AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE AND THEIR TECTONIC IMPLICATIONS D19 DECAY As a whole, the decay of the aftershock sequence conforms with the modified Omori's law (Utsu, 1969; Reasenberg and Simpson, this chapter). We determine a best-fit to Omori's law, dN/dt = K/(t + c) P, by binning the aftershocks into equal log time intervals, starting 11.5 minutes after the main shock (fig. 19). Using all events for the entire two-year catalog, we find: p = 0.94+0.01, c = 0.34+0.04, and K= 670+25. Observed rates during the first 13 hours of the sequence (first 6 intervals, figure 19) undoubtedly underestimate the true rate of M421.0 events, due to grossly incomplete report- ing at this magnitude threshold. When only events My423 are considered, p = 1.07+0.04, c = 0.04+0.03, and K= 26+3.4. The end of aftershock activity may be defined as the intersection of the Omori rate dN/dt with the rate preced- ing the earthquake. Reasenberg and Simpson (this chap- ter) estimate this time as 2.3+0.3 years, based on activity (M,21.5) within the broadly defined aftershock zone through April 17, 1990. In this study we define the "main aftershock zone" with a series of polygons and depth ranges (fig. 21B). Because the pre-earthquake rate varies widely throughout the aftershock zone, we compare the aftershock decay to the background rate (average number of M 21.0 events per day from 1980 to October 17, 1989) in three different areas: the main, dipping part of the af- tershock zone (portion of the main zone which lies in regions LPS, LPC, and LPN of figure 6; generally north of 36° 58"); the part of the main zone along the microseismically active part of the San Andreas fault at the south end of the sequence (portion of the main zone in region CSA of figure 6); and the previously active part of the Castro fault (commonly thought to be the Sargent fault-see below) to the southeast of the main shock hy- pocenter (region SAR in figure 6). Within the main dip- ping zone, aftershock activity will not reach the very low pre-Loma Prieta rate of 0.02 events/day until over 25 years have passed, or until 2015. This long time reflects the very low level of pre-Loma Prieta seismicity in this zone. Activity along the San Andreas fault immediately south- east of the earthquake would be expected to intersect the background rate of 0.16 events/day after 1=1.5 years if we model only the aftershocks prior to the April 18, 1990, Chittenden earthquake. However, this event further boosted the seismicity rate in this region with its own aftershock sequence, lengthening the actual time to decay to back- ground rate. Activity along the Castro fault met the back- ground rate after just one year. Peak Horizontal Velocity (cm/s) I I I 0 10 20 30 40 50 60 70 Epicentral Distance (km) Figure 13.-Peak horizontal velocity of S-waves recorded by strong-motion accelerographs from the first 1.0 s of rupture. Amplitudes were measured from accelerograms after integration to velocity. Measurement window restricted to direct arrivals from the foreshock and excludes arrivals from the main shock. Curves are predicted mean (solid) and +10 (dashed) peak horizontal velocity from Joyner and Boore (1988) for M=5 earthquake at 16 km depth. D20 AFTERSHOCKS AND POSTSEISMIC EFFECTS Table 4.-Foreshock and main shock arrival times (seconds after 891018 0004 15.35) Station Latitude Longitude Foreshock Main Shock N W P _ wt S$ wt P _ wt $ wt C563 37 12.60 - 121 48.18 4.51 0 7.173 2 6.13 0 9.65 2 SAOC 36 45.9 121 26.7 8.60 0 10.23 0 COO7 37 2.176 121 48.18 5.75 0 7.05 0 BSRZ 36 39.99 121 31.12 8.90 0 10.38 O0 WAHOO 36 58.38 121 59.76 3.87 0 6.01 3 5.25 0 7.172 3 LGPC 37 10.32 122 - 0.60 4.35 0 6.95 0 5.85 0 8.45 0 BRAN 37 2.82 121 59:10 3.25 0 5.40 0 4.95 0 6.65 0 BKSB 37 56.2 122 14.1 15.80 0 17.55 0 HBTM 36 51.01 121 33.04 6.62 0 7.97 0 HSFM 36 48.72 121 29.97 7.32 0 8.77 0 JBLM 37 7.69 122 10.08 5.25 0 6.90 -0 JMPM 37 27.33 122 - 9.93 9.57 0 11.12 0 HPLM 37 3.13 121 17.40 9.49 0 10.94 0 HQRM 36 50.02 121 12.76 11.11 0 12.66 0 A Santa Teresa Hills B Branciforte 0 § q hil C Walter's House D U.C. Santa Cruz o o E o - o O 0 Seconds Seconds Figure 14.-Displacement seismograms of vertical component of motion _ foreshock. Synthetic seismograms determined by inverting observed dis- from broad-band integration of digital accelerogaph stations located in the _ placements for seismic moment release rate of point source double couple near-field of the earthquake (solid line) and synthetic seismograms (dot-dash _ with orientation of main shock focal mechanism. Divergence of fit after 2 line). Bandpass of displacement filtered with 1-pole Butterworth filter at _ s reflects breakdown of point source approximation and may reflect change 0.01 Hz. Origin time for each seismogram is P-wave arrival time of the _ in radiation pattern. AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE AND THEIR TECTONIC IMPLICATIONS 1 e25 A dyne-cm 6 8 e24 e24 4 e24 2 e24 o eo 0 1 2 3 Seconds D21 O -| B io o iw ke 3 * & & s] - C fa} E o | o i -> © | t 0 1 2 3 Seconds Figure 15.-Cumulative seismic moment release versus time determined by inversion of near-field displacement seismograms (figure 14). Both the total seismic moment (A) and equivalent moment magnitude (B) are shown. Results indicate that the foreshock released more moment than would be predicted by either its amplitude magnitude of 5 (figure 13) or its triggering of strong-motion stations. I I I I T I & LW - |L <4 1000 x | /» 3 @ QH‘B o g - -< 100 |- a CC ® o _ 5 b = 0.83 4 1 i 1 !-aéaemedecec6! 0 0 1 2 3 4 5 6 7 MAGNITUDE Figure 16.-Number of earthquakes in each 0.1 magnitude-unit bin for aftershocks (excluding activity on October 18, 1989) in our catalog. The aftershock catalog is complete at M,21.0 and the b-value for the se- quence is 0.83+0.01. STRUCTURE OF THE AFTERSHOCK ZONE The overall form of the aftershock zone is dominated by the central dipping zone, or main zone, that rises from the main shock hypocenter toward the San Andreas fault to the northeast. About 75 percent of the aftershocks fall in this zone which extends from 2 to 18 km depth. The coincidence between the N51°W+2° strike and 65°+5°SW dip of the zone and the main shock fault plane leaves little doubt about the existence of a primary, causal rela- tionship between these events and the main shock. There is much more to the sequence, however, than continuing faulting in the style of the main shock disloca- tion. Both the occurrence of aftershocks within a much larger crustal volume, and the wide range of aftershock focal mechanism types demonstrate that numerous faults participated in the post-main shock process of crustal ad- justment. As first noted by Oppenheimer (1990), almost none of the aftershock focal mechanisms resemble the main shock. Indeed, with the exception of aftershocks at the base of the zone near the main shock, it can be fairly D22

? remy n.$3..qu”nu-inn.umI'm-”Qumra-u.u|-¢.uu¢u|nn¢uu oo % t" o ., * 20+ w. R +:, x ++ __ . -] +0. 00+ +o. + + + + 20.8 + *+ F < > [- & * + +00+ - *. +#. & + + .* + + + + +00* * ++ +004 + [¥- + + + & ++ * + +/ * + frt 4 ** L* % - v a0¢ . ,+ +4 *, & & + * +0 % + |- tot wroot og + * + - *to» + +o % f * +> +. + + # + + + |- + «a + + ~- «* ++ "$ +o 0C tuat to > + + + + ** + «+++ wor + * |- { 20¢00 * +4 2% .+ 2204+ - + & + % . £0® & 2 ie & .J *- & % b o 20t+ . - + so +o. + * + be = "Gf f . + *. + + &0* * * [- 1 .’,j-. + .%. *to .* eu, + + - * “Ct * + * * *t ** a. * *.* be * + + - ffi & * * * * + + + 00. 4+. £ tu .a +, 0+ + . +, 4, + + L_ + + e+ «.< +, + 0+ & : ¢ o* * ® o ut e ege 0 T. & « 244 r k l +* , * *+ + ++ *; to .+ * + -- e ++ Ca & M + & - + * + % Fa + + +2 + *, + * . #+* 08 * 400+ * +00% + 0+ + t + + * +. soy, } - '+ {A* a* +40 + 0# 4 * & + + & + & q + + *% ++ * |__ :, 44 @ ++ + + + + + _ ** l+ +08 . & &+ ** *, to <% * fl + ’!+ + fand &+ * +. .+_.' * vore # Fa + . * a 4 * P ¥ #. + ty +o t g' > t,. 0 +Titd tu, +. * gh en BP c," CC & o ket +t . *- €" % *a 2 & $ u+** * «-| + + + + m A ee thk} oa +a *%" wt t - s+ 4+ so %* +0. & + + + v4 + ++ + a + + & |__ + k. + <+ to.. . '% sof r' Gq * - 8 a * mel +t *to. . ** +, . th. e*. a *% ++". 20+ 26 e + 4 ++ *, 8 + t,t. &. *a "t *,. |- Ml #4 & #4 . + woe, *+ ~I + + £4 % .+ M + + ve . ngo "LG +44 + tert , '¢‘t +4 * -- Folk af 3 .3 Y . °""")'°f.‘ +C, .?,. . - + * + Fu to a wey Py C “é +14 + * bs: © a rtm F UI UTL O.} CALLAN T |__ + +o% + +, .+ #4 * + 4; + m _| & 2 ++ a+ * Lo. £32. _] * * + * - tGre " - + “a Q «+ (+# m'wttk'gwl l.-nn“tr-5h”--¢.rvq-|zwnu.|1-quuvvu + uy LUO OCT 18, 1989 ++ + i .‘0 + <4 .** 7 & + E x # i E- 3 - 0 O l=] l=] le] co el < sd Z! (WM) 3JONYLSIQ N 1990 1991 19 20 21 22 23 24 25 26 27 28 29 30 31 dots show positions of landmarks from figure 6. Earthquake symbols are the same as in figure 3. .-Distance versus time plots of aftershock seismicity from northwest (Z) to southeast Figure 17 ; (B) days 2-14; and (C) days 15-730. Labeled (Z') along the San Andreas fault for (A) day 1 AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE AND THEIR TECTONIC IMPLICATIONS D23 CTT -T TTT T TT T- T 7000 o - A 6000 S (V) - aA 200 |- g bes - 5000 [- c & -- Inu a f T a u - 4000 > F | a L LJ |_ g u. LL |_ 0 0 [~ Fl 3000 2 « 100 |- 8 +f UW |- c co m [_ © - g [~ E 2000 3 z |- 3 4 42 [~ :C [~ UO 1000 o L a bul p p L UU b [ p p p p p p p U 0 18 19 20 1990 1991 OCT 1989 Figure 18. -Cumulative number of events versus time for (A) the first 3 days of the sequence and (B) 2 years of the sequence. Heavy line shows all M,>3.0 events in our catalog and thin line shows all M,21.0 events. stated that no seismic displacements took place on the main shock fault once its dynamic motions stopped. Thus, the central dipping zone and the more removed parts of 10000. 1000. 100. 10. RATE (EQ PER DAY) 1.0 0.1 I | | | 0.01 0.1 1.0 10. 100. TIME AFTER MAIN SHOCK (DAYS) Figure 19.-Seismicity rate (number of earthquakes per day) versus time after the main shock (in days) for all M21.0 aftershocks in our catalog (diamonds). The line shows the best fit to Omori's Law, dN/dt = K/(t +c)", with parameter values given above. 1000. the sequence both speak to a complex process of adjust- ment that bears little resemblance to the typical San An- dreas earthquake sequence in which the aftershocks continue the main shock faulting process (for example, Eaton and others, 1970). To facilitate the examination of the aftershock zone we subdivide our discussion into five parts corresponding to the main dipping zone and four geographically distinct areas surrounding it. MAIN ZONE The main zone consists of a tabular volume of after- shock hypocenters that maintains a nearly-constant dip of 60° to 65° for a distance of 42 km, centered approxi- mately on the main shock (figs. 10, 20, 22, 23). The depth at the base of the zone averages about 18 km to the north- west of the main shock and shoals to about 8 km to the southeast at the point where it joins the San Andreas fault. The top of the dipping zone lies about 6 km below the surface, on average, and locates directly beneath the sur- face trace of the San Andreas fault. Seismicity shallower than 6 km primarily locates northeast of the San Andreas trace and defines no single large-scale structure. Specifi- cally, the main zone does not appear to cross the San Andreas fault as a continuous body. Viewed in longitudinal cross section (figs. 10, 23), the main zone is seen to be composed of numerous clusters of D24 AFTERSHOCKS AND POSTSEISMIC EFFECTS activity, generally located on the periphery of the zone. The interior of the zone is relatively devoid of activity. As has been noted by nu- merous authors, the less seismic interior of the zone correlates well with the regions that slipped during the main shock (see Spudich, 1996). The absence of aftershocks on the por- tions of the fault that slipped during the main shock is a general feature of the main shock/ aftershock process, at least for California earthquakes (Mendoza and Hartzell, 1988). Transverse cross-sections show that the main zone has an average width of 1 to 2 km (fig. 22). Because the width is resolved by the hypocentral locations (fig. 5) and has rela- tively sharply defined edges, it implies that these aftershocks map out a volume of rela- tively weak rock triggered into activity by the main shock. We cannot tell, however, on the basis of these data, where the main shock rup- ture passes in relation to the main zone. It could locate equally well within or on either edge of the body. Aftershock focal mechanisms in the dipping zone do little to clarify this picture. With the exception of a few focal mechanisms located at the base of the zone just northwest of the main shock, none of the aftershock focal mechanisms in the main zone show right- lateral, oblique-thrust motion on the main shock plane. Rather, they display a very wide range of mechanism types and orientations (fig. 20), with the only unifying trend being a tendency for the P-axes to orient approxi- mately normal to the fault plane (fig. 24). This fault-normal coordination of the P-axes is con- sistent with an approximately uniaxial stress field acting perpendicular to the main shock fault plane (Zoback and Beroza, 1993) and thus is incompatible with the stresses that drove the main shock faulting. It is possible that some aftershock mecha- nisms may have closely resembled the main shock mechanism while the grid search method of FPFIT selected a different mechanism as the optimal solution. To test this hypothesis, we re-examined each aftershock through De- cember 1989 to determine if the main shock focal mechanism (initial slip) could provide a reasonable explanation of the aftershock's ob- served first-motion polarities. Using FPFIT, we constrained the solution to be the main shock mechanism and compared the resulting misfit to the confidence region of each L T Imt * . + 90 a> ~ 0.1;3'0 >- lllllllll 1 o o C) (wx) HL430 to o% + +o + a & & & & $ (5 -s s s -N l < 2 ; 4 G 0 % 20 10 DISTANCE (KM) fault (SAF, solid triangle), Sargent fault (S, open triangle), Berrocal fault (B, hatched tri- Figure 20.-Enlargements of aftershock seismicity and selected focal mechanisms (lower angle), and coastline (arrow). Subregions SCV (A,B); LPN (C,D,E); LPC (F,G,H); LPS (LJ,K); CSA (L,M); SAR (N,0); and MON (P,Q). For all plotted mechanisms: number and letter hemisphere projections) in map view and depth section for the seven subregions shown in figure 6. Earthquake symbols are the same as in figure 3. For map views: lettered dots show labels refer to entries for the same event in appendix A (first motion data) and in table 6 (event locations and nodal planes). depth section endpoints, and unlabeled dots are the landmarks from figure 6. Faults are labeled in map view and denoted by symbols at 0 km depth on cross sections: San Andreas D25 AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE AND THEIR TECTONIC IMPLICATIONS (Wx) OL 0 Oc OL (WM) HLd49GQ panunuo}y-pz amsiq £090 §00L06 pelo 228006 OStHO pPLFPO OEE0O €260 91 LO 8010 8§00 L500 8E00 810168 I b S d N W J 1 E a O WM S AFTERSHOCKS AND POSTSEISMIC EFFECTS D26 panunuo}y-'9z amslq Agzvmoz - - 4 L - *. { - h - 295. 4 - - - 2 >: { - - Fes - - te then ~ - - n é"." 4 - - - 4 - - L 4 - - - { - { plug aa uca aus fuga a ac glug, pla ga a a a a f alg a g H H' H hs V ror - F '1 # { - f'z B - GP ols - - B - § - - - - - - - - B - - - - - - pog uaa aga | ggg uggla pag a aaa ac f asu gaga rrr rt o 10 DISTANCE (KM) Figure 22.-Continued o 10 DISTANCE (KM) D33 AFTERSHOCKS AND POSTSEISMIC EFFECTS D34 'aroy pasn euajuo juaAa 1oJ g17 angy Aytotustas ,, auoz ©1661 "L 1 1990190 YSnouy) '6861 "81 4290190 (g) PUE '6861 sag 'doup ssans eq-o¢ e Sununsse tare arnidnu 01 pajeos are sazts joqui{g '[66] "LI 499 'LI 1990190 ySnoxp '696] 'I Arenue; (y) [ urepy,, j[ng] seaipuy ueg ay) ySnouy1 ©6861 'g1 1290190 (@) pu® '68g61 'L1 12q0100 ySnory) 'g961 '1 Zrenue; (9) suoje (7) 4s 01 (7) MN wou; pareoor-3saq ayy jo suonoas yidag-'g7 amsiq (WM) JONYLSIG 08 OZ 09 0s (00 (ola 00 O L 0 ___--_—_- _ Tor P Or POP P TOT _ POT T T T TOT q-fl-Jfi-d—d TOT — Tor r r P TOT qq—fi- T —.— T TOT — Por r r P P OP TOT _ FOr P Or POP P POT [ LLOLL6 - 810168 O. ough 0 e 00 . C> m - lacs... - L 1 1 1 1 1 a 1 n- is — n 1 1 1 L Jfi 1 H — 1 1 4 h P 1 4 4 hum L_ 4 1 1 b. 1 _0P 1 h 1 .~ 1_ hL-om 1 h *j — 1 4 0 L 1 40404 ru— m 0 1 1 1 1 -0- L _ 1 1 T L w 1 T L 1 ] __-____-—_--__~ T — POP P r P P OTOT -—<-- T T OTOT _ Por P r P P P TOT — Tor r r r P T TOT — Por P r r T T TOT — FOr P Or P OT OT TOT - - fix I - e 00 e 1 [ , - 6 ] t Do, . 1 |- o -I n ' > 1 fi-b phr—W _- _-F_ 1 1 4 4 _c-_0_~_ L — 1 1 1 1 T T L 1 —m——-———_0- T _ 1 1 1 1 —Q 1 1 1 — 1 —0- 1 L L L 1 —_w_ T 1 1 -0— 1 _ L 1 L L W- 1 L -l Fals 9° ($ y § «> {0 # y _ z 0C 0 L 0C 0 L (WM) HLd3GQ (WM) AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE AND THEIR TECTONIC IMPLICATIONS D35 por -r or or-r -r T oT oT 80 Zl ‘ llllllllllllllllllIIIllllllllIIllllllllllllllllllllllllllllllllllllllllllllllll - - L A L - fa .% - {b ‘E | [nsd 4; " r F . +- - eB. 08. r- - ae o [ "& > o a pak! y- - n r- h - O i l. [ep] 29. y- - % - 5 r- C ®.. I p - I - a r- i [e o] § O ¢ - - ye F O - O a y- - fop} - i O - |- e O L & L + o cel - - . o'- p- *- O : - 0 : oy 'O llll‘rlllI'I1VIllllYlltiilllVlllIIIVIIIIIIIIVIIII + rorf r -r®*rt-t 40 DISTANCE (KM) + 0 30 + Q)» 20 J Pe 10 + Ill‘llllIII|VIIIII3I|I|IIIIYII| 8 lllllllIlllllllIlllllilIIllllljlIIlllllIlllllllIlllLlllllllllllllllllllllllllll 2. - N uu 1 f y bbb pupula | pub bba O 0 y- al 0 e- (WM) HLd3G (WM) HLd3GQ0 20 20 0 Figure 23.-Continued D36 P-axes 891018-911017 0 - 5.99 km depth "I 20' ~I 10' -I 37° J :\\ t I 50' mas \\ \ N X | L I N - . x* \ [ \_ 10 km | { E \ > 1 ud A \\\ * ~I pvc ps hs eb as ala P-axes 891018-911017 10.00 - 13.99 km depth y -r-pyres 20° ———\——\| f « - < 124 4 a J p ap a a I p p pa Pp aa a L Tr frr t 10' hat 37° C A 4 K* I -\ 4 -\ E 50° ~ PGY 10° 122° 50° 40° 30° Figure 24.-Map view projections of P-axes from well-determined mecha- nisms of M ;21.5 earthquakes. Events from October 18, 1989, to October 17, 1991, at (A) 0-6 km depth, (B) 6-10 km depth, (C) 10-14 km depth, AFTERSHOCKS AND POSTSEISMIC EFFECTS P-axes 891018-911017 6.00 - 9.99 km depth :I-vrvI'vv-|---u -r-- a a a a A p p p a J gp g p a d pga | P-axes 891018-911017 14.00 - 20.00 km depth |r q a a a A a p g g A p g a ga Appa | g & I ] ke \N I- A N A k N ~ bas » :\\\\_X\\\\\\ \\\\ x | ] E\ Y, CXX $3. | 4 »\ I‘ \\\ x, x l— f A SS N E A 10 km ~ | & \\|""."||| & a % ~ 1 J ma \ C NL LL CL 22 22 C2 CCC .. 1!..\..|....|x.?‘.p|.... ads ag 1 a g aa Pp gga J L i 9 aa o 122° 50° 40° 30° and (D) 14-20 km depth; and (£) January 1, 1969, to October 17, 1989, 0-20 km depth. (F) a few focal mechanisms of small events near the San Andreas fault, 1969 to 1989. AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE AND THEIR TECTONIC IMPLICATIONS D37 SAN ANDREAS FAULT At its southern end, the aftershock zone merges seamlessly into the activity of the creeping central seg- ment of the San Andreas fault where fault slip occurs as a mix of earthquakes and fault creep (Wesson and others, 1973b). Focal mechanisms of earthquakes on this section of the San Andreas fault are homogeneous and reflect dextral slip on the San Andreas fault. Slip in the main shock, as determined from both dynamic (Beroza, 1991; Steidl and others, 1991; Wald and others, 1991) and static models (Lisowski and others, 1990; Snay and others, 1991; and Marshall and others, 1991), extended for no more than 15 to 20 km to the southeast of the hypocenter. The seismicity rate in the southernmost section decays differ- ently from the rate in the dipping portion of the after- shock zone (Reasenberg and Simpson, this chapter). Thus, activity along the southeastern 25 km of the aftershock zone probably reflects stress transfer to it from slip in the main shock. About 10 km southeast of the main shock epicenter, the main zone begins to steepen in dip and it warps toward a slightly more northerly strike. It continues to steepen until it merges with the San Andreas fault at a point about 8 km northwest of Pajaro Gap (PG in figs. 6, 21, 23). As the zone steepens, it also narrows in width, until it has no measurable width (<0.7 km) after merging with the San Andreas fault. P-axes 690101-891017 20° 10° 37° 50° Figure 24. -Continued The southern terminus of rupture in the main shock coincides with the point where the aftershock zone reaches the San Andreas fault. The end of the rupture may be marked by the clusters of aftershocks located 20 to 21 km southeast of the main shock hypocenter. The largest after- shock, the M=5.4 Chittenden earthquake of April 18, 1990, locates immediately to the southeast of this point at the top of the aftershock zone. This earthquake triggered a major secondary aftershock sequence that extended for 10 km to the southeast along the San Andreas fault (fig. 17). This sequence filled a 7-km-long gap on the fault between 3 and 8 km depth that was visible in the pre-Loma Prieta seismicity (fig. 234). The mechanisms of these events show predominantly right-lateral strike-slip motion on northwest-trending planes. Any dip to these fault planes is believed to be an artifact caused by lateral refractions from unmodeled local velocity changes, as previously stated. CASTRO AND SOUTHERN SARGENT FAULTS Directly to the northeast of the San Andreas fault along the southeastern part of the aftershock zone (region SAR, fig. 20 N, O0), aftershock activity defines a shallow (1-4 km), near-vertical, right-lateral strike-slip fault. This fault also stands out as a prominent feature in the pre-Loma Prieta seismicity (fig. 3; Olson and Hill, 1993). This ac- tivity has been commonly assumed to be on the Sargent Focal Mechanisms 690101-891017 aia b pga b pg f alas / aia hp a a 7 AFTERSHOCKS AND POSTSEISMIC EFFECTS D38 uo (,pST 5 5,51 I) su0aa isnuyj-onbifgo 'souey© s p9-) _ pue (x) Surddip-ms uo (4551 z] aye)) suoaa difs-ayins pers squaaa anbifgo put pug '(umop-jur0d-a3ueLn ',59- 5 2421 3 ,p | [-) $JuoA2 [eun1ou 01 peuonsodoid st azts 'oyes uo spuadap ad/{j joqurdy {(dn-qjurod-afSuemn) ',¢] | 5 5 ,59) sjuaaa jsnuy} '(arenbs [- 3 oyE1 3 [-) $1u349 ay) 01 (.SPF.OET) our;d pepou auo wim suustueyoour [to. _ (g) (arenbs) Surddip-gp pue uado) surddip-Mm$ Sutaey (g1z aas) xoq 7-7 oy} uiym syooyszaye jo suonsas iday, (WM) JNYLSIQ 08 OZ 09 0s Op 0€ OZ (o O --..4-__.-_-.__._-_..--__.--._.<.__..-__..-_.__-._.-—-_.-_____-_____._ _mEhoc>RES:m:c__no_22m_-Em:D 4 isniy; Y - onbijgo jeagre|-yo| + vy m V O C v O y v O V, V V o v y ¥ yv ve Ed O <<¢ v o é —--—_0-————__——0-_-—__-—-_--m—-—-—_0-__—-h-0—-—-—-0---——-0-—--0__—_—-_“—_-— O O O [> By O O 4+ < F: f + # 6 ag ® < D % <4 < Lop aba a a a a acl ogo _--_-__—-___-_-—__--d\--_---._—-__q--—-_--—-_-__--___—-_-_fl-— o O - isruy} onbijgo jesore|-4ybiy + x - dis oyuys jeaore-4ybig IN MS aurejd Jo dig 13 % a%: el bubb apaa f gppabab abba ¥ O Cx x 2 [ WW XW+ % % xA x W ° [ G® H Co XE O o o O _ 1 1 3 4 -0 1 4 4 _ L4 OO O4 J- 1 “T_ 1 L_ 4 -%- 1 _m _L CLOG _0P 1 _ 1 1 _I —Q 1_1_4 h- _0— 140 O4 4 _ 1 w- 11 4 no— 1 _ 140 O4 ¢_ L_ 1 4 i 9° a> y $00 @ $ _ 3 z 0C O y- 0C 0 L (WM) (WM) HLd9G0 AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE AND THEIR TECTONIC IMPLICATIONS D39 fault (for example, Eaton and others, 1970). Our loca- tions, as well as those of Olson and Hill, however, lie distinctly east of and at an acute angle to the southern Sargent fault. They coincide, however, with the Castro fault, first described by Allen (1946). We believe that this seismically defined fault is most probably the subsurface expression of the Castro fault and is not associated with the Sargent fault. A few aftershocks are also scattered between the traces of the Sargent and San Andreas faults. Most of these events also show right-lateral strike-slip on northwest planes, but a few shallow thrust events occur as well. Only 4 of the events in this region are M 423.0, the largest a Myj=3.2 on December 31, 1989. As noted previously, the rate of ac- tivity along the Castro fault underwent a step-like increase at the time of the Loma Prieta earthquake, then declined to the pre-event rate within about one year. FAULTS NORTH OF THE MAIN ZONE AND EAST OF THE SAN ANDREAS The Loma Prieta earthquake triggered activity within a plexus of faults in the footwall block, principally within the southern Santa Cruz Mountains to the east of the San Andreas fault. Aside from the activity along the Castro fault (described above) the aftershock activity in this re- gion occurred on faults that were virtually aseismic dur- ing the two decades before the earthquake (figs. 3, 22, 23). Areas that had been active before the main shock were stilled by the earthquake. The overall pattern of aftershocks in this region reflects release of fault-normal compression, principally on re- verse faults. These events are generally deepest adjacent to the San Andreas fault, shallow to the northeast, and are consistent with movement on the Shannon, Monte Vista, and Berrocal faults (fig. 22). Within this general frame- work, however, the case can be made for reverse faulting on steeply northeast-dipping planes, in addition to the named northwest-dipping faults. Strike-slip events, although rare, are also observed. These occur at shallow depth (1-2 km), and directly un- derlie the northern part of the Sargent fault (fig. 20 F, H). At only slightly greater depth, reverse faulting predomi- nates within the same area. One of the clearest examples of a distinct secondary fault outlined by aftershock activity appears updip of the main shock hypocenter. This activity clusters in a near-vertical plane between 6 and 9 km depth lying be- tween the surface traces of the San Andreas and Sargent faults. This cluster was initiated by a M=4.3 aftershock on October 25, 1989, at 0127 UTC. The focal mechanism implies pure reverse motion (SW side down) on a vertical plane parallel to the San Andreas (event a, fig. 20 G, H; appendix A). The total amount of aftershock activity in the northern- most portion of the foot wall of the Loma Prieta earth- quake was relatively modest, with only four well-located aftershocks M23 (region SCV, fig. 20 A, B). The largest of these, M4=4.7, occurred 3 minutes after the main shock along the northeastern perimeter of the entire sequence, just to the west of the Shannon fault. Because P-wave arrivals for this event were obscured by the main shock coda waves, a reliable focal mechanism could not be de- termined. We strongly suspect, however, that this event was a shallow reverse-faulting event associated with the Shannon fault. Field investigations along the Monte Vista and Shan- non faults revealed systematic evidence of tectonic short- ening caused by the Loma Prieta earthquake (Schmidt and others, 1996). The events along this northernmost exten- sion of the aftershock zone (fig. 20 A, B) are consistent with minor displacement along these faults during the main shock or in the earliest moments of the aftershock se- quence. The prompt occurrence of the My=4.7 event at shallow depth just west of the trace of the Shannon fault supports this interpretation. HANGING WALL ACTIVITY Although the majority of all aftershocks locate south- west of the San Andreas fault, almost all of these lie within the main zone, and comparatively few locate in the hanging-wall block of the earthquake. The aftershock ac- tivity in the hanging wall (region MON, fig. 20 P, Q) concentrate in one large and several smaller clusters. The large cluster contains the My=4.5 event of October 19, 1989, at 1014 UTC (event 2, fig. 20 P, Q; appendix A) and its aftershocks. This sub-vertical cluster trends 143° and is centered about 10 km south-southeast of the main shock. It is about 5 km long and lies between 5 and 8.5 km depth. The focal mechanism indicates right-lateral strike-slip on a 170° trending vertical plane, with the P-axis in fault-normal coordination with the main zone. This sec- ondary fault lies near the projected position of the Zayante fault (R. Jachens and A. Griscom, written commun., 1992), but has focal mechanisms and fault plane strikes that are incompatible with their occurrence on this reverse fault. A second cluster locates 9 km northwest of the first, at a depth between 1 and 4 km. The largest event in this group also has a north-striking, right-slip fault plane solu- tion (event 11, fig. 20 P, Q; appendix A). Ten kilometers further to the northwest, strike-slip faulting in the hanging-wall activity appears again, this time with a north- west strike and reverse component (event I, fig. 20 D, E; appendix A). The fourth minor cluster locates in Monterey Bay, a few km seaward of the shoreline and directly south of the main shock epicenter. The largest event in this cluster (event 2b, fig. 20 P, Q; appendix A) shows right D40 slip on a NSO°W, near-vertical plane. No young faults have been identified at the surface in the vicinity of these other clusters. RELATIONSHIP OF PRESHOCKS TO THE MAIN SHOCK AND AFTERSHOCKS The detailed record of seismic activity afforded by the CALNET during the 20+ years before the Loma Prieta earthquake bears little resemblance to the distribution of aftershocks (fig. 3; Olson and Hill, 1993). In only three peripheral areas does it appear that the aftershocks oc- curred on the same faults displaying activity in the two decades before the shock. These are the San Andreas fault southeast of the main shock, the Castro fault to the south and east of the earthquake, and the zone of reverse fault- ing at the extreme northwestern end of the zone, generally in the area near Stevens Creek Reservior (SC, fig. 6). As noted above, all of these areas continued to be active after the Loma Prieta earthquake and each experienced an ac- celerated rate of earthquake production following the main shock (Reasenberg and Simpson, 1992, and this chapter). As discussed by Olson and Hill (1993), the main zone had very little activity for at least the 20 years leading to the main shock (figs. 3, 22, 23). While the deep activity forming the prominent "U-shaped" band of activity that was subsequently filled by the main shock (U.S. Geologi- cal Survey Staff, 1990) included the sparse main zone activity, it was principally constructed from activity to the northeast of the main zone. These northeastern areas, how- ever, were virtually silent during the aftershock sequence, as can be seen by comparing the pre- and post-earthquake event distributions (fig. 23). In general, the pre-earthquake seismicity reflects the release of a north-south compres- sive stress, as would be expected for the San Andreas fault system (fig. 24 E, F). Several important sequences of M; =4-5 earthquakes preceded the main shock. The two events near Lake Elsman in 1988 and 1989, M; =5.0 and Mr =5.2 respectively, are well-known to be the largest events near the Loma Prieta earthquake in over 20 years (Olson and Hill, 1993; Sykes and Jaumé, 1990). These events ruptured adjacent parts of deep fault at the intersection of the Sargent fault with the San Andreas fault near what was to become the northwest end of the Loma Prieta rupture. As with the other zones noted above, seismicity in the Lake Elsman zone was not activated by the main shock. In addition to the 1988 and 1989 events, several no- table sequences of M=4-5 events also occurred near the main zone in the 1960's. These include the events of No- vember 16, 1964 (M; =5.0), October 14, 1966 (M; =4.0), September 7, 1967 (M; =4.7), December 18, 1967 (M =5.3), and March 21, 1968 (M; =4.3). Although the catalog locations of these events place them in the general AFTERSHOCKS AND POSTSEISMIC EFFECTS region, they lack the precision to draw any firm conclu- sions regarding their association to the Loma Prieta earth- quake. Using the method of joint hypocentral determination (Dewey, 1972), we can accurately locate these events with respect to the modern data. Furthermore, the 1967 and 1968 events were recorded by the CALNET and can be accurately located without resort to JHD. Relocation of these events using both JHD and absolute methods shows that they are significantly more tightly clustered than the catalog locations would suggest (fig. 26, table 5). The 1964 and 1966 events locate quite near the San Andreas fault, and are plausibly on it. The activ- ity in 1967 and 1968, however, form a clear northwest trending lineation, subparallel to and removed 2 km to the northeast of the San Andreas fault at the latitude of the main shock hypocenter. This trend also contains the My =4.5 event of August 19, 1982. A reliable focal mecha- nism was determined for the largest event (December 18, 1967), and it shows right-lateral strike slip on a N50°W-striking fault. We conclude that these events do indeed fall on a single fault, possibly the Sargent fault, but probably not the San Andreas fault. As with most of the other significant concentrations of pre-Loma Prieta seismicity, activity along this zone was curtailed by the main shock. Finally, we address the question of whether or not the San Andreas fault was active adjacent to the Loma Prieta pe- 20 he -- "\ --, \ - *. \\ . \ 2 Fo \\ - hoa \\ he 4 wil \\ ~ -I r © 12 Nov 1973 1 po ~ _s - pj 8 Aug 1989 1 10° |- ~ a o & m CQ l - N ~ & 27 Jun 1988 ~ 7 pi apr 1981 ° ~ s N 1 i ~ 7 Sep 1967 \. 1 [| ~ 18 Dec 1967 _ \\ 7] \. 21 Mar 1968 \ | E hk 18 0ct1989* 18 Aug 1982 .\ 4 a { ° & 14 Oct 1966 37 16 Nov 1964 N \ * ~ < 10 Jan 1974 \ & o 1] 18 Jun 1980 1] 7Jan1981<><> 1] 27 Jan 1981 i] 50° 3 Oct 1972 28 Sep 1981 \ C 23 Sep 1972 i *, 13 Apr 1980" ! ___________ M 1 iL aaa fu 40 30° Figure 26.-Map of M, 24.0 events from 1964 through October 17, 1989; star = main shock, circle = M; 25.0, diamond = M; 24.0. Earthquakes occurring in the 1960's were located relative to a subset of later events (solid symbols, table 5) by a joint hypocentral determination (JHD). In the JHD, events after 1970 were fixed at their locations from this study, but their origin times were free to move. AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE AND THEIR TECTONIC IMPLICATIONS Table 5.-Earthquakes M, 24.0 located by joint hypocentral determination Origin Time Latitude Longitude Depth M, yrmoda hrmn _s N W km 641116 0246 41.94 37° 1.46 121° 45.39" 9.81 5.0 661014 2034 28.65 37° 0.41" 121° 43.46" 5.57 4.2 670907 1239 17.12 37° 3.46 121° 46.00 10.41 4.7 671218 1724 31.83 37° 2.85 121° 45.32 9.98 5.3 680321 2154 59.92 37° 2.20 121° 44.42" 10.69 4.3 800618 0452 26.34 36° 54.16" 121° 38.08 4.69 4.2 810425 1941 37.22 37° 6.67 121° 52.78" 12.04 4.1 820818 0843 49.65 37° 2.02 121° 44.00 10.88 4.5 880627 1843 22.44 37° 7.50 121° 53.83" 12.36 5.3 890808 0813 27.57 37° 8.62 121° 55.76 13.46 5.4 891018 0004 15.42 37° 2.01" 121° 53.08" 15.94 7.0 D41 rupture and main zone of aftershocks between 1969 and the earthquake. A critical examination of the pre-Loma Prieta activity shows that there are a few events in the central part of the zone that lie vertically below the trace of the San Andreas fault (fig. 22 D1; fig. 3A). They sug- gest the possibility of some activity on a vertical San Andreas fault. Although focal mechanisms for these small events are difficult to obtain, a number of them indicate right-lateral slip on planes parallel to the San Andreas fault (fig. 24F). If these events locate on the primary fault that moved in the 1906 earthquake, they represent only very minor subsequent movement. DISCUSSION Perhaps the most intriguing feature of the Loma Prieta aftershock sequence is its almost complete disassociation from the preceding seismicity. Only at the extreme ends of the sequence, well removed in space from the main shock slip surface, do we observe aftershock activity of the same style and at the same location as the preshocks. Nearer the source, the aftershocks activated numerous faults not seen before the earthquake, and their focal mechanisms reflect a first-order change from release of north-south compression expected for the San Andreas system to release of fault-normal compression. Indeed, it appears probable that the main shock effectively released all of the shear stress acting across its fault plane (Michael and others, 1990), leaving a residual stress field in an average state of uniaxial compression (Zoback and Beroza, 1993). Activation of the full aftershock zone proceeded very rapidly. In fact, we detected no significant change in the form of the zone from its earliest observable moments (fig. 10). A high degree of spatial clustering of aftershocks, particularly within the main zone and nearby areas north- east of the San Andreas fault, is a primary feature of the distribution. Many of these clusters represent individual faults and have been studied in detail by Seeber (written commun., 1992). These clusters commonly produced their largest locatable event very early in the sequence and thus may be related to short-wavelength variations in main shock displacements such as at the ends of the rupture. The rapid spatial variations in the style and orientation of aftershock focal mechanisms between the interior clus- ters, such as changes over short distances from fault-normal compression to fault-normal extension (fig. 25), further suggest a causal association with localized stresses cre- ated by the main shock. The general lack of correlation of the aftershock clusters with the main shock slip zones (see Spudich, 1996) thus supports Oppenheimer's asser- tion (1990) that abrupt spatial variations in main shock displacement is required to account for the observed short distance mechanism variability, particularly within the in- terior of the main zone. THE EARTHQUAKE AND THE SAN ANDREAS FAULT Finally, we examine the question of the relationship between the Loma Prieta rupture and the San Andreas fault in the southern Santa Cruz Mountains. The geologic investigations of Prentice and Schwartz (1991) and de- tailed trenching studies by Schwartz and others (1991) leave little doubt that the 1906 earthquake ruptured the main trace of the San Andreas fault adjacent to the 1989 earthquake. Furthermore, both geodetic and strong-motion analyses of the Loma Prieta main shock (Spudich, 1996) require a large component of right-lateral strike-slip mo- tion in the earthquake to the southeast of the main shock D42 epicenter. However, the upper extent of this slip did not reach the surface. We also know that the main zone of aftershocks merges seamlessly with the San Andreas fault near Pajaro Gap, virtually at the southern terminus of the 1906 fault break. If the main zone of aftershocks faithfully maps the main shock rupture, as we hypothesize, then the association between it and the San Andreas fault could have several resolutions. In one, the Loma Prieta fault truncates a ver- tical San Andreas fault at ~9 km depth. Alternatively, Olson (1990) suggests that the San Andreas truncates the Loma Prieta rupture plane. A third possibility is that they are one in the same, with the San Andreas fault following the main zone below 10 km and steepening to vertical above that depth. Neither the aftershocks nor the geodetic models of the earthquake (Lisowski and others, 1990; Snay and others, 1991; Marshall and others, 1991; Horton and others, 1996) imply that main shock faulting continued to the northeast of the San Andreas surface trace. The main zone, like the geodetic models, terminates updip below the trace of the San Andreas fault (sections B2 through F2, fig. 22). Thus, evidence for the San Andreas being cut by the Loma Prieta fault is lacking. The opposite case of the San Andreas truncating the Loma Prieta fault is at least consistent with these data. Extending the San Andreas to depths below the hypothesized truncation point at ~9 km depth, how- ever, has little to recommend it. A deep extension of the fault might be argued for on the basis of a single earth- quake with a right-lateral strike-slip focal mechanism seen before the earthquake (fig. 22 D1, deepest event). Re- garding the third possibility, slip on both the deeper, in- clined part of the fault in the Loma Prieta earthquake and on the vertical part in the 1906 earthquake is kinemati- cally possible only along the southeastern part of the 1989 rupture. North of the epicenter, the 1989 slip vector for the inclined part of the fault (predominately reverse), and the 1906 slip vector for the shallower vertical part (hori- zontal) are incompatible for a single structure and cannot be sustained in ongoing movement. Indeed, a common shortcoming of all three of these possibilities is that they fail to address the kinematics of the crust traversing the prominent left (compressional) bend in the San Andreas fault within the southern Santa Cruz Mountains. Over geologic time, flow of the crust around the bend must be accommodated by either lateral flow, subduction or mountain-building (Anderson, 1990). The simple kinematic model of Dietz and Ellsworth (1990) for the present-day geometry successfully predicts the ratio of horizontal to vertical slip on the Loma Prieta fault (fig. 27). While this model is not unique, it suggests that we are observing a rapidly evolving fault geometry which must continually adjust as new crust enters the compres- sional bend. AFTERSHOCKS AND POSTSEISMIC EFFECTS The smooth connection between the Loma Prieta rup- ture and the San Andreas fault to the southeast of the bend raises the possibility that the Loma Prieta earth- quake followed what was once the main strand of the fault, but which is now in the process of being aban- doned. Along the southern half of the earthquake where motion was predominately strike slip, the strand contin- ues to carry some of the San Andreas motion. Farther to the north, where motion was dominantly reverse, it has been abandoned as a transform fault, but it still released the compressional component of the kinematically pre- scribed slip, with the translational part presumably being accommodated on the San Andreas. Ultimately, the resolution of this problem must await a more complete understanding of the faults within the south- ern Santa Cruz Mountains. The complete absence of seis- micity on the fault that ruptured in this Mg=7.1 earthquake over at least the 20 years preceding the event and the absence of an obvious imprint left by its rupture in the geologic record should remind us that other large active faults, including the San Andreas itself, may be "hidden" within the southern Santa Cruz Mountains and will only be fully revealed when they slip in major earthquakes. ACKNOWLEDGEMENTS We thank L. Jones and R. Page for their helpful re- views of this manuscript. H= u cos0 D = u sind/cosp u = block motion vector 0 = change in fault strike § = fault dip within the bend S = slip vector on inclined plane H = horizontal slip D = dip-slip motion Figure 27.-Simple block model to describe the slip vector (S8) on a dipping fault segment within a bend of a vertical strike-slip fault. This model successfully predicts the ratio of horizontal () to vertical (D) slip observed on the Loma Prieta fault. AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE AND THEIR TECTONIC IMPLICATIONS D43 Table 6.-Event locations and nodal planes from focal mechanisms in figure 20 and appendix A EQ Origin Time Latitude Longitude Depth - M, Nodal Plane 1 Nodal Plane 2 yrmoda hrmn N W km dip-azm dip rake dip-azm dip rake A 891018 4 37° 2.01' _ 121° 53.08 15.94 - 7.1 220 - 70 130 332 44 _ 30 B 28 36° 57.32 121° 43.11' 12.39 - 3.7 200 - 45 160 304 _ 76 47 C 38 37° 9.78' - 121° 59.30 10.68 _ 4.3 230 - 45 110 23 48 _ 71 D 41 37° 10.85 122° 3.67 15.75 5.2 262 _ 81 160 355 70 - 10 E 58 37° 6.23 121° 59.26 17.43 - 3.7 65 _ 81 150 160 60 - 10 F 108 37° 11.54" - 122° 2.69 14.07 3.6 71 _ 70 169 165 80 20 G 116 37° 10.71' _ 121° 59.22 6.88 3.6 271 18 -123 125 75-80 H 145 37° 1.62 121° 47.52 9.60 3.7 255 - 50 130 22 54 - 53 I 208 36° 57.58' - 121° 43.46 12.55 3.5 215 - 80 130 317 41 _ 15 J 226 37° 2.15 - 121° 46.84' 2.06 4.0 190 - 60 100 351 31 - 73 K 302 37° 6.178' - 121° 52.16 1.67 - 3.5 230 - 85 -180 140 _ 90 _ -5 L 321 37° 6.41' _ 121° 50.32 3.62 - 3.8 215 45 100 21 46 _ 80 M 323 37° 7.90 - 121° 59.45 14.40 - 3.9 195 _ 70 -20 292 71 -159 N 330 37° 7.93% - 121° 59.40 14.43 - 4.1 190 - 60 -10 285 81 -150 O 335 37° 6.67 - 121° 52.15 3.39 - 3.5 54 - 40 172 150 85 - 50 P 414 37° 7.64' _ 121° 59.08 14.51 - 3.6 210 - 70 20 113 71 159 Q 416 37° 3.62 121° 53.39 12.48 - 3.7 241 _ 80 5 150 85 170 R 425 37° 2.55 121° 47.25 6.52 3.6 220 - 60 120 351 41 49 S 450 37° 10.09 122° 0.45 9.99 4.2 235 - 40 110 30 53 - 74 T 518 37° 1.34 - 121° 50.99 17.04 - 4.2 190 - 60 120 321 41 49 U 1022 37° 1.87 - 121° 47.64' 9.61 4.3 225 0 85 110 328 21 14 V 891019 - 953 36° 56.64' - 121° 41.08 6.88 - 4.3 67 - 46 -144 310 65 -50 W 891021 49 37° 3.14 - 121° 51.53 12.08 - 4.2 195 - 90 -120 105 30 0 X 2214 37° 3.92 121° 54.01' 13.91 4.7 72 - 80 -15 165 75 -170 Y 891025 1300 36° 54.05 _ 121° 38.43' 4.55 - 3.8 50 - 50 180 140 90 40 Z 891026 - 901 37° 3.54 121° 52.72 11.40 _ 3.7 260 - 75 160 355 71 - 16 a 891102 - 550 37° 4.05 - 121° 48.32 7.89 4.3 35 - 85 100 151 11 _ 27 b 891105 _ 130 37° 4.40 _ 121° 54.76 12.54 - 3.7 98 - 60 6 5 85 150 c 1337 37° 3.75 - 121° 53.46 12.40 - 3.9 236 - 62 23 135 70 150 d 900418 1337 36° 55.22 121° 39.35 4.61 _ 4.3 50 - 50 180 140 90 - 40 e 1341 36° 55.78' - 121° 39.77 4.170 - 4.5 52 - 51 167 150 80 40 f 1353 36° 55.86 - 121° 39.82 4.42 5.4 51 - 50 173 145 85 40 g 1452 36° 55.06 - 121° 39.63 6.88 - 4.2 67 - 46 -144 310 65 -50 h 1528 36° 56.51" - 121° 40.67 6.27 4.2 67 - 44 -150 315 70 -50 i 1536 36° 56.88' - 121° 41.14' 6.87 3.9 67 - 51 167 165 80 40 j 900422 200 36° 54.13' _ 121° 38.32 4.67 3.6 50 - 50 180 140 90 _ 40 k 2124 37° 12.33' - 122° 3.97 12.86 3.7 230 - 35 100 38 56 - 83 1 901005 - 604 37° 4.31' 122° 0.83 11.63 3.6 225 - 80 130 327 41 - 15 m 910324 _ 342 36° 57.76 - 121° 44.00 14.08 _ 4.5 15 - 85 -160 283 70 - -5 n 910919 - 906 36° 54.17 - 121° 38.36 4.176 3.9 48 _ 41 165 150 80 - 50 o 907 36° 54.63 - 121° 38.76 4.36 - 3.9 50 - 50 180 140 90 40 REFERENCES CITED 249, p. 397-401. Arnadottir, T., Segall, P., and Matthews, M., 1992, Resolving the dis- Allen, J.E., 1946, Geology of the San Juan Bautista quadrangle, Cali- crepancy between geodetic and seismic fault models for the 1989 fornia: California Division of Mines and Geology Bulletin 133, Loma Prieta, California, earthquake: Seismological Society of 111 p. America Bulletin, v. 82, p. 2248-2255. Anderson, R.S., 1990, Evolution of the northern Santa Cruz Mountains Beroza, G.C., 1996, Rupture history of the earthquake from high-fre- by advection of crust past a San Andreas fault bend: Science, v. quency strong-motion data, in Spudich, P., The Loma Prieta, Cali- D44 Table 6.-Continued AFTERSHOCKS AND POSTSEISMIC EFFECTS EQ Origin Time Latitude Longitude Depth - M, Nodal Plane 1 Nodal Plane 2 yrmoda hrmn N W km dip-azm dip rake dip-azm dip rake 1 891018 330 37° 16.21 122° 3.90 4.82 2.8 185 - 35 70 29 - 57 103 2 891019 1014 36° 58.30 _ 121° 50.13' 8.07 4.5 260 - 90 155 350 - 65 _ 0 2b 891025 1630 36° 54.50 _ 121° 52.93 1.13 - 2.8 220 - 75 160 315 - 71 16 3 891107 2342 37° 13.87 - 122° 1.54" 9.65 3.9 220 - 45 110 13 - 48 _ 71 4 891202 2002 37° 13.87 122° 1.71' 9.61 3.4 190 - 45 80 24 - 46 100 5 891215 543 36° 59.02 121° 38.71 1.80 2.6 205 - 75 -170 112 - 80 -15 6 891231 648 36° 58.15 - 121° 36.77 3.07 3.1 215 - 70 -170 122 - 81 -20 7 813 36° 58.19 _ 121° 36.92 3.33 3.2 215 - 71 -164 120 _ 75 -20 8 900216 736 37° 13.17 122° 0.17 3.21 2.3 230 - 50 120 8 - 48 59 9 900417 2126 36° 59.31' - 121° 39.56 3.56 2.4 210 - 80 150 306 - 60 12 10 900425 802 36° 57.99' - 121° 39.93' 1.74 - 2.8 220 - 81 -150 125 - 60 -10 11 900529 1343 36° 59.98' _ 121° 55.48" 3.25 2.1 85 - 65 -160 346 - 72 -26 12 900910 1701 36° 52.19 _ 121° 31.99 6.72 2.0 20 - 44 112 170 - 50 70 13 910416 _ 56 37° 19.00 _ 122° 6.43' 4.60 _ 3.1 195 45 80 29 - 46 100 fornia, earthquake of October 17, 1989-Earthquake occurrence: U.S. Geological Survey Professional Paper 1550-A, p. A9-A32. Beroza, G.C., 1991, Near-source modeling of the Loma Prieta earth- quake: evidence for heterogeneous slip and implications for earth- quake hazard: Seismological Society of America Bulletin, v. 81, p. 1603-1621. Beroza, G.C., and Zoback, M.D., 1993, Mechanism diversity of the Loma Prieta aftershock sequence: implications for the mechanics of mainshock-aftershock interaction: Science, v. 259, p. 210-213. Choy, G.L., and Boatwright, J., 1990, Source characteristics of the Loma Prieta, California, earthquake of October 18, 1989 from glo- bal digital seismic data: Geophysical Research Letters, v. 17, p. 1183-1186. Dewey, J.W., 1972, Seismicity and tectonics of western Venezuela: Seismological Society of America Bulletin, v. 62, p. 1711-1751. Dietz, LD., and Ellsworth, W.L., 1990, The October 17, 1989 Loma Prieta, California, eathquake and its aftershocks: Geometry of the sequence from high-resolution locations: Geophysical Research Letters, v. 17, p. 1417-1420. Eaton, J.P., 1992, Determination of amplitude and duration magnitudes and site residuals from short-period seismographs in northern California: Seismological Society of America Bulletin, v. 82, p. 533-579. Eaton, J.P., Lee, W.H.K., and Pakiser, L.C., 1970, Use of mi- croearthquakes in the study of the mechanics of earthquake gen- eration along the San Andreas fault in central California: Tectonophysics, v. 9, p. 259-282. Eberhart-Phillips, D., Labson, V.F., Stanley, W.D., and Michael, A.J., 1990, Preliminary velocity and resistivity models of the Loma Prieta earthquake region: Geophyiscal Research Letters, v. 17, p. 1235-1238. Eberhart-Phillips, D., and Stuart, W.D., 1992, Material heterogeneity simplifies the picture: Loma Prieta: Seismological Society of America Bulletin, v. 82, p. 1964-1968. Fletcher, J.B., and Boatwright, J., 1991, Source parameters of Loma Prieta aftershocks and wave propagation characteristics along the San Francisco peninsula from a joint inversion of digital seismo- grams: Seismological Society of America Bulletin, v. 81, p. 1783-1812. Foxall, W., Michelini, A., and McEvilly, TV., 1993, Earthquake travel time tomography of the southern Santa Cruz Mountains: Control of fault rupture by lithological heteogeneity of the San Andreas fault zone: Journal of Geophysical Research, v. 98, p. 17691-17710. Habermann, RE., and Craig, M.S., 1988, Comparison of Berkeley and Calnet magnitude estimates as a means of evaluating temporal consistency of magnitudes in California: Seismological Society of America Bulletin, v. 78, p. 1255-1267. Horton, Stephen, Anderson, J.G., and Mendez, A.J., 1996, Frequency- domain inversion for the rupture characteristics during the earth- quake, using strong-motion data and geodetic observations, in Spudich, P., ed., The Loma Prieta, California, earthquake of Octo- ber 17, 1989-Earthquake occurrence: U.S. Geological Survey Professional Paper 1550-A, p. A59-A73. Hough, S.E., Seeber, L., Lerner-Lam, A., Armbruster, J.G., and Guo, H., 1991, Empirical Green's function analysis of Loma Prieta af- tershocks: Seismological Society of America Bulletin, v. 81, p. 1737-1753. Johnson, LR., 1974, Green's function for Lamb's problem: Geophysi- cal Journal of the Royal Astronomical Society, v. 37, p. 99-131. Joyner, W.B., and Boore, D.M., 1988, Measurement, characterization, and prediction of strong ground motion: Proceedings of Earth- quake Engineering and Soil Dynamics II, GT Div/ASCE, Park City, Utah, June 27-30, 1988, p. 43-102. Kanamori, H., and Satake, K., 1990, Broadband study of the 1989 Loma Prieta earthquake: Geophysical Research Letters, v. 17, p. 1179-1182. Kissling, E., Ellsworth, WL., Eberhart-Phillips, D., and Kradolfer, U., 1994, Initial reference models in local earthquake tomography: Journal of Geophysical Research, v. 99, p. 19635-19646. Klein, FW., 1989, User's guide to HYPOINVERSE, a program for VAX computers to solve for earthquake locations and magnitudes: U.S. Geological Survey Open-File Report 89-314, 58 p. Kradolfer, U., 1989, Seismische tomographie in der Schweiz mittels lokaler erdbeben: Doktor der _ Naturwissenschaften, Eidgendssischen Technischen Hochschule Ziirich, 109 p. Lees, J.M., 1990, Tomographic P-wave velocity images of the Loma Prieta earthquake asperity: Geophyiscal Research Letters, v. 17, p. 1433-1436. Lisowski, M.J., Prescott, W.H., Savage, J.C., and Johnston, M.J., AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE AND THEIR TECTONIC IMPLICATIONS D45 1990, Geodetic estimate of coseismic slip during the 1989 Loma Prieta, California, earthquake: Geophysical Research Letters, v. 17, p. 1437-1440. Marks, S.M., and Lester, FW., 1980, Catalog of earthquakes along the San Andreas fault system in central California, April-June 1977: U.S. Geological Survey Open-File Report 80-1264, 43 p. Marshall, G.A., Stein, R.S., and Thatcher, W., 1991, Faulting geom- etry and slip from co-seismic elevation changes: The 18 October 1989, Loma Prieta, California, earthquake: Seismological Society of America Bulletin, v. 81, p. 1660-1693. McNally, K.C., Simila, G.W., and Brown, J.G., 1996, Main-shock and very early aftershock activity, in Spudich, P., ed., The Loma Prieta, California, earthquake of October 17, 1989-Earthquake occurrence: U.S. Geological Survey Professional Paper 1550-A, p. A147-A170. Mendoza, C., and Hartzell, S.H., 1988, Aftershock patterns and main shock faulting: Seismological Society of America Bulletin, v. 78, p. 1438-1449. Michael, A.M., Ellsworth, W.L., and Oppenheimer, D.H., 1990, Coseismic stress changes induced by the Loma Prieta, California, earthquake: Geophysical Research Letters, v. 17, p. 1441-1444. Michael, A.J., and Eberhart-Phillips, D., 1991, Relations among fault behavior, subsurface geology, and three-dimensional velocity models: Science, v. 253, p. 651-654. Mooney, W.D., and Colburn, R., 1985, A seismic-refraction profile across the San Andreas, Sargent, and Calaveras faults, West-central California: Seismological Society of America Bulle- tin, v. 75, p. 175-191. Murphy, J.M., Catchings, R.D., Kohler, W.M., Fuis, G.S., and Eberhart-Phillips, D., 1992, Data report for 1991 active-source seismic profiles in the San Francisco Bay area, California: U.S. Geological Survey Open-File Report 92-570, 45 p. Olson, J.A., 1990, Seismicity preceding the 1989 Loma Prieta earth- quake: Geophysical Research Letters, v. 17, p. 1429-1432. Olson, J.A., and Hill, D.P., 1993, Seismicity in the southern Santa Cruz Mountains during the 20-year period before the earthquake, in Johnston, M.J.S., ed., The Loma Prieta, California, earthquake of October 17, 1989-Preseismic observations: U.S. Geological Survey Professional Paper 1550-C, p. C3-C16. Oppenheimer, D.H., 1990, Aftershock slip behavior of the 1989 Loma Prieta, California earthquake: Geophysical Research Letters, v. 17, p. 1199-1202. Prentice, C.S., and Schwartz, D.P., 1991, Re-evaluation of 1906 sur- face faulting, geomorphic expression, and seismic hazard along the San Andreas fault in the southern Santa Cruz Mountains: Seis- mological Society of America Bulletin, v. 81, p. 1424-1479. Pujol, J., 1995, Application of the JHD technique to the Loma Prieta, California, mainshock-aftershock sequence and implications for earthquake location: Seismological Society of America Bulletin, v. 85, p. 129-150. Reasenberg, P.A., and Ellsworth, W.L., 1982, Aftershocks of the Coy- ote Lake, California, earthquake of August 6, 1979; a detailed study: Journal of Geophysical Research, v. 87, p. 10637-10655. Reasenberg, P.A., and Oppenheimer, D., 1985, FPFIT, FPPLOT and FPPAGE: Fortran computer programs for calculating and display- ing earthquake fault-plane solutions: U.S. Geological Survey Open-File Report 85-739, 109 p. Reasenberg, P.A., and Simpson, R.W., 1992, Response of regional seismicity to the static stress change produced by the Loma Prieta earthquake: Science, v. 255, p. 1687-1690. Romanowicz, B., and Lyon-Caen, H., 1990, The Loma Prieta earth- quake of October 18, 1989: results of teleseismic mantle and body wave inversion: Geophysical Research Letters, v. 17, p. 1191-1194. Schmidt, K.M., Langenheim, VE., Ellen, S.D., and Jachens, R.C., 1996, Damage patterns from the 1989 Loma Prieta earthquake along the frontal thrust-fault system, Santa Clara Valley, Califor- nia: implications for strain partitioning along the San Andreas fault: EOS, American Geophysical Union Transactions, v. 77, no. 46, p F742. Schwartz, D.P., Okumura, K., Pantosti, D., Powers, T.J., and Fuentes, J.M., 1991, Paleoseismic constraints on earthquake recurrence, San Andreas fault, Santa Cruz Mountains, California: Geological Society of America, Abstracts with Programs, v. 23, p. A98. Schwartz, S.Y., Orange, D.L., and Anderson, R.S., 1990, Complex fault interactions in a restraining bend on the San Andreas fault, southern Santa Cruz Mountains, California: Geophyiscal Research Letters, v. 17, p. 1207-1210. Simila, G.W., McNally, K.C., Nava, E., Protti-Quesada, M., and Yellin, J., 1990, Evidence of very early aftershock activity along the northwest portion of the 18 October 1989 earthquake rupture zone: Geophysical Research Letters, v. 17, p. 1785-1788. Snay, R.A., Neugebauer, H.C., and Prescott, W.H., 1991, Horizontal deformation associated with the Loma Prieta earthquake: Seismo- logical Society of America Bulletin, v. 81, p. 1647-1659. Spieth, M.A., 1981, Two detailed seismic studies in central California, part 1: earthquake clustering and crustal structure studies of the San Andreas fault near San Juan Bautista: Ph.D. Dissertation, Stanford University, 138 p. Spudich, Paul, 1996, The Loma Prieta, California, earthquake of Octo- ber 17, 1989-Main-shock characteristics: U.S. Geological Survey Professional Paper 1550-A, 297 p. Steid1, J.H., and Archuleta, R.J., 1996, Are rupture models consistent with geodetic measurements?, in Spudich, P., ed., The Loma Prieta, California, earthquake of October 17, 1989-Earthquake occurrence: U.S. Geological Survey Professional Paper 1550-A, p. A195-A207. Steid1, J.H., Archuleta, R.J., and Hartzell, S.H., 1991, Rupture history of the 1989 Loma Prieta, California, earthquake: Seismological Society of America Bulletin, v. 81, p. 1573-1602. Sykes, LR., and Jaumé, S.C., 1990, Seismic activity on neighbouring faults as a long-term precursor to large earthquakes in the San Francisco Bay area: Nature, v. 348, p. 595-599. U.S. Geological Survey Staff, 1990, The Loma Prieta, California earth- quake: an anticipated event: Science, v. 247, p. 286-293. Utsu, T., 1969, Aftershocks and earthquake statistics (I): Journal of the Faculty of Science, Hokkaido University, Japan, Ser. VII, Vol. III, no. 3, p. 129-195. Wald, D.J., Helmberger, D.V., and Heaton, T.H., 1991, Rupture model of the 1989 Loma Prieta earthquake from the inversion of strong-motion and broadband teleseismic data: Seismological So- ciety of America Bulletin, v. 81, p. 1540-1572. 1996, Strong-motion and broadband teleseismic analysis of the earthquake for rupture-process modeling and hazard assessment, in Spudich, P., ed., The Loma Prieta, California, earthquake of October 17, 1989-Earthquake occurrence: U.S. Geological Sur- vey Professional Paper 1550-A, p. A235-A262. Wesson, RL., Roller, J.C., and Lee, W.H.K., 1973a, Time-term analy- sis and geologic interpretation of seismic traveltime data from the Coast Ranges of central California: Seismological Society of America Bulletin, v. 63, p. 1447-1471. Wesson, RL., Burford, R.O., and Ellsworth, W.L., 1973b, Relation- ship between seismicity, fault creep and crustal loading along the central San Andreas fault: Stanford University Publications in Geological Sciences, v. 13, p. 303-321. Zoback, M.D., and Beroza, G.C., 1993, Evidence for near-frictionless faulting in the 1989 (M6.9) Loma Prieta, California, earthquake and its aftershocks: Geology, v. 21, p. 181-185. D46 AFTERSHOCKS AND POSTSEISMIC EFFECTS 891018 0004 891018 0116 891018 0323 891018 0450 891025 1300 Z=15.58 M=5.40 Z= 6.88 M=3.60 Z=14.40 M=3.90 Z= 9.99 M=4.20 Z= 4.55 M=3.85 891018 0028 891018 0145 891018 0330 891018 0518 891026 0901 Z=12.39 M=3.70 Z= 9.60 M=3.75 Z=14.43 M=4.15 Z=17.04 M=4.20 Z=11.40 M=3.70 891018 0038 891018 0208 891018 0335 891018 1022 891102 0550 Z=10.68 M=4.30 Z=12.55 M=3.50 Z= 3.39 M=3.55 Z= 9.61 M=4.30 Z= 7.89 M=4.30 891018 0041 891018 0226 891018 0414 891019 0953 891105 0130 Z=15.75 M=5.20 Z= 2.06 M=4.00 Z=14.51 M=3.60 Z= 6.88 M=4.35 Z=12.54 M=3.70 891018 0058 891018 0302* 891018 0416 891021 0049 891105 1337 Z=17.43 M=3.75 Z= 1.67 M=3.55 Z=12.48 M=3.75 Z=12.08 M=4.20 Z=12.40 M=3.90 891018 0108 891018 0321 891018 0425 891021 2214 900418 1337 Z=14.07 M=3.65 Z= 3.62 M=3.85 Z= 6.52 M=3.65 Z=13.91 M=4.70 Z= 4.61 M=4.30 Appendix A.-Focal mechanisms (lower hemisphere projections) from figure 20 showing compressional (+) and dilatational (open circle) first arrivals. Lettered events locate in regions LPN, LPC, LPS, and CSA while numbered events locate in SAR, SCV, and MON. Event locations and nodal planes are listed in table 6. AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE AND THEIR TECTONIC IMPLICATIONS D47 900418 1341 900822 2124 891019 1014 891231 0813 910416 0056 Z= 4.70 M=4.50 Z=12.86 M=3.70 Z= 8.07 M=4.50 Z= 3.33 M= 3.20 Z= 4.60 M=3.10 900418 1353 901005 0604 891025 1630 900216 0736 Z= 4.42 M=5.40 Z=11.63 M=3.60 Z= 1.13 M=2.80 Z= 3.21 M=2.30 900418 1452 910324 0342 891107 2342 900417 2126 Z= 6.88 M=4.20 Z=14.08 M=4.55 Z= 9.65 M=3.90 Z= 3.56 M=2.40 900418 1528 910919 0906 891202 2002 900425 0802 Z= 6.27 M=4.20 Z= 4.76 M=3.95 Z= 9.61 M=3.40 Z= 1.74 M=2.80 900418 1536 910919 0907 891215 0543 900529 1343 Z= 6.87 M=3.90 Z= 4.36 M=3.95 Z= 1.80 M=2.60 Z= 3.25 M=2.10 900422 0200 891018 0330 891231 0648 900910 1701 Z= 4.67 M=3.60 Z= 4.82 M=2.85 Z= 3.07 M=3.10 Z= 6.72 M=2.00 Appendix A.-Continued THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989; EARTHQUAKE OCCURRENCE AFTERSHOCKS AND POSTSEISMIC EFFECTS RESPONSE OF REGIONAL SEISMICITY TO THE STATIC STRESS CHANGE PRODUCED BY THE LOMA PRIETA EARTHQUAKE By Paul A. Reasenberg and Robert W. Simpson, U.S. Geological Survey CONTENTS Page Abstract D49 Introduction 49 Regional stress changes caused by the Loma Prieta earthquake 50 Regional seismicity changes after the Loma Prieta earthquake 53 Earthquake data 53 Quantifying seismicity rate changes --------------------------- 53 Pattern of seismicity rate Changes ------------------------------ 59 Comparison of regional stress and seismicity patterns ----------- 59 Temporal changes in seismicity caused by the Loma Prieta earthquake 64 DiSCUSS10N ANG CONCIUSIONS coon ono > 67 References cited 70 ABSTRACT The Loma Prieta earthquake perturbed the static stress field over a large area of central California and greatly altered the levels of regional microseismicity. We have calculated the static stress changes produced on central California faults using elastic dislocation models of the Loma Prieta rupture. We used a Coulomb failure function (CFF) to describe the proximity of these faults to failure, and compared the calculated changes in CFF after the earthquake to observed changes in microseismicity rates on these faults. The pattern of stress changes on the major faults in the region that were predicted by our models of the earthquake agree closely with the pattern of changes in the regional seismicity rate observed after the earth- quake. The agreement between stress change and seismic- ity change was detected at epicentral distances of up to 100 km. The agreement is best for models with low as- sumed values of apparent coefficient of friction (0.1 pasealoop yiim sjuawsas 'sayenbyjrea sonpord 01 4joy1p arow are (pai) 44> postarout JIm sjuoudag - t € c . '10j09 Aq pareotput st ¢'q=,M 10} juawsos jng} yora uo (sreq Ut) 449v 'sjuouunseau onap -0a3 wos; paAuap (moj[a4) arnidn1 ewo7 ay} 10J japou (0§6]) .SJoYIJ0 pUe DSMOSTT Suisn pojefnofe> penuoo uf sing; Sunuasaidar sjuow@sas sefn3urpoot uo ayenb -yurea jou ewo7 ay} {q poonpord (4457) uonouny aim|ie} quojno; ut aSueyy-'¢ amsiq 3N78 'a3XVIIH 0 D56 contrast of average seismicity rate between two time in- tervals in a specified area (Matthews and Reasenberg, 1988). We compared the rate r,, in the postseismic inter- val of duration t,, with the rate r, in the preseismic inter- val of duration r,, where r,=n,/t,, rp=n,/t,, and n,, np are the numbers of earthquakes occurring in the respective intervals. The rate change is expressed as n, - E(n B(na! Apstgs tb) = d, var(n,) where var denotes variance and E(n ,)=rpt,, is the value of n, expected under the null hypothesis of stationary ran- [ FULIIIIIIIIIlllILIllllllllllllllllllllllllllllll AFTERSHOCKS AND POSTSEISMIC EFFECTS dom occurrence. The variance was represented by that of a binomial process: var(n,,)=npt,,. We calculated B for fixed t,, and t, in overlapping 10- km-square cells covering an area 140 by 390 km for the preseismic (background) period between 17 October 1979 and 17 October 1989 and postseismic period between 18 October 1989 and 31 May 1991 (fig. 5). Gaussian spatial smoothing with halfwidth 5 km was applied to the data in each cell to minimize artifacts introduced by the square gridding. Positive values of B indicate that the postseismic rate was higher than the background rate; negative values, lower than the background rate. Significance levels for IBI estimated from its asymptotic (Gaussian) distribution are 1.96 (p=0.05) and 2.57 (p=0.01). 123° 122° VIIIIIIIIIIIIT|1||I||[mulnllluInfirlllllllll 121 ° 120 ° Figure 4. -Earthquakes (M21.5) located in the study area 10 years before and 20 months after the Loma Prieta earthquake. (A) Earthquakes occurring between 18 October 1979 and 17 October 1989. Some 21,000 events are represented. (B) Earthquakes occurring between 18 October 1989 and 31 May 1991. Approximately 5,000 events are represented. RESPONSE OF REGIONAL SEISMICITY TO THE STATIC STRESS CHANGE An asymmetry in the definition of B introduces an am- biguity in its interpretation: a relatively low postseismic rate cannot be distinguished from an abnormally high preseismic rate, as would result, for example, from an earthquake swarm or aftershock sequence in the preseismic (background) period. This limitation stems from our use of the empirical background seismicity level as a refer- ence level and presents an inherent difficulty in discern- ing rate changes in a finite sample of a point process. In our measurements of San Francisco Bay region seis- micity, several artifacts corresponding to the aftershocks of moderate earthquakes in the pre-Loma Prieta back- ground period are apparent (fig. 5). These artifacts appear as dark blue patches and correspond to the 1979 Coyote Lake (M=5.9), 1980 Livermore (M=5.9), 1983 Coalinga (M=6.7), 1984 Morgan Hill (M=6.2), 1985 Kettleman Hills |l|IllllllIllllllllllllllllllllilllllullllllll D57 (M=5.5), 1986 Mt. Lewis (M=5.7), 1986 Quien Sabe (M=5.7) and 1988 Alum Rock (M=5.1) (Reasenberg and Ellsworth, 1982, Oppenheimer and others, 1990, Oppenheimer and others, 1988, Eaton, 1990). We tried varying the choice of background period to avoid these artifacts. When the background is taken as 1969-1979, many of the aftershock zones that were blue in figure 5 became red (fig. 6), indicating that these after- shock sequences may not have fully decayed at the time of the Loma Prieta earthquake. New artifacts, including those associated with earthquake sequences near Danville in 1972, Bear Valley in 1972, and on the Busch fault in 1974, were introduced (Oppenheimer and others, 1990, Ellsworth, 1975, Lee and others, 1971). We tried to reduce these artifacts by removing after- shocks from the catalog with a computer algorithm [111111 38 ° |-- - |- 37°- |- 36 °- e - L - |_ 35 ° i e IIIIIIIIIllllllmlIlll1|l|11Tlllfil|HlH’fllllll 123° 122° 121 ° 120 ° Figure 4.-Continued D58 AFTERSHOCKS AND POSTSEISMIC EFFECTS (Reasenberg, 1985). Removal of aftershocks apparently postseismic rate (-3O0 and B>O, £, is the number of fault segments for which ACFF correspond to high values of f; and f,, low values of f, and f; and positive correlation between the stress and seismicity rate change. Signifi- cance levels for x2 are 3.84 (p= 0.05), 6.64 (p=0.01), and 10.83 (p=0.001). An advantage of this formulation of chi- squared is that it makes no assumptions about the func- tional relationship betweeen the stress changes and seismicity changes, except that they be of the same sign to be in agreement. Application of the x> test to all 141 fault segments rep- resented in figure 9 rejects the null hypothesis that B and ACFF(p') are independent (p<0.001 for 0.2: p< 0.01 for 0.121) for the 80 segments between 0 and 100 km from the source; x> just fails to exceed the p=0.05 critical point for the farthest 70 segments located 80 or more km from the earthquake. For comparison of this distance range to that of other earthquakes, we calcu- late that static stress changes of 0.1 bar are produced by M=3 earthquakes at hypocentral distances of approximately 0.4 km, and by M=5 earthquakes at distances of approxi- mately 7 km on favorably oriented planes. Thus, small earthquakes close to the fault segment may have an equally i 15 20 25 10 Seismicity Rate Change (Beta) 5 123 3 4 Change in Coulomb Failure Function (bar) Figure 9.-Seismicity change B and stress change ACFF(p') calculated on 141 model fault segments, assuming p'=0.2. Numbers (refer to fig. 3) indicate segments experiencing the largest changes (I E]. 1>2 and IACFF;120.2 bar). D64 AFTERSHOCKS AND POSTSEISMIC EFFECTS intense, but spatially limited, effect on seismicity rate as do large, distant earthquakes. In a speculative vein, the apparent sensitivity of the central California seismicity to small stress changes sug- gests that, in general, fluctuations in regional seismicity might also reveal regional, aseismic slip events such as earthquake afterslip, slow earthquakes, slip on the ductile portion of vertical faults, and slip on horizontal detach- ment surfaces. The feasibility of monitoring the seismic- ity to infer aseismic strain events obviously would depend on both the sensitivity of the seismic network and the overall level of regional seismicity, and it would certainly be an enormous challenge to future modelers to constrain the location and source parameters of such events solely from seismicity data. There are some suggestions of largely aseismic strain events in the year before Loma Prieta. Galehouse (this volume), in his creep data from the Hayward fault, finds evidence of slowdowns at various sites ranging from 0.3 to 3.6 years before the Loma Prieta earthquake. Gladwin and others (this volume) report the beginning of a shear strain change at a tensor strainmeter near San Juan Bautista in mid-1988. Chi-squared 4 6 8 10 12 0.0 0.2 0.4 0.6 Coefficient of Friction Correlation 0.4 0.5 0.6 0.7 0.0 0.2 0.4 0.6 Coefficient of Friction Figure 10.-Measures of agreement between stress changes and seis- micity changes on fault segments, shown as a function of the appar- ent coefficient of friction p' assumed for the stress calculations. Solid lines, all 141 fault segments; broken lines, fault segments ex- periencing significantly large changes (numbered points in fig. 9). (A) Chi-squared vs. friction. Chi-squared confidence levels are 6.64 (p=0.01) and 10.83 (p=0.001). (B) Correlation vs. friction. TEMPORAL CHANGES IN SEISMICITY CAUSED BY THE LOMA PRIETA EARTHQUAKE Up to this point we have considered the change in re- gional seismicity as a coseismic step change in rate in order to facilitate comparison to our calculated step changes in stress. Having seen that the sense of change in seismicity and stress agree in many areas, we now focus on the time dependence of the seismicity rate in some of these areas. The time-dependent part of the response of seismicity to the regional stress change may help con- strain present and future models of fault behavior and regional deformation. In this section we examine the rate of M>1.5 earth- quakes in several geographical zones as a function of time. We defined the zones to be large enough to contain suffi- cient earthquakes for a statistically significant analysis, yet small enough to resolve areas that appear to have responded coherently to the Loma Prieta earthquake. The T 2 0 g) * 5 $ 1.000) 'Way A 7 o 00] *' 3k“? m U. *e 3 . g ¢ ®. > ”3-5?an m 0.010 ® L "3.0 e*,0 6 O e o ®" ® .....Q 8 0.001 g & s 0 50 100 150 200 250 E Distance from Earthquake (km) 0 50 100 150 200 - 250 Distance from Earthquake (km) fa O 2 yx U 0 25 ke) < - 20 R B § 15}. ® 10 | ° . * 6 5 Pe o o a* .** > o | # o @ 0 _e e o Q2“! Pet sor /v *® & € -5 * _ © & Figure 11.-Absolute levels of stress change and seismicity rate change on individual segments, shown as a function of segment distance from the earthquake epicenter. (A) Calculated stress changes (for assumed value of apparent coefficient of friction p'=0.2). (B) Observed seismicity rate change statistic, B. Signifi- cant correlation between the stress changes and seismicity rate changes was observed at distances up to 100 km. RESPONSE OF REGIONAL SEISMICITY TO THE STATIC STRESS CHANGE D65 zones are defined in terms of subsets of the model fault segments shown in figure 3 and include all M>1.5 earth- quakes within 5 km of any segment in the subset (table 1). There is a small amount of geographical overlap among the zones. For each zone, we calculated the rate of earth- quakes in 30-day, non-overlapping intervals between Janu- ary 1979 and March 1992 (fig. 12). The choice of 30-day intervals for our rate estimates was subjectively made to provide a balance between resolution and accuracy. Seismicity during the preseismic period in all zones appears to occur at more or less constant rates punctuated by bursts of activity. For example, the largest four bursts of seismicity along the creeping section of the San Andreas fault (Zones 2N and 28) before the Loma Prieta earthquake are associated with moderate (M=4.4 to M=4.7) earthquakes near Fremont Peak in April 1980 and near Bear Valley in August and September 1982, May 1986 and July 1988 (fig. 12, table 2). The lower average preseismic rates in the other zones are similarly punctuated by smaller clusters involving smaller earth- quakes. The change in seismicity rate after the Loma Prieta earthquake varied significantly from zone to zone (table 1). For a rough measure of these differences we com- pared, for each zone, the change in seismicity rate be- tween the 10-year preseismic and 20-month postseismic periods. Of course, the seismicity rates varied greatly dur- ing this postseismic period, so our results depend strongly on the choice of intervals, but they serve for comparison. Along the southern portion of the creeping section of the San Andreas fault (zone 28) the relatively high background rate of 0.58 events/day increased 50 percent after the Loma Prieta earthquake, while farther north and closer to the rupture (zone 2N) the rate increased sevenfold. Within the aftershock zone proper (zone 1) seismicity rate in- creased nearly twentyfold, and on the southern San Fran- cisco peninsula (zone 3), seismicity rate increased threefold. We modeled the seismicity in each zone during the background period as a homogeneous (constant rate) Pois- son process. While these models generally are not very well fit (in a least-squares sense) by the seismicity rate data due to the presence of bursts of aftershocks, they provide convenient and well-defined reference levels for comparison to the postseismic period. Seismicity rate im- mediately after the Loma Prieta earthquake in zones 1 to 4 abruptly increased and then decayed in an aftershock like manner. Accordingly, we modeled the seismicity rate for the postseismic period in zones 1 through 4 as a nonhomogeneous Poisson process with rate N(#) obeying the modified Omori's law (Utsu, 1971): K NCC) = (t+e)" ' where t is time after the main shock, and K, c, and p are positive constants. We estimated these model parameters separately for the seismicity in each zone with a maxi- mum likelihood method (Ogata, 1983) (fig. 13). These models generally provided excellent fits to the first 100 days of data but were incapable of modeling the complex- ity in seismicity rate in zones 1 and 2N resulting from the strong secondary aftershock sequence that followed the M=5.6 aftershock near Chittenden on 18 April 1990. For these zones, we fit Omori models for the intervals 18 October 1989 to 17 April 1990 (table 1); in zones 3 and 4, data between 18 October 1989 and 31 March 1992 were used. Our power-law models for the aftershock zone, the southern San Francisco peninsula, and the northern part of the creeping San Andreas fault all have p-values close to the mean value of p previously found for California aftershock sequences of 1.07+0.2 (+10) (Reasenberg and Jones, 1989) (fig. 13, table 1). Our model for the decay of seismicity rate on the San Gregorio fault based on earth- quakes mostly occurring near segments 131 to 134 is not well constrained, owing to the overall low level of the seismicity response there, and the absence of recorded aftershocks during the first day of the sequence. While the estimate for p there (0.62+0.15) is lower than the Cali- fornia average, these data are too sparse to justify inter- preting this model. From the Omori models and observations of the long term background rates we estimated times at which the seismicity rate in each zone will return to the pre-Loma Prieta level. We defined the end of elevated-rate period as the time when the Omori model rate equals the 1979- 1989 background rate. We found that in the aftershock zone (Zone 1) elevated activity will have ended between 1991.3 and 1992.9. The activity on the northern creeping section of the San Andreas fault will have subsided to normal between 1990.4 to 1992.0, and activity on the San Francisco peninsula will have returned to normal between 1990.5 and 1991.3 (table 1). Because the levels of both aftershock activity and background activity vary spatially, our models (and, consequently, these estimates) depend on the particular zones we have used. The zones group together shallow and deep earthquake activity and include multiple fault strands. As a consequence, our results rep- resent subregional averages. In contrast, Dietz and Ellsworth (this chapter) have applied a similar analysis to the seismicity located within approximately 30 km from the main shock epicenter, using MP1 earthquakes in smaller and more complexely defined zones, and find sig- nificant variations in temporal decay of the aftershock activity associated with specific fault structures. The post-Loma Prieta seismicity rate along the Hayward fault (zone 5) is the lowest among the regions we have considered, and the coseismic change there was a rate D66 AFTERSHOCKS AND POSTSEISMIC EFFECTS Table 1.-Summary of seismicity in selected zones (Northern Section) Zone Region Fault Segments Preseismic Postseismic K p c Return to (1) Rate (2) Rate (3) (Events/Day) (day) Preseismic (Events/Day) | (Events/Day) Rate (Years) 75 £7 0.90 + 0.02 | 0.21 £ 0.07 69+ 1.5 (4) 1 Aftershock Zone 37-42, 193-195 0.07 1.3 121 + 20 1.12 + 0.05 | 0.57 £0.16 2.3 + 0.3 (5) 22.1 £ 2.5 0.60 + 0.02 | 0.09 +£0.09 11.2 + 3.8 (4) 2N Creeping - SA _ fault 42-44 0.15 1.15 55.6 + 16 0.95 + 0.08 1.1 £ 0.56 1.4 + 0.8 (5) 28 Creeping - SA _ fault 45-50 0.58 0.88 «-- -- - e (Southern Section) 3 Southern S.F. Peninsula 32-37, 198-199 0.08 0.27 10.3 + 1.7 0.81 £ 0.04 | 0.02 £ 0.04 1.1 + 0.4 4 San Gregorio fault 128-134 0.02 0.08 2.1 £ 1.8 0.62 + 0.15 | 1.83 £7.73 7.4 + 17.4 0.11 0.06 -- -- -- - 5 Hayward fault 90-99 0.12 (6) 0.07 (7) - - - - 3 4 5 6 7 . Earthquakes within 5 km of any of the fault segments were used. . Mean rate in the 10-year interval 17 October 1979 to 17 October 1989. . Mean rate in the 20-month interval 18 October 1989 to 31 March 1992. . Model fit to the interval 18 October 1989 to 31 March 1992. . Model fit to the interval 18 October 1989 to 17 April 1990. . Mean rate in the interval 17 October 1979 to 1 January 1988. . Mean rate in the interval 1 January 1988 to 31 March 1992. decrease, which is inherently more difficult to detect than a rate increase. As a result of these differences, the Hayward fault seismicity rate record is particularly diffi- cult to interpret, and the changes we infer for the Hayward fault are less certain than those for the other zones. After the Loma Prieta earthquake, the seismicity rate on the Hayward fault is about half the rate during the 10-year background period (fig. 12, table 1). In a previous section we interpreted this reduction in rate as support for our stress models. However, the Hayward activity appears to have undergone its most prominent rate decrease around the beginning of 1988, after which time the rate averages 0.07 events/day and does not exceed 0.17 events/day. This decrease, also on the order of 50 percent, coincides roughly with the times of observed decreases in right lateral creep rates between May 1988 and June 1989 at four of five alignment arrays situated along the Hayward fault (Galehouse, 1990, 1992; also, this chapter). While there is considerable uncertainty in the onset times of both our seismicity rate decrease and Galehouse's creep rate re- ductions, their approximate coincidence suggests a con- nection and raises the possibility that a separate episode of relaxation on part of the Hayward fault may have be- gun in this period before the Loma Prieta earthquake. When compared to the relatively low-rate period starting in 1988, the post-Loma Prieta seismicity still shows a decrease, although the size of the decrease is naturally smaller than it is in comparison to the 10-year background period. A possible interpretation of the Hayward fault seis- micity is that two relaxation events may have occurred- one related to the 1988-1989 creep retardation and the other related to the Loma Prieta earthquake. During the period since the Loma Prieta earthquake, the seismicity rate on the Hayward fault has increased slightly (fig. 12). We attempted to model this increase as a linear rate change, suggestive of a constant-rate process of reloading by tectonic forces. We had hoped to be able to estimate the loading rate from the ratio of the average coseismic stress drop calculated for the Hayward fault to the estimated time at which the seismicity rate returns to its preseismic level. This was not possible, however, be- cause of the large uncertainties in the regressed time of RESPONSE OF REGIONAL SEISMICITY TO THE STATIC STRESS CHANGE return to the preseismic rate (1,,,,,,=1994.7+112 years) that result from the low seismicity rates on the Hayward fault. DISCUSSION AND CONCLUSIONS The pattern of coseismic changes in seismicity rate in the San Francisco Bay region are positively correlated with the static stress changes calculated for central Cali- fornia faults at distances up to 100 km from the Loma Prieta rupture. The largest increases in seismicity were near the ends of the earthquake rupture on the San Fran- cisco peninsula to the northwest and along the creeping section of the San Andreas fault to the southeast. The seismicity increase on the San Francscio peninsula is con- D67 sistent with elastic dislocation models for the earthquake, which predict increases in the Coulomb failure function there and support a conclusion of the Working Group (1990) that the probability of a large earthquake on the San Francisco peninsula was increased by the Loma Prieta earthquake. The correlation between the observed seismicity rate changes and modeled stress changes is best for models involving low apparent friction (0.1n & o B 10.004 O x © 3 Z € mofi Co W o CG 0.104 CC 0. 01 T t u t < 0.01 0.10 1.00 10.00 100.00 1000.00 Day Days After Mainshock »ck 1000.00 C San Francisco Peninsula 100.007 >~ © o e w 10.004 O Ald C 3 lex £ 1.00 Co LJ «_- A & S o.10- CC e 0.01 r T t 0.01 0.10 1.00 10.00 100.00 1000.00 Days After Mainshock Figure 13.-Rate of M21.5 earthquakes after the Loma Prieta earth- quake in zones 1, 2N, 3, and 4 (see table 1) is indicated by triangles. The postseismic decay in seismicity rate in each area is described by Omori's law (curved lines). Horizontal lines represent average seismicity rates in the 10 years preceding the Loma Prieta earthquake. Vertical lines mark time of the 18 April 1990 (M=5.6) aftershock near Chittenden. (A) Af- tershock zone (Zone 1). Model fit to data between 18 October 1989 and 17 April 1990 is shown by solid curve; model fit to all of the data is D69 surrounding faults that can be similarly modeled? Using methods similar to the one reported here, one can calcu- late the static stress changes produced on Bay Area faults by all regional earthquakes for which reliable rupture mod- els can be inferred. In principle, such static models can be combined into a single, time-dependent model that may be capable of explaining a significant portion of the re- gional background seismicity. After the stress-model-in- duced seismicity is accounted for, patterns and changes in 1000.00: B Creeping San Andreas fault 1 a | 100. 00-4 > © o -< e O x © 3 len C ked C LJ 2 G CC - 0.104 0.01 t t T T 0.01 0.10 1.00 10. 00 100.00 1000.00 Days After Mainshock 1000.00 j D San Gregorio fault 100. 00-7 @a © o _ 10.007 ® © Ps & 3 IZ 1.00 T CG LJ O a 0.107 CC 0.01 : - t - 0.01 0.10 1.00 10.00 100.00 1000.00 Days After Mainshock shown by dashed curve. (B) Northern creeping segments of the San Andreas fault immediately southeast of the Loma Prieta rupture (Zone 2N). Model fit to data between 18 October 1989 and 17 April 1990 is shown by solid curve; model fit to all of the data is shown by dashed curve. (C) Southern San Francisco Peninsula (Zone 3). Model reflects data between 18 October 1989 and 31 March 1992. (D) San Gregorio fault (Zone 4). Model reflects data between 18 October 1989 and 31 March 1992. D70 AFTERSHOCKS AND POSTSEISMIC EFFECTS the residual seismicity may provide information on other possible sources of stress, including aseismic slip on or below Bay Area faults. Concerning the apparent low friction of the faults: Do all faults in the Bay Area have a low effective coefficient of friction, or is this true of only some of the faults? Our result, which relies on earthquakes along all Bay Area faults, does not distinguish among individual faults. Is the low apparent friction on Bay Area faults suggested by our study controlled by the presence of pore fluids in the seismogenic crust? If so, on which faults, and over what characteristic time, will the effective friction change as fluids re-equilibrate in response to the postseismic stresses? Continued close monitoring of the southern Hayward fault seismicity and creep rates may help answer this question. In particular, an increase in the microseismicity and creep rates on the southern Hayward fault might indicate that pore pressure has returned (increased) to the pre-Loma Prieta level and may signal the possible occurrence of a delayed, triggered large earthquake on the Hayward fault. REFERENCES CITED Armbruster, J.G., and Seeber, L., 1991, A mechanically modeled seis- micity precursor of the Loma Prieta, California earthquake: Eos, Transactions, American Geophysical Union, v. 72, p. 310. Behr, J., Bilham, R., Bodin, P., Burford, R.O., and Burgmann, R., 1990, Aseismic slip on the San Andreas fault south of Loma Prieta: Geophysical Research Letters, v. 17, p. 1445-1448. Blanpied, ML., Locker, D.A., and Byerlee, J.D., 1992, An earthquake mechanism based on rapid sealing of faults: Nature, v. 358, p. 574-576. Byerlee, J.D., 1978, Friction of rock: Pure and Applied Geophysics, v. 116, p. 615-626. Byerlee, J., 1990, Friction, overpressure and fault normal compression: Geophysical Research Letters, v. 7, p. 2109-2112. Byerlee, J.D., 1992, The change in orientation of subsidiary shears near faults containing high pore fluid pressure: Tectonophysics, v. 211, p. 295-303. Dieterich, J.H., 1981, Constitutive properties of faults with simulated gouge, in Carter, N., Friedman, M., Logan, J., D. Stearns, eds., Mechanical behavior of crustal rocks: American Geophysical Union, Washington, D.C., p. 103-120. Eaton, J.P., 1990, The earthquake and its aftershocks from May 2 through September 30, 1983, in Rymer and Ellsworth, eds., The Coalinga, California, earthquake of May 2, 1983: U.S. Geological Survey Professional Paper 1487, p. 113-170. Ellsworth, W.L., 1975, Bear Valley, California, earthquake sequence of February-March, 1972; Bulletin of the Seismological Society of America, v. 65, p. 483-506. Ellsworth, W.L., 1990, Earthquake history, 1769-1989, in Wallace, RE., ed., The San Andreas fault system: U.S. Geological Survey Professional Paper 1515, p. 152-187. Erickson, L., 1986, User's manual for DIS3D: A three-dimensional dislocation program with applications to faulting in the Earth: Stanford University, M.S. thesis, Stanford, Calif., 167 p. Galehouse, J.S., 1990, Effect of the Loma Prieta earthquake on surface slip along the Calaveras fault in the Hollister area: Geophysical Research Letters, v. 17, p. 1219-1222. Galehouse, J.S., 1992, Creep rates and creep characteristics on certain North and East Bay faults: 1979-1992; Programs and Abstracts: Second Conference on Earthquake Hazards in the Eastern San Francisco Bay Area, March 25-29, 1992, California State Univer- sity, Hayward, p. 24. Habermann, RE., and Craig, M.S., 1988, Comparison of Berkeley and Calnet magnitude estimates as a means of evaluating temporal consistency of magnitudes in California: Bulletin of the Seismo- logical Society of America, v. 78, p. 1255-1267. Harris, R.A., and Simpson, R.W., 1992, Changes in static stress on southern California faults after the 1992 Landers earthquake: Na- ture, v. 360, p. 251-254. Heaton, TH., 1982, Tidal triggering of earthquakes: Bulletin of the Seismological Society of America, v. 72, p. 2181-2200. Hudnut, K.W., Seeber, L., and Pacheco, J., 1989, Cross-fault triggering in the November 1987 Superstition Hills earthquake sequence, southern California: Geophysical Research Letters, v. 16, p. 199- 202. Jaeger, J.C., and Cook, N.G.W., 1979, Fundamentals of rock mechan- ics [3d ed.]: London, Chapman and Hall, 593 p. Jaumé, S.C, and Sykes, LR., 1992, Changes in moderate seismicity in the San Francisco Bay region and stress changes due to large and great earthquakes: Eos, Transactions, American Geophysical Union, v. 73, p. 197. Lachenbruch, A.H., and McGarr, A., 1990, Stress and heat flow, in Wallace, RE., ed., The San Andreas Fault system: U.S. Geologi- cal Survey Professional Paper 1515, p. 261-277. Lee, W.H.K., Eaton, J.P., Brabb, E.E., 1971, The earthquake sequence near Danville, California: Bulletin of the Seismological Society of America, v. 61, p. 1771-1794. Lienkaemper, J.J., Borchardt, G., Lisowski, M., 1991, Historic creep rate and potential for seismic slip along the Hayward fault, Cali- fornia: Journal of Geophysical Research, v. 96, p. 18,261-18,283. Lienkaemper, J.J., Borchardt, G., Wilmesher, J.F., Meier, D., 1989, Holocene slip rate along the Hayward fault, northern California (abstract): Eos, Transactions of the American Geophysical Union, v. 70, p. 1349. Lisowski, M., Prescott, W.H., Savage, J.C., and Johnston, M.J., 1990, Geodetic estimate of coseismic slip during the 1989 Loma Prieta, California, Earthquake: Geophysical Research Letters, v. 17, p. 1437-1440. Louderback, G.D., 1947, Central California earthquakes of the 1830's: Bulletin of the Seismological Society of America, v. 37, p. 33-74. Matthews and Reasenberg, 1988, Statistical methods for investigating quiescence and other temporal seismicity patterns: Pageoph, v. 126, p. 357-372. Mavko, G.M., Schulz, S., and Brown, B.D., 1985, Effects of the 1983 Coalinga, California, earthquake on creep along the San Andreas fault: Bulletin of the Seismological Society of America, v. 75, p. 475-489. McNutt, Marcia, and Heaton, TH., 1981, An evaluation of the seis- mic-window theory for earthquake prediction: California Geology, v. 34, p. 12-16. Michael, A.J., 1991, Spatial variations in stress within the 1987 Whittier Narrows, California, aftershock sequence: Journal of Geophysical Research, v. 96, p. 6303-6319. Michael, A.J., Ellsworth, W.L., and Oppenheimer, D.O., 1990, Coseismic stress changes induced by the 1989 Loma Prieta, Cali- fornia earthquake: Geophysical Research Letters, v. 17, p. 1441- 1444. Ogata, Y., 1983, Estimation of the parameters in the modified Omori formula for aftershock frequencies by the maximum likelihood procedure: Journal Phys.Earth, v. 31, p. 115-124. Okada, 1992, Internal deformation due to shear and tensile faults in a RESPONSE OF REGIONAL SEISMICITY TO THE STATIC STRESS CHANGE D71 half-space: Bulletin of the Seismological Society of America, v. 82, p. 1018-1040. Oppenheimer, D.H., Reasenberg, P.A., and Simpson, R.W., 1988, Fault-plane solutions for the 1984 Morgan Hill earthquake se- quence: Evidence for state of stress on the Calaveras fault: Journal of Geophysical Research, v. 93, p. 9007-9026. Oppenheimer, D.H., Bakun, W.H., and Lindh, A.G., 1990, Slip parti- tioning of the Calaveras fault, California, and prospects for future earthquakes: Journal of Geophysical Research, v. 95, p. 8483- 8498. Poley, C.M., Lindh, A.G., Bakun, W.H., and Schulz, S.S., 1987, Tem- poral changes in microseismicity and creep near Parkfield, Cali- fornia: Nature, v. 327, p. 134-137. Reasenberg, P.A., 1985, Second-order moment of central California seismicity, 1969-1982: Journal of Geophysical Research, v. 90, p. 5479-5495. Reasenberg, P.A.. and Ellsworth, W.L., 1982, Aftershocks of the Coy- ote Lake, California, earthquake of August 6, 1979; a detailed study: Journal of Geophysical Research, v. 87, p. 10,637-10,655. Reasenberg, P.A., and Jones, L.M., 1989, Earthquake hazard after a main shock in California: Science, v. 243, p. 1173-1176. Reasenberg, P.A., and Simpson, R.W., 1992, Response of regional seismicity to the static stress change produced by the Loma Prieta earthquake: Science, v. 255, p. 1687-1690. Rice, J.R., 1992, Fault stress states, pore pressure distributions, and the weakness of the San Andreas fault, in Evans, B., and Wong, T.-F., eds., Fault mechanics and transport properties of rock: a festschrift in honor of W.F. Brace: London, Academic Press p. 475-503. Rice, J.R., and Cleary, M., 1976, Some basic stress diffusion solutions for fluid-saturated elastic porous media with compressible constitu- ents: Review of Geophysics and Space Physics, v. 14, p. 227-241. Richter, C.F., 1958, Elementary Seismology: San Francisco, W.H. Freeman, 768 p. Roeloffs, E.A., 1988, Fault stability changes induced beneath a reser- voir with cyclic variations in water level: Journal of Geophysical Research, v. 93, p. 2107-2124. Rydelek, P.A., and Sacks, IS., 1990, Asthenospheric viscosity and stress diffusion: a mechanism to explain correlated earthquakes and surface deformations in NE Japan: Geophysical Journal Inter- national, v. 100, p. 39-58. Sachs, L., 1982, Applied Statistics: New York, Springer-Verlag, 706 p. Seeber, L., and Armbruster, J.G., 1990, Fault kinematics in the 1989 Loma Prieta rupture area during 20 years before that event: Geo- physical Research Letters, v. 17, p. 1425-1428. Simpson, R.W., and Reasenberg, P.A., 1994, The Loma Prieta, Cali- fornia, earthquake of October 17, 1989, Tectonic processes and models: U.S. Geological Survey Professional Paper 1550-F, p. 55- 89. Simpson, R.W., Schulz, $.S., Dietz, L.D., and Brown, R.O., 1988, The response of creeping parts of the San Andreas fault to earthquakes on nearby faults: two examples: Pageoph, v. 126, p. 665-685. Stacey, FD., 1977, Physics of the Earth [2d ed.]: New York, Wiley, p. 131. Stein, R.S., King, G.C.P., and Lin, J., 1992, Change in failure stress on the southern San Andreas fault system caused by the 1992 magni- tude=7.4 Landers earthquake, Nature: v. 258, p. 1328-1332. Stein, R.S., and Lisowski, M., 1983, The 1979 Homestead Valley earthquake sequence, California: Control of aftershocks and postseicmic deformation: Journal of Geophysical Research, v. 88, p. 6477-6490. Sykes LR., and Jaumé, S.C., 1990, Seismic activity on neighbouring faults as a long-term precursor to large earthquakes in the San Francisco Bay area: Nature, v. 348, p. 595-599. 1992, Changes in state of stress on thesouthern San Andreas fault resulting from the California earthquake sequence of April to June 1992; Science, v. 258, p. 1325-1328. U.S. Geological Survey Staff, 1990, The Loma Prieta, California, earthquake: an anticipated event: Science, v. 247, p. 286-293. Utsu, T., 1969, Aftershocks and earthquake statistics (I): Journal of the Faculty of Science, Hokkaido University, Japan, Series VII, v.3, p. 144-157. 1971, Aftershocks and earthquake statistics (III): Journal of the Faculty of Science, Hokkaido University, Japan, Series VII, v.3, p. 380-441. Working Group on California Earthquake Probabilities, 1990, Prob- abilities of large earthquakes in the San Francisco Bay Region, California: U.S. Geological Survey Circular 1053, 51 p. Zoback, M.D., Zoback, ML., Mount, V.S., Suppe, J., Eaton, J.P., Healy, J.H., Oppenheimer, D., Reasenberg, P., Jones, L., Raleigh, C.B., Wong, I.G., Scotti, O., and Wentworth, C., 1987, New evi- dence on the state of stress of the San Andreas fault system: Sci- ence, v. 238, p. 1105-1111. THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989: EARTHQUAKE OCCURRENCE AFTERSHOCKS AND POSTSEISMIC EFFECTS SPATIAL VARIATIONS IN STRESS FROM THE FIRST SIX WEEKS OF AFTERSHOCKS OF THE LOMA PRIETA EARTHQUAKE By John W. Gephart, Cornell University CONTENTS Page Abstract D73 Introduction 73 Method 74 Test 75 Results 77 Southeast of the main shock hypocenter ---------------------- 77 Northwest of the main shock hypocenter --------------------- 80 Discussion 81 Conclusions 88 Acknowledgments 89 References cited 89 ABSTRACT The spatial variation in the state of stress along the main shock rupture plane during the first six weeks after the Loma Prieta earthquake is evaluated by estimating the stress tensor from groups of aftershock focal mechanisms in each of six subregions. Among the best-fitting stresses, two systematic patterns of stress variation are observed: (1) northwest of the main shock hypocenter the inferred stresses are only moderately well constrained but every- where indicate low shear stress on the main shock fault plane, with G; essentially perpendicular to the plane in most places. The directions of G, and G, appear to vary systematically with position on the fault plane. The mecha- nisms are generally not consistent with north-south com- pression and east-west tension, as might be predicted from strike-slip motion on the San Andreas fault if the fault is assumed to be a locus of high shear stress; (2) southeast of the hypocenter the inferred stresses are well constrained and suggest that significant shear stress may have remained on the main shock fault plane after it failed, although the absolute magnitude of the shear traction cannot be deter- mined (and may be very low if the fault is weak, as com- monly assumed of the adjacent creeping section of the San Andreas fault). The best-fitting principal stress direc- tions are coincident with the main shock kinematic axes in the region close to the hypocenter but deviate signifi- cantly from these with distance to the southeast. The di- rection of shear stress on the main shock fault plane var- ies from: (1) alignment with the overall main shock slip vector near the hypocenter to (2) shallow northwest-plung- ing at the southeast end of the aftershock zone; this varia- tion in shear stress direction (counter-clockwise to the southeast when viewed in the direction of the downward fault pole) is of the same sense as the variation in main shock slip distribution determined by others, but of much smaller magnitude. The mechanisms southeast of the main shock hypocenter generally are not consistent with fault- normal compression, as proposed by some other workers and as might occur if the main shock slip relieved most of the ambient stress in the region. The pattern of stresses observed in this study apparently reflects the distribution of load on the main fault plane shortly after failure. It is possible that the two regions (northwest and southeast of the hypocenter) sample different spatial domains around the main shock fault plane: the diverse events to the north- west reflect the crust adjacent the fault zone, while the well-aligned events to the southeast reflect the fault zone itself. If this is so, then the observations above may sup- port the model of Rice (1992), which predicts spatial het- erogeneity in stress around a well-developed fault zone owing to differences in pore pressure and material proper- ties between the fault zone and the adjoining crust. INTRODUCTION Aftershocks occurring in the first six weeks following the Loma Prieta earthquake exhibit diverse mechanisms which vary widely with location around the main shock fault plane (Oppenheimer, 1990; Dietz and Ellsworth, 1990). Whereas each event indicates a partial measure of the state of stress at the time and place of its occurrence, the well-constrained focal mechanisms of these data can be used to map out variations in stress in the period im- mediately following the M=7.1 earthquake. Preliminary investigations of this kind, based on qualitative or ap- proximate quantitative means, infer strongly heterogeneous D73 D74 postseismic stresses, issuing from the observation of extreme local diversity of mechanisms, which appear to be mutually inconsistent (Oppenheimer, 1990; Michael and others, 1990). Moreover, these studies suggest that the aftershocks may reflect a condition in which the ambient tectonic stress was relieved almost entirely during main shock slip (Michael and others, 1990; Zoback and Beroza, 1993; and Beroza and Zoback, 1993). This study searches for possible spatial heterogeneity of the stress field indi- cated by 164 well-constrained focal mechanisms of early aftershocks (M=~0.9-4.0) from around the rupture zone (fig. 1), applying a technique that may resolve subtle stress variations. It appears that while a portion of the main shock fault plane is largely free of shear stress at this time, another part of the fault plane remains loaded, al- though the absolute magnitude of the deviatoric stress can- not be determined. This work may improve our understanding of the way in which stress is redistributed during large earthquakes. METHOD In this study the stress field is inferred from the direc- tions of slip on fault planes, as indicated by the focal mechanisms of earthquakes, following the procedure of Gephart and Forsyth (1984) and Gephart (1990a, 1990b) and based largely on the earlier principles of Wallace (1951), Bott (1959), and McKenzie (1969). The informa- tion embodied by focal mechanisms-the orientations of fault planes and associated slip directions-constrains four of the six numbers of the stress tensor. As discussed by Gephart (19902), these four dimensionless quantities can be expressed as three Euler angles-fixing the principal stress orientations, and one stress magnitude term, R- describing the magnitude of G, relative to G; and O;; R=(0,-6,)/(0,;-0,), such that 0< R<1. The two numbers of the stress tensor that cannot be determined by this ap- proach reflect the magnitudes of a regional normal and shear stress, omE(0'1+G3)/2 and ‘cmE(0'1—c53)/2; as these numbers have the dimensions of stress, they cannot be estimated from (dimensionless) orientation data alone. By combining the information expressed in multiple focal mechanism observations, which are assumed to reflect a common stress field, best-fitting values of the four resolv- able stress parameters can be determined. In these experi- ments all the data are grouped in spatially distinct subregions, and independent inversions are performed on each set. The variations in the best-fitting stress fields determined for all the subregions should approximate the continuous variation in stress across the region. The constraint on the four resolvable stress parameters is derived from the requirements that (1) there be no com- ponent of shear stress on any fault plane perpendicular to AFTERSHOCKS AND POSTSEISMIC EFFECTS the direction of slip and (2) the component of shear stress parallel to the direction of slip have the same polarity as the slip. Thus, if s is the (unit) slip vector, b is the (unit) B-axis vector (b=nxs, where n is the unit fault normal), 08 [Uuon09$s w ssoq 40 30 Map View Cross Section d focal mechanism (projected onto planes of section). 0 km C, I, J, K). Left, Index map. Right, Map view and two orthogonal cross sections on a grid about the -20 121° 10° California 122° 30° main shock epicenter, showing the distribution of the data relative to the hypocenter (M) an Figure 1.-Location of data and definition of subareas (events labelled A, B, 38° 00° 36° 00° SPATIAL VARIATIONS IN STRESS FROM THE FIRST SIX WEEKS OF AFTERSHOCKS D75 and T is the shear stress vector in the fault plane, then it is required that teb=0 and tes>0. The inverse procedure ef- fects a transformation of the stress tensor from these fault coordinates (unique to each datum) to regional coordi- nates (common to all data). Following Gephart (19902), this is performed based on the conventional stress trans- formation equation, cijl=BikBj16kl (applying the Einstein summation convention of summing over repeated sub- scripts), where O'ij' and G,, are the stress tensor in the two different reference frames, and Bij is the transformation matrix of angle cosines between the two coordinates sets. It is important to note the extreme non-linearity of tensor mechanics that is bound up in the stress transformation equation; this implies that the shear stress direction on a fault plane may vary widely with small variations in the plane or stress tensor, and that there are always many stress tensors that are consistent with any given fault da- tum based on the two conditions above. As a practical matter, this precludes the possibility of substituting aver- aged kinematic (P, B, and T) axes for principal stress directions. That is, the maximum and minimum stress (0) and G,) directions may not be at large oblique angles (~45°) to the fault plane, and the intermediate stress (0;,) direction may not be at a small angle to the plane, as is conventionally assumed for simplicity. Thus, as only the direction of shear stress on the fault plane is indicated by the focal mechanisms, and no measure of the magnitude of shear stress, the approach implicitly allows that any of the fault planes may be either strong or weak. Regarding the extreme non-linearity of the problem, a simple and reliable approach for inverting the fault data is by a grid search over the four resolvable stress parameters (Gephart and Forsyth, 1984). A range of combinations of four stress parameters are tested one-by-one against all of the data. For each set of four parameters (a stress model), the misfit of each datum is determined. Here the misfit is defined as the smallest difference in orientation between the observed mechanism and any one that fits the stress model-in which the slip direction matches the predicted shear stress direction on at least one of the two nodal planes. The difference in orientation between any such observed and predicted mechanisms is equivalent to a ro- tation in space about some arbitrary axis. The fitness of a particular model relative to all the data is given by the average misfit rotation for the whole data set. The opti- mum stress model is the one which has the smallest aver- age misfit, and confidence limits can be defined based on the values of the mean misfits using statistics for one- norm procedures (Gephart and Forsyth, 1984). In the course of the inversion, it is necessary to select the fault plane from among the two nodal planes of each focal mechanism. In the absence of an objective means for iden- tifying the true fault plane based on observation, this is done by testing each nodal plane independently against each stress model and selecting the one with the smaller rotation misfit as the preferred fault plane relative to that stress model (the result for the alternate plane is neglected). To regard the physical properties of particular fault planes (for example, "strong" versus "weak"), it is useful to consider the stress vector that acts on the fault plane. The relation between a stress tensor and a fault geometry can be succinctly illustrated using the Mohr Sphere dia- gram presented by Gephart (1990a) (fig. 2). This is a generalization of the conventional Mohr Circle diagram for stress (for example, Jaeger and Cook, 1979), which is a plot of normal-versus-shear stress. The Mohr Sphere diagram differs from the Mohr Circle diagram in that it regards two orthogonal components of shear stress, paral- lel and perpendicular to the observed slip direction, rather than a single component. In both the Mohr Circle and Sphere figures, admissible fault data are represented as points which indicate the magnitudes of normal and shear stresses acting on a fault plane. Whereas the locus of all possible fault orientations in the Mohr Circle diagram is the area between the one large and two small circles which intersect at three crossing points on the normal stress axis (marking the principal stress magnitudes), the locus of all possible fault geometries in the Mohr Sphere diagram is the volume between the corresponding one large and two small spheres. (That no admissible fault geometries are represented by points in the regions within the two smaller circles/spheres or outside the largest one is an intrinsic property of the stress tensor.) Projections of the Mohr Sphere onto planes parallel to the normal stress axis re- semble the Mohr Circle diagram. Two orthogonal projec- tions of the Mohr Sphere, each plotting normal stress versus one component of shear stress, are sufficient to illustrate the distribution of fault geometries in this three- dimensional figure. Each fault geometry appears as a single point in each of the two projections. For a particular fault geometry to be consistent with the stress tensor in ques- tion, based on matching shear stress and slip directions, the corresponding point must plot above the abscissa in the upper figure (tes>0) and on the abscissa in the lower one (teb = 0), indicating that the shear stress and slip vectors are aligned and of the same polarity. TEST To test for spatial variations in stress, the focal mecha- nisms of 164 well-constrained Loma Prieta aftershocks were divided among several subsets based on location (fig. 1). To the extent possible, divisions between the sub- regions were chosen to correspond with distinct spatial gaps among the aftershocks. Some experimentation was applied to establish appropriate domain sizes and bound- aries, such that meaningful stress variations could be D76 AFTERSHOCKS AND POSTSEISMIC EFFECTS determined; however, the separation was defined solely by event location-no consideration was given to the ori- entations of the mechanisms (and none were rejected as inconsistent, in spite of remarkable local heterogeneity). Three subregions-A, B, and C-were defined along the strike of the main shock rupture zone in the region below and to the southeast of the main shock auxiliary plane (see fig. 1). Subregion A essentially coincides with the northwestern end the central creeping section of the San Andreas fault (U. S. Geological Survey Staff, 1990). The aftershock zone appears to steepen, and its maximum depth shallows, with increased distance to the southeast from a point near the main shock hypocenter; thus the main shock fault plane may vary in orientation and vertical dimension from subregions C to A. Three subregions-I, J, and K- were defined above and to the northwest of the main shock auxiliary plane. Subregion I comprises a group of events immediately northwest of the main shock hypocenter at Mohr Sphere Two orthogonal projections admissible orientations Figure 2.-Mohr Sphere construction: a plot of normal versus two components of shear stress (here parallel to the slip vector, s, and B axis, b, on any fault plane). Fault geometries falling on the [tes>0, Teb=0] half-plane relative to any given stress model are admissible, in that shear stress and slip vectors agree in both direction and polarity. SPATIAL VARIATIONS IN STRESS FROM THE FIRST SIX WEEKS OF AFTERSHOCKS D77 depths >10 km, subregion J comprises a cluster of rela- tively deep events farther to the northwest, and subregion K comprises all shallow events. The orientations of aftershock focal mechanisms are illustrated by the P and T axes at the left side of figure 3. The P and T axes of the main shock mechanism are shown for reference on each plot; however, the main shock da- tum was not included in the inversions of the aftershock data sets. These figures indicate generally modest varia- tion in the character of mechanisms in adjacent subareas, except across the main shock hypocenter where the varia- tion is marked. As noted by Oppenheimer (1990), many of the events southeast of the main shock hypocenter (sub- areas A, B, and C) are aligned along a steeply-dipping plane and have mechanisms indicating right-lateral strike- slip on a northwest-trending fault, consistent with slip on the San Andreas fault. In contrast, aftershocks to the north- west of the main shock hypocenter (subareas I, J, and K) are distributed widely around the main shock fault plane with mechanisms that vary widely and generally do not resemble that of the main shock, suggesting that these events occurred on a variety of secondary structures (Oppenheimer, 1990). Independent tests were performed on each of these data subsets in order to observe variations in the stress field across the region. For each data set, an exhaustive search over the ranges of four stress parameters was made to find the model that is most nearly consistent with the data, based on matching observed and predicted slip di- rections on one of the two nodal planes of each focal mechanism. RESULTS The results of the experiments for each of the six sub- regions are illustrated in the right side of figure 3, which shows the best-fitting stress models, as indicated by the orientations of G; and G, axes on a stereonet, a histogram of R values, and the associated uncertainties of each. The main shock focal mechanism is superimposed on each stereonet for reference. The principal stress directions and value of R of the best-fitting models of each subregion are listed in table 1. Figures 4 and 5 illustrate the variations of the best-fitting stresses among the subregions and em- phasize the relation between the inferred stresses and the main shock mechanism. The misfit rotations and the ob- served and predicted (nearest perfectly fitting) focal mecha- nisms of each event relative to the best-fitting stress model in each region are given in table 2; this listing indicates the chosen (preferred) fault plane from among the two nodal planes of each mechanism. Based on these results, there appears to be a marked difference in the nature of the stresses on either side of the main shock hypocenter. Thus, in the following discussion the results of the three southeastern and three northwestern subsets are presented separately. SOUTHEAST OF THE MAIN SHOCK HYPOCENTER To the southeast of the main shock, the best-fitting stresses fit the data very well, with average rotation mis- fits of <5°; this suggests that the stresses may be reason- ably uniform within each group, as assumed in performing the analysis. Although the aftershock mechanisms of each of the three subsets appear to be qualitatively similar to one another, based on P- and T-axis orientations, the ranges of acceptable stresses vary systematically from one subre- gion to the next with distance along the main shock fault trace (fig. 3). In subarea C, near the main shock hypo- center, G, is nearly horizontal north-south and G, plunges moderately to the east. In subarea A, at some distance from the main shock hypocenter, 6, plunges moderately to the north and G;, is nearly horizontal east-west. In sub- area B, located between subareas C and A, the principal stresses are intermediate in orientation to those on either side. Although the formal uncertainties in stress orienta- tions are relatively small, the results do not absolutely require that the stresses are smoothly varying from subar- eas C to A, as there is some overlap at the 95-percent confidence level; however, as the variation in the best- fitting stresses are systematic, the following discussion assumes that this observed trend is meaningful. In all three cases, the acceptable G; directions (within the 95-percent confidence limit) are nearly confined to the main shock dilatational quadrants, and the G, directions to the com- pressional quadrants. The preferred value of R varies among the subregions from 0.3 to 0.7, but shows no clear systematic spatial trend, as with the stress directions. The orientations of the best-fitting stresses across the aftershock region appear to be systematically related to the focal mechanism of the main shock, as illustrated in figure 4. In subarea C (adjacent to the main shock hypo- center) the inferred G;, G,, and G, directions are very nearly aligned with the P, B, and T axes, respectively, of the main shock focal mechanism. Progressively to the southeast of the main shock hypocenter, in subareas B and A, the G, axis remains parallel to the main shock fault plane, but of increasing rake from a northwest strike line. Concomitant variations in the the other two principal stresses are also systematic in space: G, steepens progres- sively northward, and G, shallows progressively eastward, from subareas C to B to A. Based on these inferred stress models, the predicted slip (shear stress) direction on the main shock fault plane varies from alignment with the observed main shock slip direction in subarea C, to shal- low southeast-plunging (with a small reverse dip-slip AFTERSHOCKS AND POSTSEISMIC EFFECTS D78 Jo} j01d yoea uo umoys st wstueysour feoo; yo0ys urew ay, 443u soddn ay) e umoys st siystur uonr01 agesoAt Jo apmugeuw Surpuodsanod ay) juaorad-gg pug juaorad-g9 oy} Jo sogue1 pur '(..%0,,) JopOW 1saq ay} areotput Surpeys Jo saoiSaq 'sonfea y Jo ypoys ureu & - $% syooysioye o - m to=:o J wss () «oe to@ - wee © u FOB %0 ' yi to =p m to@ vee ©) «.se 'ol = xo @ -> to =o e pue suonsanp to pue 'o Jo e 4q pareotput 'sopour afqerdaoote ay} umoys are 14311 ay} uo '(suotsJoAut ay} ut pasn jou sem tnjep yooys ureu-Ajuo adoualojoi JoJ) yooys ureur ay} Jo asoy} JIM Suoje 'stustueyoou [290] yo0YsI3}JE aU) JO Soxe J, pUE q oJ UMOYS are JJJ ay} uo 'asea yoea Uj '(Y 'f 'I 'D 'g 'v) suot8aigns xis Jo yora 10} pure ejegqg-'¢ amnsiy yoous -& - $% syooysiaye o - ® to=°0 m 'o=°o to= ip m nofi $3. «t's 'of] yo @ .or to=to y 'o =to wss ) .02 to mw %89 ‘ 20 'of _ wo @ - SPATIAL VARIATIONS IN STRESS FROM THE FIRST SIX WEEKS OF AFTERSHOCKS D79 Table 1.-Best-fitting stress models of each subarea [Principal stress directions indicated by plunge and azimuth] Subarea Principal Stress Directions R 91 02 03 A 27 - 2 63 187 2 - 93 0.7 B 9 _ 10 64 261 24 104 0.3 C 4 190 47 284 43 96 0.6 I 40 298 29 180 36 66 0.8 J 22 36 13 301 64 183 0.4 K 9 - 37 56 140 33 301 0.4 component) in subarea B, to shallow northwest-plunging (with a small normal dip-slip component) in subarea A. It can be shown that if the main shock fault plane in subar- eas B and A is steeper than shown in figure 4 (but with a slip vector of the same rake), as suggested by the align- ment of aftershocks in that region, then the mismatch be- tween the observed and predicted main shock slip is increased in that region; thus we cannot easily associate the along-strike differences among the inferred stresses with variations in main shock fault orientation. Relative to the best-fitting stress tensor in each subre- gion, the traction vector on each inferred fault plane is illustrated in the Mohr Sphere diagrams of figure 5 (again, aftershock fault planes were selected from among the two nodal planes of each event in the course of the inversion, so as to minimize the difference between observation and model). That the observed and predicted slip directions are nearly in agreement on most of the aftershock faults is indicated by a clustering of poles on the [tes>O, teb=0] half-plane. For most of the aftershocks in all three subre- gions, the magnitude of shear stress approaches the maxi- mum for the region (near the top of the largest circle in the plot of tes versus G); this suggests that the aftershock fault planes may be relatively strong, as they apparently fail only under relatively high shear stress. Of course as described above, the absolute shear stress magnitude is unknown, as the magnitude of t,,, (the radius of the Mohr Sphere) is unconstrained in the present analysis. Also shown in figure 5 is the traction on the main shock fault plane for the best-fitting stress models of each of the subregions. In all three southeastern subregions the total relative magnitude of shear stress on the main shock is high (nearly the regional maximum), but the magnitude of the components parallel and perpendicular to the observed slip direction vary considerably with location. The main shock fault geometry is consistent with the inferred stresses in subregion C, immediately to the southeast of the main shock hypocenter, but it is progressively less consistent with distance to the southeast in regions B and A, as indicated by the increasing divergence of the main shock datum from the abscissa, [teb=0], in the lower figures. This is consistent with a spatially varying inferred shear stress direction on the main shock fault plane, as shown in figure 4. It is interesting that, while in all three south- eastern subregions the main shock P and T axes fall within the scatter of those of the aftershocks (fig. 3), apparently the main shock and aftershock stress tensors are consis- tent only in subregion C; that is, in terms of the inferred stress tensors, the main shock mechanism is clearly dis- tinct from the aftershocks in subregions B and (especially) A, as indicated in the Mohr Sphere diagrams. NW of mainshock hypocenter Subareas J K I "ie J (R= 0.4) %s K (R=0.4) % I1 (R=0.8) SE of mainshock hypocenter Subareas C B A m C (R=0.6) =, B (R=0.3) *~ A (R=0.7) ; Figure 4.-Summary of best-fitting stress models for each subregion and comparison with main shock focal mechanism. Top, Three north- western subregions: principal stress axes (0;, 9;,, ;) tend to align paral- lel and perpendicular with the main shock fault plane. In all cases the resolved shear stress on the main shock fault plane is very small relative to the regional maximum. Bottom, Three southeastern subregions: prin- cipal stress directions in subregion C are aligned with main shock kine- matic axes, but drift away in subregions B and A; in all cases the G;, axis is nearly parallel to the main shock fault plane. The predicted shear stress direction on the main shock fault plane varies smoothly between subareas C, B, and A, as shown. D80 AFTERSHOCKS AND POSTSEISMIC EFFECTS NORTHWEST OF THE MAIN SHOCK HYPOCENTER To the northwest of the main shock hypocenter the av- erage rotation misfits among the best-fitting stress models in each subarea are on the order of 6-9° (fig. 3)-greater than to the southeast but less than the uncertainty in the focal mechanism determinations; thus, the analysis does not preclude the possibility that stresses are essentially uniform within each subregion. Moreover, the stresses appear to be less uniform between subareas I, J, and K than among the subareas to the southeast. e main shock As shown in figure 4, the stresses northwest of the main shock hypocenter appear to bear a very different relationship to the main shock fault plane than to the south- east; here the principal stress directions are nearly parallel and perpendicular to the fault. In subarea I, near the main shock hypocenter, the stresses are especially poorly re- solved, although the best-fitting G; and G;, directions are nearly aligned with the B and T axes of the main shock, respectively. In an experiment similar to the present one (using a different algorithm to invert for stress), Michael and others (1990) were unable to identify acceptable homogeneous stresses in the region of subarea I, and con- « aftershocks Figure 5.-Two orthogonal projections of the Mohr Sphere construction for the six subregions, corresponding to the best-fitting stresses in each case. Fault geometries that are consistent with the stress model fall on the abscissa in the lower projection and above the abscissa in the top one. Both aftershock (small open circles) and main shock (large filled circles) mechanisms are shown. SPATIAL VARIATIONS IN STRESS FROM THE FIRST SIX WEEKS OF AFTERSHOCKS D81 cluded that the stresses there were extremely heteroge- neous. In subareas J and K the inferred stress orientations ap- pear to be similar at high confidence levels (for example, 95-percent), with acceptable G; axes clustering about the pole of the fault plane and G;, axes distributed in orienta- tions parallel to the fault plane; among the best-fitting stresses, the G; directions of the two subareas are nearly identical but the G;, directions are nearly orthogonal to one another. This is illustrated more fully in figure 6, which shows the results of tests based on only a very limited set of stress models-those with G; perpendicular to the main shock fault plane. Among these models, fig- ure 6A shows the range of acceptable G;, directions paral- lel to the main shock fault plane, in the same format as in figure 3; the best G, directions are oriented differently along the fault plane for subareas J versus K. This de- scription involves some inherent ambiguity, as each G;, direction is associated with several models, spanning a range of values of R (and each value of R is associated with a range of G;, directions). Figure 6B is an unambigu- ous illustration of the full range of models with G; per- pendicular to the main shock fault plane, with each point indicating a single model (a unique combination of stress directions and R). At the 95-percent confidence limit, there is nearly a complementary distribution of acceptable mod- els for the two subregions-generally, models acceptable for J are not acceptable for K, and vice versa. Thus, while certain elements of the ranges of acceptable stresses are common to the two regions (G, direction), other elements appear to be mutually exclusive (the combination of 0, direction and R). In the Mohr Sphere plots in figure 5, there is consider- able scatter of the aftershock data about the [tes>0, teb=0] half-plane among subareas I, J, and K, indicating that the fit to the optimum stress models are only moderately good. In the [tes versus G] projections, most aftershocks plot at relatively low shear stress magnitudes (at considerable distances from the top of the largest circle), suggesting that the aftershock fault planes may be weak. In all three of these subregions, the main shock fault geometry plots very near to a point of intersection of two of the spheres, indicating that the fault plane is nearly a principal plane- and thus experiences a very low relative shear stress mag- nitude at the time of the aftershocks. In such cases, the direction of shear stress on the fault plane is of little sig- nificance (and so is not shown at the top of figure 4). DISCUSSION The above results indicate a systematic spatial variation in stress along the main shock rupture zone, with mark- edly different expressions on the northwest and southeast sides of the main shock hypocenter. To the northwest of the hypocenter, principal stress axes exhibit large differ- ences in orientations between subregions (an exchange of axes among the principal stresses) but in a manner that everywhere yields relatively low shear stress on the main shock fault plane. To the southeast of the hypocenter, principal stress axes exhibit only minor differences be- tween subregions (small deflections of axes) but in all areas so as to yield relatively high shear stress on the main shock fault plane, and with a shear stress vector of smoothly varying rake with position along the fault. This distinct spatial and geometric association of the main shock and stress inferred from aftershocks implies a physical connection between the main shock failure and the postseismic state of stress. Presumably, during the main shock the ambient (preseismic) stress was reduced in magnitude and redis- tributed, although perhaps not completely or evenly; the stress determined from the early aftershocks must reflect the stress state that exists following the main shock stress drop. The marked difference in stress observed on either side of the hypocenter reflects the way that stress was reorganized on the rupture plane. To the northwest, the main shock fault plane appears to be mostly relieved of stress, suggesting that the stress drop was nearly complete (as suggested by Zoback and Beroza, 1993, and Beroza and Zoback, 1993), while to the southeast the fault plane appears to remain loaded, possibly owing to incomplete stress drop on that portion of the fault (in apparent con- flict with the interpretation of these workers). It is interesting that the pattern of the shear stress direc- tion on the main shock fault plane inferred between sub- areas C and A (fig. 4) bears some resemblance to the interpreted variations in the main shock slip distribution from geodetic-leveling and strong-motion data. The latter, with minor differences among various workers (for ex- ample, Beroza, 1991; Hartzell and others, 1991; Marshall and others, 1991; Steidl and others, 1991), indicate pre- dominant thrust slip to the northwest of the main shock hypocenter, and predominant right-lateral strike slip to the southeast (although Wald and others, 1991, find a more uniform slip rake). The inferred postseismic stresses from this study indicate a similar counter-clockwise rota- tion of predicted slip (shear stress) on the main shock fault plane with distance to the southeast, though of much smaller magnitude than observed in coseismic slip, and limited to a region only southeast of the hypocenter. As- suming that the slip distribution reflects shear stress di- rections (as implicit in the stress inversion procedure), it is possible that a large spatial variation in preseismic stress, which is reflected in the two disparate domains of main shock slip, evolved into a more subdued and localized variation in postseismic stress observed among the after- shocks. AFTERSHOCKS AND POSTSEISMIC EFFECTS D82 U4 - C 6L 8'LS l'EFP 965 I U - 0°08 0 09 O' IF L'I9T £05 (€1T) E6 9 I U 8 S8 L'LT9 L'6p U O'S8 0 09 T OS CT FSI P IF 0'L (€TH) 8'I SI T - €tp1 L'pL Clt 7 - O'0S O' StI 8 pL 8 IF 9° TS 6°97TC (€°81) 6°0 ll To CISO 9 PPI g'pL __ 7 - O0'SS _ LEL - TBE - stots (0'st) T9 E1 To _ 6'Spl LTB _ 80s 7 009 _ O'Stl P'Is _ 00s €€€ 0 (ToD 6'I TI 7 SSS 8 LPT 9 pL 6 9¢ 7 9° IS 0° SL 0 0s 6'9¢ (TET) L's L1 T7 - O'6¢ 9°9€ I €'18 6'°8€ T7 - O'St O 'Of I 6°T8 6'TF 6'9¢ L'pEE (STH T9 OI 7 6°08 L'8#1 0 66 L'9t N UMA O° SEI 9' IS L'TE 0 6¢ 8'rOT (TTE) 891 6 ¥ 809 0 __ 008 _ ¥ Ssto9 _ LOST _ 0'08 _ 0'ss o'og _ 6'06E _ (O'FT) £50 8 7T 9°09 9 6p I €18 L'pS 7 - 0'09 0 OS I P I8 0° SS p Og T (0O'€1) L0 L U 8 68 ESET 8 tt 'St U - 0 06 0° SS 0° St O' SFI s'0s 'f (TST) 9 U 988 O'LEZT 9°8t 8 StI ¥ - 006 O° SEZ 00s O' SFI OSF T II (8°OID 6°T S AM 8° $9 6'81T L'T9 S°SII L _ 0'59 O OTT I'€9 L'PS L'TS (TIU S' a 7 - 009 OOF I 0 06 00s TI - 0'09 O 'Of I 0 06 0 0s L'TS L'OpE (POI) 0'0 € U $79 8'ELT S'LL T LLI U L'l9 S'ELT 0° SL O° SLI L'O€ S'IIT (P61) S'e CT U P 6L p OS 8 PC 0° 9SI U - 0'08 0°SS SIE L191 O' LS SCEE (9°St) T9 I 'q eoxueqnsg U 9° L8 'STT 0'Stf P LLE U - O°S8 O'OET €C0Op 6'STE CT EF T 601 (0'8E) 9'L 61 TI - TPS L'I91 S58 6° L9 TI - O'0OS 0 091 P TS $69 I'SE T'ITI (TTD L9 8 I U LOL 9 O° SL O'SEI U LOL 966 O° SL O° SEI 6°8T L'TOI (PS) 00 LT U 6'°6L ELS LTE ¥ - O0°08 0°sS S'I€ L'I9T 9°8S O'6fI (T'PT) PT 9 I 7T 6°S59 9°9€ I €FP8 Ott 7 L'OL 8'9EI 0°S$8 O' St I'8 9 PCI (601) ¥P'¢ SI U4 - £89 £595 PSS I'Z91 U O0 $9 0 0s P FPS S'6SI 9° TS 0 98 (S61) 89 pI 7 1°08 T STI S19 LOTT 7 6°08 L'OET 0 $9 O° STT O rOT (8:98) T9 € L - O'0S O'°SEI S'L9 P SPT L _ O'0§ O° SEI S'L9 PSPC TLL O' TSI (POT) 0'0 TI 7T 8 ° 95 L TPI €T8 L'LET TI - 0'09 O' SFI P I8 O Opi T Of L'TET (PITT E'¢ IT U L:O8 8 IPC L'69 €'0OpE U _ 9°08 S'IPC 0 OL O'OrE T St p S91 (8°E¥P) F#'0 OI JL 1'S$9 S°8TI 6°€S €°8EC L 0 'OE I P PS S'6ET S'L9 9 6T (9'6T) P' 6 U SP8 SPECT £095 T 8TE U O'S8 O' OET 1©09 6'TTE 6°0S P FEI (0'SE) T'9 8 7 L'S9 O' SFI T I8 O'IS 7 0 $9 O' SFI 6°08 LOS $6 P 611 (9°L) _ 80 L U P S8 8°SS 9° TSI U 0°S$8 0 0s p Og 9 °8¥FI €°5S 8'9EI (8'97) L9 9 U S18 L'p$S O'Opf I U 8° I8 CT PF 0°sS O 'Of I 9°E€ 6° 68 (60) S'0o S L L'69 8'9TI 9° L9 9 STT L _ 0'$9 O°SII I'€9 L'S1T SCS 6'O0LI (091) O'TI t U $98 P TIT O' EFP T IIC U - O°S8 O' STT €'0p 6'0TE¢ €'0p 9901 (660) Cpr € L CTBt 8°SS 1©96 L P'8t 8 STC 0 09 0° sOT I'ES (TOT) L'8 T L _ 099 LIZ 9°€S S'EO1T JL 8° €°E1CT 0°sS 0° sOT Slt 891 I (PET) 6'1 I 'y eoIeqng da _ oxins dig _ - oxins - {dig 9yI1§ - (11V) Jol vmzw 'xny ney neq vazm 'xny ney ney sxy sxy # 10¥ __ 10¥X _ (30% 'sq0 _ Ccsq0 _ '$g0 _ '$90 10X -_ JOY juaag wstueyoa]Af [2904 wstueyoa]Af [2504 paaAIasq() 10199 \ UOnEJOY [saaisap ut [[V] baingns yova fo ssaus Sun1f-152aq 10f pup aqeL D83 SPATIAL VARIATIONS IN STRESS FROM THE FIRST SIX WEEKS OF AFTERSHOCKS - m m Q bom Z q Q m Q md md Z mA 0 beopé md be fe pd 9 Q Mod mom a md Bom m md md mr bob rm m 0 y 6° 68 6' It PS61 T - 0'06 0° S8T O' Of 0 S61 T 6 0 861 (66) 0°C I 'J coxreqng 8 T9 L897 6 p9 8 p91 Uo L897 0 $9 0° S91 I'SI 6'°EEZT (8°S1) £50 8C T 99 SCOTT 6'6TI U - 0 S9 O° OTT 0 06 0 °OET 0'€ I©88 (€10) 8'I LT p 96 P IPT 0 09 6'6TI L_ T 96 8 TPT 0 09 0 OEI O' EF 8'9¢ (69) r'0 9T $69 P P8T I'tt P'St L __ 0'€9 EP8T O' SF P 8T P 6TT (FID O' SC (ArAXA 8° S8 U - TOs 8'°STC 0 °$8 0 OTE 8°0 I'E€97 (ETT) 891 PC O' L8 L'9TT CT pL S'LTE U4 - 0 06 0 0s 0 08 Uras 8°ST L'O67 (8°L1) ECL €T P LL 8 9T I TLE T - 0'08 O° SEI O'It L'9€T L'ES P I61 (SLU 0'6 TC $'09 9 PS 6°7T9 L'LOE U - 0 09 0° SS €'p9 6'°80¢€ P Of I'SO€ (LT) 6 IT P P8 ESTE T EL O' LS T - 0'06 O° SSI 0 09 0 's9 €:E9T (6°SE) 9'LI OCT 9°€9 O TPZ P 88 T ISI U - 009 O' Opi 0 06 0 8 91 T b9T (9°LT) TF 61 6°9L 6°08 L'S8TE U - 0O'08 L'6ZT 0° $8 0° O7TE I'81I T I8T (PIT) 9°S 8 I LLE L'88T LSL P 6L1I U - CEE l'p8T O° SL 0 OLI €'LS T LST (T'6T) 9%6 LT €'p9 L'S9 1'9Ss 6°91¢ L_ 0O'$9 0° $9 P PS S'SIE 6'6¢ 'EIE (58) _ ET 9 I 0 6L 9°OSI 6°88 t 09 T - 0'08 0° OSI 0 06 0 09 T SI I'E8T (TST) S' SI 9 65 L'6TT P' 6L U - CTOs (AL 0°S8 O° SEI (Ml 0 '9L (0'ST) 6'01 tI 8°€8 L'I8T I'L8 TLT T - O'S8 0 O8 I 0 08 6'OLT I©8 8°08T (TLT CL € I L'9t 9 PLT 6°8L U - LOS E€'8LT 0 08 0 O8 I L'Os 6'TST g'9 TI 6°T8 O TII O'EF €'p1 U - 6°7T8 TTH O' St O SI 1'6 CPC (P910 ITT LT P S9 LTL P EIT TIT - 0'$9 OSTI 6' TIL L'ETZT P LLE (TPT) 5:0 OI €TL O'LST P Lt 6' 6p I U - SFL CT EST 0 0s 0 °OSI 9°E€ TIZ (66) T's 6 €6L SPIT 0 0s 9°E1T TIT - 0'08 O SII LOS €°E1IT L'8T (€) _ CTU 8 9°08 T IFI $08 9 '6f TI - 7T 08 8'ItP1 0 08 0 0s I'EE€ EPLT (€'8T) 80 L T 69 9°pPIT 888 I'SO€E U - O'0L O' SIT 0 06 0 so€ L'T L'T6 (LTT) S' 9 ELE 9° PEC P S9 9° LOT L_ CT IEZ 0 $9 0° sOT €TL P' LE (610) TC s 9° £9 T 6EI 9° €9 SEPT L_ 0O'$9 O 'Of I I'€9 L'EPT L'O€ T TEI (PE) _ 9'I 12 LOB 8°8IT T OIE U - T 08 T8BIT 0 08 0 OIE 6p I €18T (T6T) E' € 9 TI L'9S L'T8 9° I8T «U - O'OI 0 0s 0°S$8 P 067 0 69 P F6T (TSE) c'6 CT 6°7T9 TC ITE £559 8 L_ €1TE 0 $9 0 $9 €C0Op 6'°SST (69) - r'o I i_) PoIeqng 9 98 T 91 L'S8 $901 U - 0 °OT 0 08 6'OTI P SC SSZLT (PL0 TL TC 0'Of L't 9°0S 9961 N - O'Of 0s LOS 0 861 1©'09 T OLT _ T' IE 9° L9 €6TT 9° 19 TTEE U - O'0L O' 0 79 CT IPE 8°SL €°SST (9°E€€) 9°01 0 € $98 6 LS L'St €'IS1 U - 006 0 $9 0 'Of O° SSI 9 Lt S6 (8°ET) 66 6T T 95 CT PHI 9°T8 CT lt T - 0'09 O Op I P I8 0'St 8°SC (P61 8p 8T 8 9p CT PSI U - 008 0° SF S'I€ LIST P P 6I¢ (TLD ET LT 9°8S CT lL¢ L'ISI U - O°S8 0°sS T Os T 6pI 8°St 6°S (66) E's 9T $89 9° T9 I 6°EL 0 99 T - LOL t 091 0° SL 0° 8°CC (S°81) TC ST 0 LS T 6€I £598 8 T _ 1:09 6' LET 0'St ETE L'8EE (TLD 6'€ PC CTPF I'SE S'SLT N _ 0 'Of £086 T 991 6° 95 L'ETE (8°T1 EL €CT p 09 T S61 O I8 0 001 U _ 1:09 6°T61 0° $8 0 001 00 0 °O8T (LS) _ TCT L'pL T 191 l'pPE I'SLT L_ O'SL 0 091 CT EE U'pPLT €'1S sCOLT (TPC) E1 IT O' 6f 9° 9€ I £18 6°8C T - O Of I 6°T8 6 TF 6 9p l'pEE (STH) T9 OT SOL P OLT 9 PL SPLIT U4 - LOL P OLT 0° SL O SLT S8 I P SOT (8°81) P0 61 t 68 8°IET 0 6€ LIPI U - 0 06 0 0s O'Of O 'Of I I'8$ (TID ITT 8 I PS9 O'OfI L'T8 (AATA T - 0'S9 0 Of I 6°08 EPEC S's (5T - P'0 LT ponunuop-7 apgey, \ a m e a a - o s a O 00 Nl - \p nl O w co to ~om som co s m e < oO & o Cal AFTERSHOCKS AND POSTSEISMIC EFFECTS t -am on O t D- OWD -- ON Z tod m Z m Z Z mm q q Em Z - F m rM zZ q q q que E Erm fr d Ere - E d d M E -E o on \ co e co 0, teb=0]. A weak plane of this orientation may be at risk of failure under the stress at subarea A, as it experi- ences relatively high shear stress and low normal stress, and thus may exceed conventional failure envelopes. SPATIAL VARIATIONS IN STRESS FROM THE FIRST SIX WEEKS OF AFTERSHOCKS D89 ACKNOWLEDGMENTS I am grateful to David Oppenheimer and others at the U. S. Geological Survey for compiling and making avail- able the focal mechanism data used in this study and for helpful discussions. David, Andy Michael, and Paul Reasenberg provided thoughtful reviews. This work was supported by USGS Grant 14-08-0001-G1841. INSTOC Contribution 193. REFERENCES CITED Atkinson, BK., 1987, Introduction to fracture mechanics and its geo- physical applications, in Atkinson (ed.), B.K., Fracture mechanics of rock: London, Academic Press, p. 1-26. Beroza, G.C., 1991, Near-source modeling of the Loma Prieta earth- quake: Evidence for heterogeneous slip and implications for earth- quake harzard: Bulletin of the Seismological Society of America, v. 81, no. 5, p. 1603-1621. Beroza, G.C., and Zoback, M.D., 1993, Mechanism diversity in the Loma Prieta aftershocks and the mechanics of main shock-after- shock interaction: Science, v. 259, p. 210-213. Bott, M.H.P., 1959, The mechanics of oblique slip faulting: Geological Magazine, v. 96, p. 109-117. Dietz, LD., and Ellsworth, W.L., 1990, The October 17, 1989, Loma Prieta, California, earthquake and its aftershocks: Geometry of the sequence from high-resolution locations: Geophysical Research Letters, v. 17, no. 9, p. 1417-1420. Gephart, J.W., 1990a, Stress and the direction of slip on fault planes, Tectonics, v. 9, p. 845-858. 1990b, FMSI: A FORTRAN program for inverting fault/slick- enside and earthquake focal mechanism data to obtain the regional stress tensor, Computers & Geosciences, v. 16, p. 953-989. Gephart, J.W., and Forsyth, D.W., 1984, An improved method for de- termining the regional stress tensor using earthquake focal mecha- nism data: Application to the San Fernando earthquake sequence, Journal of Geophysical Research, v. 89, p. 9305-9320. Hartzell, S.H., Stewart, G.S., and Mendoza, C., 1991, Comparixon of L, and L, norms in a teleseismic waveform inversion for the slip history of the Loma Prieta, California, earthquake: Bulletin of the Seismological Society of America, v. 81, no. 5, p. 1518-1539. Jaeger, J.C., and Cook, N.G.W., 1979, Fundamentals of rock mechan- ics: Chapman and Hall, New York, 593 p. Marshall, G.A., Stein, R.S., and Thatcher, W., 1991, Faulting geom- etry and slip from co-seismic elevation changes: The 18 October 1989, Loma Prieta, California, earthquake: Bulletin of the Seismo- logical Society of America, v. 81, no. 5, p. 1660-1693. McKenzie, D.P, 1969, The relationship between fault plane solutions for earthquakes and the directions of the principal stresses: Bulletin of the Seismological Society of America, v. 59, p. 591- 601. Michael, A.J., Ellsworth, W.L., and Oppenheimer, D.H., 1990, Coseismic stress changes induced by the 1989 Loma Prieta, Cali- fornia earthquake: Geophysical Research Letters, v. 17, no. 9, p- 1441-1444. Oppenheimer, D.H., 1990, Aftershock slip behavior or the 1989, Loma Prieta, California, earthquake: Geophysical Research Letters, v. 17, no. 8, p. 1199-1202. Oppenheimer, D.H., Reasenberg, P.A., and Simpson, R.W., 1988, Fault plane solutions for the 1984 Morgan Hill, California, earth- quake sequence: Evidence for the state of stress on the Calaveras fault: Journal of Geophysical Research, v. 93, p. 9007-9026. Reasenberg, P.A., and Simpson, R.W., 1992, Response of regional seismicity to the static stress change produced by the Loma Prieta earthquake: Science, v. 255, p. 1687-1690. Rice, J.R., 1992, Fault stress states, pore pressure distributions, and the weakness of the San Andreas fault, in Evans, B., and Wong, T.-F., eds., Fault mechanics and transport properties of rock: London, Academic Press, p. 475-503. Steid1, J.H., Archuleta, R.J., and Hartzell, S.H., 1991, Rupture history of the 1989 Loma Prieta, California, earthquake: Bulletin of the Seismological Society of America, v. 81, no. 5, p. 1573-1602. U.S. Geological Survey Staff, 1990, The Loma Prieta event: An antici- pated event: Science, v. 247, p. 286-293. Wald, D.J., Helmberger, D.V., and Heaton, T.H., 1991, Rupture model of the 1989 Loma Prieta earthquake from inversion of strong-mo- tion and broadband teleseismic data: Bulletin of the Seismological Society of America, v. 81, no. 5, p. 1540-1572. Wallace, RE., 1951, Geometry of shearing stress and relation to fault- ing: Journal of Geology, v. 59, p. 118-130. Zoback, M.D., and Beroza, G.C., 1993, Evidence for near-frictionless faulting in the 1989 (M 6.9) Loma Prieta, California, earthquake and its aftershocks: Geology, v. 21, p. 181-185. THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989: EARTHQUAKE OCCURRENCE AFTERSHOCKS AND POSTSEISMIC EFFECTS LOMA PRIETA AFTERSHOCK RELOCATION WITH S-P TRAVEL TIMES FROM A PORTABLE ARRAY By Susan Y. Schwartz and Glenn D. Nelson, University of California, Santa Cruz CONTENTS Page Abstract D91 Introduction 91 Method and data 92 Synthetic examples 93 Loma Prieta aftershOCK 10CatiONS ------------------------------------- 96 DiSCUSSiON ANd 101 Acknowledgments 102 References cited 102 ABSTRACT Aftershocks of the Loma Prieta earthquake are located using S-P arrival time measurements from stations of the PASSCAL aftershock deployment. We demonstrate the effectiveness of using S-P arrival time data in locating earthquakes recorded by a sparse, three-component net- work. Events are located using the program QUAKE3D (Nelson and Vidale, 1990) and a three-dimensional P- wave velocity model developed independently for this re- gion. Both a constant and a variable vp/vs ratio model are used to generate S-wave velocities. The dense coverage of the area around the Loma Prieta rupture zone by instru- ments of the California Network (CALNET) has allowed the U.S. Geological Survey (USGS) to find P-wave earth- quake locations, which we compare with our solutions. We also perform synthetic calculations to estimate realis- tic location errors resulting from uncertainties in both the three-dimensional velocity structure and the timing of ar- rivals. These calculations provide a comparison of loca- tion accuracies obtained using S-P arrival times, S and P arrival times and P times alone. We estimate average ab- solute errors in epicentral location and in depth for the Loma Prieta aftershocks to be between 1 and 2 km using S-P phase data and the sparse PASSCAL instrument cov- erage. The synthetic tests show that these errors are much smaller than those predicted using P-wave data alone and are nearly the same as those predicted using S and P phase data separately. This suggests that future aftershock recording deployments with sparse networks of three- component data can retrieve accurate event locations even if absolute timing is problematic using S-P times. INTRODUCTION Seismic studies using local networks have revealed im- portant information about earthquake dynamics, fault zone structure, and the effects of near-surface geology on the amplification of strong ground motion. Determination of accurate hypocentral locations is an essential step in analy- sis of local earthquake data; however, accurate locations are difficult to obtain in the common situation of sparse instrument coverage. Location difficulties arise from sev- eral sources: the mathematical instability of solutions due to the interdependence of origin time and depth, inad- equacies in the velocity model used in the location pro- cess, and uncertainties in arrival times caused by both reading and clock errors. In this paper, we demonstrate the effectiveness of utilizing S-P differential arrival-time data in locating earthquakes recorded by a sparse three- component network. The use of S-P times for earthquake location removes the trade-off between depth and origin time as well as arrival-time uncertainties arising from clock errors. Furthermore, locations derived from S-P arrival times are less sensitive to errors in the velocity model, provided that the Poisson ratio remains relatively constant in the region under consideration. We use the finite-difference location program QUAKE3D (Nelson and Vidale, 1990) to locate after- shocks of the Loma Prieta earthquake using data recorded by stations of the PASSCAL aftershock deployment (Lerner-Lam and others, 1990; Schwartz and others, 1990). The program QUAKE3D locates earthquakes in regions with complex three-dimensional velocity heterogeneity. The P-wave velocity structure in the source region of the Loma Prieta earthquake has been well studied, and both two-dimensional (Dietz and Ellsworth, 1990) and three- dimensional (Eberhart-Phillips and others, 1990a; Lees, 1990) velocity models have been computed. The Loma Prieta earthquake occurred in a region of dense CALNET D91 D92 AFTERSHOCKS AND POSTSEISMIC EFFECTS instrument coverage, providing accurate locations for most of the aftershocks that we compare with our solutions. Therefore, the aftershock sequence of the Loma Prieta earthquake provides an excellent opportunity to evaluate the effectiveness of using S-P arrival times to locate earth- quakes in a realistic three-dimensional velocity structure. In this study, we are constrained to use differential arrival times due to uncertainties in the absolute timing of the seismic phases recorded by the PASSCAL instruments. Because such timing uncertainties are often encountered in seismic field deployments, earthquake location tech- niques that require only the relative timing of phases are useful. The application of an S-P location scheme can also substantially reduce the time required to locate earth- quakes for studies involving large volumes of data, since time-consuming clock calibrations and corrections are not necessary. METHOD AND DATA The location algorithm QUAKE3D has been described elsewhere (Nelson and Vidale, 1990) and will only be summarized briefly here. QUAKE3D is a two-step loca- tion procedure where the first step is the discretization of the source volume and the computation of travel times from all recording stations to all points in the grid using the finite-difference method of Vidale (1990). S-wave travel times are produced from P-wave travel times by assuming a constant Poisson ratio throughout the source volume or by using a three-dimensional model of vp/vs ratios determined for this region by Thurber and Atre (1993). Next, the earthquake hypocenters are determined by finding the position within the source volume that yields the smallest travel time residuals for a set of arrival times for each earthquake. Since QUAKE3D is a grid searching algorithm rather than a least-squares iterative inversion, as are most earthquake location programs, the L1 or L2 norms can be employed with equal ease to determine earth- quake locations. Our solutions are computed using the L1 norm, since it is less sensitive to data outliers. In this study, we use S-P times to locate aftershocks of the 1989 Loma Prieta earthquake rather than only P times, as in Nelson and Vidale (1990). The S-P arrival times are picked by hand from recordings at stations of the Loma Prieta IRIS-PASSCAL aftershock deployment. The record- ing instruments consisted of Reftek 16-bit digitizers with L-22 (2 Hz) three-component geophones. They were de- ployed in the northern section of the Loma Prieta after- shock zone from October 20 through November 21, 1989. We locate 50 aftershocks with magnitudes between 1.5 and 3.0 that occurred in the region of dense instrument coverage between October 22 and November 4. In this period, between 10 and 18 widely distributed stations were in operation. Between 4 and 10 stations yielded high- quality P- and S-wave arrivals for each event. This num- ber is significantly lower than the total number of stations deployed (21) for several reasons: (1) only 10 stations were established during the first week, with instrument coverage expanding southward to a total of 20 stations by the end of the second week; (2) not all stations recorded all events; (3) some stations had instrument problems; and (4) some arrivals were not impulsive enough to accu- rately read P- and S-wave times. Figure 1 shows a map of the PASSCAL station locations and the USGS locations (Dietz and Ellsworth, 1990) of the events we analyze. We locate these events using a three-dimensional ve- locity model derived through inversion of CALNET P- wave arrival-time data from Loma Prieta aftershocks (Eberhart-Phillips and others, 1990b). The three-dimen- sional velocity model was interpolated from the published model to produce a uniform Cartesian grid (horizontal directions perpendicular and parallel to the strike of the San Andreas Fault and the vertical direction) consisting of 58 x 100 x 56 points with a 0.5-km spacing (29 x 50 x 28 km) covering the source area. The approximate bor- ders of this travel-time grid are indicated in figure 1. P- and S-wave travel times were then computed from each station to every one of the 324,800 grid points to be evalu- ated as potential earthquake sites during the location pro- cedure. S-wave travel times were computed either using a constant vp/vs ratio of 1.73 or the three-dimensional vp/ vs ratio model of Thurber and Atre (1993). The PASSCAL instruments recorded both high- and low-gain channels of three-component data that triggered when the ratio of the short term average amplitude (1 s) to the long-term average amplitude (30 s) exceeded a value of 4.5. Once triggered, 60 s of data were recorded, includ- ing a pre-trigger length of 10 s. The data sample rate was 200 samples per second during the first week of opera- tion, after which it was reduced to 100 samples per sec- ond. We read P- and S-wave arrival times from the vertical and horizontal components respectively of the high-gain channel. When data recorded on this channel were clipped, we read arrival times from the low-gain channel. In gen- eral, data quality was very high; however, for several events, S-wave arrival times could not be determined reli- ably and these events were not located. We estimate our reading errors to be less than 0.05 s and 0. 1s for the P and S$ waves respectively. Figure 2 shows examples of P and S$ waveforms from different earthquakes recorded by sev- eral of the PASSCAL stations. In many cases, both P and S$ waves are quite complicated; however, initial arrival times can be read very accurately from the vertical and horizontal components, respectively. It is apparent from figure 2 that attempts to read S-wave arrival times from the vertical component seismograms would introduce large errors, sometimes exceeding 0.2 s. LOMA PRIETA AFTERSHOCK RELOCATION WITH S-P TRAVEL TIMES FROM A PORTABLE ARRAY D93 SYNTHETIC EXAMPLES The advantages of using S phases to help constrain earth- quake location have been recognized for some time (for example, James and others, 1969; Buland, 1976; and Gomberg and others, 1990). Much of the benefit of in- cluding S$ phases in the determination of earthquake hypo- centers arises from the substantial increase in the range of partial derivatives of travel time with respect to the three spatial parameters of the hypocenter compared with P waves. The expanded range of the partial derivative with respect to depth can result in a reduction of the trade-off between depth and origin time, which plagues location schemes that use only P waves (Gomberg and others 1990). In situations where exact origin times of earthquakes are unnecessary, the use of S-P times for locating earthquakes removes the trade-off between depth and origin time, al- lowing depth determinations to be better constrained. Also, we show that S-P arrival times are less sensitive to uncer- tainties in velocity models than are P and S times sepa- rately. However, because uncertainties in both velocity structure and timing of arrivals are greater for S waves than for P waves, the actual benefit of incorporating S waves into earthquake location schemes may be substan- tially diminished compared with noise-free theoretical ex- pectations. To evaluate the effectiveness of using S-P times for earthquake location and to approximate the errors expected from our earthquake locations, we first performed com- prehensive synthetic experiments. The first experiment is designed to estimate location errors resulting from uncer- tainties in the assumed velocity model. Using the same receiver geometry as our data and the three-dimensional velocity model of Eberhart-Phillips and others (1990b), we computed exact P - and S-wave arrival times from the locations we determined for the 50 events shown in figure 1. We then located these events in a model generated by adding Gaussian-distributed random-velocity perturbations with an RMS value of 5 percent with scale lengths of about 5 km to the model of Eberhart-Phillips and others (1990b) and using the following combinations of phases: S$ and P, S-P, and P alone. When using both P- and S- wave data to locate earthquakes, we divided the P- and S- wave residuals by weights of 1 and vp/vs, respectively, prior to choosing the location that minimizes the L1 norm of the residuals. The weighting was applied to equalize 20 10° - 122° 50° Figure 1.-Map showing sites of PASSCAL stations (triangles) and USGS (Dietz and Ellsworth, 1990) locations of the earthquakes analyzed (solid circles). The boundaries of the finite-difference travel-time grid are indicated by the outer box. SAF is the San Andreas fault and SF is the Sargent fault. Earthquakes within the inner box are the subset of events included in the synthetic calculations. D94 the constraint imposed on the solution by the S and P phases. Figure 3 shows cross sections through the original and perturbed three-dimensional velocity models. The in- clusion of 5-percent random-velocity perturbations to the original velocity model is expected to adequately reflect realistic uncertainties in crustal velocities. If the true crustal structure contains sharp lateral-velocity contrasts, pertur- bations to a smoothly varying three-dimensional model may not adequately reflect velocity uncertainties. Further investigation of this possibility is beyond the scope of this paper. For several of the simulations, P- and S-wave station delays were computed for each station by averaging the station residuals from all events, as is often done in real earthquake location. These station delays were then sub- AFTERSHOCKS AND POSTSEISMIC EFFECTS tracted from the arrival-time data and the events were located again. The station delays were applied to account for imperfections in the assumed velocity model. Although the application of station delays had a small effect on the event locations and slightly reduced the L1 residuals, it did not significantly change the estimates of the location errors. Therefore, the results of the synthetic experiments do not include station delays. Since the 50 earthquakes are distributed over a large area, we also compute loca- tion errors from a subset of 19 closely spaced events lo- cated near the center of our travel-time grid (figure 1). In addition to calculating absolute errors in location result- ing from erroneous assumptions about the velocity model we also estimate relative errors. Relative errors are deter- mined by calculating the epicentral distances and depth 10/23/89 13:20 10/25/89 00:27 10/30/89 14:04 h=3.5 km mag=2.0 h=14.0 km mag=1.8 h=3.0 km mag=2.0 DBWN WSWD SAOS V P s P s S | + Figure 2-Vertical (V) and horizontal (N and E) component seismograms for three typical earthquakes recorded by six PASSCAL stations. P- and S-wave arrival times are marked on the vertical components. S-wave arrival times can be accurately determined from the horizontal components. h=earthquake depth and mag=magnitude, both determined by the USGS (Dietz and Ellsworth, 1990). LOMA PRIETA AFTERSHOCK RELOCATION WITH S-P TRAVEL TIMES FROM A PORTABLE ARRAY differences between all events in the original synthetic configuration of earthquake locations, subtracting them from the distances and depth differences between events after new locations are computed and averaging these val- ues. If the relative earthquake pattern does not change after the location procedure, then the average of these differences will be zero. The average lengths of the abso- lute and relative horizontal and depth mislocation vectors for all 50 events and for the subset of nineteen events are summarized in table 1 for the different combinations of phases. A small component of the error listed in table 1 is due to errors in the finite-difference calculation of travel times and in the interpolation of travel times prior to earthquake location. From a synthetic example with a similar size grid, Nelson and Vidale (1990) estimated average loca- tion errors of 0.2 km in the horizontal plane and 0.3 km in depth arising from the computation of finite-difference travel times. The use of either S and P or S-P phase data yield aver- age lengths of the horizontal and depth mislocation vec- tors of about 1.2 km and 0.5 km, respectively. The average mislocation increases only slightly when computed for events having observations from only four or five stations and decreases only slightly when computed for events with observations from six or more stations. Only three events have both a horizontal mislocation greater than 2 km and a depth mislocation greater than 1 km. These SW _- 3-D Velocity Model - NE M -I 5k 10 } _ 6 ~~ 15K 20 Tre 6.5 -- 25“? 7 ; ; --- -15 -10 --5 O 5 10 Distance (km) Figure 3-Cross sections through the 3-D model of Eberhart-Phillips and others (1990b) and the velocity model generated by introducing Gaussian- distributed random-velocity perturbations, with an average value of 5 percent over scale lengths of about 5 km, to this model. The cross sec- D95 events are recorded by four stations having a highly unfa- vorable distribution (either lying along a single line or clustered into two groups). Thus it appears that given rea- sonable station coverage, nearly comparable location ac- curacy can be obtained for events recorded by four or five stations as for events recorded by six or more stations using either S and P or S-P phase data. Locations calcu- lated using P phases alone produce much larger errors, especially in depth. This is not surprising considering the locations using only P arrivals have half the number of observations as locations using S$ and P data. The average horizontal and depth mislocation for events located with P waves from four or five stations is over twice the aver- age of events located using P arrivals from six or more stations. As noted by previous workers, the inclusion of S-wave phase data drastically reduces the number of sta- tions necessary to obtain accurate earthquake locations. The location errors are slightly reduced when only a subset of events with fairly uniform station coverage are considered (table 1). The average lengths of the relative error vectors are larger than the lengths of the absolute error vectors when all earthquakes are considered. The lack of improvement in the relative errors over the abso- lute errors is due to the rather large region containing the earthquake hypocenters as well as the lack of a consistent set of stations recording each event. The average lengths of the relative horizontal error vectors are smaller than the absolute error vectors for the more closely spaced SW - 3-D Perturbed Model _- NE O 5M 6 10 © 15 20 0 J 7.5 255ij a -~_ 1 wwf ]} -15 -10 --5 O 5 10 Distance (km) tions were taken near the intersection of the San Andreas fault with the Sargent fault. Contours are P-wave velocities in km/s and distances are measured from the San Andreas fault (0 km). D96 AFTERSHOCKS AND POSTSEISMIC EFFECTS Table 1.-Synthetic location errors All Events Subset of Events Absolute Errors | Relative Errors | Absolute Errors | Relative Errors Phases Error | xy sd z sd | xy sd z sd | xy sd z sd | xy sd z sd Source S and P V 12 0.9 0.5 0.6]1.5 1.3 0.7 0.8]0.9 0.4 0.3 0.2]0.6 0.4 0.4 0.3 S-P V 1.2 0.8 0.4 0.5| 1.4 1.2 0.6 0.6|1.0 0.4 0.2 0.2]0.7 0.5 0.4 0.3 P V 1.8 1.5 1.8 2.0|2.2 2.2 2.4 2.3]1.4 0.9 1:3 1.2] 1.1 0.9 1.8 1.4 Sand P V&T | 1.2 1.0 0.5 0.6] 1.5 1.4 0.8 0.8]|0.9 0.4 0.4 0.2]0.7 0.5 0.6 0.4 o V&T 18 2.4 0.8 2.2 1.1 1.3]1.0 0.7 0.5 0.3|0.9 0.7 0.7 0.5 P v&T|2.2 2.5 2.3 2.5|2.9 3.4 3.0 2.9] 1.7 1.7 2.0 2.1]1.6 1.7 2.6 2.5 Efror sources result from locating earthquakes in a different velocity model from the one used to calculate the phase data (V) and from the addition of noise to the arrival time data proportional to our estimated reading errors (T). xy=average length of the horizontal error vector, z=average length of the depth error vector and sd=standard deviation of the error. All errors are in km. subset of events; however, the absolute depth errors are still smaller than the relative depth errors. Although we expected the inclusion of S phases in the location determination to greatly improve their accuracy, the equal success of S and P and S-P arrival times is somewhat surprising considering that the range in the par- tial derivative of the S-P travel time with respect to the hypocentral parameters is substantially decreased compared with S or P arrival times (Gomberg and others, 1990). The success of S-P arrival times is partly due to the di- minished sensitivity of this differential measure to the absolute velocity model assumed. Although we located the synthetic earthquakes in a velocity model perturbed from the one used to compute the arrival-time data, both P- and S-wave velocities were perturbed in the same man- ner using a constant vp/vs ratio of 1.73. If we locate the synthetic earthquakes using an S-wave velocity structure having a vp/vs ratio that differs from that used to compute the arrival time data, the error estimates increase in pro- portion to the difference between the correct and assumed ratios. In general, the vp/vs ratio is better known than the absolute velocity structure, so the use of S-P differential times can provide nearly as accurate locations as those obtained using S and P arrival times. The magnitude of the location errors resulting from a 3-percent difference between the assumed and actual vp/vs ratios, using S-P arrival times, is on the order of 1 km both horizontally and in depth. This error does not increase when 5-percent random velocity perturbations are added to the original model. To estimate realistic errors to be expected in our after- shock locations, we repeated the first experiment after adding random time delays with an average amplitude determined from our estimated reading errors (0.05 s for P and 0.1s for S) to the arrival-time data. The results, averaging location errors determined for three separate simulations of random reading errors, are listed in table 1. The inclusion of reading errors had little effect on loca- tions determined using S and P data but did increase the errors, as expected, when using S-P or P data. These syn- thetic calculations indicate that if absolute timing of S phases is unavailable, S-P arrival times can locate earth- quakes with only slightly less accuracy than the use of S and P phases separately. Previous experiments on syn- thetic datasets including $ phases indicated that epicentral parameters were far less sensitive to data inadequacies than depth (Gomberg and others, 1990). Here, we find that the inclusion of S phases in earthquake location re- sults in consistently smaller depth mislocation than epi- central mislocations. LOMA PRIETA AFTERSHOCK LOCATIONS The synthetic calculations indicated that we can expect average absolute errors in QUAKE3D locations of the Loma Prieta aftershocks to be about 1.8 km in the hori- zontal plane and 0.8 km in depth using S-P phase data and the sparse PASSCAL instrument coverage, provided that the average vp/vs ratio is reasonably well known. We use the three-dimensional velocity model of Eberhart- Phillips and others (1990b) to relocate 50 aftershocks of the Loma Prieta earthquake. We compare QUAKE3D so- lutions obtained using this velocity model and a constant vp/vs ratio to locations obtained by Eberhart-Phillips and others (1990b) using only P-wave arrivals. Associated with all seismicity maps, we show three cross-sections; one parallel to the strike of the San Andreas Fault and two perpendicular to it. Figures 4 and 5 show locations of 50 aftershocks deter- mined with QUAKE3D using S-P arrival time data from LOMA PRIETA AFTERSHOCK RELOCATION WITH S-P TRAVEL TIMES FROM A PORTABLE ARRAY D97 15" - 10° 5° S5° 50° B x x> [al 1 T 1 i 1 L 1 1 1 I L 1 T 1 1 1 1 1 1. 1 T 1 1 1 1 T T T 1 1 1 1 1 L 1. 1 Si [ x X® . R x k R x - X (3 - x x - .Xx |- J y y L ace: _ 4 gig ¢ x n = | x J x | ra | o x _ _® | / O T -10"" e x o} o x - F- | L 8:1 & X‘ * x * X A 1 a “a? “X % | - x p x |- | hes e x * % x |- e bat g< { x o . J % .>< ’( L | xX _x o o e |. P 3 x . “20 T T T T T T T ffi—| T T T T T T T T l T T T T T T T T T I T T T T T T 0 10 30 Figure 4.-(A) Map view comparing locations determined for the 3-D velocity model using S-P finite-difference travel times from the PASSCAL stations (x) and locations from Eberhart-Phillips and others (1990b) (solid dots). The events marked by the boxes and triangles are 20 DISTANCE (KM) the two events with very large location discrepancies. Lines X-X", A-A and B-B' indicate the location of cross sections. (B) Cross section X-X" along the length of the San Andreas fault. D98 AFTERSHOCKS AND POSTSEISMIC EFFECTS the PASSCAL aftershock deployment (x's) and determined __ work (CALNET). The QUAKE3D locations are indicated by Eberhart-Phillips and others (1990b) from P-waves re- in table 2. The average difference in epicentral location corded by stations of the dense California Seismic Net- _ and in the absolute value of depth between the two loca- A 1 SAF SF o A 1 1 1 1 1 1 1 L 1 l I 1 1 1 1 1 I 1 1 l 1 1 1 1 1 1 1 1 O { x L - X o. X - - . x x |- - x 8a, x - - 20 kts [ ® e -I x @ e a - - * |- E X x %o G R . X |- ¢ ~107 i- U .- a x ® e* x A { » W X |- - ex - - ® y oe |. J - x - J G 1 L L 1 L 1 m 1 1 l T T 1 L 1 T L 1 1 I 1 1 T L T 1 1 x - - ® x X L E - X . e \ R x . int xX © y fo -1047 xX |- R o i- - p 7 x X uJ .- A 7 % o - R x L X X ).‘ x - o L {4 xo, 8y L Be oe | _ $ x . -2O T T T T I T T T I I T T T T T T T T T I T T T T T T T 0 10 20 DISTANCE (KM) Figure 5.-Cross section A-A" (A) and B-B" (B) perpendicular to the San Andreas fault (SAF) showing the seismicity in figure 4. Sections A-A" and B-B' include all events north and south of the intersection of the SAF with the Sargent fault (SF), respectively. LOMA PRIETA AFTERSHOCK RELOCATION WITH S-P TRAVEL TIMES FROM A PORTABLE ARRAY Table 2.-Loma Prieta aftershock relocations Event Latitude Longitude - Depth Mag __ N* yr mo dy hrmn_ deg __min deg __ min __ km 891023 1320 - 37 11.61 121 58.01 2.0 _ 2.0 6 891023 1402 - 37 6.450 122 3.12 8.3 _ 2.3 4 891023 1812 37 14.62 122 3.69 8.4 _ 2.0 4 891023 - 2127 37 8.40 121 54.00 8.6 _ 3.0 5 891024 _ 0336 37 10.27 122 3.12 14.0 1.8 4 891024 _ 0403 37 8.67 121 57.86 11.8 2.1 5 891024 _ 0702 37 8.78 121 55.24 2.9 - 2.8 7 891024 _ O855 _ 37 11.34 121 56.05 4.1 _ 2.5 7 891024 _ 1003 37 10.69 121 58.96 7.1 1.8 6 891024 _ 1003 37 10.85 121 57.15 3.5 1.8 6 891024 _ 1444 37 5.69 121 57.01 16.5 2.3 4 891024 _ 1832 37 8.71 121 58.87 11.9 1.9 5 891024 _ 2225 37 11.15 121 56.77 4.0 - 2.5 5 891024 _ 2341 37 10.73 121 56.34 5.3 - 2.3 8 891025 - 0026 37 8.97 122 3.88 13.2 1.8 4 891025 - 0305 37 10.73 122 1.97 12.3 1.8 6 {891025 - O314 37 11.46 122 4.69 1.6 _ 2.3 4 {891025 1223 37 2.90 121 49.60 3.1 _ 2.1 5 891025 1515 37 10.73 121 59.30 6.5 2.0 4 891025 _ 1926 37 8.78 121 59.35 13.2 2.0 4 891026 - OS518 37 10.08 121 58.87 7.9 _ 2.1 6 891026 _ 1100 37 6.23 121 57.20 15.9 1.9 4 891026 _ 1208 37 9.89 121 57.48 9.7 - 1.9 4 891026 _ 1326 37 2.22 121 54.47 14.0 2.4 4 891026 _ 1508 37 5.35 121 54.38 7.9 2.0 6 891026 _ 1544 37 11.84 121 59.92 5.3 - 2.1 4 891028 - 0001 _ 37 2.90 121 54.95 15.9 2.1 5 891028 1120 _ 37 - 9.43 122 5.22 11.3 2.0 7 891028 _ 1947 37 8.17 121 54.67 10.0 2.1 7 891029 _ 1310 37 5.58% 121 55.72 16.0 2.9 4 891029 _ 1918 37 3.17 121 51.37 17.7 2.3 6 891029 _ 2043 37 - 5.65 121 53.52 13.0 2.2 10 891029 _ 2155 37 3.67 121 54.09 9.6 _ 2.9 8 891030 _ 0452 37 4.47 121 54.62 14.5 2.5 7 891030 _ 0710 37 11.00 121 59.54 6.9 _ 1.5 5 891030 _ 1258 37 6.38 121 55.00 11.6 2.1 7 891030 _ 1403 37 11.07 121 58.20 0.8 _ 2.0 10 891030 _ 1541 37 7.91 121 53.76 0.5 _ 2.4 6 891030 _ 1740 37 11.91 122 1.06 11.4 1.7 6 891031 _ 0218 37 7.03 121 56.00 9.7 _ 2.1 7 891031 _ 0302 37 10.58 122 3.12 15.0 2.3 7 891031 _ 0332 37 - 8.44 121 54.52 11.1 2.3 9 891031 _ 0457 37 13.82 122 5.74 2.6 _ 2.0 5 891031 _ 0649 37 7.91 121 59.39 12.5 2.2 7 891031 1811 _ 37 - 4.350 121 53.52 14.4 2.5 5 891101 _ 2300 37 7.10 121 55.33 13.6 2.5 5 891102 - 0423 37 10.35 121 59.11 6.5 _ 2.0 8 891102 - 1010 37 6.38 121 58.06 16.3 2.1 5 891102 - 1548 37 3.93 122 3.79 7.2 2.0 6 891103 _ _0257 _ 37 10.23 122 0.68 11.0 _ 1.7 10 *N is number of S-P phases used in location {Events with large location discrepancies compared with USGS locations. tion sets is 2.3+1.4 km and 1.3+1.7 km, if two events with large location differences are excluded (these events are indicated by triangles and boxes in figure 4). Eberhart- Phillips and others (1990b) determined locations for both of these events using arrival time data determined with a real time processor (RTP), and it is possible that the RTP missed an emergent first arrival. One of the events (box in figure 4A) was located with P-wave arrival times from D99 only nine stations; the smallest number of arrivals for any of the 50 events. This event also had the largest rms re- sidual and the largest error estimates of the 50 events. We located these two events with observations from only a small number of stations with a less-than-ideal distribu- tion. Although our synthetic tests indicated that locations using only four or five S-P observations were, in general, as accurate as those obtained using many more observa- tions, the largest mislocation vectors determined for our tests were obtained for events with observations from few, poorly distributed stations. Although there are significant differences in the earth- quake locations determined in this study and those deter- mined by Eberhart-Phillips and others (1990b), it is difficult in such a geologically complex region to apply any simple criterion to decide which locations are more accurate. Our synthetic experiments predict that locations determined with S-P times should be better constrained than those obtained from P-wave data if the event-station geometries are constant and our assumption and assign- ment of a constant average vp/vs ratio is fairly accurate. However, the locations determined by Eberhart-Phillips and others (1990b) from P-wave data used many more P- wave observations from more distant stations than the num- ber of nearby S-P observations that we used. Also, recent work suggests that our assumption of a constant vp/vs ratio for this region may not be valid. Thurber and Atre (1993) inverted S-P arrival times from Loma Prieta after- shocks recorded by stations of the PASSCAL aftershock array to determine a three-dimensional model of vp/vs variations in the Loma Prieta rupture zone. Their results indicate that variations in the vp/vs ratio may be as large as 10 percent in this region. To determine how a variable vp/vs ratio would effect our locations, we calculate earth- quake locations using the three-dimensional P-wave ve- locity model of Eberhart-Phillips and others (1990b) and the vp/vs model of Thurber and Atre (1993). A compari- son of these locations with those determined using a con- stant vp/vs ratio of 1.73 are shown in figures 6 and 7. The average difference in position of earthquakes located using the variable vp/vs ratio compared with the constant ratio is 1.5% 0.9 and 1.2+ 0.9 km in epicentral location and in depth, respectively. The depth of events located using the variable vp/vs ratio is systematically shallower than those located using a constant ratio. This is because the vp/vs ratio (1.73) of the constant model is smaller than the average ratio in the variable model. Increasing the constant vp/vs ratio would increase the focal depths and eradicate this depth bias. The use of Thurber and Atre's (1993) variable vp/vs ratio to locate Loma Prieta aftershocks does not reduce the travel time residuals compared with a constant vp/vs ratio and results in earthquake locations that are only slightly closer in loca- tion to those determined by Eberhart-Phillips and others (1990b). D100 AFTERSHOCKS AND POSTSEISMIC EFFECTS 15" “- 5‘ é : 10° S5" 122° 35° 50° G 1 1 T 1 L T 1 1 l L i T 1 1 1 1 1 I 1 T i 1 1 1 1 i L 1 1 1 T 1 o % F x x 4 . . x - & | x XX. # e L 4 x .. - | x x * - x e x" . % 4 28s L 4 © o X o L X x x X = | # %o ® | S | x x ° x | x o 4 ~107 ® *e |- i- | a f o © @ x XX. * o =] 1 ® __ Xe x '. x x |- 4 X e X - o @ o ® x - . . a x - ® ® - # y . } 1 % - x B x - | bat L "20 T T T 7 T T T T T 1 T T T T T T T : T l T T T T T T T T l T T T T T 0 10 20 30 DISTANCE (KM) LOMA PRIETA AFTERSHOCK RELOCATION WITH S-P TRAVEL TIMES FROM A PORTABLE ARRAY < Figure 6.-(A) Map view comparing locations determined for the 3- D velocity model using S-P finite-difference travel times from the PASSCAL stations and a variable vp/vs ratio model (x) with locations computed using a constant vp/vs ratio of 1.73 (solid dots). (B) Cross section X-X" along the length of the San Andreas fault. D101 DISCUSSION AND CONCLUSIONS Although several previous studies have addressed the sensitivity of earthquake locations to uncertainties in assumed velocity models and in the timing of phases, these studies have almost exclusively considered layered velocity structures only. Changes in the lateral velocity A A SAF SF A (s, 1 L 1 1 1 1 I 1 1 I 1 1 1 11 O1 1 l 1 1 l 1 1 | 1 1 1 i 1 | o | s * x o [ | X | x | ** 0 . | 4 x Ax L e I y * | | i L C of . & x q 10-7 X e |- -- a 1 e fix. - L A 1 a' 052. bt ~ - s e* _ I . - Ce | | B B1 O 1 1 L 1 1 1 L 1 1 I 1 1 1 1 1 L 1 L l 1 1 1 1 L 1 1 e - XX - e © a | s L 2 { ® * - z | x x xx ® | qo ® _ e e * |- i- - E | wo 00 - e [ 4 xX X - | x & I ® ®, L x 4 & M L "3 | yo ® | | ® L "20W T T T T T T T T Tj T T T T T 1 I T T I I T T 0 10 20 DISTANCE (KM) Figure 7.-Cross sections A-A" (4) and B-B' (B) perpendicular to the San Andreas fault (SAF) showing the seismicity in figure 6 in the same manner as in figure 5. D102 gradients will cause raypaths to curve. The effects of ray curvature on travel times are diminished when lateral gra- dients are not allowed, as in the case of propagation through a layered structure. This can result in an underes- timation of location errors when they are assessed in plane- layered media. Here we estimate location error that results from random perturbations in a three-dimensional veloc- ity model, although the model we consider possesses only smooth velocity gradients. Our methods correctly account for ray curvature. We evaluate location errors using the event-station geometries provided by Loma Prieta after- shocks recorded at the PASSCAL instruments. An aver- age of only six S-P observations were used in the location determinations. The results are summarized in table 1. Since earthquake locations are often computed from sparse data, an estimation of the location accuracy that can be obtained from so few observations is important. We found that in the absence of reading errors, S-P arrival-time data can locate earthquakes with comparable accuracy as S and P data separately. While the inclusion of realistic timing errors slightly degrades the accuracy of the S-P earthquake locations, the differential measure can still lo- cate earthquakes with small uncertainties (< 2 km in epi- central location and <1 km in depth). The relocation of 50 Loma Prieta aftershocks using S-P arrival time data from the PASSCAL aftershock de- ployment are consistent with several other studies indicat- ing complex faulting at depth within the Loma Prieta rupture zone. Although the locations suggest the exist- ence of many fault segments, a definitive image of the faulting geometry at depth for this structurally complex region requires earthquakes to be located with greater ac- curacy than we can presently achieve. Such accuracy may be possible if present uncertainties in the velocity struc- ture can be substantially diminished. Rapid response is essential for the success of aftershock deployments such as those initiated after the 1988 Arme- nian and 1989 Loma Prieta earthquakes. The installation, testing, and operation of sophisticated clock systems is both expensive and time consuming. Therefore, the assur- ance of accurate absolute timing can significantly decrease the amount of data recorded during a field operation. In the absence of reliable absolute timing, as was the case for the PASSCAL Loma Prieta aftershock data, where there were many clock problems, we have shown that accurate earthquake locations can be obtained using S-P differential travel times from a sparse, three-component network. The Loma Prieta earthquake occurred in a re- gion of very dense pre-existing instrument coverage that was able to provide accurate earthquake locations for the aftershock sequence independent of the PASSCAL de- ployment. Although the locations we obtained in this study using S-P data from the sparse, three-component PASSCAL network may or may not have improved the AFTERSHOCKS AND POSTSEISMIC EFFECTS previous determinations, their accuracy demonstrates the usefulness of S-P times to locate earthquakes in more remote regions. In remote locations, such as Armenia, field deployments consisting of sparse instrument cover- age may be the only means available to obtain accurate earthquake locations. In these situations the location method discussed here can be very useful. ACKNOWLEDGMENTS We are grateful to John Vidale for his excellent advice throughout the course of this project and for his thorough review of this manuscript. We also appreciate and ben- efited from discussions with Thorne Lay. We thank Cliff Thurber, Lynn Dietz and Donna Eberhart-Phillips for re- viewing this manuscript and Zhi Zhang for her help with data processing. This work was supported by USGS Grant no. 14-08-0001-G1836, IRIS Grant SC90115, and a grant from the W.M. Keck Foundation. Contribution 180 of the W.M. Keck Seismological Laboratory and the Institute of Tectonics. REFERENCES CITED Buland, R., 1976, The mechanics of locating earthquakes: Bull. Seismol. Soc. Am., v. 66, p. 73-187. Dietz, LD. and Ellsworth, W.L., 1990, The October 17, 1989, Loma Prieta, California, earthquake and its aftershocks: Geometry of the sequence from high-resolution locations: Geophys. Res. Lett., v. 17, p. 1417-1420. Eberhart-Phillips, D.M., Labson, V.F., Stanley, W.D., Michael, A.J., and Rodriguez, B.D., 1990a, Preliminary velocity and resistivity models of the Loma Prieta earthquake region: Geophys. Res. Lett, v. 17, p. 11235-1238. Eberhart-Phillips, D.M., Michael, A.J., Fuis, G., and Luzitano, R., 1990b, Three-dimensional crustal velocity structure in the region of the Loma Prieta, California, earthquake sequence from inver- sion of local earthquake and shot arrival times: Seismol. Res. Lett., v. 61, p. 48. Gomberg, J.S., Shedlock, K.M., and Roecker, S.W., 1990, The effect of S-wave arrival times on the accuracy of hypocenter estimation: Bull. Seismol. Soc. Am.,v. 80, p. 1605-1628. James, D.E., Sacks, I.S., Lazo, E., and Araracio, G., 1969, On locating local earthquakes using small networks: Bull. Seismol. Soc. Am., v. 59, p. 1201-1212. Lees, J.M., 1990, Tomographic P-wave velocity images of the Loma Prieta earthquake asperity: Geophys. Res. Lett, v. 17, p. 1433- 1436. Lerner-Lam, A., Simpson, D., Menke, W., Kim, W.Y., Hough, S., Pacheco, J., and Estabrook, C., 1990, Initial results of waveform analyses of IRIS/PASSCAL Loma Prieta aftershock data: Trans. Am Geophys. Union, v. 71, p. 289. Nelson, G.D. and Vidale, J.E., 1990, Earthquake locations by 3-D fi- nite-difference travel times: Bull. Seismol. Soc. Am., v. 80, p. 395-410. LOMA PRIETA AFTERSHOCK RELOCATION WITH S-P TRAVEL TIMES FROM A PORTABLE ARRAY D103 Schwartz, S.Y., Velasco, A., Protti, M., Bonamassa, O., Nelson, G.D., Thurber, C.H., and Atre, S., 1993, Three-dimensional vp/vs variations McNally, K.C., Vidale, J.E., Lay , T., and Flatté, S., 1990, A along the Loma Prieta rupture zone: Bull. Seismol. Soc. Am., v. PASSCAL instrument aftershock deployment along the rupture 83, p. 717-736. zone of the 18 October 1989 Loma Prieta Earthquake: Trans. Am. Vidale, J.E., 1990, Finite-difference calculation of travel time in 3D: Geophys. Union, v. 71, p. 290. Geophysics, v. 55, p. 521-526. e THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989: EARTHQUAKE OCCURRENCE AFTERSHOCKS AND POSTSEISMIC EFFECTS EMPIRICAL GREEN'S FUNCTION STUDY OF LOMA PRIETA AFTERSHOCKS: DETERMINATION OF STRESS DROP By H. Guo, A. Lerner-Lam, and W. Menke, Lamont-Doherty Earth Observatory of Columbia University, and S.E. Hough, U. S. Geological Survey CONTENTS Page Abstract D105 Introduction 105 Data 105 Analysis 108 Determination of corner frequency ---------------------------- 108 Calculation Of Stres$ COP 115 Stress drop results 115 ConclUusion$ @Md GiSCUSSiOM 115 Acknowledgments 119 References cited 119 ABSTRACT This paper presents the estimation of stress drops of aftershocks of the Loma Prieta earthquake recorded by an array of 22 PASSCAL instruments. Using an empirical Green's function analysis, corner frequencies of 98 of the aftershocks are first determined. Brune-model stress drops, AG, are calculated from the corner frequencies and mo- ments. A dependence of stress drop on moment is ob- served. INTRODUCTION Estimates of earthquake source spectral parameters can be contaminated by effective attenuation along the source- receiver path, frequency-dependent propagation, and near- receiver site response. The shaping of the spectrum of an arrival by attenuation is known to bias estimates of corner frequency and thus its interpretation in terms of stress drop (Hough and Anderson, 1988). To avoid this bias, a reference measurement technique must be employed when parameterizing the source spectrum, especially in a local aftershock survey. In this paper, we apply an empirical Green's function (eGf) analysis (Frankel and others, 1986; Geller and Mueller, 1980; Hartzell, 1978; Hough and oth- ers, 1991; Hutchings and Wu, 1990; Li and Thurber, 1988; Mori and Frankel, 1990; Mueller, 1985; Xie and others, 1991) to aftershock doublets of the Loma Prieta sequence recorded on the IRIS/PASSCAL digital seismographs. The smaller event in each doublet provides the reference sig- nal for correcting the spectral characteristics of the larger source for spectral shaping by attenuation. In principle, eGf methods provide an efficient technique for quickly parameterizing large amounts of digital data. In practice, however, one has to be careful about the ap- plication of the method in the case of inadequate signal- to-noise ratio for the smaller reference signal. We develop a methodology for identifying and mitigating such low- signal problems and show how to assess the possibility of bias in the corner-frequency estimate. In addition, we per- form sensitivity tests to demonstrate that the estimate of these spectral parameters is robust with respect to varia- tions in fitting range and noise level. The combination of extensive coverage of the Loma Prieta aftershock zone with IRIS/PASSCAL instrumenta- tion, coupled with the locating capabilities of CALNET, provides a unique data set for waveform work and eGf analysis of spectral characteristics. In this paper, a frequency-domain deconvolution using the empirical Green's function method is used to deter- mine the corner frequencies and the stress drops of 98 aftershocks of the Loma Prieta earthquake. DATA The Loma Prieta earthquake and its aftershocks are an exceptionally well-recorded earthquake sequence. The af- tershocks were recorded not only by the permanent CALNET stations but also by a temporary network. Fol- lowing the Loma Prieta mainshock, an array of 22 IRIS/ PASSCAL (Incorporated Research Institutions for Seis- mology/Program for Array Seismological Studies of the D105 D106 Continental Lithosphere) stations was deployed in the epi- central area in the Santa Cruz mountains (see fig. 1). More than 2,000 events were recorded by three-component L- 22D 2-Hz velocity sensors paired with IRIS/PASSCAL RefTek model 72A recorders. Two separate geophones, recorded at different gain levels and sampled at 200 samples per second, were deployed at each site. These geophones were calibrated after the Loma Prieta deploy- ment by Menke and others (1991). A total of 763 of these aftershocks, with local magni- tude, M;, ranging from 1.5 to 4.5 and recorded from Oc- tober 20 to November 25, 1989, have been time-associated with CALNET data. These associated aftershocks were relocated relative to preliminary CALNET hypocenters by inverting CALNET phase data with a one-dimensional velocity model and a modified HYPOINVERSE program with station corrections dependent on hypocentral loca- tion (Seeber and Armbruster, 1990). This method has been shown to give better relative locations (Seeber and Armbruster, 1990; Pujol, 1992). AFTERSHOCKS AND POSTSEISMIC EFFECTS Candidate pairs for eGf analysis were selected from the relocated hypocenters. The fundamental assumptions of eGf analysis require that (1) the paired events must be closely spaced so that source-receiver paths are nearly coincident, (2) the event pair should have similar focal mechanisms and have similar waveform complexities, and (3) the rupture duration of the smaller or reference event should be short relative to the larger event so that the reference corner frequency is resolvably different from the corner frequency of the large event. This last criterion has implications in terms of signal-to-noise ratio of the quotient spectrum. It is most convenient to interpret this requirement in terms of relative magnitude difference, AM;, The AM; values must be large enough to resolve the dif- ference in corner frequencies but not so large as to vitiate the assumption that attenuation and near-receiver effects are linear for two events. Our criteria for the relocated Loma Prieta aftershocks were (1) the hypocenters must be within 1 km of each other, (2) time- and frequency-domain signals must be 37.3 LL LL CC A (5 LL Q u Q _J E 37.0 t _J A A A F 3 A \A A A 36.8 122.2 -122.0 121.5 LONGITUDE, DEGREE Figure 1.-Portion of the aftershock zone of the Loma Prieta earthquake. Main shock (diamond), 22 PASSCAL stations (circles), 53 closest CALNET stations (triangles), San Andreas Fault (SAF), and Sargent fault (SF) are shown. A total of 124 events recorded by the PASSCAL array with magnitudes larger than 1.5 are used for empirical Green's function analysis and denoted by plus symbols. EMPIRICAL GREEN'S FUNCTION STUDY OF LOMA PRIETA AFTERSHOCKS: DETERMINATION OF STRESS DROP visually similar, and (3) AM; >0.7. Comparable values of this lower bound for AM; have been used in other studies (Li and Thurber, 1988: 0.5; Mueller, 1986: 0.7; Xie and others, 1991: 0.5). The largest AM; for out data set is 2.9, which is small enough to maintain linearity. Of these cri- teria, the visual check is the most subjective. In the time domain, the comparison emphasizes the first motion di- rection, the detailed shape of the first several oscillations, and the overall shape of the envelope of the whole seis- mogram (for example, fig. 2). In the frequency domain, the comparison emphasizes spectral content below 30 Hz D107 because of spurious geophone resonances detected by Menke and others (1991). Only those event pairs that sat- isfy the above criteria on 80 percent of the seismograms are suitable for further analysis. Seismograms with re- cording errors or signal-to-noise ratios less than approxi- mately 20 dB were discarded. Low-gain data were used only if the signals were clipped on the high-gain data streams. Of the 763 relocated aftershocks with AM, >1.5, 347 pairs were found to meet the hypocentral and restric- tion, and 226 event pairs were found to meet the subjec- 0.1 I | I T y I | II 0.1 i |- z N g |- e 0.14 F O 1.7e-3 Z C | “4 1.6e-3 F < N N“ - g L 3 2.7e-3 | -- E - | | | | 0 2 4 6 8 10 TIME, SECOND Figure 2.-Three-component seismograms of a typical event pair (events 107 and 106). Amplitude scales differ between seismograms. D108 tive waveform similarity criteria. The larger events in each pair have magnitudes between 2.2 and 4.4, and the smaller events have magnitudes between 1.5 and 2.9. ANALYSIS DETERMINATION OF CORNER FREQUENCY The P-wave spectra were computed for a 2.5-second window starting just before the arrival. The length of the window was derived after experimentation to determine the trade-off between the successful characterization of the long-period signal and the inclusion of too much scat- tered energy. The spectra were computed using 4mt-pro- late multitapers, which yield a smooth spectral estimate with good resistance to leakage for relatively short time windows (Park and others, 1987). The paired spectra were divided to obtain spectral ra- tios, which should depend only on the relative source char- acteristics. Spectral ratios from different components and different stations were averaged to reduce the noise. This averaging procedure neglects the effects of directivity, which is not thought to be a large effect. The parameterization of the spectral ratio in terms of source characteristics is achieved by assuming a Brune spectrum, A(f), for each event (see fig. 3): T -TTIIm] T-TTTtmmn to TT Tm] T Tm] LOG SPECTRAL AMPLITUDE T T TTT] tc2 = 20 Hz LOL t | 1 Lo p uu 0 1 1 1 o LOG FREQUENCY, HZ Figure 3. Corner frquencies of a hypothetical event pair, showing spec- tra as a function of frequency for large event (top), small event (middle), and ratio (bottom). fc1=5 Hz and fcz=20 Hz are the corner frequencies of the larger and smaller events, respectively. 1 2 AFTERSHOCKS AND POSTSEISMIC EFFECTS AP=A,01 +f") where A, is the amplitude scaling and f, is the corner frequency. The spectral domain deconvolution yields the spectral ratio R(f): 14, 2)! where subscripts 1 and 2 (01, o,) denote the larger and smaller members of the event doublet, respectively. Un- less fc2 is high enough relative to fC , or A,,, is below the ambient noise, all four parameters must be included in the representation of the spectral ratio. We do this by finding the amplitude ratio A, /A 03 that minimizes the 2-norm of the residual R (f)—R(f) where Rf) is the observed spec- tral ratio, while f,. and f,,, are varied on a grid. The values of fc and fC that yield the smallest residual are then taken to be estlmates of the event corner frequencies. The cor- ner frequencies are first gridded at 1-Hz increments be- tween 1 Hz and a maximum of 100 Hz and 0.1-Hz increments for lower corner frequencies. R(f) is fitted be- tween the instrument corner at 2 Hz and an upper noise- free frequency, which is determined independently for each pair. Poor choice of the fitting range yields obvious insta- bilities. The morphology of the residual manifold is as- sessed visually for each event pair studied. Two examples of this procedure, displayed in figure 4, show that this methodology can be problematic under cer- tain circumstances. Figure 4A shows the residuals for the event pair (20, 24). In this case, the corner frequencies (f L. ) (3, 16 Hz) are well-resolved. This figure also shows the bias that is possible if we assume the smaller event to be a delta function and the search is gridded only over the corner frequency f,. . Figure 4B shows the residu- als for the event pair (58, 60). In this case no closed residual low is observed, and the corner frequency of the smaller event is not resolved at all. However, fcl has a small variation relative to the poor resolution of fez. This is generally true for event pairs where the smaller event magnitude is less than 2 or where the ambient noise is high. In order to get the best estimate of f we fix f to a value consistent with the average value found in prev1ous studies (Molnar and others, 1973; Boatwright and others, 1991; Fletcher and Boatwright, 1991). Because f in this case is assumed, it is indicated with a blank arrow in figure 5 and excluded from the averaging. Of the 226 pairs analyzed, 102, which include 124 indi- vidual events, are fitted well by the Brune ratio model, shown in figure 5. Averaging of common events and fur- ther exclusion of artifacts yielded 98 individual measure- ments of P-wave corner frequencies, with 53 from the larger events (see table 1). These corner frequencies fall in the range 3.8 to 37.5 Hz, with corresponding events having local magnitudes between 1.5 and 4.4. A so 40 CORNER FREQUENCY Fc2 CORNER FREQUENCY Fe2 30 20 10 100 90 80 70 60 50 40 30 20 0 10 EMPIRICAL GREEN'S FUNCTION STUDY OF LOMA PRIETA AFTERSHOCKS: DETERMINATION OF STRESS DROP pair 20/24 tc1 = 3 Hz, fo2 = 16 Hz 1 20 30 CORNER FREQUENCY Fei 40 pair 58/60 tfc1 = 8 Hz, fo2 = 30 Hz 10 20 30 40 50 60 70 80 CORNER FREQUENCY Fai 90 N N 1 1 1 1 1 CL usual g g 1 50 100 D109 Figure 4.-Typical least-squares residuals as functions offc1 and f.. White represents the minimum residual, and black indicates five times the minimum residual. Variation is linear in logarithmic residual, giving 256 grey scales. (A) Fitting for pair 20/24. Low residuals are centered in a very narrow frequency range and give us well-resolved corner frequencies off =3 Hz and f, -l6 Hz. (B) Fitting for pair 58/60. The rCSIdual contour is not closed. The residual increases with increasing f.., but J. v changes less with increasing £.. An artifact limit, 30 Hz, is assumed forf and thenf] =8 Hz is resolved in this case. AFTERSHOCKS AND POSTSEISMIC EFFECTS D110 "sured juaAaa zQJ| asoy} UI poAjOAULT are Jo [E01 v 'red yoeo 10} 1ou109 1amof ay) ut umoys are sarouanbay} Jour0a om) ay} Jo sanfea ouf, qjuaAa Joffews ay} 1oj Xouanbayy sour jotjrure Ue sjuasaida1 more yuejq pure 'Aouanbayy sour0a e sjuasaidat more pJJO§ qyuaaa Joffews ay) Jof 1481 oy; pue juoaa ay) JoJ yoj oy) 'sorouanbayy Jou10s Jo uonisod ay} apeaitput smoury 'juoAa Joffews oy) 10J soqunu 131 pur juoAa Jo3.e| oy) JoJ soquinu yoj YJIM | 3iqe) ut soquinu juaaa ay} Aq poreotput st red yoeq '(aur; pjog) sopow arenbs-esauro jo one1 ay} 0) 1; Sumoys '(oulf pijos) sired juaaa z1| 10} Aouanbay; yo uonoun} e se ones penoadg-*g omslq ZH 'ADN3NOIJY4A 901 2 . 0 2 . 0 2 L 0 2 L 0 2 . 0 0 pm- 0 gomo--t-mmmtt-| 0 ZH O6 = 294 { ZH Of = 294 ?HP2=294 - HTL { ZH 06 = 294 ZH TL = 19 ZH IL = 19} ZH LL = 194 7 ZH p = 19} 7 ZH E1 = 19} 1 _| immnninen «- O v- 1 _I Sp/61 | | . 12/91 Luu ad lad L8/c L Luu -- 1 T Ap to d 08/9 o |mmmininan o gmmmmt-r--pmmmror-t 0 0 ZH O6 = 29} ZH 06 = 294 ZH 06 = 294 ZH €1 = 19} ZH LL = 19} ZH OL = 194 c-=t> moder- i_ uuu f a in ou «- 0 0 L7 ZH SL = 291 ZH 12 = 294 ] 7H Z = 19} 7 ZH E1 = 19} 0 L L a 1 1 lua 3 I | 1 4 e t | 28/81 3 ¢ SL/ZLL 8 SOL/LL _ 32 SS/G 3 ¢ L/6t 2 FTT -T ym -T- 0 fmTTTT-T-T prmmToT-T i grmTT-T TT-T a frt [Fr T -T L= L= zH og =2op - Hos =2> ] H EE =2o ] rH 92=2o ] zH 08 = 294 ] ZH GL= 194 1 IHP =19p 3 H 3 ZH OL = 194 3 ZH 9; = 19op 3 7 3 o 3 0 3 o 3 0 1. - - : - { 3 | 3 | 3 | 3 | 1 4 1 | po 3 t O4 t 4 1 4 GL/91 3 ? PL/9Q9ZL - #: 8/6 3 : /G 3 ¢ L/St 3 ¢ [rP TOTO ~::__~ T 0 TTT OT OT ______~_ I D T TOT frr r T O . 3414344111 0 zH og =294 - zH 06 =294 - ZH O8 {4 ZH pL = 204 - E2 m ZH ZL = 19} m ZH 91 = lo} m ZH p = 19} o m i. i, i, 1. | | | + | t 4 3 1 4 E 1 3 h 3 L9/9L _ 4: S161 - # : L/Q0S 3 : OL/E 2 LPP 3 : OILV¥H TVHLOT4S 9071 D111 EMPIRICAL GREEN'S FUNCTION STUDY OF LOMA PRIETA AFTERSHOCKS: DETERMINATION OF STRESS DROP ZH 'ADN3NOIJH4 9017 panunuoy-'g amnsiq 2 a 0 2 . 0 f . 0 2 . 0 2 u 0 -T --mmTT TOT -- 0 MTTTT TOT [TTT T T -T 0 OT OT rT T TOT -T [TTT T OT frmo - ?H ZL =294 - ZH 06 = 294 - ZH E61 =294 - ZH 61 =294 - ZH 12 = 294 ZH E61 = 19 7 ?H 2k = 19 7 ZH $ = 193 7 ZH 8 = 19} 7 = 19} F 3 3 i 3 1 3 E pe/os _ 2 CO/PO0L _ =: So/lo6 - 2: Cco/lto - > mwBN fTrrrT-T-T frmmrT-r-t pitt L= frmmmTT-T imm -T 0 fmm -T H - ?He6 =294 - ZH LL =294 - ZH 81 =294 - ZH Lp = 29} ZH OL = 194 7 ?H6= 19} 7 2H Z= 194 7 ZH LL = 194 7 TH S = 19} \J\<§U// I 4 ° 2x); 1 ' 4 E - = 66/16 _ 3: Bo/te - 2: _ > vam rrr -T fmTTrT-T-T 0 itm -T 0 fTTrrT-T-T mmmrT-r-t ZH 06 = 294 - ZH 66 = 294 - ZH 61 =294 - 2H P2 =294 - ZH 6 = 294 m 2H ZL = 194 7 ZH EL = 194 7 ?H 8 = 19} 7 ZH ZL = 19} 7 THE = 194 3 3 3 3 3 2 \?{Vir - j 1" |1' | . PE/St - = pE/62 i: - i: _ = €6/100 - & frrtrT-r-T -T rrrrT-Tr-T a TTrTT-T T a -A rmT-T- ZH EL =294 - ZH £2 = 294 CL = 29 ZH PL = 291 ZH 91 = 294 - L= 19} 7 ZH 6 = 19} 2 19 ZH8= 194 | ?H E6 = 194 7 3 2 2 - 3 & & _._|:—-_ TOT —::__ TOT O —::—_ T O4 _::__ T T F freer T I T ______— T 1 O 141.4 TOT ____:_ T T O _:____ TOT _:___fl_ I H - ZH 61 =294 - ZH Of = 294 - IH O8 =291 { ?HO=204 4 ZH P= 19} | THP = 194 3 ZH 91 = 19} 3 ZH OL = 194 J ZH ZL = 194 ] {j | ]" 1 1* f 5 99/06 - 2: eo Ace LL/L¥V 3 2 CP/LOL _ 2 : L6/LOL _ 2 —::___ I _::___ T F ___:__ T OT _____4_ T 4 O _._|:_~—_ T _::__ TOT O ____:_ TOT _______ T T O jfljflq. T —::__ T T ZH OL =294 - H 12 - ZH O6 =294 { ?HB82=294 { ZH 16 = 294 - p=193 ] ZH 8 = 194 7 ZH SL= 19 7 ?H 9 = 19} 7 ZH 21 = 194 7 | Eff | | | fa m 99/78 3 8 pPG/LL - 3: _ 2 ? C8/LP _ =: 96/PlLL _ > j—«aq- TOT fi:___ I I O _i:_— T T ~::__ TOT O _::__ TOT _::- T I P —_‘:~—_ T T —::__ TOT r .:____ TOT __._:_ I T IH 12=204 - ZH 60 = 29} TH Of = 294 - ZH Of = 29} ZH O6 =294 - 2H 9 = 19} 7 ZH 6 = 19} ; 2H ZL = 194 7 ZH SL = 194 _| ZH 91. = 19 7 EIR 7 3 ; - 3 g/ _ 3 { 7 { - 3: £€9/09 £ - 3s OP/ZLOL _ 2 : SE/6H - > OILVH 901 D113 EMPIRICAL GREEN'S FUNCTION STUDY OF LOMA PRIETA AFTERSHOCKS: DETERMINATION OF STRESS DROP ZH ©0171 panunuo)y-'¢ amsiy C . 0 2 L 0 2 L 0 2 . 0 2 0 fer TUT [EPP TOP N T OCO rer TG O T TOT fpr T F PPT TOT peed T T O [rT T TOO _::__ T H 02 =294 - ZH Of = 294 m ZH €1 = 294 HOE =24 - ZH 92 = 294 ZH $ = 194 7 ZH / = 19} W 1 ZH 8 = 194 | ZHZL = 194 7 ZH 9 = 19} EC 7 { 3 1 _ 3 ; | _ 08/00L _ a» 6L/LL - As 69/00L _ 3 z 99/20 3 z €L/09 __:____ T _::__ I I O _::___ I _:_:__ T O ________ T fl_______ T F ________ T —::_flfl T F. _______— T d::—_‘fi T TH 16 = 294 - ZH Of = 294 - { ZH Of = 29} m ZH Of = 294 ?H6= 194 7 ZH 91 = L9} u ?H 9 = 194 7 ZH §. = L9} m o ZH OL = 19} 3 ; 15 3 1 3 z 1 _ _ | E 3 Q » E » 3 ¢ m Q 08/66 3 z 0Z/O6 3 z 19/001" a _ i: 09/69 ___:__ T T ____:_ TOT O —::___ T _::__ T T O ___:___ T _::__ T T O 22... T T _:____ I T O _::-— T _______4 T ZH £0 = 29 m ZH 66 = 29} m ZH 02 = 29} ZH E1 =204 - ZH Of = 29 ZH OL = 19}; 3 ZH CL= 3 ZH CL =19J | ?H S = 194 7 ZH 8 = 19} 43 | 4 | i 3 3 { 4 1 x5 43 z 3 z | 7 ol/bl. 3 : €01L/69 _ as: 19/66 1. - =: 88m _::__ T T _::_~ TOT O _______ T T _______ T I O _:_:~ TOT T—Z—q TOT w: _::__ T -T _::__ TOT O _::~— T T _::__ T ZH 06 = 29} ZH Of = 294 ZH 02 = 294 ZH 20 = 29} ZH Of = 29 ZH OL = 19} 1 ZH LL = 19} ZH Z = 19} o ZH 8 = 19} ZH L = 19} . | L 2 L i. 86/ LZ 6 LO L/69 2 19/69 2 b9// ¢ z 32mm & OlLYH 71¥H19348 901 u AFTERSHOCKS AND POSTSEISMIC EFFECTS D114 ZH _>OZm_DOmEn_ ©9071 panunuo}y-'g amily z T o 2 L o 2 o z 0 2 ( o L- grmmmmm-t--pmmmm-t 0 prrmmmT-t ___:__ T-t 0 . zH 92 = 2o1 mHos=zo ] mH 8 = 20 ZH $1 = 204 ZH pL = 20 mo=io 3 | mrl= io 3 | TH 8= 19 2H9= 19; 194 | 3 L 2 { 1 3 ¢ | LLL/9L } 2 90L/Z0L #s mafia 2 601/26 C 68/06 ~ gmmmTt-t-pmmmmT- 0 i gimm-T-t--pmmmr-t a 0 -- ZH be = gol mss=zop ] zH of = 291 ZH 21 = 291 mel =2op - ZH 6 = 19} top 3 | ZH 91 = 19} 2H /= 19} ZH p= dof - L f w 0 L 3 3 | | Prim”; 2 cO0L/0OdL i: SLL/tP6 } 80L/26 2 88/60L - = [PET TT TOT w::x_‘_ TOT O EPE TOT T __ZJfl— TCT O [PEEP T ET T TOT L (EPT OT T [PEET OT T D [FEET TCT T T:_—— T T zH ge = gol zH g2 = gol ZH pL = go) ZH 62 = go} TH 91 = go) TH 6 = 19} ZH LL = Lop TH 9= 19 ZH P1 = 19} 2H 9 = 19} 7 1 2 . g . ©€01/66 2 C0L/€66 £ O LL/L6 2 98/26 __:__. T T ___:_aq‘ T o —_:.___ I _|_}_\_w_qfi_ I F _——__—_~ T _:_—-— T O —_~_____ T f:_—__ T F _xqalflfla. I T _______ I I H 12 = go} 2H 51 = gop TH 81 = go zH 02 = gop 2H 82 = 201 ZH 21 = los ZH 2 = 19} TH 2 = 19} 2H 8 = 19} ZH 8 = 19} 1 1 2 PolL/ZLO} 2 66/00 L 2 801/66 2 68/00 L C 18/98 OILYVH T1VHLOT4S 901 EMPIRICAL GREEN'S FUNCTION STUDY OF LOMA PRIETA AFTERSHOCKS: DETERMINATION OF STRESS DROP CALCULATION OF STRESS DROP The determination of stress drop for a small earthquake is generally based on a source determination derived from either the corner frequency of the displacement spectrum or the pulse width of the waveform. We model a small earthquake as a circular rupture and estimate the stress drop from the corner frequency, {.;, using the formula (Brune, 1970; Brune, 1971; Kanamori and Anderson, 1975): Ao=M,,(f /0.49v,])} where the rupture velocity v, is taken to be 3.5 km/s (Dahlen, 1974; Xie and others, 1991; Hough and others, 1991). The seismic moment, M0, is calculated for each event (table 1) from the long-period displacement spectral lev- els using M,=(4r0 proMFR oy) where a is the P-wave velocity, p is the density, r is the hypocentral distance, 2, is the long period spectral level, F is the free surface correction, and R,,, is the average correction for radiation pattern. We use =6 km/s, p=2.7 gm/cm3, F=2, and R ¢W=O.7. We also calculated moment taking into account exact radiation terms as determined from the focal mechanisms, but found these estimates to be less consistent (among moment estimates for the same event at different stations) than those obtained with a con- stant radiation term. The corner frequency is shown as a function of moment in figure 6. STRESS-DROP RESULTS The Brune-model stress drops are found to be highly variable, ranging from approximately 1 to nearly 800 bars N _C sd ~ 3 --* i 3 \ es y *~ Z r- OU LL | sam Y- t n foud n r- - al ~ T 0 n T 3 r- a 9 - OC 5 3 8 B i sy | -~ se | ( § inane mnie a paco 4 o (l) vlw OILYVH TVHLOTd4S 901 Gg D115 (see table 1 and figure 7). The dashed lines on figure 7 indicate the resolution limits corresponding to the fre- quency range over which we can generally resolve corner frequencies, 1 to 40 Hz. It appears that the apparent upper trend in figure 7 may in fact be an artifact of the high frequency resolution limit. However, the methodology and data should be sufficient to resolve low-stress-drop larger events, if they existed. A dependence of stress drop on moment can be inter- preted in terms of a critical slip distance resulting from fundamental characteristics of established rate and state variable friction laws (for example, Dieterich, 1986). The critical slip distance to achieve dynamic rupture corre- sponds to a minimum earthquake rupture dimension (for example, Dieterich, 1986; Scholz, 1990), thus leading to a breakdown in the similarity of earthquake rupture pro- cesses. It is also possible to explain the inferred stress drop scaling in terms of tectonic factors. Global compilations (Abercrombie and Leary, 1993) show no resolved system- atic scaling of stress drop with moment, implying that globally, there is no preferred earthquake rupture dimen- sion (in other words, breakdown in similarity). However, within a limited tectonic region, it is plausible that a char- acteristic length scale will exist. If, for example, after- shocks in the Loma Prieta segment are interpreted as activation of cross faults within the fault zone (L. Seeber, 1994, unpub. data), then their rupture dimensions may preferentially correspond to the width of the fault zone. CONCLUSIONS AND DISCUSSION It has been demonstrated that eGf analysis is a quick and effective method for estimating the source character- istics of a large number of small- to medium-sized earth- quakes with good resolution. The frequency-domain deconvolution removes the effects of travel path, site, and the instrument and isolates the source characteristics of the two events. The Brune spectral ratio model appears sufficient to parameterize the spectral ratio, provided that the differences in magnitude are not too small and the ambient noise at high frequencies is not too large. It is important, moreover, to specifically address the proper- ties of the fitting residual as a function of the gridding variables. The estimate of the large-event corner frequency can be biased by improper selection of the small-event corner frequency, and so it is important to grid addition- ally over small-event corner frequency. The grid search- ing yields a well-resolved estimate of the corner frequency of the large event and sometimes of the smaller event as well. The resolution of the corner frequency of the larger event will be much better than that of a single spectrum parameterization of source and path effects if we properly D116 Table 1.-The 124 events used for empirical Green's function analysis [The 124 events involved in the 102 event pairs shown in figure 5. Of these, 98 event pairs have resolved corner frequencies, with 53 from the larger events (indicated in column L). The other 26 events are generally smaller events with magnitudes between 1.5 and 2.0. Their corner frequencies are not well-resolved from the ratio fitting and are excluded from the corner frequency averaging and the calculation of stress drops.] AFTERSHOCKS AND POSTSEISMIC EFFECTS Event Date Time Long. Lat. Depth ML Moment fc Stress L (yymmdd ) (Gmt) (W) (N) (Km) (dyne*cm) (Hz) (bar) 1 891020 0406: 04.39 121.957 37.177 6.38 1.7 5.66e+18 21.0 10.39 2 891020 0513 : 51.66 121.803 37.060 4 . 62 2.3 1.58e+19 3 891020 0617 :24.49 121.973 37.174 3 . 54 2.6 7.94e+19 4 . 0 1.01 X 4 891020 0646: 07.10 121.806 37.076 9 . 08 1.7 6.2le+18 26.0 21.64 5 891020 0752: 49.06 121.809 37.076 9 . 42 2.7 - 4.05e+19 8 .5 4.93 X 6 891020 0812 :54.07 122.070 37.181 14.13 3.6 - 1.95e+21 4.0 24.74 X 7 891020 0926: 16.93 122.070 37.190 13.47 1.6 1.03e+19 8 891020 1144: 00.45 121.919 37.049 16.92 1.8 3.4le+18 33.0 24.29 9 891020 1153 :29.14 121.921 37.048 17.57 2.7 - 5.63e+19 8.0 5.71 X 10 891020 1348 :28.20 121.974 37.178 2.98 1.7 3.55e+18 14.0 1.93 11 891021 0011:11.27 121.975 37.108 15.11 2.5 8.69%e+19 10.0 17.23 X 12 891021 0102: 45.64 121.851 37.053 12.45 2.6 8.75e+19 11.0 23.09 X 13 891021 0257 :51.95 121.863 37.103 4.23 1.5 1.66e+19 14 891021 0320 : 24.43 121.640 36.924 5.99 2.4 3.98e+19 30.0 213.09 15 891021 0459: 28.71 121.992 37.171 7.86 1.7 1.99e+19 30.0 106.52 16 891021 0832: 19.49 122.000 37.121 14.96 2.7 4.76e+20 13.7 242.65 X 17 891021 1057 : 04.52 121.987 37.165 8.41 2.6 3.22e+20 17.0 313.62 X 18 891021 1452 :24.27 121.881 37.105 3.59 2.5 2.62e+20 13.0 114.11 X 19 891021 1558 :45.58 121.860 37.105 4.56 2.6 2.5le+20 12.0 85.99 X 20 891021 2214 :56.52 121.905 37.061 15.27 4.4 7.08e+21 3 .8 77.02 X 21 891021 2311:19.86 121.957 37.177 6.09 1.7 1.28e+19 18.7 16.59 22 891021 2329: 34.16 121.900 37.065 14.36 1.8 1.34e+19 19.0 18.22 23 891022 0031: 16.48 121.900 37.066 14.41 2.5 2.47e+20 7.0 16.80 X 24 891022 0254: 29.59 121.904 37.063 14.36 2.4 6.58e+19 11.5 19.84 X 25 891022 0338 :31.35 121.809 37.077 8.79 2.1 4.54e+19 13.0 19.77 26 891022 0341: 09.58 121.842 37.090 4.96 1.6 5.25e+18 23.0 12.66 27 891022 0433 : 23.13 122.001 37.126 14.36 1.7 3.00e+19 28 891022 0641:51.31 121.897 37.067 14.17 1.6 1.04e+19 11.5 3.14 29 891022 0800 : 53.26 121.959 37.177 6.02 2.3 1.73e+20 13.0 75.35 X 30 891022 0817: 08.47 122.074 37.183 13.65 2.6 2.23e+20 14.0 121.31 X 31 891022 1215: 09.50 121.948 37.059 17.21 2.4 1.83e+19 9 .0 2.64 X 32 891022 1351: 03.20 121.932 37.066 15.19 1.5 3.18e+19 33 891022 1627: 18.20 121.903 37.063 14.60 2.3 1.1l16e+20 9 .0 16.76 34 891022 1710: 32.43 121.961 37.180 6.57 1.5 2.00e+19 23.0 48.24 35 891022 1720: 52.24 121.801 37.051 4.15 1.6 - 3.75e+19 36 891022 1938: 54.69 121.903 37.062 14.15 1.5 6.42e+18 31.0 37.92 37 891022 2148 :31.54 121.667 36.944 10.97 2.3 1.46e+20 20.0 231.55 38 891023 0347: 04.23 121.838 37.111 4.47 1.9 1.07e+19 17.0 10.42 39 891023 0541: 50.90 121.941 37.059 17.57 1.7 4.6le+19 32.0 299.47 40 - 891023 0552 : 52.55 121.666 36.945 10.97 1.8 - 6.60e+19 41 891023 1514 :22.93 121.655 37.005 2.71 3.0 - 6.0 11.65 42 891023 1805:59.85 121.665 36.941 10.05 1.6 - 1.85e+19 43 891024 0448: 11.99 121.803 37.056 4.97 3.2 4.79e+20 14.5 44 891024 0702 :23.52 121.959 37.179 6.14 2.8 5.33e+20 ~ 45 891024 0856: 12.93 121.959 37.179 6.04 2.5 1.86e+"°° 46 891024 1825: 32.76 121.668 36.945 11.07 2.1 1.07e1 Table 1.-Continued EMPIRICAL GREEN'S FUNCTION STUDY OF LOMA PRIETA AFTERSHOCKS: DETERMINATION OF STRESS DROP D117 Event Date Time Long. Lat. Depth ML Moment fc Stress L (yymmdd) (Gmt) (W) (N) (Km) (dyne*em) (Hz) (bar) 47 891024 1930 : 37.71 121.653 36.935 6.71 2.4 8.95e+19 16.0 72.68 X 48 891024 2213 :10 .54 121.865 37.104 4 . 46 1.5 9.29e+18 49 891024 2226:13.11 121.962 37.179 5.76 2.5 1.6le+20 11.3 46.05 X 50 891024 2342 :16.15 121.964 37.175 5.94 2.3 2.82e+19 13.0 12.28 X 51 891025 0051 : 33.46 121.670 36.946 10.37 1.7 - 6.10e+18 52 891025 0127 :26.35 121.820 37.081 10.20 4.2 - 5.50e+21 9.0 794.87 Xx 53 891025 0538 : 42.96 121.825 37.085 10.42 2.8 2.46e+20 23.0 593.37 54 891025 0915: 14.31 121.708 36.966 13.99 1.6 1.57e+19 21.0 28.82 55 891025 0923 : 31.94 121.806 37.075 9.16 1.7 1.67e+19 16.0 13.56 56 891025 1314 :16.26 121.809 37.075 9 .0 6 1.9 4.96e+19 12.5 19.21 57 891025 1508: 33.00 121.882 36.903 4.31 1.9 3.64e+19 21.0 66.83 58 891025 1510 : 54.73 121.881 36.904 4.13 2.6 2.12e+20 7.0 14.42 X 59 891025 1630 :51.19 121.882 36.903 4 .10 2.8 3.12e+20 7.5 26.09 X 60 891025 1808: 58.44 121.880 36.903 4.10 1.6 - 7.82e+19 61 891025 1926: 36.98 122.001 37.120 14.74 2.0 - 4.23e+19 62 891025 2201:49 .43 121.797 37.004 17.13 3.7 3 6.0 144.74 X 63 891025 2244: 41.96 121.883 36.902 4 .21 2.6 2.38e+20 10.0 47.18 X 64 891026 0459 : 19.07 121.808 37.075 9.11 2.0 4.22e+19 17.5 44 . 84 65 891026 0727 :25.69 121.642 36.920 6.91 1.8 1.10e+19 66 891026 0840 : 59.66 121.663 36.942 10.12 1.7 - 3.55e+18 67 891026 1303 :51.32 121.794 37.068 8.76 1.7 1.94e+19 22.7 44.99 68 891026 1519 :24.88 121.803 37.056 4.57 1.7 7.04e+18 28.0 30.64 69 891026 1613 :25.46 121.793 37.066 9.47 2.9 1.74e+20 10.8 43.45 X 70 891026 1715: 11.20 121.904 37.120 4 . 48 1.6 1.10e+19 71 - 891027 1329 :51.30 121.710 36.969 14.52 3.0 - 2.16e+20 8 .3 24.48 X 72 891027 1540 : 03.73 121.754 36.998 15.15 1.6 2.5le+l18 23.0 6.06 73 891027 2206:21.22 121.804 37.002 16.61 1.6 2.79e+19 26.0 97.21 74 891027 2215 :06.83 121.758 37.005 14.84 2.9 2.24e+20 10.0 44.38 X 75 891028 0228 : 00 . 68 122.000 37.125 14.48 1.7 - 3.13e+19 76 891028 2127 :49.47 121.636 36.917 6.22 3.3 - 8.91e+20 4.0 11.31 X 77 891029 0133 : 20.15 121.654 36.931 6.75 1.5 3.28e+19 78 891029 0315: 17.60 121.903 37.065 13.82 1.8 2.43e+19 19.0 33.04 79 891029 0347: 37.49 121.705 36.969 13.73 1.8 2.71l1e+19 80 891029 0602: 04 . 27 121.793 37.069 8.94 1.5 2.56e+19 25.5 84.15 81 891029 1107 :40.27 121.852 37.054 11.97 1.5 1.62e+19 24.0 44 . 40 82 891029 1441 :32.57 121.882 37.107 3 . 58 1.6 1.34e+19 83 891030 0006: 20.04 121.655 37.000 2.16 1.7 1.74e+19 28.0 75.72 84 891030 0216 :06.12 121.910 37.056 15.10 1.6 1.15e+19 37.5 120.23 85 891030 0452 : 24.59 121.902 37.057 15.25 2.5 1.35e+20 14.5 81.59 x 86 891030 0946: 34.71 121.808 37.075 9.16 1.7 - 3.55e+18 15.0 2.37 87 891030 1117 :13.54 121.807 37.070 9 . 66 3.6 3 .47e+21 5.3 102.42 Xx 88 891030 1142 :36.59 121.925 37.063 14.47 1.8 3.63e+19 13.0 15.81 89 891030 1353 : 59.86 121.804 37.075 9.34 1.5 2.69%e+19 17.0 26.20 90 891030 1541 121.907 37.118 3.78 2.4 9.08e+19 16.0 73.73 X 91 891031 0650 : 07.51 121.996 37.135 12.62 2.2 1.30e+20 14.0 70.72 X 92 891031 0834: 48.82 121.804 37.071 8.93 3.3 7.23e+20 5.4 22.57 X 93 891031 0834: 51.34 121.803 37.069 8.55 3.3 - 7.23e+20 6.3 35.84 Xx 94 891031 1729 :39.36 121.745 36.996 15.19 2.4 1.2l1e+20 16.0 98.25 Xx 95 891101 0323 : 30.97 121.809 37.062 4.95 1.6 1.00e+19 96 891101 0616: 03.79 121.836 7.104 4 . 63 2.4 1.17e+20 8.0 11.88 97 891101 0803 : 17.43 121.827 37.105 4 . 62 3.7 - 1.37e+21 4.5 24.75 X 98 891102 0505 : 37.90 121.715 36.976 14.26 1.7 - 3.77e+19 99 891102 0512 : 34.29 121.796 37.069 8.93 2.5 3.91le+20 11.8 127.36 X D118 Table 1.-Continued AFTERSHOCKS AND POSTSEISMIC EFFECTS Event Date Time Long. Lat. Depth ML Moment fc Stress L (yymmdd ) (Gmt) (W) (N) (Km) (dyne*cm) (Hz) (bar) 100 891102 0550 : 10 . 87 121.800 37.069 9 . 08 4 .3 1.33e+22 6.6 758.04 X 101 891102 0636: 45.27 121.787 37.069 9 .35 1.6 1.96e+19 102 891102 0737 :19.50 121.792 37.070 7.40 1.5 1.63e+19 33.0 116.13 103 891102 0801 :57.17 121.791 37.067 8 . 62 1.5 1.12e+19 28.0 48.74 104 891102 1011 121.930 37.064 15.40 2.3 2.29e+20 12.0 78.45 Xx 105 891102 2019 :20.05 121.974 37.107 15.13 1.5 1.46e+19 106 891103 1014 :01.24 121.664 36.944 10.11 1.7 2.17e+19 107 891103 1047: 56.48 121.666 36.944 10.65, 3.1 1.65e+21 13.0 718.66 X 108 891104 1613 :26.32 121.806 37.070 9.19 1.9 4.03e+19 17.5 42.82 109 891105 0130: 41.94 121.921 37.065 14.31 3.8 2.19e+21 4.0 27.79 X 110 891105 1406: 39.68 121.997 37.129 12.90 1.5 1.35e+19 29.0 65.27 111 891106 0750 : 16.88 121.887 37.114 4.31: 2.0 3.03e+19 23.5 77.96 112 891106 0827: 05.01 121.887 37.112 4.16 2.9 2.47e+20 9 .0 35.70 X 113 891106 0827 :15.28 121.889 37.113 3.55 2.9 2.47e+20 9 .0 35.70 _ X 114 891106 1548: 18.42 121.907 37.064 14.42 2.4 1.94e+20 12.0 66.46 X 115 891111 2116 121.741 36.996 14.56 1.6 1.83e+19 116 891114 2041 121.838 37.089 5.27 2.8 1.13e+20 6.0 4.84 X 117 891114 2053 : 10.33 121.840 37.090 5.33 1.5 1.38e+19 26.0 48.08 118 891114 2116 :42.70 121.840 37.089 5.53 3.4 5.23e+20 9.0 75.59 X 119 891117 1932: 18.88 121.787 37.064 8.12 2.6 7.94e+19 6.0 3.40 X 120 891117 1935: 19.16 121.789 37.066 8.13 2.3 5.18e+19 20.0 82.15 X 121 891117 2116 :59.77 121.790 37.069 7.43 1.5 1.78e+18 34.0 13.86 122 891120 1739 :57.81 121.648 36.923 7.23 2.8 7.4le+19 15.0 49.58 X 123 891122 1213 :52.32 121.805 37.075 9.16 1.6 7.63e+18 14.0 4.15 124 891122 1757 :11.80 121.666 36.944 10.13 2.0 1.38e+19 21.0 25.34 1000 50 T -T 11|u|| -T- [- ut LU $14” 1d C; l ‘&$IHIC)I N - C A jz, AA 7 E - /'IA O8 > A mo - / - A A O 20 } a * li F 0 o a A24 pan 0 G = 100 } / o 0 3 A, mace 9 o & - AA‘QLOOMO o O |_ - Oo - Aim O 0 W 1° E ° o, 4 &. 9 1 o C fl A" B % LL |- ap ° o 1 O _,'AA 000009 - 00 o O 8 - W / A A CC = 5 | O 1 -C a 4 o & u 0 - i0 L0 A a A O _ o Z - o 0 © op _ 1 - : C_ _L - F A o : Q 8 - a _ O° 0 O , | 3 - 4G _ 9 ¢ + _] 4 1 4 4 Luu] llquy 1 ta iu] llluflul 1 4 Liu] L OOO 0 a 0p aul 4 18 19 20 21 2p 18 19 20 21 22 LOG MOMENT, DYNE*CM Figure 6.-Corner frequency as a function of seismic moment for 98 Loma Prieta aftershocks. Circles indicate the larger events and triangles indicate the smaller events. limits corresponding to fixed corner frequencies of 1 Hz and 40 Hz. LOG MOMENT, DYNE*CM Figure 7.-Brune-model stress drop as a function of seismic moment for 98 Loma Prieta aftershocks. Circles indicate the larger events and tri- angles indicate the smaller events. The dashed lines are the resolution EMPIRICAL GREEN'S FUNCTION STUDY OF LOMA PRIETA AFTERSHOCKS: DETERMINATION OF STRESS DROP choose the event pair. Although the empirical Green's function method is limited to available event pairs, we have shown that, within a dense aftershock sequence, it can be used for a large-scale investigation of source prop- erties. In this study, we obtain a total of 98 well-resolved corner frequencies for events with magnitudes between 1.5 and 4.4. ACKNOWLEDGMENTS We thank the many people who contributed to the suc- cessful deployment of the IRIS/PASSCAL network, in- cluding David Simpson, Larry Shengold, Robert Busby, and James Fowler. We thank Professor Lynn Sykes for providing helpful comments and suggestions. We thank Leonardo Seeber and John Armbruster for providing relo- cation data. We also thank David Oppenheimer for sup- plying CALNET phase data used for relocation. The research was supported by U.S. Geological Survey under Grant 14-08-001-G-1831, and the aftershock data collec- tion was supported by the Incorporate Research Institu- tions for Seismology and by the National Science Foundation. The PASSCAL aftershock data are available from the IRIS Data Management Center or from the au- thors at Lamont-Doherty. This is Lamont-Doherty contri- bution number 5479. REFERENCES CITED Abercrombie, L. and P.C. Leary, 1993, Source parameters of small earthquakes recorded at 2.5 km depth, Cajon Pass, California: Im- plications for earthquake scaling: Geophysical Research Letters 20, p. 1511-1514. Brune, J.N., 1970, Tectonic stress and the seismic shear waves from earthquakes: Journal of Geophysical Research, v. 75, no. 26, p. 4997-5009. Brune, J.N., 1971, Correction: Journal of Geophysical Research, v. 76, no. 20, p. 5002. Boatwright, J., Fletcher, J.B., and Fumal, TE., 1991, A general inver- sion scheme for source, site, and propagation characteristics using multiply recorded sets of moderate-sized earthquakes: Seismologi- cal Society of America Bulletin, v. 81, no. 5, p. 1754-1782. Dahlen, A., 1974, On the ratio of P-wave to S-wave corner frequencies for shallow earthquake source: Seismological Society of America Bulletin, v. 64, no. 4, p. 1159-1180. Dieterich, J., 1986, State variable fault constitutive relations for dy- namic slip, in Earthquake Source Mechanics: American Geophysi- cal Union Geophysical Monograph 37, p. 311-318. Frankel, A., Fletcher, J., Vernon, F., Haar, L., Berger, J., Hanks, T., and Brune, J., 1986, Rupture characteristics and tomographic D119 source imaging of ML~3 earthquakes near Anza, Southern Cali- fornia: Journal of Geophysical Research, v. 91, no B12, p. 12633- 12650. Fletcher, J.B., and Boatwright, J., 1991, Source parameters of Loma Prieta aftershocks and wave propagation characteristics along the San Francisco peninsula from a joint inversion of digital seismo- grams: Seismological Society of America Bulletin, v. 81, no. 5, p. 1783-1812. Geller, R.J., and Mueller, C.S., 1980, Four similar earthquakes in cen- tral California: Geophysical Research Letters, v. 7, no. 10, p. 821- 824. Hartzell, S.H., 1978, Earthquake aftershocks as Green's functions: Geophysical Research Letters, v. 5, no. 1, p. 1-4. Hough, S.E., and Anderson, J.G., 1988, High frequency spectra ob- served at Anza, California: Implications for Q structure: Seismo- logical Society of America Bulletin, v. 78, no. 2, p. 692-707. Hough, S.E., Seeber, L., Lerner-Lam, A., Armbruster, J.G., and Guo, H., 1991, Empirical Green's function analysis of Loma Prieta af- tershocks: Seismological Society of America Bulletin, v. 81, no. 5, p. 1737-1753. Hutchings, L., and Wu, F., 1990, Empirical Green's functions from small earthquakes: a waveform study of locally recorded after- shocks of the 1971 San Fernando earthquake: Journal of Geo- physical Research, v. 95, no. B2, p. 1187-1214. Kanamori, H., and Anderson, D.L., 1975, Theoretical basis of some empirical relations in seismology: Seismological Society of America Bulletin, v. 65, no. 5, p. 1073-1095. Li, Y., and Thurber, CH., 1988, Source properties of two mi- croearthquakes in Kilauea volcano, Hawaii: Seismological Society of America Bulletin, v. 78, no. 3, p. 1123-1132. Menke, W., Shengold, S.L., Guo, H., Hu, G., and Lerner-Lam, A., 1991, Performance of the short-period geophones of the IRIS/ PASSCAL array: Seismological Society of America Bulletin, v. 81, no. 1, p. 232-242. Molnar, P., Tucker, B.E., and Brune, J.N., 1973, Corner frequencies of P and S waves and models of earthquake sources: Seismological Society of America Bulletin, v. 63, no. 6, p. 2091-2104. Mori, J., and Frankel, A., 1990, Source parameters for small events associated with the 1986 North Palm Springs, California, earth- quake determined using empirical Green's functions: Seismologi- cal Society of America Bulletin, v. 80, no. 2, p. 278-295. Mueller, C.S., 1985, Source pulse enhancement by deconvolution of empirical Green's function: Geophysical Research Lettters, v. 12, no. 1, p. 33-36. Park, J., Lindberg, C.R., and Vernon, FL., 1987, Multitaper spectral analysis of high-frequency seismograms: Journal of Geophysical Research, v. 92, no. B12, p. 12,675-12,684. Pujol, J., 1992, Joint hypocentral location in media with lateral veloc- ity variations and interpretation of station correction: Physics of the Earth and Planetary Interiors, v. 75, no. (1-3), p. 7-24. Scholz, C.H., 1990, The mechanics of earthquakes and faulting: Cam- bridge, Massachusetts, Cambridge University Press, 441 p. Seeber, L. and Armbruster, J.G., 1990, Fault kinematics in the 1989 Loma Prieta rupture area during 20 years before that event: Geo- physical Research Letters, v. 17, no. 9, p. 1425-1428. Xie, J., Liu, Z., Herrmann, R.B., and Cranswick, E.D., 1991, Source processes of three aftershocks of the 1983 Goodnow, New York, earthquake: high-resolution images of small, symmetric ruptures: Seismological Society of America Bulletin, v. 81, no. 3, p. 818-843. - ens , * - hit iss u hos tm f m : - 6 i - - - - - t R - - - & - R 5 \ - - $ a , - - - - is Baw i THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989; EARTHQUAKE OCCURRENCE AFTERSHOCKS AND POSTSEISMIC EFFECTS U.S. GEOLOGICAL SURVEY AFTERSHOCK GROUND-MOTION DATA By Leif Wennerberg, U.S. Geological Survey CONTENTS Page Abstract D121 Introduction 121 GEOS data 121 Large-length-scale seismic-wave-propagation studies ------ 122 Influence of local geology on ground shaking --------------- 122 Collocations with strong-motion instruments ---------------- 125 DR-200 data 125 Santa Cruz 126 Moss Landing POWeL PIANt 126 Los Gatos 126 Liquefaction sites 127 Topographic effects SUdies ------------------------.--..-..-.-.-- 127 Collocations with strong-motion instruments ---------------- 128 USGS seismic CaSSEtte I@COTG@rS 128 Acknowledgments 128 References 130 ABSTRACT Portable seismographs were deployed to record after- shocks of the Loma Prieta earthquake for studies of earth- quake source processes, effects of local geology on ground motions, and crustal and basin wave-propagation effects. This report provides maps and descriptions of specific experimental objectives of the aftershock deployments. Thirty-eight digital seismographs were deployed at 195 sites during the period from the day after the October 17, 1989, main shock until mid-January 1990. Aftershock re- cordings were obtained at or near 26 sites where the main shock was recorded by permanent strong-motion instru- ments. Fifteen aftershock recording instruments were at, or near, locations where previous USGS studies have been done; nuclear-explosion recordings and/or borehole geo- logic and velocity data are available for these sites. There was also a brief deployment of 60 USGS refraction-sur- vey analog instruments in the epicentral area on the night after the main shock. INTRODUCTION Aftershock sequences provide a reliable source of earth- quake ground motions which can be used to address is- sues relating to the physics of the seismogenic process and to address practical problems of strong-ground-mo- tion prediction, such as the relevence of weak-motion records to strong-ground motions or the effects of local geologic conditions on ground motion. Following the Loma Prieta earthquake, several groups from the USGS recorded aftershocks in an effort to further our understanding of earthquake hazards in the San Francisco Bay area. The digital data sets include a total of more than 5,000 three- component records from 195 sites in the greater San Fran- cisco Bay area, and have been collected on two CD-ROMs (see Mueller and Glassmoyer, 1990, and Carver and oth- ers, 1990, for data access). In this paper I summarize the USGS data-collection efforts and describe the experimental objectives. Certain recording sites could be used for more than one purpose, some of which may not have been considered, so the ob- jectives listed here cannot be exhaustive. As I discuss the experiments, I refer to any published results I know of that have used these data. I do not discuss the results of these studies. In the "References" section I also list a few analyses not referred to explicitly in the text. There is still a significant amount of work that could be done using these data. This discussion provides an overview and introduction to the data sets, and supplements Mueller and Glassmoyer's (1990) and Carver and others' (1990) presentations of the digital data. No report has been written specifically de- voted to the analog data set. Mueller and Glassmoyer's report (1990), while providing a comprehensive descrip- tion of a significant portion of the digital data set, pro- vides a much briefer discussion of experimental objectives than presented here. It contains no discussion of previous work or published analyses. My discussion of the data presented in Carver and others (1990) briefly summarizes their report and includes references to relevant work that has appeared since their report was published. D121 D122 For the purposes of this report, I divide the USGS af- tershock investigations into three parts corresponding to three instrumentation groups, one of which collected only a small and quite specialized data set, but which is in- cluded for the sake of completeness. The two main data- collection efforts were undertaken by a group using GEOS recorders (Borcherdt and others, 1985; Mueller and Glassmoyer, 1990) and a group using Sprengnether DR- 200 recorders (Carver and others, 1986; Carver and oth- ers, 1990; Cranswick and others, 1990). The data collected by the GEOS group are mostly from Santa Clara County and the San Francisco peninsula, with some data recorded in Monterey, San Benito, Santa Cruz, and southern Alameda Counties. These data are described in detail in Mueller and Glassmoyer (1990). The DR-200 data are mostly from Santa Cruz and Santa Clara Counties, with some records obtained in Monterey County and are de- scribed in Carver and others (1990) and King and others (1990). The GEOS records digital data with 16-bit resolu- tion. The DR-200 has a 12-bit digitizer, but is gain-ranged to give an effective 16-bit recording range. Both types of instruments sampled at 200 samples per second. Three- component velocity transducers with natural periods be- tween 1 and 2 Hz were used. Most GEOS recorders were also deployed with three-component force-balance accel- erometers. See Mueller and Glassmoyer (1990) and Carver and others (1990) for details on sensors. In this paper I discuss the GEOS experiments, present- ing station maps and a table of sites and experiments, and note related work done by the USGS. Then I discuss the DR-200 data, presenting maps and discussing experiments and selected sites. Finally, I briefly describe the refrac- tion-survey-instrument data set. GEOS DATA Twenty-seven GEOS recorders were deployed at 94 sites during the aftershock studies. 2,615 three-component ve- locity records were collected, and 2,592 accelerograms were collected from sensors collocated with velocity trans- ducers. Figures 1 through 8, taken from Mueller and Glassmoyer (1990), show the sites occupied by GEOS recorders. Table 1 lists the sites where data were obtained and indicates the reason(s) for each site's occupation. The main shock was recorded by permanent strong-motion in- struments at twenty-three GEOS sites. Mueller and Glassmoyer (1990) give more detailed information on site locations and characteristics, instrument parameters, and aftershocks recorded. The following is a somewhat ideal- ized description of the aftershock studies undertaken. Mueller and Glassmoyer (1990) give a clearer indication of how much data relevant to each study were actually obtained. AFTERSHOCKS AND POSTSEISMIC EFFECTS The topics of interest during the GEOS deployments can be placed in three categories: large-length-scale wave- propagation effects, local site effects, and collocations with strong-motion instruments. The large-scale studies were intended to clarify crustal propagation effects, or the pos- sibility of a distinct seismic response of the sedimentary basin of the south San Francisco Bay. Studies of local site effects were primarily motivated either by damage or by an interest in the shaking properties of common geologic units. These studies typically consisted either of instru- ments deployed on bay mud and/or alluvium sites with simultaneously recording instrument(s) "nearby" on one or more rock sites, or of dense arrays designed to study wave propagation and (or) the small-scale variability of ground motions. The collocations with strong-motion in- struments were intended to collect small-earthquake data that could be used for an empirical-Green's-function analy- sis of the main shock (for example, Hartzell, 1978), or for information on the relationship between small-earthquake records and large-earthquake ground motions. LARGE-LENGTH-SCALE SEISMIC-WAVE- PROPAGATION STUDIES The main large-scale deployment was a line of stations along the San Francisco peninsula from the aftershock zone to San Francisco. The stations are listed in table 1 under the experiment PP. The main interest of this de- ployment was crustal wave propagation, with a secondary interest in local site response. Results of analysis of the data are presented by Fletcher and Boatwright (1991) and Boatwright and Fletcher (1991a). McGarr and others (1991) also used some of these stations to analyze wave propagation up the peninsula. Two shorter transverse lines, oriented roughly southwest-northeast, were deployed in the south bay: a line from the southern peninsula to the southern east bay flatlands (from SRL on the west to FLB and POR on the east, fig. 4) and a line crossing the north- ern Santa Clara Valley (from BMT on the west to TUL on the east, fig. 4). The stations involved are listed in table 1 as PE and SV, respectively. These lines were deployed to observe possible basin response. INFLUENCE OF LOCAL GEOLOGY ON GROUND SHAKING Most of the aftershock studies focused on the effects of local geology on ground motion. Two particularly signifi- cant efforts were studies of the Marina District of San Francisco and of the San Francisco Airport area (Mb and Ms, and SFO in table 1; figs. 7 and 5, respectively). The Marina suffered dramatic damage, which was surprising U.S. GEOLOGICAL SURVEY AFTERSHOCK GROUND-MOTION DATA given its distance from the main shock, and was the ob- ject of quite thorough instrumentation. The USGS was requested by the airport authorities to conduct a study to interpret damage patterns at the facility. Analyses of the Marina data have appeared in Boatwright and others (1991b, 1992). McGarr and others (1991) analyzed the airport data. A study of San Francisco (SF in table 1) focused on the correlation of site amplification and damage with geology D123 in the city. An analysis of these data appears in Seekins and Boatwright (1994). Instruments were also deployed near Foster City on the bay margin (FC in table 1, fig. 4). These data have been discussed in Celebi and McGarr (1991a, b). The trans-south-bay line (PE in table 1, fig. 4) was also designed to sample a variety of alluvial units to determine if significant variations in ground shaking oc- curred. These data were analyzed by Margheriti and oth- ers (1994). Six instruments were deployed on or near the 40" 30° 20' 10' 122° 50 40' 30' 20' 10' 121° 38° -[lljllllIllIIllIlllllllllllllllllllll{ll nulnullull:1”lunlulllnulunlnulnu'unlllll 38° E \§\ \\ \ E El A\ \FIG5| | F in I NB N F 50' -A Eso- 40' 4 F- 40 3 F 3 F 30° < -- 30° 5 E 20' 7 - 20° 10' —: E— 1ol 37°] E- 37° 3 F 50" < F- so" 40" < - 40" ]. g IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIllll|IlTllllll|llllIllll|llIl'llllllllllllll|lllIlllllllllrlllllllill 40" 30° 20' 10' 122° 50° 40" 30° 20' 10' 121° Figure 1.-GEOS station locations (dots) in greater San Francisco Bay area. Main faults are indicated by solid and broken lines. Approximate Loma Prieta aftershock zone is indicated by shaded area. Boxes outline detail maps in figures 2, 3, 4, and 5. Map from Mueller and Glassmoyer (1990). D124 Oakland Bay Bridge in an effort to gather data relevant to the bridge closure (BB in table 1, fig. 6). Many instru- ments were located at sites of particular interest due to damage or geologic conditions. These site features are described under the heading U in table 1. Fifteen instrument locations were the same as, or less than a kilometer away from, sites that have been studied previously by the USGS; one location was the site of a small array (ST 1+, figs. 4, 8). These are referred to as Old sites in table 1. The earlier studies began with compari- sons of the response of different sites around the bay to nuclear explosions at the Nevada Test Site (Borcherdt, 1970; Gibbs and Borcherdt, 1974; Borcherdt and Gibbs, 1976) and included geologic and velocity logs to 30 meters in depth at selected sites (Gibbs and others, 1975, 1976a, and 1976b). A data analysis indicating that the velocity- log data were useful for interpreting the amplitudes of the explosion data was given by Borcherdt and others (1978). A deeper (180 m) borehole was drilled at station RAV, which has been the subject of quite detailed analysis (Warrick, 1974, and Joyner and others, 1976). Table 2 lists the aftershock-instrument sites that are near to (sig- 122 ° 50° 40° 10 baa f gag ap ab f gp pae fap apap ae f p aa f apap upa f up pape f apap uae | up F a u AFTERSHOCKS AND POSTSEISMIC EFFECTS nificantly less than a kilometer), or the same as, sites where nuclear explosions were recorded. It also lists spe- cifically the correspondence between Borcherdt and Gibbs's recordings and the aftershock-instrument sites. The GEOS sites BMT and COY are noteworthy because they were also reference sites for the earlier work. Also listed in table 2 are sites for which near-surface velocity data are available. Most of these sites were used in the nuclear explosion work. GEOS stations AP7, MAL, and SF1 were not included in the nuclear studies but are the sites of strong-motion accelerometers that recorded the main shock. Recently, a 150-m velocity log was ob- tained at station SF1 (Gibbs and others, 1992). In addition to a borehole log at SWS (Kayen and others, 1990), si- multaneous downhole/uphole small-earthquake data have been recorded there (Liu and others, 1992a, b). Four sub-kilometer-scale arrays recorded data relevant to spatial variations in ground motions. Three instruments were deployed in the Santa Cruz mountains as part of an attempt to observe topographic effects (SC in table 1 and KKO+ in figs. 3, 8, 14). This deployment was concurrent with DR-200 instruments, as discussed below. Two arrays 30° 20° 10 50° 40° 10 KM «|- LOO] BV2\ 122 ° 50° 40° T I T I I T T T I I I I T T I I TOT T I T I I T I I T I I I T T I I | I TOT T I I I I T I T T I I 30° 20' 10 Figure 2.-GEOS stations in Monterey and San Benito Counties (southernmost region in figure 1). Main shock was recorded at MON, SAG, and TOM. Map from Mueller and Glassmoyer (1990). U.S. GEOLOGICAL SURVEY AFTERSHOCK GROUND-MOTION DATA were intended to study the propagation of surface waves in deeper sediments: a three-component array deployed in Palo Alto (SP in table 1, DUV+ in figs. 4, 8) and a four- component array was deployed in the north central Santa Clara Valley (SA in table 1, AG1+ in figs. 4, 8). A four- component array was deployed near Stanford University, mainly as a test of array instrumentation techniques (ST in table 1, STI+ in figs. 4, 8. See also table 2). COLLOCATIONS WITH STRONG-MOTION INSTRUMENTS Twenty-three GEOS recorders were located where the main shock was also recorded. These sites are designated as CS in table 1. All stations but one were located with strong-motion accelerometers operated by the state of Cali- fornia (Shakal and others, 1989) or by the USGS (Maley and others, 1989). For further information on these strong- motion collocations, see Mueller and Glassmoyer (1990). AMF was deployed across the street from the Hyatt Hotel D125 in Burlingame, where accelerometers operated by the build- ing owners recorded the main shock (Celebi and McGarr, 1991a, b). The high-rise building in which AMF was lo- cated has been razed, apparently because of earthquake damage (Gelebi, oral commun., 1993). DR-200 DATA During the aftershock deployment, 11 DR-200 record- ers were deployed at 101 sites. A total of 2,926 three- component velocity records were recorded. The main focus of these studies was on local site effects on ground mo- tion. In special studies, data were recorded to investigate topographic effects and sites where liquefaction occurred. Five instruments were collocated with, or very near, sites where the main shock was recorded. Two of these latter sites were also occupied by GEOS recorders. Information analogous to table 1, describing the DR-200 deployments, is available in the detailed reports given by Carver and others (1990) and King and others (1990). Figures 9 10° 122 ° 50° 40° 30° [1111|1l|1|1111'1111I1111]|1||l||1|l||11|11¥11|1|1 \ \ he \\ | ® SAR ~, __ |- \ ~~ | 10 10 KM \ _10’ L. DMD [~ \\ | \ : \ . \\ L- DD \ - j | CRG‘K \ i- 37 ° \A x, aa, ® cAo0 |. \ __ a he & 3 gA \ \\\ \\ \ | T I I T I T TCT I I I I I T I \ TCT I UCI I T T I I T T T I I T I T T I T TOTO I I T TOTO I I I T 10° 122 ° 50° 40" 30° shock was recorded at CAP, CRG/KOI (collocated GEOSs), DMD, GAQ/ GA2/GAV (GEOS's in and around building where main shock was re- corded), and SAR. Map from Mueller and Glassmoyer (1990). Figure 3.-GEOS stations in Santa Cruz and southern Santa Clara Coun- ties (south-central region in figure 1). KKO+ marks a three-component, sub-kilometer-scale array that was part of a topographic experiment that also included eight DR-200 recorders (see figures 8 and 14). The main D126 through 14, taken from Carver and others (1990), show the locations of the sites they studied. SANTA CRUZ Thirty-seven sites were occupied in the Santa Cruz area (fig. 10; Carver and others, 1990). The main objective was to characterize the response of different geologic units in the city. Part of the deployment was a systematic study of site response of the flood plain of the San Lorenzo River (fig. 11). This study was presented by Cranswick and others (1990). Brief comments on some of the sites occupied during this deployment are listed in table 3. In addition to aftershock recordings, King and others (1990) analyzed small-scale refraction studies of 10 sites in the Santa Cruz area. Their results are relevant to velocities at depths of up to approximately 20 to75 meters. See also Williams and others (1994). AFTERSHOCKS AND POSTSEISMIC EFFECTS MOSS LANDING POWER PLANT Three DR-200's were deployed at and in the power plant (fig. 9) to gather information regarding this signifi- cant site. One instrument was in the free field and two were deployed in the control room in the upper floors of the power plant. These stations are discussed in Carver and others (1990). The Electric Power Research Institute also deployed instruments at the same time as the DR- 200's (Schneider, 1990). LOS GATOS Twenty-eight sites were occupied to study the effects of local site geology on ground motion in the city of Los Gatos (fig. 12). Stations LG3 and LG6 were located in the area of heaviest damage in the old business district along Santa Cruz avenue. Stations L12 and LEX, which was collocated with the CSMIP strong-motion instrument on 20 10' 122 ° 50° 40° L capa apap aab | pup f uaa I bapa uuu l aa \ __ 7 Aca e |- -_ e coy CNT ® 10 KM LOO - ® POR .- ESC ® \ - MDC @ [~ 30° - \\ |- 30° | TUL | 7 ® ED [ - aer. @ ask * USSD - - \ _ - ® SUN [ HAL 20' - ® PKT \ \\ HA2 |- 20' - \ _ i ~ \\ - - ~ \ - \ _ @ SAR }« __ \ |_ -| w _ \ ~ T I T T { TOT T T I T T T T l T T T I | T T T T I T T T T 1 T T T T ‘ T T T I I TOT T T I T T TCT 20° 10' 122 ° 50° 40° Figure 4.-GEOS stations in San Mateo, southern Alameda, and north- ern Santa Clara Counties (north-central region in figure 1). Main shock was recorded at AMF, AP2, AP7, ASH, BBP, HAL/HA2, MAL, POR, and SAR. AG1+, DUV+, and ST1+ mark sub-kilometer-scale arrays (see figure 8). Nuclear explosions have been recorded at or near stations AP2, BMT, COY, PKT, RAV, STI+, and USG. 30-meter velocity logs are available from boreholes near AP7, BMT, MAL, and STI+. A 180- meter velocity log is available for RAV. See table 2 for references to these data. Map from Mueller and Glassmoyer (1990). U.S. GEOLOGICAL SURVEY AFTERSHOCK GROUND-MOTION DATA the left abutment of Lexington Dam, served as hard-rock reference sites. LIQUEFACTION SITES Seven instruments were installed at sites in Monterey and south Santa Cruz Counties where liquefaction occurred during the main shock. They are sites AIR, ART, BEL, MIL, MOL, MOS, and ORD and are shown in figure 9. TOPOGRAPHIC EFFECTS STUDIES Two deployments were made in the Santa Cruz moun- tains to study variations of ground motion with topogra- phy over distances of less than a kilometer. These studies were motivated by dramatic damage to structures along ridge tops. The site locations and topography are indi- cated in figures 9, 13 and 14. D127 The study of the ridge topped by Rebecca Drive, east of Boulder Creek (fig. 9), included 15 stations installed along the top and face of the ridge and in the canyons on either side of the ridge (fig. 13). Stations MID and REB were located near minor damage to residences. TOP was placed in an area of significant structural damage, which seemed largely due to its position at the head-scarp of a landslide (King, written commun., 1990). In addition to variations in topography, there were significant variations in site geology among the stations in this study. The second study was of a ridge topped by Robinwood Lane above Hester Creek off of Old Santa Cruz highway (fig. 9). The ridge appeared to be geologically homoge- neous. Eight DR-200's and three GEOS recorders were deployed along the ridge top and descending approximately 45 meters down slope from the site of severely damaged houses near stations LP5 and LP6 (fig. 14). The GEOS sites are denoted by the letters KK in the station name, and the DR-200 sites are denoted by LP. 40° 30° 20° 10° 122 ° 50° lllIIILIllllllllllllllllllllllIllllllllllllllllll \ - \ \ - \ L 10 KM \ [ LOLOL LU] | 50° |- 50° |- \ - \ - \ | L |- a - &, \ | 40° # |- 40° a | t Aca o \” MAL CNT o A\ | Irllllllv‘q—Y—TTW' 40 30° 20' 10° 122 ° 50° Figure 5.-GEOS stations in San Mateo, San Francisco, and Alameda Counties (northernmost region in figure 1). Area in box is detailed in figure 6. Main shock was recorded at DIA, FPO, MAL, and SF1. A 150-meter velocity log is available for station SFI (see table 2). No data were obtained from CRA. Map from Mueller and Glassmoyer (1990). D128 COLLOCATIONS WITH STRONG-MOTION INSTRUMENTS Instruments were deployed at five sites where, or close to where, the main shock was recorded. Three of the sites are shown in figure 9: HAL, SAL and SAR. Instruments were also collocated with a strong-motion accelerometer on the left abutment of Lexington Dam (LEX in fig. 12) and with the strong-motion accelerometer at UC Santa Cruz (Station LOE in fig. 10). With the exception of SAR (see Maley and others, 1989), the strong-motion instru- ments were operated by CSMIP (Shakal and others, 1989). USGS SEISMIC CASSETTE RECORDERS On the night following the main shock, 60 cassette re- corders were deployed in the epicentral area to study crustal structure (fig. 15). These analog recorders were AFTERSHOCKS AND POSTSEISMIC EFFECTS deployed with 2-Hz vertical geophones. See Healy and others (1983) for a description of the instruments. The recorders do not operate in a triggered mode, so they were programmed to turn on for six time intervals lasting 12 to 13 minutes each. The time intervals began at 0600, 0630, 1000, 1030, 2300, and 2300 universal time on Oc- tober 19, 1989. These data have been digitized and ana- lyzed by Eberhart-Phillips and others (1990). They are also discussed in this professonal paper (Eberhart-Phillips and others, this chapter). No report has been written spe- cifically describing these data. ACKNOWLEDGMENTS Paul Spudich suggested that this report would be of use to future workers. Jack Boatwright provided particularly thorough comments on the GEOS data set. Dave Carver and Charles Mueller were helpful by providing copies of 25° 20' m 1 | | 1 | 1 | 1 1 EM1 o 50° - |- 50° @ RBT OAB 1 4 om BBS |- vYBI uso e LEA o Aup 8&£5" CAL o cos eSNP ® a RIN,RI2 RUS o aro e ch@ @* e Kis ® CWH | 5 KM L | | I 1 | T T I T T T T I T T 25 20 Figure 6.-GEOS stations in San Francisco and Alameda Counties. Marina district is outlined and detailed in figure 7. Main shock was recorded at CAL, EMT, and RIN/RI2Z. Nuclear explosions have been recorded at or near stations AUD, BLX, EMT, RIN, RI2, and SND (see table 2). No data were obtained at CCS or YBI. Map from Mueller and Glassmoyer (1990). U.S. GEOLOGICAL SURVEY AFTERSHOCK GROUND-MOTION DATA D129 SAN FPRANCISCO BAY ( | gwfimmm mle ND I | L] meters LOMBARD ST leo _b HH DIvISAD VAN NESS Figure 7.-GEOS stations in the Marina District of San Francisco. Borehole geologic, velocity, and downhole seismic data are available at SWS (see table 2). Map from Mueller and Glassmoyer (1990). ® AG2 @ AG3 ® st4 STI o sti @ S9 ® Ac1 AG4 @ ® KK2 @ DAW Duv o pas e KK1 @ © kio Figure 8. -Details of small GEOS arrays indicating dimensions and relative station postions. North is up. See figures 3 and 4 for array locations. Figure 14 shows topographic detail for, and DR-200 stations recording 1 KM simultaneously with, the array KKO, KK1, and KK2. \ ) Map from Mueller and Glassmoyer (1990). D130 their figures as well as review comments on an earlier draft. An unpublished report of Ken King's was helpful. Andy Michael provided timing and location information for the cassette recorder data. I would also like to thank Mehmet Celebi, Chris Dietel, Art Frankel, Jim Gibbs, Gary Glassmoyer, Tom Hanks, Tom Holzer, Art McGarr, and Linda Seekins for their helpful input. REFERENCES Boatwright, J., Seekins, LC., and Mueller, C., 1990, Ground motion amplification in the Marina District, in Effects of the Loma Prieta earthquake on the Marina District, San Francisco, California: U. S. Geological Survey Open-File Report 90-253, P. F1-F24. Boatwright, J., Fletcher, J.B., and Fumal, T., 1991a, A general inver- sion scheme for source, site and propagation characteristics using multiply recorded sets of moderate-sized earthquakes: Bulletin of the Seismological Society of America, v. 81, no. 5, p. 1754-1782. Boatwright, J., Seekins, L.C., Fumal, TE., Liu, H-P., and Mueller, C., 1991b, Ground motion amplification in the Marina District: Bulle- tin of the Seismological Society of America, v. 81, no. 5, p. 1980- 1997. Boatwright, J., Seekins, L.C., Fumal, TE., Liu, H-P., and Mueller, C., 1992, Ground motion amplification, in The Loma Prieta, Califor- nia, earthquake of October 17, 1989- Marina District: U.S. Geo- logical Survey Professional Paper 1551-F, p. 35-50. Borcherdt, R.D., 1970, Effects of local geology on ground motion near San Francisco Bay: Bulletin of the Seismological Society of America, v. 60, no. 1, p. 29-60. Borcherdt, R.D., and Gibbs, J.F., 1976, Effects of local geological con- ditions in the San Francisco Bay region on ground motions and the intensities of the 1906 earthquake: Bulletin of the Seismologi- cal Society of America, v. 66, no. 1, p. 467-500. Borcherdt, RD., Gibbs, J.F., and Fumal, TE., 1978, Progress on ground motion predictions for the San Francisco Bay region, Cali- fornia, in Proceedings of the Second International Conference on Microzonation for Safer Construction-Research and Application, San Francisco, California: National Science Foundation, UNESCO, American Society Civil Engineers, EERI, SSA, Univer- sities Council for Earthquake Engineering Research, p. 241. Borcherdt, R.D., Gibbs, J.F., and Lajoie, KR., 1975, Prediction of maximum earthquake intensity in the San Francisco Bay region, California, for large earthquakes on the San Andreas and Hayward faults: US Geological Survey Miscellaneous Field Studies Map MF-709, scale 1:125,000, 11 p. Borcherdt, R., Fletcher, J., Jensen, E., Maxwell, G., VanShaak, J., Warrick, R., Cranswick, E., Johnston, M., and McClearn, R., 1985, A general earthquake observation system (GEOS): Bulletin of the Seismological Society of America, v. 75, no. 6, p. 1783- 1825. Carver, D.L., Cunningham, D.R., and King, K.W., 1986, Calibration and acceptance testing of the DR-200 digital seismograph: U.S. Geological Survey Open-File Report 86-340, 28 p. Carver, D.L., King, K.W. , Cranswick, E., Worley, D.M., Spudich, P., and Mueller, C., 1990, Digital recordings of aftershocks of the October 17, 1989, Loma Prieta, California, earthquake: Santa Cruz, Los Gatos, and surrounding areas: U.S. Geological Survey Open-File Report 90-683, 203 p. Celebi, M., and McGarr, A., 1991a, Discussion of site-response at Fos- ter City and San Francisco Airport, in Proceedings of the Fourth International Conference on Seismic Zonation, August 25-29, AFTERSHOCKS AND POSTSEISMIC EFFECTS 1991, Stanford, California: Earthquake Engineering Research Institute, v. III, p. 367-373. Celebi, M., and McGarr, A., 1991b, Site-response at Foster City and San Francisco Airport & Loma Prieta studies, in Soil Dynamics and Earthquake Engineering V, Computational Mechanics Publi- cations: Southampton and London, England, Elsevier Applied Sci- ence, p. 35-46. Clark, J.C., 1981, Geologic map and sections of the Felton-Santa Cruz area, Santa Cruz County, California: U.S. Geological Survey Pro- fessional Paper 1168, plate 2. Cranswick, E., King, K., Carver, D., Worley, D., Williams, R., Spudich, P., and Banfill, R., 1990, Site response across downtown Santa Cruz, California: Geophysical Research Letters, v. 17, no. 10, p. 1793-1796. Eberhart-Phillips, D.M., Michael, A. ., Fuis, G., and Luzitano, R., 1990, Three-dimensional crustal velocity structure in the region of the Loma Prieta, California, earthquake sequence from inversion of local earthquake and shot arrival times: Seismological Research Letters, v. 61, no. 1, p. 48. Fletcher, J.B., and Boatwright, J., 1991, Source parameters of Loma Prieta aftershocks and wave propagation characteristics along the San Francisco peninsula from a joint inversion of digital seismo- grams: Bulletin of the Seismological Society of America, v. 81, no. 5, p. 1783 -1812. Gibbs, J.F., and Borcherdt, R.D., 1974, Effects of local geology on ground motion in the San Francisco Bay region, California-a continued study: U.S. Geological Survey Open-File Report 74- 222, 146 p. Gibbs, J.F., Fumal, TE., and Borcherdt, R.D., 1975, In-situ measure- ments of seismic velocities at twelve locations in the San Fran- cisco Bay region: U.S. Geological Survey Open-File Report 75- 564, 87 p. 1976 a, In-situ measurements of seismic velocities in the San Francisco Bay region; part II: U.S. Geological Survey Open-File Report 76-731, 145 p. Gibbs, J.F., Fumal, TE., Borcherdt, RD., and Roth, EF., 1976 b, In- situ measurements of seismic velocities in the San Francisco Bay region; part III: U.S. Geological Survey Open-File Report 77-850, 142 p. Gibbs, J.F., Fumal, TE., Boore, D.M., and Joyner, W.B., 1992, Seis- mic velocities and geologic logs from borehole measurements at seven strong-motion stations that recorded the Loma Prieta earth- quake: U.S. Geological Survey Open-File Report 92-287, 139 p. Hartzell, S., 1978, Earthquake aftershocks as Green's functions: Geo- physical Research Letters, v. 5, no. 1 p. 1 -4. Healy, J.H., Mooney, W.D., Blank, H.R., Gettings, ME., Kohler, WM., Lamson, and R.J., Leone, LE., 1983, Saudi Arabian seis- mic deep-refraction profile: final project report: U.S. Geological Survey Open-File Report 83-390, 360 p., 9 plates. Helley, E.J. and LaJoie, K.R., 1979, Flatlands deposits of the San Francisco Bay Region, California: their geology and engineering properties and their importance to compreshensive planning: U.S. Geological Survey Professional Paper 943, 88 pp. Joyner, W.B., Warrick, RE., and Oliver, A.A. III, 1976, Analysis of seismograms from a downhole array in sediments near San Fran- cisco Bay: Bulletin of the Seismological Society of America, v. 66, no. 3, p. 937-958. Kayen, RE., Liu, H-P., Fumal, TE., Westerlund, RE., Warrick, RE., Gibbs, J.F., and Lee, H.J., 1990, Engineering and seismic proper- ties of the soil column at Winfield Scott School, San Francisco, in Effects of the Loma Prieta earthquake on the Marina District, San Francisco, California: US Geological Survey Open-File Report 90- 253, p. G1-G18. King, K., Carver, D., Williams, R., Worley, D., Cranswick, E., and Meremonte, M., 1990, Santa Cruz seismic investigations following U.S. GEOLOGICAL SURVEY AFTERSHOCK GROUND-MOTION DATA D131 Table 1.-GEOS data sites described in Mueller and Glassmoyer (1990) and main experimental objectives Abbreviations: BB Oakland Bay Bridge study CS Colocated with a USGS, CSMIP or private strong-motion accelerograph that recorded the mainshock FC Foster City study of bay-margin response IT Instrumentation test Mb Surface instrument near subsequent hole drilled in Marina District of San Francisco Ms Site study in the Marina District of San Francisco NS Colocated with a strong-motion accelerograph without a mainshock record Old In previously published site response study. See table 2. PE Line from southern peninsula to the southern East Bay PP Peninsula wave-propagation study SM Sub-km-scale joint deployment with DR-200 recorders for Santa Cruz mountain topographic study SA Sub-km-scale four-station array on alluvium in the central Santa Clara Valley SV Line crossing northern Santa Clara Valley SF San Francisco site-response study SFO Few-km-scale array at San Francisco Airport SP Sub-km-scale three-instrument array on the alluvium in the southern peninsula U Geologic unit of specific interest. Alluvial unit descriptions from Helley et al (1979) BB__CS FC IT _ Mb Ms NS Old PE PP __ SA SM SV SF SFO SP _ U ACQ .... lll oal lll all a s ** a kkk kkk kkk ek. k... ..}. gravel quarry AG2 ** ... (same location as ASH) l. ** AG3 220®® L, ** AG4 2. 2 Ls ** AMF sok 22. 22. .. bay mud, damaged high-rise AP2 sok | sok (| xk _ 220k. . bay mud ASH ** ... (same location as AG2) ... *% AUD .... sok _ , . near damaged freeway BB4 ** BB5 ** BBP .... ** . bay mud BBT ** BCC .... M BEA .... 20% ,, . hydraulic fill the October 17, 1989 Loma Prieta earthquake: U.S. Geological Survey Open-File Report 90-307, 59 p. Liu, H-P., Warrick, RE., Westerlund, RE., Sembera, E. D., and Wennerberg, L., 1992a, Observation of local site effects at a downhole-and-surface station, in The Loma Prieta, California, earthquake of October 17, 1989-Marina District: U.S. Geological Survey Professional Paper 1551-F, p. 51-74. Liu, H-P., Warrick, R.E., Westerlund, RE., Sembera, E.D., and Wennerberg, L., 1992b, Observation of local site effects at a downhole-and-surface station in the Marina District of San Fran- cisco: Bulletin of the Seismological Society of America, v. 82, no. 4, p 1563 -1591. Maley, R., Acosta, A., Ellis, F., Etheredge, E., Foote, L., Johnson, D., Porcella, R., Salsman, M., and Switzer, J., 1989, U.S. Geological Survey strong-motion records from the northern California (Loma Prieta) earthquake of October 17, 1989: U.S. Geological Survey Open-File Report 89-568, 85 p. Margheriti, L., Wennerberg, L., and Boatwright, J., 1994, A compari- son of coda and S-wave spectral ratios: estimation of site response in the south San Francisco Bay, Bulletin of the Seismological So- ciety of America, v. 84, no. 6, p. 1815-1830. McGarr, A., Celebi, M., Sembera, E., Noce, T., and Mueller, C., 1991, Ground motion at the San Francisco International Airport from the Loma Prieta earthquake sequence, 1989; Bulletin of the Seismo- logical Society of America, v. 81, no. 5, p. 1923-1944. Mueller, C. and Glassmoyer, G.,1990, Digital recordings of after- shocks of the 17 October 1989 Loma Prieta, California, earth- quake: U.S. Geological Survey Open-File Report 90-503, 147 p. Schneider, J., 1990, Aftershock recordings of the Loma Prieta earth- quake: project RP 3181-1, EPRI Technical Report. Seekins, LC., and Boatwright, J., 1994, Ground motion amplification, geology, and damage from the 1989 Loma Prieta earthquake in the City of San Francisco: Bulletin of the Seismological Society of America, v. 84, no. 1, p. 16-30. Shakal, A., Huang, M., Reichle, M., Ventura, C., Cao, T., Sherburne, R., Savage, M., Darragh, R., and Peterson, C., 1989, CSMIP strong-motion records from the Santa Cruz mountains (Loma Prieta), California earthquake of 17 October 1989: California De- partment of Conservation, Division of Mines and Geology, Office of Strong Motion Studies Report OSMS 89-06, 195 p. Warrick, RE., 1974, Seismic investigation of a San Francisco Bay mud site: Bulletin of the Seismological Society of America, v. 64, no. 2, p. 375-385. Williams, R.A., Cranswick, E., King, K.W. , Carver, D.L., and Worley, D.M., 1994, Site-response models from high-resolution seismic reflection and refraction data recorded in Santa Cruz, California, in Borcherdt, R.D., ed., Strong ground motion: U.S. Geological Survey Professional Paper 1551-A, p. 217-242. D132 Table 1.-Continued AFTERSHOCKS AND POSTSEISMIC EFFECTS BB _CS__FC IT Mb Ms NS Old PE PP _ SA SM SV SF SFO SP __ U sok BLX =.... 20** . e .... near damage BMT .... eer feces BVF £**® 2 0 ull ull ... .... (BVF inside shed with BV2 Lol. sl caa. ** 2 ull lll e.. .... bad resonance. BV2 outside) CAL 2. ** . - ** CAP .... ** CGC .... ** CHB .... 2. 2 2 ull. .... bay mud CNT .... 2 ake ** y lull alll all a... .... Holocene coarse-grain alluvium COY .... 2}. o r r CRG .... ** . . ** ._ (Guralp accelerometer collocated with KOI) CWH .... 2 0** 2 ull. .... near liquefaction DAS .... L lull aula all lal ol.. .... ** .. Holocene fine-med. alluvium DAW .... aon aaa aaa aaa ** y gull lala. all ok. +.. .... ** .. Holocene fine-med. alluvium DEM .... ... Lu Luu lll lak laksa ae. o.. o... .... .... on (near?) hydraulic fill DIA .... ** . 2% , ** DMD .... , DUV .... , lull alll agl lal ol.. .... ** .. Holocene fine-med. alluvium EDS .... !. ee . ** ESC L lull alll lll lll «+k «... .... Holocene med. grain alluvium FER 2. 2 0** 2 u... .... building on pilings FLB .... . L lull alll ak. lee e.. .... .... near-fault, Pleistocene alluvium FOX .... 222 ** ae . bay margin, near damage? FPO .... ** . oe 2 0** 2 lll. .... near damage GAO |... ** _. .. (GAV and GAZ were GAV .... ** . sequentially deployed GAZ .... ** . in and near one structure) GRO .... 2... lll eel eee eee eee} 2 0** . .... .... dune sand near damage HAL .... **. . (HAZ same location as HAL, HA2 , ** . but with different sensors) IVE a** L lull alll aaa a... .... Pleistocene alluvium JSQ 20** 2 ull. .... bay mud near undamaged unreinforced masonry KIS 220k. 2 0** . 2... .... dune sand KKO , ** KK1 , ** Table 1.-Continued U.S. GEOLOGICAL SURVEY AFTERSHOCK GROUND-MOTION DATA D133 KK2 KOI LEA LMS .... MAL .... MAS .... MA2 MDC .... MON .... NPT OAB PKT POR PUC RAV 2.222 22. ..**; s RIN RJ2 . RUS SAG .... SAR SFI SF2 SF3 SF4 SND .... SRL STI ST2 ST3 ST4 STQ SUN .... Ssws 2... TOM .... TUL USG .... BB _CS FC IT Mb Ms NS Old PE PP tok 2, sok , ** tok .A..** - SBM .... rok tok Sok Sok sok sok *% + ** - - wok - Hook lok Hook sok Sok tok Sok sok tok sok sok sok tok sok tok sok sok *% sok tk sok tok ... (MAS and MA2 were sequentially . near each other on hard rock) SA SM SV SF SFO SP U sok . ** . (Collocated with Guralp, station CRG) . . ** - sok R Hock . (RIN relocated to _.. RI2. Same building) 2. sok . "ok L2 )| . lila &+ [osx 0 - sok - .. «0+ [ sok . Hook sok | Sok sok [ *% ad ** ad - ** «> sok . near (but off?) hydraulic fill . Holcene fine-med. alluvium . close to greatest damage; fire . . off fill, building renovated 1906 . bay mud . near damage . hard rock . near damage . near damage . damage, Pleist. alluvium D134 AFTERSHOCKS AND POSTSEISMIC EFFECTS Table 2.-GEOS aftershock recording locations at or near previous USGS studies [The left column indicates analyses of seismic data from sites collocated with or nearby the listed GEOS aftershock sites. Letter-number combinations denote particular site and nuclear explosion combinations as discussed in Borcherdt (1970), Gibbs and Borcherdt (1974), Borcherdt and others (1975), and Borcherdt and Gibbs (1976). Two more site-specific seismic studies, which analyze borehole data, are explicitly listed. The right column lists velocity and geologic log data sources for the GEOS sites.] Nuclear explosion, and borehole seismic studies AP2 15¢c (less than 1 km away) AP7 AUD Q5, P7 BLX T5 (less than 1 km away) BMT $3, $4, $5, $16, J17, J18, S19 COY KI, K12, $13, S14, $15 EMT H18 (less than 1 km away) MAL Gibbs and others (1977) PKT P18 RAV I11, Joyner and others (1976) RIN - J5, J7 RIZ - J5, J7 SFI SND R7 (less than 1 km away) STl L3 ST2 - L3 ST3 - L3 ST4 - L3 SWS Liu and others (1992a, b) USG J1 Velocity and geologic log data "Pulgas Tunnel" in Gibbs and others (1976) "Black Mountain" in Gibbs and others (1975) "Audobon School" (~ 1 km away) in Warrick (1974) "SF Airport" in Gibbs and others (1992) "Page Mill" in Gibbs and others (1977) "Page Mill" in Gibbs and others (1977) "Page Mill" in Gibbs and others (1977) "Page Mill" in Gibbs and others (1976b) Kayen and others (1990) U.S. GEOLOGICAL SURVEY AFTERSHOCK GROUND-MOTION DATA D135 Central California Study Region San A S AR Jose & HAL Los Gatos Study Area Boulder Creek Re Rid Study 'Af8§ : | Robinwood L a Study Are: O & oF elton . CG - m e Mainshock < 37° N o ~a O L3 Santa Suz - w Study Area 1 [4 I I I Gilroy > O Ma W ats o viIIeBEL 2,1 o m Mmi C tial A AIR =s (8 & A 6 -n- < . é, mos @aMoss Landin P | & \ é A MOL ‘ 0 2 0k m I A ART ‘3 ke oir (o ORD aA Monterey SAL ® Salinas 122° W Figure 9.-Locations of DR-200 recorders in Santa Cruz, Monterey and southern Santa Clara Counties. Boxes outline areas detailed in figures 10 to 14. Liquefaction occurred during the main shock at stations AIR, ART, BEL, MIL, MOL, MOS, and ORD. Main shock was recorded at HAL, SAL, and SAR. Map from Carver and others (1990). D136 AFTERSHOCKS AND POSTSEISMIC EFFECTS Santa Cruz Study Area 1 Mile f----------_-- O 1 Kilometer Tertiary Sedimentary Rocks Crystalline Rocks Marine Terrace Deposits 122° Figure 10.-Locations of DR-200 recorders (triangles) in the city of Santa Cruz. Surficial geology adapted from Clark (1981). Main shock was recorded at LOE. Map from Carver and others (1990). U.S. GEOLOGICAL SURVEY AFTERSHOCK GROUND-MOTION DATA Santa Cruz Flood Plain Array Santa Cruz Harbor ALLUVIUM w £5324;va SEDIMENTARY MARINE-TERRACE - | K bepodrs [«] enystacLine rocks SBR TRE ap l 53 wasce2 Lau Lav 88 ¥. SHIL3W O€ 2 1 _ KILOMETERS 0 Figure 11.-Locations of DR-200 recorders (triangles) for the study of seismic response of the flood plain of the San Lorenzo River (Cranswick and others, 1990). These sites were occupied simulta- neously. Bottom of figure shows idealized geologic cross section, a portion of which is offset and drawn to a different scale than the rest of the figure. Map from Carver and others (1990). D137 D138 AFTERSHOCKS AND POSTSEISMIC EFFECTS Table 3.-Noteworthy selected sites in Santa Cruz study [See Carver and others (1990) and King and others (1990).] ALM BAR BLA CE2 CED KAL LOE MAI MON PAC WAL Marine terrace, minor chimney damage Hard rock Heavy damage Heavy damage Heavy damage Hard rock Mainshock recorded at site Marine terrace Heavy damage Heavy damage Heavy damage 122° | Los Gatos Study Area | 37°13! N Figure 12.-Locations of DR-200 recorders (triangles) in the city of Los Gatos (roads shown). The shaded areas denote lakes. Highway 17 is indicated. Highway 9 intersects highway 17 at cloverleaf. Main shock was recorded at LEX. Map from Carver and others (1990). U.S. GEOLOGICAL SURVEY AFTERSHOCK GROUND-MOTION DATA D139 Rebecca Ridge Study Area 37°07' Nf seal é; @ *s SSV Contour Interval 40 feet 0 0.5 Mile lee ~ 77) a Qflfi an Ce | ) ) BXV v‘ W COMA UN --- (NOR a Cailin \ 1/7 C , 0.5 Kilometer Figure 13.-Locations of DR-200 recorders (triangles) used in the Rebecca Ridge topographic study. See figure 9 for array location. Map from Carver and others (1990). D140 AFTERSHOCKS AND POSTSEISMIC EFFECTS Robinwood Lane Study Area 121°57'W 12 1°57'W Contour Interval 40 feet . 2.5 Mile I 0.5 Kilometer Figure 14.-Locations of DR-200 and GEOS recorders used in the Robinwood Lane topographic study. GEOS instruments are denoted by KK. See figure 9 for array location. Map from Carver and others (1990). U.S. GEOLOGICAL SURVEY AFTERSHOCK GROUND-MOTION DATA D141 Loma Prieta Cassette recorder deployment 10' 37° - | 20 KM _\\ \\\\ \ \ III 1 d [1—111Ll pO SG I‘I L \\ A \N\ L J \\ \\ \ \ L A No *t i sol u T T L l U u T u I T T T T I T T T U | u T T T I L u I U l T U u u 5° 122° 50' 40° 30° Figure 15.-Locations of USGS seismic cassette recorders deployed the night of October 19, 1989 (triangles). Recordings were made continuously for 12- to 13-minute intervals starting at the following times: 0600, 0630, 1000, 1030, 2300, and 2330 universal time. For comparison, note that the main shock source zone is shown in figure 1 along with the same geologic features as in this figure. THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989; EARTHQUAKE OCCURRENCE AFTERSHOCKS AND POSTSEISMIC EFFECTS RESPONSE OF U.S. GEOLOGICAL SURVEY CREEPMETERS TO THE LOMA PRIETA EARTHQUAKE By K.S. Breckenridge and R.W. Simpson, U.S. Geological Survey CONTENTS Page Abstract D143 Introduction 143 Data from eight creepmeters closest to the earthquake ---------- 147 Preseismic slip deficit: precursory retardation or drought? ---- 147 Signal versus noise in the Creep data -----------------------------.-- 152 Coseismic steps and static stress changes -------------------------- 152 Creep rate changes and static stress changes ---------------------- 155 Coefficient Of aPParent friCtiON 159 CIEEP TAt@ VETSUS St@SS IAW cen eee 160 Models exploring depth-origin of Loma Prieta afterslip --------- 164 Retardation and anomalous behavior before the earthquake? --- 166 Conclusions 171 Acknowledgments 175 Appendix A: Site notes 176 References cited 177 ABSTRACT At the time of the Loma Prieta earthquake, 18 U.S. Geological Survey creepmeters in central California re- corded coseismic steps ranging in size from -0.35 to +6.8 mm. Five of the closest instruments on the San Andreas fault and three on the Calaveras fault also displayed long- term rate changes in the months after the earthquake. The coseismic steps seem to bear little relation in magnitude or sense to static stress changes calculated using disloca- tion models of the earthquake rupture, but 1-year average creep-rate changes (faster on the San Andreas and slower on the Calaveras) do correlate well with static stress changes. This correlation favors low values of apparent coefficient of friction. Observed advances and deficits in cumulative slip at the closest sites caused by the positive and negative rate changes are in fair agreement with de- formation predicted by a three-dimensional dislocation model that requires anomalous slip to extend to depths in excess of 10 km. Rate changes observed at several creepmeters in the years before Loma Prieta may be pre- cursory but are difficult to interpret unambiguously be- cause they fall in a period of extended drought. A 2-year period of retarded creep at the Cienega Winery site before the Loma Prieta earthquake contains an interesting inter- val with normal right-lateral creep events superposed on unusual left-lateral background drift. Assuming that this behavior was tectonic in origin and not drought- or instru- ment-related, it can be explained by stresses imposed from two different sources. One source could be slip on the San Andreas fault below the creepmeter, causing the right- lateral creep events. The second could be slip on the subparallel Calaveras fault or other local structures which were caught up in regional tectonic adjustments following the 26 January 1986 Tres Pinos earthquake, causing the left-lateral movement and retarding the cumulative progress of the instrument. In this scenario, the retardation at Cienega Winery could have been a precursor to Loma Prieta in the sense that it too was a manifestation of re- gional tectonic changes that would eventually trigger the Loma Prieta earthquake. INTRODUCTION At the time of the Loma Prieta earthquake, 27 U.S. Geological Survey (USGS) creepmeters were operating within 205 km of the epicenter (fig. 1A). Eighteen of these instruments recorded coseismic steps ranging in size from -0.35 to +6.8 mm, and eight instruments located closest to the epicenter recorded significant changes in slip rate in the months after the earthquake (table 1). For several creepmeters, the Loma Prieta signal is the largest anomaly recorded in up to 25 years of operation. Possible precur- sory rate changes were observed on several instruments (Breckenridge and Burford, 1990). The USGS creepmeters are of several different designs, and data are collected at various intervals ranging from minutes to months, depending upon the configuration of a particular instrument (table 1). Descriptions of the instru- ments and sites have been published in Schulz and others (1982, 1983) and more recently in Schulz (1989). Eigh- teen sites are equipped with satellite telemetry that samples D143 D144 the creepmeter every 10 minutes (Silverman and others, 1989), while micrometer dial readings at the remaining 9 sites are taken quarterly. Additional creepmeters on the San Andreas fault in- stalled by researchers from the University of Colorado are AFTERSHOCKS AND POSTSEISMIC EFFECTS discussed by Behr and others (1990) and by Behr and others (this chapter). Alignment arrays on the Calaveras, Hayward, and other Bay Area faults maintained by re- searchers at the University of California at San Francisco are described by Galehouse (1992, and this chapter). 123° 122° 121° 120° I I 38° [ + + % Figure 1B 37° [+ 4. | Pacific | Ocean 36° |- p q € _.... .n ___, JKM y . | Figure 1.-A, Locations of USGS creepmeters in central California op- erating at the time of the Loma Prieta earthquake. The epicenter of the earthquake is shown by a star. Filled triangles indicate creepmeters which recorded definite coseismic steps; larger triangles indicate creepmeters which experienced long-term rate changes (table 1). B, Lo- cations of eight creepmeters lying within 60 km of the Loma Prieta rupture. Epicenters of six earthquakes discussed in the text are also shown as stars. RESPONSE OF U.S. GEOLOGICAL SURVEY CREEPMETERS TO THE LOMA PRIETA EARTHQUAKE Numerous instances of triggered slip on fault segments at considerable distances from an earthquake rupture are reported in the literature. For example, such slip occurred after the 1968 Borrego Mountain earthquake (Allen and others, 1972), the 1979 Imperial Valley earthquake (Fuis, 1982; Sieh, 1982; Bilham and Williams, 1985), the 1981 Westmorland earthquake (Sharp and others, 1986a), the 1983 Coalinga earthquake (Mavko and others, 1985; Schulz and others, 1987), the 1984 Morgan Hill earth- quake (Schulz, 1984), the 1986 Tres Pinos earthquake (Simpson and others, 1988), the 1986 North Palm Springs earthquake (Fagerson and others, 1986; Sharp and others, 1986b, Wesson and others, 1986; Williams and others, 122° 12130" D145 1986, 1988), the 1987 Superstition Hills earthquake (Sharp, 1989; Hudnut and Clark, 1989; McGill and others, 1989), and the 1989 Loma Prieta earthquake (Behr and others, 1990; Galehouse, 1990; McClellan and Hay, 1990). If the triggered slip persists, it may be difficult to dis- tinguish from afterslip, although traditionally "afterslip'" has been used to describe postseismic slip occurring on the same fault that the earthquake occurred on and local- ized in the rupture region or in its immediate vicinity (for example, Smith and Wyss, 1968; Scholz and others, 1969; Burford, 1972; Burford and others, 1973; Cohn and oth- ers, 1982; Wesson, 1987; Bilham, 1989; Marone and oth- ers, 1991). 121° 37°30 ~ ~ ~ \\\ N e ~ ~ K* Loma Prieta °" [ _ (891018, M71) Me as Chittenden Monterey (900418, MS. N* 0 20 KM A Thanksgiving Day \\ (741128, M5.1) \ ¥ _ Tres Pinos \\ (860126, M5.3) 36°30° Figure 1.-Continued D146 AFTERSHOCKS AND POSTSEISMIC EFFECTS Table 1.-Summary table of all USGS creepmeters in operation at the time of the Loma Prieta earthquake [Asterisk (*) in first column indicates a site on the Southwest Fracture near Parkfield. Asterisk (*) in the coseismic change column indicates initial RL movement which was offset by subsequent LL movement within hours to days of the earthquake. Question marks in the coseismic change column indicate lack of data for determining coseismic movement. Negative values indicate left-lateral coseismic steps; positive values, right-lateral.] Site Fault Origin __ Sample Rate Distance from Coseismic Post-LP (mo/yr) Loma Prieta (km) _- Change (mm) _ Rate Change(?) XSJ San Andreas 11/74 10 minute 39 5.2 * Y XHR San Andreas 9/70 _ 10 minute 50 4.3 Y CWC San Andreas 10/68 10 minute 54 6.8 Y XFL San Andreas 4/73 _ 10 minute 68 -0.35 Y XMR San Andreas 6/69 10 minute 79 2.6 ? MRW _ San Andreas 10/72 _ quarterly 80 ? Y BIT San Andreas 7/69 _ quarterly 107 ? N XMP San Andreas 6/69 _ quarterly 133 ? N XSC San Andreas 6/69 _ 10 minute 155 0.2 N XMM _ San Andreas 9/79 _ 10 minute 171 0.7 ? XMD San Andreas 7/86 _ 10 minute 174 1.8 N XVA San Andreas 4/87 10 minute 177 3.1 N XRS* San Andreas 5/87 10 minute 178 -0.01 N XPK San Andreas 9/79 _ 10 minute 180 1.5 N XTA San Andreas 9/85 10 minute 182 0.14 N WKR San Andreas 9/76 _ 10 minute 186 0.8 N XHS* San Andreas 6/87 10 minute 184 0.1 N CRR San Andreas 6/66 10 minute 190 0.03 N XGH San Andreas 6/69 _ 10 minute 192 0.06 N X46 San Andreas 8/86 10 minute 204 0.2 N HWR Hayward 4/68 _ quarterly 73 2 N HWE Hayward 4/68 _ quarterly 72 ? N HWW _ Hayward 4/68 _ quarterly 72 ? Y HWP Hayward 5/70 _ quarterly 71 2? 2? XSH Calaveras 6/71 10 minute 40 5.1 * Y HLC Calaveras 4/70 _ quarterly 47 ? Y HLD Calaveras 4/70 _ quarterly 48 ? Y Some of these triggered responses, especially coseismic steps recorded on creepmeters, are most likely produced by the shaking of fault and instrument during passage of the seismic waves (King and others, 1977; Allen and oth- ers, 1972; Fuis, 1982; Williams and others, 1988; McGill and others, 1989). Such shaking probably triggers the re- lease of a backlog of slip that had been held in the Earth's near-surface or in the creepmeter itself by friction. Other longer-term responses, including alterations in creep-rate and occasionally in creep direction (Mavko, 1982; Mavko and others, 1985; Simpson and others, 1988), may be caused by static changes in the stress field produced by the distant earthquake's fault offset. The reality of such static stress changes in the earth is not in question, be- cause sensitive strainmeters have detected them at great distances from earthquakes (for example, Johnston and others, 1987; Shimada and others, 1987; Johnston and others, 1990). Microseismicity rate changes on central California faults also appear to correlate with calculated static stress changes after the Loma Prieta earthquake (Reasenberg and Simpson, 1992, and this chapter), sug- gesting that these faults can and do react to small stress perturbations at seismogenic depths. In the following sections, we present data for the eight USGS creepmeters that were situated nearest to the Loma Prieta epicenter. One of the largest uncertainties in the interpretation of creepmeter data is introduced by sea- sonal and rainfall-induced fluctuations. Therefore, rainfall records are also presented, and we explore the possible effects of weather on creepmeter behavior. Because tec- tonic accelerations or retardations in creep rates may oc- cur before certain earthquakes (for example, Nason, 1973; Burford, 1988), establishing the tectonic as opposed to the rainfall-related significance of creep-rate changes is important, although difficult to accomplish (Goulty and Gilman, 1978, Langbein and others, 1993). We attempt to relate the observed coseismic steps and longer-term creep-rate changes that occurred at the time RESPONSE OF U.S. GEOLOGICAL SURVEY CREEPMETERS TO THE LOMA PRIETA EARTHQUAKE of the Loma Prieta earthquake to static stress changes calculated from simple dislocation models of the rupture in an elastic half-space. To put the Loma Prieta observa- tions into perspective and to improve the statistics, we have compiled creepmeter data for four other moderate- to-large Bay Area earthquakes that occurred before 1989, in addition to the 1990 Chittenden earthquake, which was a large aftershock to the Loma Prieta earthquake. The results described here suggest that although the sense (right-lateral, RL, or left-lateral, LL) of coseismic steps at creepmeter sites seems to bear little relation to the im- posed static shear-stress changes, there is a statistically significant relation between the sense (RL or LL) of long- term average creep-rate changes and the sign of coseismic shear-stress changes. For the Loma Prieta earthquake, the magnitudes of these quantities are also well correlated. We also describe three-dimensional boundary element models used to estimate the total anomalous slip advance or deficit that can be expected at creepmeter sites as a result of the Loma Prieta creep-rate changes. The models suggest that, under some instruments, complex spatial dis- tributions of stress can exist, and the effects of RL stress changes imposed at the surface might be superseded over time by larger LL stresses imposed at depth. These mod- els suggest that for the five creepmeters on the San An- dreas fault, cumulative advances ranging from 12 to 60 mm can be expected as a result of the Loma Prieta earth- quake, whereas at the three sites on the Calaveras fault, cumulative delays of -6 to -29 mm might be expected. Although post-Loma Prieta adjustments at the creepmeters are not yet complete, these estimates appear to agree with extrapolations of observed advances or deficits in long- term pre-earthquake trends. DATA FROM EIGHT CREEPMETERS CLOSEST TO THE EARTHQUAKE In this paper we concentrate our discussion on the re- sponses of the eight closest USGS instruments on the San Andreas and Calaveras faults (fig. 1B). Data for all of the operating instruments is available to interested investiga- tors. Long-term data from the five closest instruments on the San Andreas fault are shown in figure 24. Data from a 3- year time window centered about the Loma Prieta earth- quake date are displayed in figure 2B. These plots were made using daily data (1 point per day) selected (for the most part) from 10-minute interval data by an automatic algorithm. These daily data have been adjusted to agree with quarterly micrometer readings where data gaps exist or calibrations are in question. Such adjustments are usu- ally unnecessary or small, and none of the significant varia- tions in creep rate presented here are likely to result from instrument calibration problems. D147 The most obvious features in the long-term data in fig- ure 2A are the post-Loma Prieta rate increases at XSJ, XHR, and CWC. Post-seismic changes at XFL and XMR are smaller compared with the other three sites and with other variations recorded at those sites over the years. At the bottoms of figure 2A and 2B, plots of daily rainfall at Paicines, California, located near the fault between creepmeters CWC and XFL, show that the Loma Prieta earthquake occurred during a drought lasting from 1986 to about 1991 (U.S. Dept. of Commerce). Figure 3A shows cumulative data for the three creepmeters along the Calaveras fault near Hollister. Two traces are plotted for the Shore Road creepmeter; XSH for daily data selected from 10-minute telemetered data and XSHM for approximately quarterly micrometer measure- ments to facilitate comparison with dial readings from Central Avenue (HLC) and D Street (HLD) in Hollister. The most obvious feature is the slip-rate decrease since the Loma Prieta earthquake at all three sites. There are indications that this decrease began before the earthquake and may be related to the M=5.6 Tres Pinos earthquake on 26 January 1986. These declining slip-rates occur dur- ing a period of drought, as shown by the plot of daily rainfall from Paicines, which casts some uncertainty on the tectonic significance of these changes. Galehouse (this chapter) has reported slower creep rates at a number of sites on the Hayward and southern Calaveras faults after the Loma Prieta earthquake. PRESEISMIC SLIP DEFICITS: PRECURSORY RETARDATION OR DROUGHT? Small, pre-Loma Prieta rate changes occurred at XHR, CWC, and XMR in the 2-3 years before the earthquake (figs. 2A, B). It is of some interest to know whether these changes are of tectonic origin, representing precursors to the Loma Prieta earthquake. Burford (1988), for example, discussed the possible significance of retardations in creep rate as precursory signals for magnitude 4-6 earthquakes. Burford also speculated that a larger, more distant shock might be associated with slip deficits across a broader region. The rate reductions seen at several creepmeters before the Loma Prieta earthquake are tantalizing in that respect. However, creepmeters are also known to respond to rainfall, and the prolonged drought that occurred from 1986 to 1991 might also play a role in preseismic creep rates. Schulz and others (1983) examined the impact of rain- fall at XSH, XSJ, and XMR during the 1970's, which includes the drought period from 1976 to 1978. Review- ing 8-11 years of creep data, they found no significant correlation between rainfall and either the total amount of annual creep, the number of annual episodic events, or AFTERSHOCKS AND POSTSEISMIC EFFECTS D148 'ypoysioye __ J0J tiep dooro oanejnump 'g 'g1 amy ut umoys pue jxa1 ay} ut possnosip soyenbyuea oy} pue ayenbyjrea tJoug ewo7 oy) Jo soum oy) yrew q pue q jeoruoA jod elWwojife; fenuoo aSre| 0} ageJopour xis Jo saum oy} yew 4 ySnouy} y soulf jeoJA ay) Jo wonoqg oy) je umoys osfe are soouaund00 |[eJurey 'oyenbyrrea ewo7 ay} yord ay) jo wonoq ay} je umoys osfe are saouauno00 J[EJUIEY '€66] 01 0L6I Wol} joue sreak sory} 01 arojaq sreak sory} wojj jng} seaipuy ueg ay} uo siojowudasato aay - J[NCJ Staipuy uteg o} UO simowudsaara aay wep daor> aanenum; 'y-'z ainsi Z661T T661 0661 6861 8861 L861 0661 , G861 | 0861 , SL61T | OL6T ~ Tm - tp:o z rin 3:4: z _ Eco a YIVI ] § -f A YIVd 3 a E a m3 " " Od TO CC O* T OG T -U0H S _<_._<_.-.- _~q "TCG" —4- T -<-< S Manama. Hmfliomfl 6861 8861 L861 06 .H_.:m$:.:::fi.:mm fliotfi [D "*+ 0 I J [T I | 1 TZ | S ] \\\|\ “M 7 ot { ra g 1 st I ~ fo \\ | - \\\\K \\ 0¢ - \l - + | y- 8 o { 6 SZ np A cee / - 3 HWX - ,/ sr tot z - \ _ <- ] _ jE 3 14% { \\\LL It 3 | _/ Pa m k L CQ 7 r |_" l [~ fam | p el a SC 1 l (l & 1 a .\.\\. H omo _ P! 14% fl if" rl { 1 fixixk. A 5000 ep l. -~ | [9 T l\\.\‘_,1\ l “R15 it Mo \\ Lil\|\\\\ FOS 7 8T l»" |f o- ~ a YHX | | MHX 55 [SX m Cu (5% 1L\ < ror-yr-r-r-ro-r- ' -T TT {Q9 1 3 j J - a J q 'v D149 RESPONSE OF U.S. GEOLOGICAL SURVEY CREEPMETERS TO THE LOMA PRIETA EARTHQUAKE 'oxenbyirea uapuanty; a pue ayenbyrrea oug ewo7 ay) Jo saum oy) yreu 4 pue q sour reoruoA ay} jo woyoq ay} je umoys oste are saouannd00 Iyyurey '[7TH put ;y7H wou} sSurpeot aSne$ perp ay} iim uostreduroa Jatsea 10} ats 32 sSurpear 1ojour010tu Ap1oprenb squasardar d0€1 J 'oxyenb -yueo ay) Joye sreoX sory} 01 arojoq sreak sory} wos} j[ng] C661 T661 O66T 6861 8861 L861 xii ___ £1.33, T T T T T T U C661 T66T O66T 6861 8861 L861 HAOO SIHONI HA O R _L -- ropes n - 'g 'g| omy ut umoys pug jxa1 ay} ut possnostp soyenbyprea penuss aie] 01 arerapour xis yo saum ay} yreur 4 ySnouy} y soutf reoruoa 01d ay} Jo wopnoq ay} je umoys osfe are soouai -1n090 [[EJUIEY °C TH pu® 10j posn sSurpear sojowor>tu ay) yiim uostredur0s Jotsea 10}; pojuasaid 'Hgy ouls 12 sSurpeoat saj3wo19tuL SI HSK £661 01 OL6I wor; jng; seroatfeq) ay uo siojowdsaro sary 10; eep door> aanenuunyp 'y-'g amiq 066T G86T O86T SL - 1 0 L6T mzfiézse, Tm 2:45, A 0661. SB§1., O8§1 . Si§1.. OL r aH If]\l\|\ta\){£al mm 4 xx“ H 1\\k\ 1\\.) WHSX ~ zlzfi fix HSX | \- J 7, I 13... C V T7 TI -T- mT 11 Q J d V "TIT DE SIHONI SYTLIWILLN3ID D150 the total creep accumulated from annual episodic events. The study identified slight RL or LL increases at XSJ and XMR associated with rainfall, and at XSH, oscillatory events of about 1-day duration accompanying rain. No obvious seasonal trends are present at XHR, CWC, or XFL. Figures 44 and 4B depict residual creep data with residual rainfall from Paicines, located just south of the CWC creepmeter. An annual average rate of 15.4 inches AFTERSHOCKS AND POSTSEISMIC EFFECTS per year has been removed from the rain record to visu- ally enhance seasonal changes. Drought periods, from 1976 to 1978 and from 1986 to 1991, appear as downward trends, and are marked with horizontal bars above the traces. For San Andreas fault creepmeters in this study, there is no apparent seasonal trend associated with rainfall ex- cept at XMR, where strong LL movement occurred dur- 50 a_ 4 i_} M i L 4 alld ald aan | u ug a pu Py gy gpu ina d i i {xsd __ F Z p ~a.er\—F\ L a _ R L 1970 1975 1980 1985 18990 50 a a d i i 1 i i i i L ua p u g a 1.—l 1 u g i i i 1 po ull £ :XHR C -m- fry-”J Z 4 Nw Pay E Oi WV \\’\ manny an. 1 1970 1875 1980 1985 {o 1990 _ 50 i i i i l i i F a | i_ i 4a al jCWC a Z A Pap INA y a «p W_ 3 E ' a or -r -T M * - 1970 - 1975) ~- 19800 - 1985) - 19900 50 i M M 1 1 F i i_ i 1 _, pull i N i M pula 1 . i 1 jJXFL L a} L 1970 975 1980 1985 1990 50 i N i 1 1 i 1 M 1 | 1 i M 1 M i 1 1. L u p u g _ *= aq 3 1970 __ 1975 | - _ 1980 1985 { - 1990 50 1 M i i 1 M i i i | N ca i | F i i i | N F t j] PAIR ooo p I 4 ~ U f - Z q f 1970 - - - = _ 1975) - 1980 - 1985) - 1990 - Figure 4.-A, Detrended creep data for five creepmeters on the San Andreas fault for the period 1970-1993. Trend was determined by cal- culating best-fitting least-squares line. Horizontal bars denote periods of creep retardation. Bottom plot shows detrended cumulative rainfall record. B, Detrended creep data for three creepmeters on the Calaveras fault for the period 1970-1993. Trend was determined by calculating best-fitting least-squares line. XSHM is micrometer readings at site XSH, presented for easier comparison with the micrometer readings used for HLC and HLD. Horizontal bars denote creep retardation. Bottom plot shows detrended cumulative rainfall record. RESPONSE OF U.S. GEOLOGICAL SURVEY CREEPMETERS TO THE LOMA PRIETA EARTHQUAKE ing the 1991 and 1992 rainy seasons. The average creep rate at XMR for those years remains consistent with the long-term average of about 18 mm/yr despite these sea- sonal fluctuations. On the Calaveras fault, rainfall signals are more pronounced at XSH after 1986, when the instru- ment was rebuilt. The oscillatory character of the signal remains. Appendix A discusses the reconstruction at XSH in more detail. During the 1976 drought, XSH recorded a slip deficit before the Coyote Lake earthquake in 1979. When normal rainfall resumed in 1978, there was no change in the character of the retardation signal in re- sponse to increased precipitation. Schulz and others (1983) note that while lack of rain may delay onset of a creep event, stress build-up from slip at depth will eventually override conditions in the shallow soil to produce a creep event. We find no strong relationship between changes in creep rate and periods of drought or deluge for the creepmeters in this study. D151 Turning to creep retardation, horizontal bars above creep data in figures 4A and 4B mark slip deficits identified by Burford (1988), updated through December 1992. Although the definition is subjective, as a general rule we look for rate decreases of at least 30 percent compared to the long- term average rate, using a duration of at least 12 to 18 months to avoid seasonal effects. By this rule, the deficit at XHR in 1988-89 does not qualify as retardation due to its brevity, but we include it nevertheless because of its uniqueness in the long-term data set. On the San Andreas fault, four out of five creepmeters recorded slip retarda- tions of some duration before the Loma Prieta earthquake, though there is little consistency in the onset. However, even in hindsight, it is not clear that these retardations are precursors to the Loma Prieta earthquake or that they could have been used to somehow predict the earthquake. Table 2 lists retardation parameters for these episodes. At two of these sites, XHR and CWC, rate reductions end just apa pag a g u s f u gp alg ug ula un f uy pal pag | pall a ug u y | a i XSH 50 warm SC wn 2 CT &_ ‘m\\f'\”" F 0 s f r ter pare pene r - r--r-- 1970 1975 1980 1985 1990 aod ug pag 1 i i i 1 | i i i 1 T i i i i | 4a lg 50XSHM | Z WN- Z 0 ¥ T "oof T C POCOCOP T T * T T T T T" T T M T T U 1970 1975 EFF] 1985 1990 i i i i 1 i i i i | i i i i L u g i i M | J_ alda y 5q HLC . Z M 8 T- TOT -tr-r-rt-r-r-r- T-mr-r 1970 1975 1980 1985 1990 i i i i L i i i i 1 i i i i | palg i i | i i SOHLD Z __ g i_ k 9 2 AR- Z 0 U T T u " T M T * T T T T T T 1970 1975 1980 1985 1990 50 i 1 i 1 1 1 i i i | i F i i i i i M IL . a 1 YIu’ {PAIR . 3 O - j E Z a F 1970 - - 0 / i1§7500 C - 1980 1985 _ 1990 __ Figure 4.-Continued D152 Table 2.-Possible creep retardations before the earthquake AFTERSHOCKS AND POSTSEISMIC EFFECTS [The retarded rate and the background rate were determined by fitting least square lines to the intervals of interest. The background interval extends from the origin of the respective meter to 17 October 1989.] Site From To Duration (months) _ Rate (mm/yr) Background Rate (mm/yr) XSJ August 1982 April 1990 92 3.8 6.4 XHR - September 1988 _ September 1989 12 6.1 7.5 CWC _ August 1987 October 1989 26 6.9 10.1 XMR _- August 1988 May 1991 36 12.3 18.0 XSH May 1989 continuing 30 -0.3 11.8 before or with the quake. At XSJ and XMR however, slip rates continued to be lower than the long-term average until 6 months and 19 months after the quake. Of the Calaveras creepmeters, the rate at XSH began to drop in May 1989, while at HLC and HLD slip deficits are completely postseismic effects. These retardation epi- sodes amount to only a shadowy precursor, given the vari- ability of response across the network. However, since the Loma Prieta is the first M=7 quake in proximity to USGS creepmeters in almost 25 years of monitoring, it is still premature to discount these deficits as local, transi- tory effects without further rigorous examination. SIGNAL VERSUS NOISE IN THE CREEP DATA As is indicated in the above discussion, separating tec- tonic signals from rainfall-induced signals and noise is an important problem. Langbein and others (1993) have esti- mated the 1/f2 noise of the creepmeters near Parkfield, California, to have a standard deviation of 6 mm/yr. We have attempted to estimate the noise levels for the eight instruments discussed in this report by calculating aver- age creep rates using 6-month and 1-year averaging win- dows moved across four intervals of interest (table 3). The standard deviations in table 3 give some indication of noise levels and how large rate changes need to be at particular sites before they can be regarded as significant. Although our analysis is simpler and not directly compa- rable to that of Langbein and others (1993), the results for XHR, CWC, and XFL show standard deviations consis- tently below the 6-mm/yr level, suggesting that these in- struments may have higher signal-to-noise ratios than typical Parkfield instruments. On the contrary, XSJ, XMR, XSH, and HLC have standard deviations that are typically worse than 6 mm/yr, at least for some intervals. Another factor that helps convince us that rate changes at the time of Loma Prieta are real is the occurrence of similar changes on a number of instruments. Even though the changes on individual instruments might be called into question because of noise problems, the agreement of a suite of instruments in the sense and magnitude of rate change can be convincing. COSEISMIC STEPS AND STATIC STRESS CHANGES Coseismic steps were recorded at the six closest USGS creepmeters on the San Andreas and Calaveras faults with sampling intervals short enough (10 minutes) to detect such steps. These "coseismic" responses occurred within 10-20 minutes of the earthquake origin time, the uncer- tainty in timing being caused by the coarseness of the sampling interval. Table 1 lists coseismic steps at all creepmeters in the network, and figure 5 shows raw data for the six closest instruments. Figure 6 compares the coseismic steps at the six creepmeters with the calculated Loma Prieta stress changes, based on a dislocation model of Lisowski and others (1990) used to match geodetic measurements of coseismic sur- face displacements. Static stress changes at the creepmeter sites were calculated using equations derived by Okada (1992) for dislocations in an elastic half-space. We show only horizontal shear stress changes and normal stress changes because the vertical shear changes are 5-10 times smaller on these vertical faults (Simpson and Reasenberg, 1994). Coseismic steps could not be inferred for HLC or HLD because these instruments are read manually at quar- terly intervals. Of six observed coseismic steps, one (at XFL) is LL and all of the others are RL. No strong rela- tion exists between either sign or magnitude of the steps and sign or magnitude of the calculated stress changes, especially in view of the tendency of coseismic steps in RL fault systems to occur in a RL sense. Five earthquakes other than the Loma Prieta with mag- nitudes greater than 5 have occurred in the region since 1974 (fig. 1B); namely, Thanksgiving Day (M=5.1), Coy- ote Lake (M=5.9), Morgan Hill (M=6.2), Tres Pinos (M=5.3), and Chittenden (M=5.4). For all six earthquakes, 4 LL and 17 RL coseismic steps were observed (table 4, figure 7). Of the 4 LL steps, two occurred at sites where model calculations predict LL shear-stress changes, and RESPONSE OF U.S. GEOLOGICAL SURVEY CREEPMETERS TO THE LOMA PRIETA EARTHQUAKE Table 3.-Average creep rates for five creepmeters on the San Andreas Fault and three creepmeters on the Calaveras fault [Average rate is determined by least square fitting of a line to data in either a 6-month window (A) or in a 1-year window (B); the windows are advanced by 3-month and 6-month increments, respectively, over the period of interest, and the resulting rates are averaged. The first interval is a background period which ends more than 3 years before the earthquake so as to avoid possible precursory Loma Prieta effects. The second interval includes possible precursory Loma Prieta effects. The third interval is for 1 year after the earthquake. The fourth interval is for 2 years after the earthquake. Negative values indicate left-lateral rates; positive values, right-lateral.] A. 6-Month Average Creep Rates, in mm/yr [The 6-month sampling interval advances in 3-month steps. First number is average creep rate in mm/yr; second number after + is one standard deviation; value in parentheses is the number of six-month intervals within the overall interval that was used to calculate average and standard deviation.] Interval Dates: 01/01/77-01/01/86 01/01/86-10/01/89 1/01/90 1/01/91 XSJ 6.1 £ 7.1 (35) 4.1 £ 5.8 (14) 15.1 + 7.7 (3) 14.5 + 6.0 (7) XHR 7.7% 3.7 (25) 8.5 + 4.0 (12) 22.0 + 3.3 (3) 19.2 + 4.4 (7) CWC 10.3 + 3.8 (30) 8.1 + 3.9 (14) 18.4 +£4.1 (3) 15.8 + 5.1 (7) XFL 6.3 + 3.7 (28) 7.3 + 3.1 (14) 15.0 + 3.1 (3) 11.3 + 4.2 (7) XMR 17.4 + 8.0 (34) 18.1 + 9.3 (14) 12.1 + 2.0 (3) 13.4 + 5.3 (7) XSH 13.1 +£15.9 (33) 9.5 £15.2 (12) 0.1 + 5.7 (3) -1.6 £10.5 (7) HLC 12.0 +12.1 (19) --- (0) -1.6 + 5.3 (3) 0.6 + 4.5 (6) HLD 8.0 + 3.1 (2) --- (0) -0.3 + 1.9 (3) -1.0 + 1.5 (6) B. 1-Year Average Creep Rates, in mm/yr [The 1-year sampling interval advances in half-year steps. First number is average creep rate in mm/yr; second number after + is one standard deviation; value in parentheses is the number of six-month intervals within the overall interval that was used to calculate average and standard deviation.] Interval Dates: 01/01/77-01/01/86 _ 01/01/86-10/01/89 1/01/90 1/01/91 xSJ 6.0 + 4.9 (17) 4.3 + 3.4 (6) 15.0 +-- (1) 14.3 + 1.0 (3) XHR 7.9 + 2.4 (14) 8.5 + 2.3 (6) 23.0 +-- (1) 20.2 + 4.2 (3) Cwe 104 + 1.4 (17) 8.0 + 1.6 (6) 18.2 +-- (1) 15.1 + 2.7 (3) XFL 6.6 + 2.1 (14) 7.6+ 1.3 (6) 14.8 +- (1) 11.0 + 3.3 (3) XMR 17.0 + 5.0 (17) 17.7 + 7.7 (6) 12.8 +- (1) 13.6 + 1.6 (3) XSH 12.7 +11.1 (17) 6.6 + 9.4 (5) 0.1 +-- (1) -0.7 + 4.1 (3) HLC 10.2 + 6.3 (15) 4.6 + 4.1 (6) 1.3% (1) 0.7 + 1.8 (3) HLD 1.9 + 2.3 (14) 1.5+ 1.8 (6) 0.3 +- (1) -0.8 £ 1.0 (3) two occurred at sites where calculations predict RL shear- stress changes. Conversely, at the six sites with coseismic steps where LL stress changes are predicted, only two showed LL coseismic steps. D153 These results imply that the sense (RL or LL) of coseismic steps at the six creepmeter sites bears little re- lation to the imposed static shear-stress changes. If we adopt the null hypothesis that the direction of a coseismic D154 step is independent of the sign of imposed static stress change, then a two-sided chi-square test (for example, Sachs, 1982) applied to the data displayed in figures 6 and 7 does not allow us to reject the null hypothesis at even the 50-percent confidence level. We conclude that the sense of observed coseismic steps bears no obvious strong relation to the calculated static stress changes. Al- though all but one of the large coseismic steps did occur for RL static stress changes, there was only one instance of a large imposed LL static change, so the geometry of faults and earthquakes in the region has not lent itself to testing the hypothesis in question. It does appear from visual inspection of figure 7 that the 26 occurrences of increased RL shear were more likely to have a sizable coseismic step than the 19 occurrences of added LL shear, many of which had a zero step and are not plotted in figure 7. The larger amplitude and the larger number of RL steps presumably reflects the fact that all of the creepmeters are installed across faults that nor- mally creep in a RL sense. AFTERSHOCKS AND POSTSEISMIC EFFECTS More often than not, the coseismic step probably re- flects the triggered release of a backlog of slip either in the near-surface of the earth or in the instrument itself by the shaking that accompanies the passage of seismic waves. The magnitude of the dynamic stress changes associated with the passage of seismic waves is, in general, much larger than the magnitude of the static stress changes (Spudich and others, 1995). On a right-lateral fault, such a backlog would normally be right-lateral, although a few sites might have a left-lateral backlog resulting from sea- sonal variations in slip direction of the instrument caused by rainfall, local soil conditions, and installation configu- ration. We cannot rule out the possibility that, because of radiation patterns and ground conditions, those sites with RL shear added also might have been the sites that expe- rienced the greater amount of dynamic shaking during the earthquake. Nevertheless, the results shown in figure 7 suggest that added RL static shear encourages the triggered release of a backlog of slip in a RL system, while added LL static shear can discourage such a triggered release. il i 1 1 i i i 40} Fo xSH B&] C-- F 1 Fo xsd 301 a E- sxur g 25 LLJ 7 L - 3 a LL L Z 3 t ~ 20_ t- - CWC Jat E E 15} 3 10] ~ C o_ XFL 5- J - _ XMR 0: T T TT T2 C ' TCO I M 17 18 19 OCTOBER Figure 5.-Coseismic response to the Loma Prieta earthquake for six creepmeters. XSH is on the Calaveras fault, while the others are on the San Andreas fault. Time extends from one day before to one day after the Loma Prieta earthquake. RESPONSE OF U.S. GEOLOGICAL SURVEY CREEPMETERS TO THE LOMA PRIETA EARTHQUAKE Williams and others (1988) suggest that the amount of slip triggered on part of the southern San Andreas fault by the 1986 North Palm Springs earthquake agreed well with the size of the slip deficit at the site. McGill and others (1989) suggest that slip deficit is probably only part of the story, because some sites showed triggered events associated with both the 1989 Elmore Ranch and Superstition Hills earthquakes, separated by only 11 hours, whereas one might expect the deficit to have been shaken out by the first shock. We looked for a correlation be- tween the magnitude of the coseismic step and the slip deficit at our eight San Andreas and Calaveras sites, where slip deficit is defined as the difference between the long- term rate at a site and the average rate in the 1-year pe- riod preceding the earthquake (fig. 8). (The two points in figure 8 with large negative deficits come from long- term rates before the Chittenden earthquake-an after- shock to the Loma Prieta earthquake. These negative defi- cits are caused by faster-than-normal rates at the two sites caused by Loma Prieta afterslip.) Although there is a sug- D155 gestion that the largest steps correlate with the largest deficits as defined here, the strength of the correlation is not great. We could find no significant agreement between sign of coseismic steps and sign of 1-year average rate changes (fig. 9). The two-sided chi-square test did not allow us to reject the null hypothesis that no relation existed with a confidence level of any greater than 40 percent. The ab- sence of any such a relation is consistent with the idea that, in most cases, coseismic steps largely reflect the release of a backlog of slip rather than a clean response to newly imposed stress changes, although it is not possible to rule out a combination of the two effects. CREEP RATE CHANGES AND STATIC STRESS CHANGES Average 1-year slip rates for the five creepmeters on the San Andreas fault increased at the time of the Loma T T T T T T 3 - z 2 |- - a. L 5 a O im |- CWC _ 3 * ® . a XSH . XHR xSJ & 3 XMR O0 o o XFL | | | | 1 1 -600 -400 200 0 200 400 600 HORIZONTAL SHEAR STRESS CHANGE, IN MBAR T T T T T T ps 3 2 2 |- - a- LW (I7) ® O m |- CWC _ 3 * . A ® XSH xsj XHR m A & $ s O0 o o XFL 1 1 | 1 1 N -600 ~400 -200 0 200 400 600 NORMAL STRESS CHANGE, IN MBAR Figure 6.-Coseismic steps for Loma Prieta earthquake at six creepmeters compared with calculated static stress changes. Coseismic steps are + for RL, - for LL. Shear stress is horizontal component, + for RL, - for LL. Normal stress is perpendicular component, + for tension, - for compression. D156 Table 4.-Observed coseismic steps and 1-year average rates for eight creepmeters and six earthquakes AFTERSHOCKS AND POSTSEISMIC EFFECTS [The earthquakes in column 1 are ch = Chittenden, Ip = Loma Prieta, tp = Tres Pinos, mh = Morgan Hill, cl = Coyote Lake, td = Thanksgiving Day. For coseismic steps in column 3, '-' indicates no step, and *?" indicates that the data were not available. 1-year average rates were calculated by fitting a least-squares line to the data in the one year interval and ignoring any coseismic offset. Fractional rate change in column 6 is calculated from columns 4 and 5 as (rate before - rate after)/(rate before). Stress changes in columns 7 and 8 were calculated from dislocation models described in table 3 and in the text. Negative values indicate left-lateral changes or rates; positive values, right-lateral.] Earthquake _ Creep- _ Coseismic Step _ 1-Year Rate Before _ 1-Year Rate After Fractional Shear Stress _ Normal Stress meter (mm) (mm/yr) (mm/yr) Rate Change (bars) (bars) ch XSH -0.04 7.0 2.1 -0.70 -0.07 -0.17 ch HLC ? 12.2 1.5 -0.88 -0.03 -0.01 ch HLD ? 7.1 -1.1 -1.15 -0.02 0.00 ch XSJ - 8.3 13.6 0.64 0.09 -0.02 ch XHR 5.40 27.6 22.7 -0.18 0.02 0.00 ch CWC 3.80 27.3 13.5 -0.51 0.01 0.00 ch XFL 0.15 8.6 9.0 0.05 0.00 0.00 ch XMR - 15.6 13.2 -0.15 0.00 0.00 Ip XSH 4.40 2.3 0.8 -0.65 -0.27 -0.69 Ip HLC ? 1.0 -1.3 -2.30 -0.19 -0.17 Ip HLD ? 1.5 0.3 -0.80 -0.15 -0.11 Ip XSJ 4.00 2.0 14.2 6.10 0.64 -0.19 Ip XHR 4.30 6.3 22.8 2.62 0.32 -0.12 Ip CWC 6.80 6.9 18.8 1.72 0.25 -0.10 Ip XFL -0.35 7.5 14.9 0.99 0.12 -0.03 Ip XMR 2.60 8.7 12.5 0.44 0.08 -0.02 tp XSH - 21.8 10.8 -0.50 -0.02 0.00 tp HLC - 1.5 -0.7 -1.47 -0.07 -0.03 tp HLD ? 6.9 1.5 -0.78 -0.06 -0.06 tp XSJ -0.05 3.7 3.9 0.05 -0.03 -0.02 tp XHR 2.68 9.2 4.2 -0.54 0.05 -0.05 tp CWC 1.54 9.3 8.8 -0.05 0.09 0.00 tp XFL -0.05 7.7 7.6 -0.01 0.01 -0.05 tp XMR 0.35 13.4 22.4 0.67 0.02 -0.01 mh XSH 13.00 5.0 21.0 3.20 0.18 0.01 mh HLC ? 1.0 15.9 14.90 0.07 0.01 mh HLD ? 1.0 1.3 0.30 0.06 0.00 mh XSJ - 5.8 1.0 -0.83 -0.10 0.06 mh XHR 0.50 3.9 6.0 0.54 -0.02 0.00 mh CWC 0.31 9.4 8.3 -0.12 -0.01 -0.01 mh XFL - 3.6 5.1 0.42 0.01 -0.01 mh XMR 0.27 12.2 14.1 0.16 0.01 0.00 cl XSH 8.90 2.4 11.6 3.83 0.31 -0.01 cl HLC ? 1.7 11.2 5.59 0.08 0.00 cl HLD ? 0.2 0.1 -0.50 0.06 0.00 cl XSJ - 3.6 5.2 0.44 -0.10 0.07 cl XHR - 3.0 7.8 1.60 -0.01 -0.01 cl CWC - 9.9 11.6 0.17 0.00 -0.01 cl XFL - 5.8 20.0 2.45 0.01 -0.01 cl XMR 17.1 11.8 -0.31 0.01 0.00 td XSH ? 18.2 23.8 0.31 -0.19 0.08 td HLC ? 16.3 13.0 -0.20 -0.15 -0.04 td HLD ? 1.1 3.4 2.09 -0.10 -0.01 td XSJ 0.30? 9.0 12.0 0.33 -0.05 -0.13 td XHR ? 9.0 7.8 -0.13 0.03 -0.01 td CWC - 8.8 9.5 0.08 0.01 0.00 td XFL - 9.9 8.0 -0.19 0.00 0.00 td XMR - 19.0 20.7 0.09 0.00 0.00 RESPONSE OF U.S. GEOLOGICAL SURVEY CREEPMETERS TO THE LOMA PRIETA EARTHQUAKE D157 Prieta earthquake, whereas rates for the three sites on the Calaveras fault decreased (figs. 2, 3; tables 3, 4). To com- pare creep-rate changes with calculated static stress changes, three different dislocation models of the Loma Prieta rupture were used, based on geometries and slip distributions proposed by Lisowski and others (1990), COSEISMIC STEP, IN MM 5 I | | 1 1 | -600 -400 -200 0 200 400 600 HORIZONTAL SHEAR STRESS CHANGE, IN MBAR I I I T I I ps S 2 2 |- a & L F.- o O i |- ) S * a w m & & s o) L O0 o - | 1 1 1 1 1 -600 -400 -200 o 200 400 600 NORMAL STRESS CHANGE, IN MBAR Figure 7.-Coseismic steps for six earthquakes, including Loma Prieta earthquake, compared with calculated static stress changes. Sign conven- tions as in figure 6. I 10 I COSEISMIC STEP, IN MM 5 / I -20 15 -10 -5 SLIP DEFICIT FOR ONE YEAR, IN MM 15 20 Figure 8.-Coseismic steps for six earthquakes compared with one-year slip deficit (defined as difference between the long term rate and the one- year average rate prior to the earthquake). D158 Marshall and others (1991), and Beroza (1991). Simpson and Reasenberg (table 1, 1994) tabulate the details of these three Loma Prieta models. For one site, changes in the rupture orientation from model to model changes the sign of calculated shear-stress because this site is near a node in the stress field. In general, the stresses calculated at distances of several rupture lengths from the epicenter are quite comparable model to model. Figure 10 compares the fractional change in 1-year av- erage creep rates before and after the Loma Prieta earth- quake with calculated static stress changes from the three Loma Prieta models. The fractional change in average rate is defined by AV /V, where AV is the change in average rate over the given time window and V, is the average rate before Loma Prieta over the given time window. To see if there was any significant relationship between the signs of the rate changes and the signs of the stress changes, we tested the null hypothesis that these quanti- ties were independent by again applying a two-sided chi- square test to the respective fourfold tables. For shear stresses, the null hypothesis that the quanti- ties are independent can be rejected at the 99-percent, 96- percent, and 83-percent confidence levels (x> values of 8, 4.4, 1.9) for the Lisowski, Beroza, and Marshall models, respectively, when the stresses are calculated at the sur- face. The confidence limits are even better if the stresses are calculated at greater depths (for example, 10 km) be- cause the horizontal shear-stresses change sign under HLC and HLD in the Marshall model (see fig. 15 and table 6). For normal stresses, the null hypothesis that the quanti- ties are independent can be rejected at the O-percent, 52- percent, and 96-percent confidence levels (x> values of 0, 0.5, 4.4) for the Lisowski, Beroza, and Marshall models, respectively. The sense of relation for normal stresses sug- gested by the Marshall model would yield an increase in creep rate in response to negative values of normal stress which, in the convention used here, would imply more AFTERSHOCKS AND POSTSEISMIC EFFECTS compression. If a relation of this sort exists, it would be counter-intuitive given that Coulomb's law predicts greater friction on the fault as a result of greater compression. We interpret these results to mean that for the three models tested, there is a good relation between slip-rate change and shear stress change, but that a relation be- tween slip-rate change and normal stress change is un- likely. In an effort to put this relationship on a firmer footing, we repeated the statistical tests using data for four earlier earthquakes and the Chittenden aftershock, as well as for the Loma Prieta earthquake (table 4). Model ruptures for each earthquake were made using dislocation rectangles positioned and oriented from main shock and aftershock locations and from focal-mechanism information (table 5). Amounts of slip were assigned to these dislocation surfaces to yield the observed seismic moments. By ex- amining the results of six earthquakes together, we hoped that model-dependent effects would be either made appar- ent or minimized in the statistics. The data for all six earthquakes are shown in figure 11. For shear stress, there is a relation at the 95-percent confidence level (x2: 3.88). For normal stress, the null hypothesis of independence cannot be rejected, implying that no strong relation exists. A slightly sharper result is obtained if the lower calcu- lated values of stress change are discarded. For example, Reasenberg and Simpson (1992, and this chapter) reported that significant correlation exists between shear-stress changes and microseismicity rate changes down to stress levels of about 0.1 bar. If we repeat the statistical tests, eliminating those data points for which the shear-stress changes were less than 0.05 bar (fig. 12), then for shear stresses, there is a relation at the 99-percent confidence level (X2: 6.58). These results are summarized in table 6. Similar statistical tests applied to each of the earth- quakes individually show that, not unexpectedly, the larg- est earthquakes (Loma Prieta and Morgan Hill) give the 10 T COSEISMIC STEP, IN MM 5 T | | | 1 L 1 | I 1 I I | 1 | -7 -6 -5 -4 -3 -2 41 0 1 2 3 4 5 6 7 FRACTIONAL CHANGE IN 1-YEAR AVERAGE RATE Figure 9.-Coseismic steps for six earthquakes compared with fractional change in 1-year average creep rates. Fractional change is defined as the 1- year average rate before the earthquake minus the 1-year average rate after the earthquake, divided by the 1-year average rate before the earthquake. RESPONSE OF U.S. GEOLOGICAL SURVEY CREEPMETERS TO THE LOMA PRIETA EARTHQUAKE most definitive results. The results of such tests are also dependent upon the choice of averaging interval. We used a 1-year interval for the tests described above. For a 3- year window, XMR yields a slower average rate after Loma Prieta compared to the rate before, which worsens the statistical results because the calculated static-stress changes require an increase in rate at XMR. It would be desirable to discover some objective criterion for choos- ing the averaging interval, but in the absence of such a criterion perhaps the best that can be done is to try a range of reasonable intervals to demonstrate that any re- sults obtained do not depend strongly on the choice of interval length. The best overall relationship between fractional 1-year average rate changes and applied static stress changes from table 6 can be expressed as AV/V,=(8 bars At, where At is in bars, although multipliers obtained range from 4 to 9 bars=". This is not a very general relationship because it is tied to a 1-year averaging window, but the data on the whole do not seem adequate to support a more complex relation. The data from XHR and CWC might be good D159 enough to warrant a fit to some of the empirical relations described below, but we have not done this. COEFFICIENT OF APPARENT FRICTION In correlating microseismicity rate changes with Cou- lomb failure function changes, Reasenberg and Simpson (1992, and this chapter) found that the best correlation was obtained for low assumed values of the apparent co- efficient of friction u'. (In the terminology used by Reasenberg and Simpson (this chapter), the apparent co- efficient of friction is the value inferred by neglecting pore fluid pressure changes.) The absence of any signifi- cant relationship between creep-rate changes and calcu- lated static stress changes reported in the previous section is consistent with a low value of u'. As another way to examine this, we correlated the Cou- lomb stress for different values of u' with the fractional rate change. Values of correlation coefficient p are weakly FRACTIONAL CHANGE IN CREEP RATE 2 |_ HR L-CWC 0 ~ M +XSH; Y f- +HLC 7 Eu | 1 | 1 ~1500 ~1000 500 0 500 1000 1500 HORIZONTAL SHEAR STRESS CHANGE, IN MBAR L -- é co a. W i |- * XSJ CC O 2 Y |- 8 a 'XHB $ Tee © < A 35 o Bln g *XSH +HLC C Y f +HLC O & 7 tu 1 1 | | 1 L_ -1000 500 0 500 1000 1500 A 1500 NORMAL STRESS CHANGE, IN MBAR Figure 10.-Comparison of the fractional change in 1-year average creep rate at the time of Loma Prieta earthquake with calculated static stress changes for the three Loma Prieta slip distributions: A, Lisowski and others (1990), B, Marshall and others (1991), C, Beroza (1991). Best fit line passing through the origin is also shown (see table 4 for slopes). D160 but systematically peaked at low values of the apparent coefficients of friction ' (fig. 13). The correlation coeffi- cient p improves slightly if we delete all values of Cou- lomb stress less than 0.05 bars. In both cases the correlation is significant at the 95-percent confidence level, although the slight differences for neighboring values of ' in fig- ure 13 are not likely to be very significant. CREEP RATE VERSUS STRESS LAW It would be very desirable to have a rheological law that would allow us to predict how a creeping fault would respond to applied stresses. The calculated stress distribu- tions discussed in the next section suggest that inferring information about such a law from the response of a single creepmeter to a single earthquake will not be easy. AFTERSHOCKS AND POSTSEISMIC EFFECTS Nason and Weertman (1973) pointed out that although the geometry of fault creep events requires an upper yield point behavior on the part of the fault gouge, a unique constitutive law cannot be inferred from the shape of the creep events alone. Nonetheless, a number of authors have attempted to infer parameters for various plausible types of creep laws by looking at creep events or afterslip decay. Crough and Burford (1977) used a power law fault- zone rheology to relate the shapes of individual creep events with stress. The power law is dU/dt=CT, and the resulting displacement for creep events is given by U()= Uf[1 -1/{Ci(a- Du + 1}”("'”], *X *XSH 1 | I *HLD HLC | 1 | FRACTIONAL CHANGE IN CREEP RATE -1500 -1000 -500 0 500 1000 1500 HORIZONTAL SHEAR STRESS CHANGE, IN MBAR * XSJ *XHR CWC +XSH | | | +HLD +HLC | | 1 FRACTIONAL CHANGE IN CREEP RATE 2 -1000 -500 0 500 1000 1500 NORMAL STRESS CHANGE, IN MBAR Figure 10.-Continued RESPONSE OF U.S. GEOLOGICAL SURVEY CREEPMETERS TO THE LOMA PRIETA EARTHQUAKE where U(t) is the displacement at time ¢ after the onset of the creep event, T is the driving stress, U, is the final displacement, C is a constant of proportionality, and n is the power law exponent. Wesson (1987, 1988) used this law for his discussions of fault dynamics and Bilham (1989) used it to fit afterslip data to the 1987 Superstition Hills, California, earthquake. Although the shapes of the Loma Prieta perturbation at XHR and CWC (fig. 14) ap- proximate the shapes of the classic creep event, the shapes of the perturbation at the remaining six sites are far from classic. Wesson (1988) has simulated composite events with somewhat similar shapes to those seen at XSJ and XMR in a model using multiple interacting slip patches, each obeying a power-law rheology. We note that the power law rheology as written, if ap- plied to the fault at all depths, does not take into account the possibility of a sign change in T at some depths that D161 might temporarily reverse the creep direction. A better version might be written aU idt=Ct|"C. For small changes in stress, the power-law yields AV (Ar) --=nl -) V T where AV is the change in creep rate and At is the change in driving stress, which motivated our use of the frac- tional creep rate change in the previous section. Compari- son “iith observations (table 6) yields the result that n/t~8 bars~'. *X +XSH | | *«HLD FRACTIONAL CHANGE IN CREEP RATE 2 I -1500 -1000 -500 0 500 1000 1500 HORIZONTAL SHEAR STRESS CHANGE, IN MBAR * XSJ c $0h 1 1 | *xXSH *HLD «HLC FRACTIONAL CHANGE IN CREEP RATE -1000 -500 0 500 1000 1500 NORMAL STRESS CHANGE, IN MBAR Figure 10.-Continued D162 AFTERSHOCKS AND POSTSEISMIC EFFECTS Table 5.-Values used in dislocation models of the six earthquakes [Negative horizontal-slip value indicates left-lateral offset; positive values, right-lateral. Negative dip-slip value indicates reverse faulting. Earthquake information from Oppenheimer and others (1990).] date name M no. lon lat depth _ len. ht. strike dip horiz dip (deg) (deg) (km) (km) (km) (deg) (deg) _ slip (m) _ slip (m) 741128 _- Thanksgiving Day _ 5.1 1 _ 121.46 36.92 4.5 1 1 33 90 -1.5 0 790806 _ Coyote Lake 5.9 1 - 121.53 - 37.10 6.0 13 8 335 90 0.2 0 840424 _ Morgan Hill § 6.2 324 - 121.61 - 37.22 - 27 12 328 90 0.24§ 0 860126 _ Tres Pinos 5.3 1 - 121.28 - 36.80 7.5 1 1 353 83 3.0 0 891018 _ Loma Prieta t 7.0 1 121.91 37.06 _ 11.2 37 13.3 136 70 1.66 -1.19 900418 _ Chittenden 5.4 1 - 121.64 36.94 4.8 1 1 312 90 4.2 0 M = local magnitude no. = number of elements in dislocation model depth = depth to center of model len. = horizontal length of rectangular dislocation surface ht. = down-dip dimension of rectangular dislocation surface strike, dip = orientation of rectangular dislocation surface horiz. slip = component of slip in horizontal direction dip slip = component of slip in down-dip direction § Model based on Harzell and Heaton (1986) used by Oppenheimer and others (1988). Slip value is average. t This is the Lisowski model described in Simpson and Reasenberg (1994) We can infer a value for the "average" ambient stress on the fault plane under the creepmeters by substituting likely values for n. Crough and Burford (1977) report values of n between 1.0 and 2.5 with an average of 1.6 inferred from typical creep events. For such values, T would lie between 0.1 and 0.3 bars, which seems quite small but which may represent average stress levels at the shallow depths most likely to be reflected in the first year of creep rate change. Sharp and Saxton (1989) proposed an empirical law to describe the time-evolution of afterslip observed on the Superstition Hills fault after two earthquakes in Novem- ber 1987 B c U(t)=Uf(l+:3t) where U(#¢) is displacement at time t after the earthquake, and Uf, B, and c are constants. (L. Wennerberg, oral commun., 1993, has proposed a refined and better-fitting version of this empirical law.) Boatwright and others (1989) discuss an inversion method to infer the three pa- rameters in the law from field data. Scholz (1990) and Marone and others (1991) assumed a relation based on constitutive laws developed by Dieterich (1979, 1981), Ruina (1983), and Rice and Gu (1983) to explore the nature of afterslip curves. Their start- ing constitutive relation is Tss = T. + (A- B)In(V/V,), where T,, is steady state frictional strength, V is sliding speed, A-B is the friction rate parameter, and T, is the strength for steady state sliding at speed V,. The afterslip displacement Up(t) at time t after the earthquake is given by k _ s . U,(t) = ulnfifi/fijt + I} +V,1, A- B where k is the thickness averaged stiffness for the veloc- ity strengthening region, VCSS is the thickness-averaged coseismic slip velocity within the velocity strengthening region, and V, is the long-term slip rate. There is a difficulty in applying the constitutive rela- tion when V becomes zero or negative. A more general RESPONSE OF U.S. GEOLOGICAL SURVEY CREEPMETERS TO THE LOMA PRIETA EARTHQUAKE formulation of the constitutive relation suggested by Dieterich (for example, 1992) can allow for these possi- bilities. At depths of several kilometers, we would expect that the imposed shear-stress changes under our creepmeter sites (typically tenths of bars or less) are small compared with the usual shear-stress levels driving these faults. If so, then a linearized form of the above laws relating creep- rate changes and stress changes may be appropriate, and the data provided by the creepmeters may not be adequate to distinguish among the various laws or, in fact, to dis- tinguish them from a linear viscous response. The fact that the three sites that were close to regions where the fault surface had LL shear-stress applied actually went LL for varying periods after Loma Prieta suggests that the usual shear-stress levels in the upper meters or kilometers are normally quite small, and that the rheological laws governing this depth range on the fault can probably not be linearized. The behavior of the Middle Mountain creepmeter near Parkfield after the Coalinga earthquake (Mavko and others, 1985; Simpson and others, 1988) pro- vides another example of such behavior. D163 Clearly, more work needs to be done to propose and test creep-rate laws with the appropriate depth-dependent behaviors, in order to explain the perturbations at these creepmeters. One benefit of such a law is the ability to calculate synthetic creep records at various sites based on the influences of nearby earthquakes. When such a law is calibrated, it can also be used to infer the existence of stressing not obviously associated with seismicity, but per- haps caused by the passage of aseismic tectonic waves. Gwyther and others (1992) describe a post-Loma Prieta shear-strain anomaly near San Juan Bautista recorded on a tensor strainmeter that could perhaps be used in con- junction with creep data to constrain a fault rheology. If indeed the creep rates are responding to static stress changes caused by earthquakes, including earthquakes on other faults, this raises the question as to whether the effect is shallow or extends to depth. The creepmeters, if they are sitting over low-friction shallow cracks, could just be acting as sensitive strainmeters. The observation that microseismicity rates on Bay Area faults also re- sponded to static stress changes after the Loma Prieta earthquake (Reasenberg and Simpson, 1992), suggests that FRACTIONAL CHANGE IN CREEP RATE 5 I o & - <>< 1 1 | | 1 1 ~1500 1000 500 0 500 1000 1500 HORIZONTAL SHEAR STRESS CHANGE, IN MBAR W & w --I T T T T T 7 a. W fra o 2 |- - Z W o * Z i |- -I Fe T & _] $ o o p 0 é | | | | | | LL 1500 ~1000 500 0 500 1000 1500 NORMAL STRESS CHANGE, IN MBAR Figure 11.-Comparison of the fractional change in 1-year average creep rate at the times of six earthquakes with calculated static stress changes for simple dislocation models of the earthquakes. D164 some part of the observed changes in surface creep-rates reflect changes in slip rate at seismogenic depths. We have not attempted to fit any of these laws to the Loma Prieta perturbations displayed in figure 14, partly because these perturbations are not very cleanly defined, and partly because we believe that the perturbations rep- resent a composite response to sometimes complex stress distributions. We will instead attempt to use some simple dislocation models to put bounds on the total amount of perturbed slip that might occur at the various sites, and to estimate the depths from which the slip might be coming as a function of time. MODELS EXPLORING DEPTH-ORIGIN OF LOMA PRIETA AFTERSLIP Although it is difficult to infer fault-zone rheology from the available "afterslip" observations, it is possible to es- timate the depth to which the earthquake-induced anoma- lous slip extends. It might be the case, for example, that AFTERSHOCKS AND POSTSEISMIC EFFECTS the afterslip recorded at the creepmeters was a superficial phenomenon, confined to the upper kilometer or two of the crust. To bound the possible depth to which anomalous slip might extend, we used Okada's (1992) dislocation sub- routines and the Loma Prieta slip distribution of Marshall and others (1991) to make a model of the San Andreas and Calaveras faults (fig. 15). Stress changes were calcu- lated at the centers of 2 km by 2 km square dislocation patches extending down to 20 km. We permitted the dis- location squares between the Earth's surface and some chosen depth to slip freely in response to the earthquake- induced static stress changes, so that the stress at their centers was canceled. All squares below the chosen depth were not allowed to slip. No squares farther than 9 km north of creepmeter XSJ were allowed to slip in response to the stress changes. We assumed that the total slip in response to the stress changes would occur instantaneously, although, in fact, the slip must occur in viscous fashion over the space of several years. The model contains an interruption in the Calaveras fault between HLC and HLD. This discontinuity, although 15 T 10 FRACTIONAL CHANGE IN CREEP RATE 5 I 1500 ~1000 -500 0 500 1000 1500 HORIZONTAL SHEAR STRESS CHANGE, IN MBAR 15 10 I | 1 FRACTIONAL CHANGE IN CREEP RATE 5 I ~1500 1000 -500 0 500 1000 1500 NORMAL STRESS CHANGE, IN MBAR Figure 12.-Comparison of the fractional change in 1-year average creep rate at the times of six earthquakes with calculated static stress changes greater than 0.05 bars in amplitude for simple dislocation models of the earthquakes. RESPONSE OF U.S. GEOLOGICAL SURVEY CREEPMETERS TO THE LOMA PRIETA EARTHQUAKE D165 Table 6.-Statistical tests to determine significance of relation between changes in one-year average creep rates and calculated changes in static stress Earthquake Stress Depth _ Two-Sided - Confidence Rho N Conf _ Slope0 Slope Intercept Model(s) Component (km) Chi-sq (%) (%) (bar—l) (bar-l) L horiz. shear 0 8.00 99 0.96 8 99 8.38 8.16+0.97 _ 0.19+0.29 B " " 4.44 96 0.90 8 99 4.16 4.14+0.83 _ 0.03+0.48 M " 1.90 83 0.87 8 99 4.75 4.74+1.08 _ 0.01+0.53 L normal 0 NA 0 0.18 8 _ <50 -1.32 2.10+4.80 1.39+1.29 B " " 0.53 53 -0.71 8 95 -6.55 -7.04+2.88 1.19+0.70 M " " 4.44 96 -0.50 8 _ <80 -5.50 _ -4.7743.35 _ 0.53%0.92 L horiz. shear 10 8.00 99 0.88 8 99 6.50 6.34+1.36 _ 0.88+0.46 B " " 8.00 99 0.86 8 99 4.38 4.17+1.02 _ 0.714+0.51 M " " 8.00 99 0.85 8 99 5.05 4.79+1.21 0.66+0.52 L normal 10 NA U 0.30 8 ~50 -0.03 2.48+3.22 1.54%1.16 B " " 0.53 53 0.13 8 _ <50 0.28 0.78+2.50 1.14£1.06 M 1.90 83 0.08 8 _ <50 0.74 0.57+2.79 1.12+1.10 ALL horiz. shear 0 3.88 95 0.48 _ 48 99 9.35 8.81+2.36 _ 0.60+0.34 ALL>0.05 " " 6.58 98 0.48 _ 25 98 9.39 8.38+3.23 1.06+0.64 ALL normal 0 0.30 41 0.07 _ 48 _ <50 -0.82 1.58+3.50 _ 0.83+0.41 ALL>0.05 " " 0.88 65 0.00 14 0 -0.88 0.02+3.16 _ 0.35+0.68 ALL horiz. shear 10 6.50 99 0.45 _ 48 99 8.51 8.09+2.36 _ 0.66+0.34 ALL>0.05 " " 9.38 99 0.46 _ 20 96 8.48 7.72+3.52 1.45+0.79 ALL normal 10 0.23 36 0.08 _ 48 _ <50 0.24 1.48+2.57 _ 0.82+0.39 ALL>O0.05 " " 0.20 34 0.13 11 __ <50 0.28 1.06+2.64 __ 0.52+0.84 Models: L=Lisowski, B=Beroza, M=Marshall, ALL=Lisowski plus 5 other earthquakes, ALL>O0.05 = same as ALL except with stress changes less than 0.05 bar omitted. Depth = depth at which static stress changes were calculated. Two-sided Chi-sq = Confidence that the fractional creep rate and change in stress component are not independent based on the two- sided Chi sqaured test applied to the respective four-fold table (Sachs, 1982). Rho = Correlation coefficient for fractional creep rates and static stress change values. N = number of samples. Conf = Confidence in best fit line. Slope0 = Slope of best fit line forced to pass through the origin. Slope, Intercept = Parameters for best fit line not forced to pass through the origin. small, will reduce the total slip estimated for nearby sites because of the fault connectivity effect of Bilham and Bodin (1992). (Sites on model faults that are distant from fault ends can slip farther than points close to ends of discontinuous segments.) The results (table 7, fig. 16) show, for example, that at CWC, a maximum of 5 mm of anomalous surface slip will occur if only the upper 2 km of the fault are able to respond to the stress changes. Figure 14 shows that slip at CWC had exceeded the amount expected at the pre-earth- quake creep rate by this amount within several months of the earthquake, even if the coseismic step is not included. Compared to the creep rate in the year before the earth- quake, at the beginning of 1993, anomalous excess slip of 30-35 mm had occurred at CWC (fig. 14), requiring anomalous slip to have occurred to depths in excess of 10 km in our model. Note that anomalous slip at some sites is highly dependent on the choice of background interval (compare last two columns in table 7). Figure 16 suggests that the total anomalous slip at the surface would not get much larger if even deeper levels were allowed to slip. We anticipate that the results ob- tained from a model where slip could extend to 50 km or 100 km would not be greatly different from the results D166 observed at 20 km. As of 1 July 1992, observed anoma- lous slip advances and deficits were estimated to range from -25 to +45 mm for the 8 creepmeters on the San Andreas and Calaveras faults (table 7). These observa- tions fall within the calculated extremes of -28 to +60 mm in the 20-km column of table 7. Because the anoma- lous slip is continuing, further comparisons will be needed. The three-dimensional distribution of earthquake-in- duced shear-stress in our model (fig. 15) suggests that efforts to infer fault-zone rheology from afterslip behav- ior recorded at the Earth's surface need to take into ac- count the geometry of the stress field both laterally and with depth on the fault surface. For example, using the slip distribution of Marshall and others (1991), the stresses imposed by the earthquake close to HLC and HLD are RL in the upper 6 km and LL from 6 to at least 20 km (fig. 15). The expected signal at these sites would initially be an increase in creep rate as the upper levels of the fault move faster in response to the added RL shear, followed by a slowing creep rate as the LL shear imposed at greater depths and to the north retards creep in these regions- which eventually propagates to the creepmeter sites. One can see suggestions of such behavior in figure 14 at sites HLC and HLD, although the noise level in these records is large enough to cast some doubt on this interpretation. AFTERSHOCKS AND POSTSEISMIC EFFECTS RETARDATION AND ANOMALOUS BEHAVIOR BEFORE THE EARTHQUAKE? A number of possible precursors to the Loma Prieta earthquake have been suggested. For example, Fraser- Smith and others (1990) reported anomalous electromag- netic radiation in the days and hours before the earthquake. Gladwin and others (1991) described a strain anomaly near creepmeter XSJ that began mid-way through 1988. Galehouse (this chapter) suggested that some alignment arrays along the Hayward fault might have slowed down in the months before Loma Prieta, and Reasenberg and Simpson (this chapter) reported a possible slowdown in Hayward fault seismicity beginning in 1988. We have listed five possible creep retardations (table 2, fig. 4) that might have foreshadowed the Loma Prieta earthquake. We examined the retardation at CWC in de- tail because the behavior of CWC (fig. 17) had been so consistent from about 1975 to 1987 that the retardation appears more convincing than do some of the others. We cannot rule out the possibility that drought conditions caused the slowdown at this site, but in the following discussion we assume a tectonic origin and see where that leads. n> gere $_ ‘|——1\1 O 2 \1 5 \ a 1 0 o \ - 5 LJ O 0 z O <3 E 44 2km 4km 6km 10km 20km Observed(1) _ Observed(3) XSJ 18.8 30.8 39.1 49.4 60.3 30 25 XHR 6.2 12.1 17.7 27.5 41.8 45 33 CWC 4.7 9.0 13.5 21.9 35.4 30 33 XFL 1.9 3.5 5.4 9.7 19.4 7 7 XMR 1.2 2.1 3.1 5.6 12.2 20 -10 XSH -7.0 -11.4 -14.6 -19.7 -28.6 -7 -25 HLC 0.1 -1.2 -2.8 -5.9 -10.5 5 -15 HLD 0.2 -0.7 -1.6 -3.3 -5.7 -5 -5 D169 o/" _ _- I | | XSH 0 5 10 15 DEPTH OF SLIP, IN KM 20 Figure 16.-Plot of total post-Loma Prieta anomalous slip predicted at eight creepmeter sites as a function of the depth to the bottom of the freely slipping layer in model. D170 the Tres Pinos earthquake. Again, slip in the proper re- gions of such a structure could probably induce LL stress at CWC. In all of these scenarios we suggest that the Tres Pinos earthquake caused or accompanied regional stress changes in the triangular wedge between the San Andreas and Ca- laveras faults which could have been recorded at CWC as LL drift. Thus CWC retardation might have reflected large- 6961 OL61 LL61 CL61 C161 §L61 9161 LL61 8161 6161 AFTERSHOCKS AND POSTSEISMIC EFFECTS scale stress changes that could also have triggered the Loma Prieta earthquake. Perhaps it is significant that the region south of San Juan Bautista in which most creep retardations have been described (Burford, 1988) contains subparallel fault strands (San Andreas and Paicines/Calaveras). If RL slip were to alternate on these strands, then creepmeters on the mov- ing strand would speed up, while those on the other would 0861 L861 2861 €861 P861 S861 9861 L861 8861 6861 0661 L661 2661 €661 N- 320 280 1 240 200 SLIP, IN MM 160 120 80 x 40 BZLLbL - £20169 - £06094 921098 022088 - 819006 6961 OL61 LL61 CL61 C161 pL61 §L61 9161 LL61 8161 6161 0861 L861 2861 €861 P861 S861 9861 L861 8861 6861 0661 L661 C661 €661 Figure 17.-Plot of creep record for CWC showing times of earthquakes that might have had an effect at this site. Earthquakes were selected from the CALNET catalog by scaling their moment by distance", which is the fall-off of maximum stress from a point dislocation, disregarding orientation. Numbers at tops of lines compare potential impact of the earthquake's static stresses (without regard to orientation information, so the value is an upper bound) to the impact expected from an optimally oriented magnitude 4.0 earthquake at a distance of 10 km. RESPONSE OF U.S. GEOLOGICAL SURVEY CREEPMETERS TO THE LOMA PRIETA EARTHQUAKE slow down. Perhaps when the slowed strand begins to move again the fault might have become more brittle be- cause it has had a chance to heal, so that if sufficient stress has accumulated to produce an earthquake, then one would be more likely to occur at the end of a retarda- tion period than at other times. Burford (1988) suggested fault interactions of this sort as a possible explanation for the observed retardations. He also offers other explanations for the phenomenon, including growth of asperities, strain hardening, stress waves, and fluctuation in driving stress. There appears to be some hope of evaluating these possibilities using simple dislocation models to explore plausible sources of im- posed stresses. D171 CONCLUSIONS Four mechanisms have been proposed to explain trig- gered slip on faults (Allen and others, 1972; Fuis, 1982; Williams and others, 1988; McGill and others, 1989): (1) static stress changes produced by the earthquake rupture, (2) dynamic stresses from the passage of seismic waves, (3) creep migrating from the earthquake source region, and (4) a regional strain event that produces both aseismic slip on some faults as well as earthquakes on others. The Loma Prieta earthquake produced coseismic steps on many of the central California creepmeters. We think that these steps were caused by shaking of the sites and -t _ -s _-. «. «© e «© el « co Co Co Co el & ~ Co © o 50 | 40 _| | dd < 1 J = 30 | < & 5 4 m -I 20 _| 10 | c co | |a © 0 a = 0 b ro a 1~ 3 d 0 pbb papa g apap f paca apap f guapa ap ws _i _x _s el el © el «© C Co Co Co «© & ~ Co © O Figure 18.-CWC creep record from 1986-1989 showing pre-Loma Prieta retardation beginning in mid-1987. Times of the Tres Pinos and Loma Prieta earthquakes are indicated by vertical lines. D172 AFTERSHOCKS AND POSTSEISMIC EFFECTS p- T T T I -0.2 -0.1 0.0 0.1 0.2 NUMBER 15 10 5 o - CREEP STEP SIZE, IN MM r my T T J NUMBER 10 15 5 o - -0.2 -0.1 0.0 0.1 0.2 CREEP STEP SIZE, IN MM Figure 19.-Histogram of creep step sizes smaller than 0.25 mm from daily data between 1 October 1987 and 1 October 1988 (bottom) compared with a similar histogram for the period from 1 January 1981 to 1 January 1982 (top). RESPONSE OF U.S. GEOLOGICAL SURVEY CREEPMETERS TO THE LOMA PRIETA EARTHQUAKE the instruments because they do not seem to correlate very well in either size or direction with calculated static stress changes, favoring explanation (2) above. The Loma Prieta earthquake produced significant changes in average creep rate at a number of sites on the San Andreas, Calaveras, and Hayward faults (Galehouse, D173 this chapter). These changes are generally consistent in magnitude and sign with the static shear-stress changes, and statistically significant correlation exists for three dif- ferent models of the Loma Prieta rupture, although the quality of the correlation varies to some degree from model to model. The change in horizontal shear-stress appears to 36°50 ° 36°40° ~ PACIFIC OCEAN LB Figure 20.-A, Earthquakes in the area of creepmeter CWC during the period from 1983-1988 from the CALNET catalog. B, Earthquakes for the same area from 1 January 1987 to 1 September 1988 during the period of retardation observed on creepmeter CWC when left-lateral steps were most obvious. Earthquakes do not lie exactly under the fault traces, probably in part because of deficiencies in the velocity models and in part because the faults may not be vertical. A star marks the epicenter of the 26 January 1986 Tres Pinos earthquake. D174 be the significant variable. Changes in calculated normal stress do not seem to correlate at significant confidence levels, which suggests that coefficients of apparent fric- tion are low for creep on these faults. A comparison of correlation coefficients for various assumed values of ap- parent coefficient of friction finds the best correlation for low values in the range 0.0-0.3. These observations seem consistent with explanation (1) above. Explanation (3) might explain rate changes for creepmeters on the San AFTERSHOCKS AND POSTSEISMIC EFFECTS Andreas fault, but hardly seems to explain the rate changes on the Calaveras or Hayward faults. A three-dimensional boundary element model was used to examine the depths that were being sampled by surface slip. This model can only put approximate limits on the depths that slip is "coming from" but suggests at CWC, for example, that by 1 year after the earthquake at the latest, slip from 10 km depth was being sampled at the surface. Total expected anomalous slip at the creep sites 12130" 12120 o \- . y c B A\ g T \ e . \@9 Tp \ ~A. n, + R $19.6 «%\ f = f‘ . ¢, \ HLC a o &: | \ Cp . 2.0 t ‘o’:’f_-_. 4. xSJ 'A e 3650 bl, + As 4 & \ ~ H * aus": LD '.' . ~ "s 36° 40° 0 10 KM R ___ L p py PACIFIC _ ocean _ \ Figure 20.-Continued RESPONSE OF U.S. GEOLOGICAL SURVEY CREEPMETERS TO THE LOMA PRIETA EARTHQUAKE can also be estimated from this three-dimensional model. The estimates range in value from -6 mm to +60 mm, which are in fair agreement with observed slip deficits and advances observed at the sites. Rainfall-induced and seasonal variations in creepmeter behavior are considerable and raise the possibility that some rate variations that might be interpreted as precur- sors to earthquakes are weather related. Because the drought conditions starting in 1987 were especially se- vere, creep retardations observed at four sites are suspect to some degree. We attempted to use creep behavior dur- ing the earlier drought years of 1976-1977 to calibrate the more recent drought but found no strong link between climate and slip fluctuations. Burford (1988) has proposed that creep retardations can occur at creep sites before local earthquakes and has tabu- lated 25 instances of possible retardations on the San An- dreas and Calaveras faults between 1957 and 1983. Assuming that a possible retardation at CWC that began in 1986-1987 and ended with the Loma Prieta earthquake might be of tectonic origin, we considered some possible tectonic causes of this retardation. The instrument contin- ued to record large RL events during this interval but small LL slip events in the intervals between the large RL events slowed the total creep-rate. This behavior is simi- lar to that described by Bilham and Behr (1992), who ascribe the large creep steps and the background creep at creepmeter sites on the Superstition Hills fault to differ- ent sources. If such an explanation holds here, it would seem likely that the large RL steps that represent fairly normal behavior at CWC come from slip on nearby parts D175 of the San Andreas fault. We propose three scenarios for the origin of the LL drift: slip triggered on the Paicines fault, slip on a nearby structure to CWC revealed in the seismicity, or slip on a regional subhorizontal detachment. In all three scenarios, this movement would be triggered by regional tectonic adjustments following the 26 January 1986 Tres Pinos earthquake in the complex triangle (Burford and Savage, 1972) between the San Andreas and Calaveras faults. If this is correct, then the retardation at CWC could be regarded as a precursor to the Loma Prieta earthquake if these same adjustments ultimately brought that earthquake closer to fruition. This scenario would also seem to give some credence to explanation (4) above as a viable mechanism. The ability of the creepmeters to respond to Loma Prieta stress changes in predictable ways suggests to us that other regional stress-change information is contained in the creepmeter signals. It becomes increasingly important to understand the effects of weather and seasons on the in- struments so that the true signals of tectonic origin can be extracted and interpreted. ACKNOWLEDGMENTS This paper was greatly enhanced by discussion with Paul Reasenberg, Evelyn Roeloffs, John Langbein, Roger Bilham, Bob Burford, and Jeff Behr. We are also grateful to Rich Liechti, whose expertise in creepmeter mainte- nance ensures a foundation for this and other studies, and to Jon Galehouse and his students at San Francisco State 3 4 5 INFLUENCE ON CWC, IN BARS/M Figure 21.-Oblique view from the northeast with the Calaveras fault in the foreground and the San Andreas fault in the background, showing influence coefficients relative to the rectangular patch under creepmeter CWC. Red indicates fault patches where RL slip would induce LL stress changes at CWC; blue indicates patches where RL slip would induce RL stress changes at CWC. Note that on the parts of the San Andreas fault shown in the figure, LL slip would be required to produce LL stress at CWC, whereas on the Calaveras fault, there are regions (red) where RL slip can produce LL stress at CWC. Each small rectangle is approxi- mately 2-km by 2-km in size. The fault planes in this model extend vertically to a depth of 10 km. D176 University for reading the creepmeters manually during their surveying expeditions. APPENDIX A: SITE NOTES The following section provides further detail on condi- tions influencing the data from different creepmeters. These aspects are typically site-specific and modulate the way we interpret the long-term record. As such they form the basis of several assumptions in this paper. XHR: COMBINED DATASETS The record for XHR used in the study is a composite of data from XHR1 and XHR2. Zero was lost on creepmeter XHR1 when it was destroyed in 1984 (Schulz 1989). To estimate a projected starting point for XHR2 data we cal- culated the amount of movement in a 245 day window, which is the duration of the gap between instruments, sliding the window every 10 days. The greatest change in a window with continuous data was 9.4 mm for the period ending 811102. We chose instead a correction value of 5.4 mm, which occurred most often in the series of win- dows. Comparison with creep data from Cienega winery, 3 km southeast of the site supports this as a reasonable adjustment for the level of activity at the time. CWC: INSTRUMENT CHARACTERISTICS AND RAINFALL At the winery, creepmeters actually measure offset of adjacent concrete floor slabs separated by the fault trace. Individual rainstorms produce only nominal changes in creep at this site. It is possible that a sustained drought might be expressed as a decrease in the background slip pattern such that episodic events, an indication of slip at depth, would be of normal amplitudes, while inter-event slip could be more sensitive to changing conditions in the shallow soil. De-coupling between the instrument and the fault at this site makes the association between drought and slip deficit tentative at best. In fact, we could argue that the retardation seen at the winery from 1987 to 1989 is caused by the same phenomenon that produced the slip deficit at XHR from late 1988 to September 1989. Per- haps both sites were responding to drought. Alternatively, these instruments may have been sensing a local perturba- tion in fault activity that was overridden by the Loma Prieta earthquake. Another instrument at the winery, CWN, is not included in this study. At this creepmeter, located about 30 meters northwest of CWC, an obstruction gradually developed AFTERSHOCKS AND POSTSEISMIC EFFECTS inside the instrument enclosure which inhibited the amount of movement recorded. Approximate onset of this condi- tion is difficult to determine, and the slip released when the problem was corrected in February 1990 was insuffi- cient to resolve the discrepancy between what was re- corded on CWN and what was observed on CWC. Prior to 1987 the two instruments tracked each other very well, both in rate of slip and in creep event characteristics onset, duration, and amplitude. XSH: SITE MODIFICATIONS AND RAINFALL In spring 1986 a new creepmeter was installed at Shore Road, slightly south of the original site and spanning an additional five meters of fault zone (Schulz 1989). The anchor pier of the new instrument is now within a few meters of an adjacent slough embankment. The stronger RL movement occurring during winter and spring may be related to increased movement of the anchor pier during wet weather. The fact that this pattern of activity was not seen at the site before 1987 adds credence to this account. REFERENCES CITED Allen, C.R., Wyss, M. Brune, J.N., Grantz, A., and Wallace, RE., 1972, Displacements on the Imperial, Superstition Hills and the San Andreas faults triggered by the Borrego Mountain earthquake, in The Borrego Mountain earthquake of April 9, 1968: U.S. Geo- logical Survey Professional Paper 787, p. 87-104. Behr, J., Bilham, R., Bodin, P., Burford, R.O., and Burgmann, R., 1990, Aseismic slip on the San Andreas fault south of Loma Prieta: Geophysical Research Letters, v. 17, p. 1445-1448. Beroza, G.C., 1991, Near-source modeling of the Loma Prieta earth- quake: Evidence for heterogeneous slip and implications for earth- quake hazard: Bulletin of the Seismological Society of America, v. 81, p. 1603-1621. Bilham, R., 1989, Surface slip subsequent to the 24 November 1987 Superstition Hills, California, earthquake monitored by digital creepmeters: Bulletin of the Seismological Society of America, v. 79, p. 424-450. Bilham, Roger, and Behr, Jeffrey, 1992, A two-layer model for aseismic slip on the Superstition Hills fault, California: Bulletin of the Seismological Society of America, v. 82, p. 1223-1235. Bilham, Roger, and Bodin, Paul, 1992, Fault zone connectivity: Slip rates on faults in the San Francisco Bay Area, California: Science, v. 258, p. 281-284. Bilham, Roger, and Williams, Patrick, 1985, Sawtooth segmentation and deformation processes on the southern San Andreas fault: Geophysical Research Letters, v, 12, no. 9, p. 557-560. Boatwright, John, Budding, K.E., and Sharp, R.V., 1989, Inverting measurements of surface slip on the Superstition Hills fault: Bulletin of the Seismological Society of America, v. 79, p. 411- 423. Breckenridge, K.S., and Burford, R.O., 1990, Changes in fault slip near San Juan Bautista, California before the October 17, 1989 Loma Prieta earthquake-A possible precursor?: Eos, Transac- tions, American Geophysical Union, v. 71, p. 1461. RESPONSE OF U.S. GEOLOGICAL SURVEY CREEPMETERS TO THE LOMA PRIETA EARTHQUAKE Burford, R.O., 1972, Continued slip on the Coyote Creek fault after the Borrego Mountain earthquake, in The Borrego Mountain Earthquake of April 9, 1968: U.S. Geological Survey Professional Paper 787, p. 105-111. 1988, Retardations in fault creep rates before local moderate earthquakes along the San Andreas fault system, central Califor- nia: Pageoph, v. 126, p. 499-529. Burford, R.O., Allen, $.S., Lamson, R.J., and Goodreau, D.D., 1973, Accelerated fault creep along the central San Andreas fault after moderate earthquakes during 1971-1973, in Kovach, RL., and Nur, Amos, eds., Proceedings of the Conference on Tectonic Problems of the San Andreas Fault System: Geological Sciences, Volume XIII, School of Earth Sciences, Stanford University, p. 275-285. Burford, R.O., and Savage, J.C., 1972, Tectonic evolution of a crustal wedge caught within a transform fault system: Geological Society of America Abstracts with Programs, v. 4, p. 134. Cohn, S.N., Allen, C.R., Gilman, R., and Goulty, N.R., 1982, Preearthquake and postearthquake creep on the Imperial fault and the Brawley fault zone, in The Imperial Valley, California, Earth- quake of October 15, 1979; U.S. Geological Survey Professional Paper 1254, p. 161-167. Crough, S.T., and Burford, R.O., 1977, Empirical law for fault-creep events: Tectonophysics, v. 42, p. 53-59. Dieterich, J.H., 1979, Modeling of rock friction, 1. Experimental re- sults and constitutive equations: Journal of Geophysical Research, v. 84, p. 2161-2168. 1981, Constitutive properties of faults with simulated gouge, in Carter, N.L., Friedman, M., Logan, J.M., and Stearns, D.W., eds., Mechanical Behavior of Crustal Rocks: American Geophysical Union, Geophysical Monograph 24, p. 103-120. 1992, Earthquake nucleation on faults with rate- and state-de- pendent strength: Tectonophysics, v. 211, p. 115-134. Fagerson, S.H., Louie, J.N., Allen, C.R., Sich, K.E., 1986, Measure- ments of triggered slip on the southern San Andreas fault associ- ated with the North Palm Springs earthquake: Eos, Transactions, American Geophysical Union, v. 67, p. 1090. Fraser-Smith, A.C., Bernardi, A., McGill, P.R., Ladd, ME., Helliwell, R.A., and Villard, O.G., Jr., 1990, Low-frequency magnetic field measurements near the epicenter of the Ms 7.1 Loma Prieta earth- quake: Geophysical Research Letters, v. 17, p. 1465-1468. Fuis, G.S., 1982, Displacement on the Superstition Hills fault triggered by the earthquake, in the Imperial Valley, California, Earthquake of October 15, 1979; U.S. Geological Survey Professional Paper 1254, p. 145-154. Galehouse, J.S., 1990, Effect of the Loma Prieta earthquake on surface slip along the Calaveras fault in the Hollister area: Geophysical Research Letters, v. 17, p. 1219-1222. 1992, Creep rates and creep characteristics on certain North and East Bay faults: 1979-1992; Programs and Abstracts, Second Conference on Earthquake Hazards in the Eastern San Francisco Bay Area, March 25-29, 1992, California State University, Hayward, p. 24. Gladwin, MT., Gwyther, RL., Higbie, J.W., and Hart, R.G., 1991, A medium term precursor to the Loma Prieta Earthquake?: Geo- physical Research Letters, v. 18, p. 1377-1380. Goulty, Neil R., and Gilman, Ralph, 1978, Repeated creep events on the San Andreas Fault near Parkfield, California, recorded by a strainmeter array: Journal of Geophysical Research, v. 83, no. B11, p. 5414-5419. Gwyther, RL., Gladwin, M.T., and Hart, R.G., 1992, A shear strain anomaly following the Loma Prieta earthquake: Letters to Nature, v. 356, p. 142-144. Hartzell, S.H., and Heaton, TH., 1986, Rupture history of the 1984 D177 Morgan Hill, California, earthquake from the inversion of strong motion records: Bulletin of the Seismological Society of America, v. 76, 649-674. Hudnut, K.W., and Clark, M.M., 1989, New slip along parts of the Coyote Creek fault rupture, California: Bulletin of the Seismologi- cal Society of America, v. 79, p. 451-465. Johnston, M.J.S., Linde, A.T., and Gladwin, M.T., 1990, Near-field high resolution strain measurements prior to the October 18, 1989, Loma Prieta Ms 7.1 earthquake: Geophysical Research Letters, v. 17, p. 1777-1780. Johnston, M.J.S., Linde, A.T., Gladwin, M.T., and Borcherdt, R.D., 1987, Fault failure with moderate earthquakes: Tectonophysics, v. 144, p. 189-206. King, C.Y., Nason, R.D., and Burford, R.O., 1977, Coseismic steps recorded on creep meters along the San Andreas fault: Journal of Geophysical Research, v. 82, p. 1655-1662. Langbein, John, Quilty, Eddie, and Breckenridge, Katherine, 1993, Sensitivity of crustal deformation instruments to changes in secu- lar rate: Geophysical Research Letters, v. 20, no. 2, p. 85-88. Lisowski, M., Prescott, W.H., Savage, J.C., and Johnston, M.J., 1990, Geodetic estimate of coseismic slip during the 1989 Loma Prieta, California, Earthquake: Geophysical Research Letters, v. 17, p. 1437-1440. Marone, C.J., Scholz, C.H., and Bilham, R., 1991, On the mechanics of earthquake afterslip: Journal of Geophysical Research, v. 96, p. 8441-8452. Marshall, G.A., Stein, R.S., and Thatcher, Wayne, 1991, Faulting ge- ometry and slip from co-seismic elevation changes: the 18 Octo- ber 1989, Loma Prieta, California, earthquake: Bulletin of the Seismological Society of America, v. 81, p. 1660-1693. Mavko, G. M., 1982, Fault interaction near Hollister, California: Jour- nal of Geophysical Research, v. 87, p. 7807-7816. Mavko, G.M., Schulz, S., and Brown, B.D., 1985, Effects of the 1983 Coalinga, California, earthquake on creep along the San Andreas fault: Bulletin of the Seismological Society of America, v. 75, p. 475-489. McClellan, P.H., and Hay, E.A., 1990, Triggered slip on the Calaveras fault during the magnitude 7.1 Loma Prieta, California, earth- quake: Geophysical Research Letters, v. 17, p. 1227-1230. McGill, S.F., Allen, C.R., Hudnut, K.W., Johnson, D.C., Miller, W.F., and Sich, K.E., 1989, Slip on the Superstition Hills fault and on nearby faults associated with the 24 November 1987 Elmore Ranch and Superstition Hills earthquakes, southern California: Bulletin of the Seismological Society of America, v. 79, p. 362- 375. Nason, R.D., 1973, Fault creep and earthquakes on the San Andreas fault, in Kovach, RL., and Nur, Amos, eds., Proceedings of the Conference on Tectonic Problems of the San Andreas Fault Sys- tem: Geological Sciences, Volume XIII, School of Earth Sciences, Stanford University, p. 275-285. Nason, R.D., and, Weertman, Johannes, 1973, A dislocation theory analysis of fault creep events: Journal of Geophysical Research, v. 78, p. 7745-7751. Okada, 1992, Internal deformation due to shear and tensile faults in a half-space: Bulletin of the Seismological Society of America, v. 82, p. 1018-1040. Oppenheimer, D.H., Bakun, W.H., and Lindh, A.G., 1990, Slip parti- tioning of the Calaveras fault, California, and prospects for future earthquakes: Journal of Geophysical Research, v. 95, p. $483- 8498. Oppenheimer, D.H., Reasenberg, P.A., and Simpson, R.W., 1988, Fault plane solutions for the 1984 Morgan Hill, California, earth- quake sequence: Evidence for the state of stress on the Calaveras fault: Journal of Geophysical Research, v. 93, p. 9007-9026. D178 Reasenberg, P.A., and Simpson, R.W., 1992, Response of regional seismicity to the static stress change produced by the Loma Prieta earthquake: Science, v. 255, p. 1687-1690. Rice, J.R., and Gu, Ji-cheng, 1983, Earthquake aftereffects and trig- gered seismic phenomena: Pageoph, v. 121, p. 185-219. Ruina, AL., 1983, Slip instability and state variable friction laws: Journal of Geophysical Research, v. 88, p. 10359-10370. Sachs, Lothar, 1982, Applied statistics-A handbook of techniques: New York, Springer-Verlag, 706 p. Scholz, C.H., 1990, The mechanics of earthquakes and faulting: Cam- bridge, U.K., Cambridge University Press, 439 p. Scholz, C.H., Wyss, M., and Smith, S.W., 1969, Seismic and aseismic slip on the San Andreas fault: Journal of Geophysical Research, v. 74, p. 2049-2069. Schulz, S$.S., 1984, Triggered creep near Hollister after the April 24, 1984, Morgan Hill, California, earthquake, in Bennett, J.H., and Sherburne, R.W., eds., The 1984 Morgan Hill, California Earth- quake: California Division of Mines and Geology Special Publica- tion 68, p. 175-182. Schulz, $.S., 1989, Catalog of creep measurements in California from 1966 to 1988: U.S. Geological Survey Open-File Report 78-203, 193 p. Schulz, S.S., Burford, R.O., and Mavko, B., 1983, Influence of seis- micity and rainfall on episodic creep on the San Andreas fault system in central California: Journal of Geophysical Research, v. 88, p. 7475-7484. Schulz, S.S., Mavko, G.M., Brown, B.D., 1987, Response of creepmeters on the San Andreas fault near Parkfield to the May 2, 1983, Coalinga earthquake: U.S. Geological Survey Professional Paper 1487, p. 409-417. Schulz, S.S., Mavko, G.M., Burford, R.O., and Stuart, W.D., 1982, Long-term fault creep observations in central California: Journal of Geophysical Research, v. 87, p. 6977-6982. Sharp, R.V., 1989, Pre-earthquake displacement and triggered dis- placement on the Imperial fault associated with the Superstition Hills earthquake of 24 November 1987: Bulletin of the Seismo- logical Society of America, v. 79, p. 466-479. Sharp, R.V., Rymer, M.J., and Lienkaemper, J.J., 19862, Surface dis- placements on the Imperial and Superstition Hills faults triggered by the Westmorland, California, earthquake of 26 April 1981: Bulletin of the Seismological Society of America, v. 76, p. 949- 965. Sharp, RV., Rymer, M.J., and Morton, D.M., 1986b, Trace-fractures on the Banning fault created in association with the 1986 North Palm Springs earthquake: Bulletin of the Seismological Society of America, v. 76, p. 1838-1843. AFTERSHOCKS AND POSTSEISMIC EFFECTS Sharp, R.V., and Saxton, J.L., 1989, Three-dimensional records of sur- face displacement on the Superstition Hills fault zone associated with the earthquakes of 24 November 1987: Bulletin of the Seis- mological Society of America, v. 79, p. 376-389. Shimada, Seiichi, Sakata, Shoji, and Noguchi, Shin'ichi, 1987, Coseismic strain steps observed by three-component borehole strainmeters: Tectonophysics, v. 144, p. 207-214. Sich, K.E., 1982, Slip along the San Andreas fault associated with the earthquake, in The Imperial Valley, California, Earthquake of Oc- tober 15, 1979; U.S. Geological Survey Professional Paper 1254, p. 155-159. Silverman, S., Mortensen, C., and Johnston, M., 1989, A satellite- based digital data system for low-frequency geophysical data: Bul- letin of the Seismological Society of America, v. 79, p. 189-198. Simpson, R.W., Schulz, S.S., Dietz, L.D., and Burford, R.O., 1988, The response of creeping parts of the San Andreas fault to earthquakes on nearby faults: Two examples: Pageoph, v. 126, p. 665-685. Smith, S.W., and Wyss, Max, 1968, Displacement on the San Andreas fault subsequent to the 1966 Parkfield earthquake: Bulletin of the Seismological Society of America, v. 58, p. 1955-1973. Spudich, Paul, Steck, LK., Hellweg, Margaret, Fletcher, J.B., and Baker, LM., 1995, Transient stresses at Parkfield, California, pro- duced by the M7.4 Landers earthquake of June 28, 1992; Observa- tions from the UPSAR dense seismograph array: Journal of Geo- physical Research, v. 100, p. 675-690. U.S. Department of Commerce, National Oceanic and Atmospheric Administration. Climatological Data, California. Ashville, NC: National Climatic Data Center, 1970-1992. Wesson, RL., 1987, Modelling aftershock migration and afterslip of the San Juan Bautista, California, earthquake of October 3, 1972; Tectonophysics, v. 144, p. 215-229. 1988, Dynamics of fault creep: Journal of Geophysical Re- search, v. 93, p. 8929-8951. Wesson, RL., Jibson, R.W., Morton, D., Campbell, R.H., and Nicholson, C., 1986, Interpretation of surface cracks and other fault-line surface deformation associated with the North Palm Springs earthquake of July 8, 1986: Eos, Transactions, American Geophysical Union, v. 67, p. 1090. Williams, P., Fagerson, S., and Sich, K., 1986, Triggered slip of the San Andreas fault after the July 8, 1986 North Palm Springs earthquake: Eos, Transactions, American Geophysical Union, v. 67, p. 1090. Williams, P.L., McGill, S.F., Sich, KE., Allen, C.R., and Louie, J.N., 1988, Triggered slip along the San Andreas fault after the 8 July 1986 North Palm Springs earthquake: Bulletin of the Seismologi- cal Society of America, v. 78, p. 1112-1122. THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989; EARTHQUAKE OCCURRENCE AFTERSHOCKS AND POSTSEISMIC EFFECTS INCREASED SURFACE CREEP RATES ON THE SAN ANDREAS FAULT SOUTHEAST OF THE LOMA PRIETA MAIN SHOCK By Jeff Behr, Roger Bilham, and Paul Bodin, University of Colorado, Boulder; Kate Breckenridge, U.S. Geological Survey; and Arthur G. Sylvester University of California, Santa Barbara CONTENTS Page Abstract D179 Introduction 179 Postseismic surface slip above the southeastern Loma Prieta Rupture zone 181 Aseismic surface slip on the San Andreas fault southeast of the LOMA Pri@tA FUPtUIG een 181 Elastic model estimates for aseismic slip in central California associated with the Loma Prieta earthquake ---- 186 Creepmeter and borehole strainmeter data ------------------------- 187 Conclusions 191 Acknowledgments 191 References cited 192 ABSTRACT Surface creep rates increased substantially on the north- western 50 km of the creeping section of the San Andreas fault following the Loma Prieta earthquake. Although postseismic creep southeast of the rupture zone initially resembled previously observed earthquake afterslip, creep rates between October 1990 and September 1992 have remained nearly constant at rates higher than pre-event values. Two-dimensional elastic boundary element mod- els indicate that if the San Andreas and Calaveras faults were frictionless, vertical dislocations, then more than 20 cm of surface slip would be induced on the northwesternmost 20 km of the creeping section in re- sponse to strain changes accompanying the Loma Prieta event. Induced additional postseismic surface slip nowhere exceeds 4 cm, indicating that much of the potential fault displacement in the northwestern creeping section has not occurred. Whereas postseismic displacements up to Sep- tember 1992 have attained only 10 to 20 percent of mod- eled values, the spatial distribution of postseismic surface slip rate increases reflects the results of these simple mod- els. In particular, potential slip and creep rates are pre- dicted to be much reduced south of the San Andreas intersection with the Calaveras fault. Changes in borehole strain and creep rates recorded along a 3.1-km-long sec- tion of the northwestern San Andreas creeping section provide strong indications of the depths of pre- and post- event shallow surface creep but satisfy a wide range of models in which the fault may be slipping at depth. The substantial difference in creep rate in that region suggests that observed strain changes may be locally enhanced by elastic inhomogeneity or local fault-slip processes. INTRODUCTION The Loma Prieta earthquake was not associated with a throughgoing surface rupture even though numerous dis- continuous surface cracks were manifest near the surface fault trace (U.S. Geological Survey Staff, 1990). It is pos- sible that distributed shear could have absorbed surface displacements over a several kilometer wide zone, although geodetic data indicate that the Loma Prieta rupture did not reach the surface (Lisowski and others, 1990). In sev- eral recent strike-slip earthquakes in California, the am- plitude of surface fault displacement has increased in the weeks to years following the main shock (Smith and Wyss, 1968; Burford, 1972; Cohn and others, 1982; Bilham, 1989; Sharp and others, 1989; Williams and Magistrale, 1989). Anticipating the possible development of post-event surface slip, we installed three digital creepmeters (Bilham, 1989) within 2 to 17 days after the Loma Prieta earth- quake. Two of these, Madonna Road and Chittenden Bridge, were installed over the southeastern rupture zone, whereas the other, at Nyland Ranch, was installed near the northwestern end of the San Andreas fault creeping section (Behr and others, 1990) (fig. 1). Although the two instruments over the rupture zone have not recorded significant post-event tectonic slip, the D179 D180 Nyland Ranch creepmeter and several USGS creepmeters installed on the San Andreas fault creeping section southeast of the rupture zone (creepmeters XSJ2, XHR2, CWC3, XFL1 and XMR1) have recorded substantial changes in creep rate since the event. The form of this creep signal exhibits a decay in slip rate characteristic of postseismic slip observed following other earthquakes (Smith and Wyss, 1968; Sylvester, 1986; Marone and others, 1991). Although the term "afterslip" has hitherto been used to describe this characteristic surface slip be- havior over the rupture zone, it conveniently describes the impulsive initiation of slip and subsequent decay in slip rate observed on adjacent fault segments. In the case of the Loma Prieta earthquake, however, the increased sur- face slip rates on the northwestern 50 km of the creeping section have not continued to decay toward pre-event levels, but have remained elevated and roughly constant AFTERSHOCKS AND POSTSEISMIC EFFECTS during the =500 to 1,100 days following the main shock. The response of the creeping section of the San An- dreas fault to previous earthquakes was examined by Mavko and others (1985), Burford (1988), and Simpson and others (1988). The northwestern creeping section, which terminates a few kilometers northwest of San Juan Bautista, defines an area of high surface creep rate on the San Andreas fault. The Loma Prieta earthquake epicenter was located approximately 50 km northwest of San Juan Bautista, and its aftershocks define a rupture zone that extends to within approximately 10 km of that town (fig. 1). In this paper, we describe the nature of fault creep observed in the northwestern 50 km of the creeping sec- tion and estimate the additional fault slip induced by the Loma Prieta earthquake. Using two-dimensional bound- ary element methods, we examine the possible effect of 37° Santa Cruz CU Creepmeter USGS Creepmeter Loma Prieta main shock |. Aftershock limits Creeping section 36° Figure 1.-Location map showing University of Colorado (CU) digital creepmeters and U.S. Geological Survey (USGS) creepmeters used in this study in relation to the Loma Prieta main shock and aftershock zone (shaded), the San Andreas fault creeping section (hachured), and the principal San Francisco Bay area faults. The inset map shows the relationship between creepmeters at Nyland Ranch and XSJ2, and the borehole strainmeter, SJT, near San Juan Bautista; San Juan Bautista Mission is marked by a cross (t). INCREASED SURFACE CREEP RATES ON THE SAN ANDREAS FAULT SOUTHEAST OF THE MAIN SHOCK variable San Andreas and Calaveras fault geometries on the amplitude and spatial distribution of surface creep southeast of the Loma Prieta rupture. We further estimate the potential amplitude of slip induced by the Loma Prieta main shock on the San Andreas fault using simple fric- tionless models of the San Andreas and Calaveras faults. Finally, we attempt to reconcile changes between pre- and postseismic creep and strain rates near San Juan Bautista with changes in the depths of shallow and deep slip on the creeping fault . POSTSEISMIC SURFACE SLIP ABOVE THE SOUTHERN LOMA PRIETA RUPTURE ZONE The longest of the discontinuous surface fractures in the San Andreas fault zone associated with the Loma Prieta earthquake was at Madonna Road (Rymer, 1990). This crack, which displayed right-lateral offset of =2 cm, was instrumented 2 days after the event with a 10-m-long digi- tal creepmeter. Even though right-lateral slip increased by 8 mm on this fracture during the 10 months following the earthquake, slip since September 1990 has displayed sig- nificant apparent left-lateral displacement (fig. 24). We interpret this apparent left-lateral signal as contraction during periods of heavy rain, followed by extension, or right-lateral motion, during dry spells. We suspect that much of the signal monitored by this instrument is related to hillslope instability. The Chittenden rail bridge crossing the Pajaro River at Pajaro Gap was instrumented in early November 1989 on the assumption that the structure crosses the 1906 fault break (Behr and others, 1990). Since mid-1990 the instru- ment has monitored no cumulative displacement other than an annual signal considered to be thermoelastic response of the bridge box-girder. If we consider that signal to be typical annual behavior, then during the first several months of instrument operation this creepmeter recorded a cumulative bridge contraction of approximately 3 cm (fig. 2B). Several mapped strands of the San Andreas fault do not pass beneath the bridge and, although it was pulled from its abutment during the 1906 earthquake, it is pos- sible the Chittenden Bridge may not cross the main active fault trace (Prentice and Schwartz, 1991). ASEISMIC SURFACE SLIP ON THE SAN ANDREAS FAULT SOUTHEAST OF THE LOMA PRIETA RUPTURE A wooden fence at Nyland Ranch (1.4 km northwest of San Juan Bautista Mission), offset at a rate of 8 mm/year between 1942 and 1978 (Burford and Harsh, 1980), repre- D181 sents the northwesternmost location of aseismic creep ob- served on the San Andreas fault. A 43-m-long alignment array was established adjacent to this wooden fence in 1967 (Nason and Tocher, 1970). The periodic measure- ment of this array was supplemented by a biaxial rod extensometer with a dial gauge read-out installed 5 m southeast of the alignment array, and by a continuously operating creepmeter (USGS instrument SJN1) that was discontinued in 1985 due to repeated flooding of the in- strument. Infrequent measurement data from the align- ment array reveal an approximately linear slip rate of 7.0 mm/year prior to 1989, although over short periods the sense of motion fluctuates in rate and sign (fig. 34). Be- tween 1968 and 1985 the creepmeter SJN1 recorded some- what more rapid creep at 8.1+0.2 mm/year, with rare creep events of amplitude approaching 1 mm (Schulz and oth- ers, 1982; Schulz, 1989). An M; =4.6 earthquake near the site in 1972 resulted in accelerated creep with a duration of several months and with a decaying rate typical of earthquake afterslip (Burford and others, 1973). Surface cracks were not present near the alignment ar- ray at Nyland Ranch a few hours after the Loma Prieta main shock, although they were evident on the fault through the town of San Juan Bautista to the southeast. Measurement of the alignment array a few days after the main shock revealed a small (=1 mm) left-lateral signal. On October 22, 1989, we installed a digital creepmeter (Bilham, 1989) next to the east-west arm of the dial-gauge creepmeter to provide a continuous record of creep at this location. During the three years following the Loma Prieta earthquake, the creep rate at Nyland Ranch has averaged more than 20 mm/year (fig. 3). This increased slip rate did not begin until 44 days after the main shock; in the first 10 days of operation the creepmeter recorded less than 0.1 mm of left-lateral signal (possibly a response to instrument installation), subsequently replaced by slow right-lateral displacement of less than 10 microns per day. An abrupt increase in creep rate, which began 44 days after the main shock, exhibits the characteristic decay in rate described as earthquake afterslip. However, the creep rate has not continued to decay in the manner of previ- ously described earthquake afterslip but instead has estab- lished an approximately linear 13.4-mm/year rate between 480 and 1,120 days after the main shock (fig. 3B). If the mean pre-earthquake creep rate of 7 mm/year is subtracted from the mean post-earthquake rate of nearly 20 mm/ year, then an additional 3.9 cm of surface slip is inferred to have been induced by the Loma Prieta main shock in the three years since the event. Rainfall clearly modulates the short-term creep rate at Nyland Ranch. At times of rainfall, the single-component creepmeter displays significant shortening (1-5 mm). Fol- lowing these periods, however, the instrument recovers and resumes recording dextral slip. These rainfall-induced transients have been removed from the record (fig. 3B) to RIGHT LATERAL SLIP, MILLIMETERS BOX GIRDER EXTENSION, MILLIMETERS D182 allow for more precise least-squares estimation of the over- all creep rate. In an attempt to suppress signals related to soil expansion induced by rainfall, the digital creepmeter was supplemented by a differential creepmeter in March 1990 (Bilham and Behr, 1992). Data from the differential AFTERSHOCKS AND POSTSEISMIC EFFECTS creepmeter are significantly less contaminated by rainfall signals. However, the continuous records from this instru- ment and from the single creepmeter have been inter- rupted by flooding of the instruments at times of high water table, a problem responsible for discontinuing op- T Oj TU [_ A IIIIIIIIIIII|II co won e n n n n oar % lllllllllllllllllIIIIIIIIIIIIIL MADONNA ROAD OCT. 20, 1989 TO MARCH 22, 1992 Illllllll}=ul=‘=_fz—IIALIII _2 ac f ppp f ap paps f pp F p aupap f ppp f bu Loba b (u a- 1/1/90 5/1/90 9/1/90 1/1/91 5/1/91 9/1/91 1/1/92 DATE [T L TTT t -f t 1] 40F i B e CHITTENDEN BRIDGE - a i NOV. 4, 1989 TO MARCH 23, 1992 1] 30 [- M - 20 |- - 10 |- 0 |- a 10f- f i [C f ue t u |\ L upa f p pap f ppp p p pab f pupae pupae ful, J ue uc | p pp p p pae f pu t ud 1/1/90 5/1/90 9/1/90 Figure 2.-A, Digital creepmeter record from Madonna Road, 20 Octo- ber 1989 to 22 March 1992. Right-lateral displacement is positive. The cumulative displacement is believed to result from ground failure of the sediments in which the instrument is installed. B, Digital extensometer record from Chittenden Bridge, 4 November 1989 to 23 March 1992. Extension of the bridge and corresponding right-lateral displacement is 1/1/91 5/1/91 9/1/91 1/1/92 DATE positive. The high-amplitude diurnal and annual signal is due to thermal expansion of the 110-m-long box girder to which the instrument rod is attached. The vertical hachured lines in both A and B mark the April 1990 Loma Prieta aftershock sequence. Gaps in the data are caused by instrument malfunction and are constrained by jaw separation readings of the instrument's digital caliper sensor. INCREASED SURFACE CREEP RATES ON THE SAN ANDREAS FAULT SOUTHEAST OF THE MAIN SHOCK eration of USGS creepmeter SJN1 in 1985. The data seg- ments in figure 3 represent a composite record constructed from discontinuous data from both creepmeters. Surface displacements during gaps in the recorded data are recov- D183 ered by measuring changes in the physical separation of the jaws of the digital caliper-displacement sensors. We examine data from the five northwesternmost USGS creepmeters on the San Andreas fault (fig. 1) from the w 200 _— A c2 [ NYLAND RANCH ALIGNMENT ARRAY E 1967-1992 LL] - E - Z] 150 |-- a . bs [ Z [ a 100 |- A - V - [ L U L 55 50 |- r |- D . se) L 1/1/68 1/1/72 1/1/76 I 10d 10" U I Tod I I T Td I I 1d JO E" I I tod | F r I Tod I I P0P 0d I 1d " Od U O'e‘I/Iel/lllllllllllllIllllllllllllllllllIllllllllllllll_ Nyland 7 Composite 7 Loma Prieta - M 7.1 ~ 1/1/80 1/1/84 1/1/88 1/1/92 DATE 60 E a [_- B 5 [ NYLAND RANCH COMPOSITE 50 [- § CF 2 f 3 40}- a . [ / 2 30 / ~ B [ 1] - M . m - 20|- 3 R < L i- 9 o 2 - 2 E A 3 z x w A0 L- § O 3 5 | -< 90 L " 5 € I 10 & F | L I_} | IL |__ 1__| J L_ | L_ I _ L_ |_| 1 I 1 1-1 L1 I_ | i L 1 | L_ 1 | IL-1 l L| L_ 1 L J_ 1 g 0 10 20 30 40 | so - 7 DISTANCE FROM NORTHERN END OF CREEPING SECTION, KILOMETERS Figure 7.-Estimates of potential slip and observed creep rate changes on the northwestern 50 km of the San Andreas fault. Potential slip is estimated using three boundary element models involving different Ca- laveras fault geometries: Calaveras locked, Calaveras approaching San shear | load | from remote antisymmetric displacement N h i // ___ creepmeter \ 1 strainmeter e a Andreas to within 3 km, Calaveras intersecting San Andreas at kilome- ter 27.5. Estimates of the change in creep rate following the Loma Prieta earthquake (column 9 of table 1) are shown as vertical bars. J J [J 0 ia) - las J Figure 8.-Block diagram representing an elastic plate of thickness H undergoing antisymmetric shear displacement with a locked patch between depths b and H-a, after Tse and others (1985). This model is used in the analysis of surface creep and strain rates observed near San Juan Bautista. D190 ciple, by estimating the elastic strain changes induced by the Loma Prieta event at Nyland and XSJ2, it is possible to determine the increment in loading rate and thereby limit the possible range in creep depth on the subsurface fault. We do not pursue this here because the starting geometry is determined by the background loading rate, which is insufficiently established. Certainly additional strain measurements would permit a substantial improve- ment in our ability to interpret the ranges of fault slip at depth. An important feature in the XSJ2 data is the absence of episodic creep events in the data in the 18 months prior to the main shock. Episodic creep is defined as an abrupt increase in slip velocity with a duration of a few hours to a few days, commonly manifest as a transient acceleration followed by a gradual decay in slip rate. In previous years, the slip rate at XSJ2 was formed from a combination of episodic creep and background creep, a slow, continuous slip of the surface fault. Bilham and Behr (1992) pro- posed that steady background creep and episodic creep may be generated at different depths on a creeping fault, and that the ratio of the rates of background to episodic creep is approximately proportional to the ratio of the depths to which these processes occur. The ratio of cumu- lative background creep to cumulative episodic creep re- AFTERSHOCKS AND POSTSEISMIC EFFECTS corded by XSJ2 is 0.29 for the period 1983-1989, and 0.37 after the Loma Prieta main shock, yielding a mean value of 0.33. That is, the long-term contribution to creep from episodic events is three times greater than the con- tribution from background creep. This implies that the depth of episodic creep is three times greater than the depth of background creep. If one of these depths is known, it is possible to estimate the other. From the preceding estimate of depth of shallow creep at XSJ2, the lack of episodic creep in the 18 months prior to the main shock suggests a maximum depth of =500 m for background creep. Following the main shock, episodic creep resumed and we inferred a creeping depth of =1.5 km. Thus, the ratio of depths derived from the long-term ratio of epi- sodic and background creep agrees with the ratio of depths determined from reconciling the strain and creep rates before and after the main shock. We conclude that the creep retardation near San Juan Bautista in the late 1980's, as noted by Sylvester and others (1990) (fig. 4), was caused by the cessation of episodic creep in the depth range 0.5 to 1.5 km. We interpret the return of episodic creep at XSJ2 after the Loma Prieta main shock as the resumption of surface creep to a depth of 1.5 km. Several factors affecting the fault zone could be responsible for arresting episodic creep prior to the main shock, including an in- r-- - 22 surface creep (b) i b T g=1.9 ustrain/year creep 12.9 mm/year \ BASE OF SURFACE CREEP, b, UPPER DEPTH OF SUBSURFACE CREEP, H-a, KILOMETERS (A) RECENT NYLAND AND XSJ2 0 1 2 3 0 Cor- -r or- (B) NYLAND PRE-MAINSHOCK APPLIED STRAIN, MICROSTRAIN PER YEAR - ----@> 1 2 0 1 2 __ surface creep (b) - surface creep (b) g=1.2 pustrain/year creep 7.0 mm/year L creep 3.3 mm/year (C) XSJ2 PRE-MAINSHOCK Figure 9.-Analytic solutions for applied strain and depths of surface (b) and subsurface (H-a) creep that satisfy observed creepmeter and bore- hole strain rates (indicated). The three curves in each graph correspond to the depth of surface creep (b) and the upper depths of a subsurface creeping zone (H-a) for crustal thickness (H) of 10 and 15 km. The vertical bar in 9A corresponds to a fault creeping above and below a region locked between 1.49 km and 4 km in a 15-km-thick elastic plate. In 94, a shear strain loading rate of 1.32 jpustrain/year generates creep of 12.9 mm/year with strain of 1.9 ustrain/year observed 1.13 km from the fault. The same load rate generates similar creep and strain rates in a 10- km-thick elastic plate with a narrower locked region. Examples B and C correspond to the pre-Loma Prieta creep rates observed at Nyland Ranch and XSJ2, respectively, and an observed strain rate of 1.9 pustrain/yr. INCREASED SURFACE CREEP RATES ON THE SAN ANDREAS FAULT SOUTHEAST OF THE MAIN SHOCK crease in fault normal stress, a decrease in pore pressure, or other physical changes in the rheology of the fault zone. Despite the persuasive consistency of the above results and the potential existence of bimodal slip at XSJ2, sev- eral alternative explanations for the observed strain and creep data may be considered. Alignment array data re- corded between 1968 and 1977 at Mission Vineyard Road, site of the XSJ2 creepmeter, indicated a diffuse shear zone (Burford and Harsh, 1980) in which the primary slip zone spanned by XSJ2 (13-m fault normal aperture) exhibited a creep rate of 7.6 mm/yr, while the alignment array (79- m aperture) recorded an average of 13.4 mm/yr of shear deformation. It is therefore possible that the observed pre- event slip deficit at XSJ2 results from a lack of coverage of the entire deformation zone. Whether or not this is the case, XSJ2 has been recording all slip within + 40 m of the primary fault trace since Loma Prieta. This seems especially to be the case during a period of significantly increased surface slip from December 1992 through June 1993, when records from XSJ2 and the Mission Vineyard Road alignment array indicate 16.8 mm and 16.9 mm of right-lateral shear, respectively (Galehouse, oral commun., 1993). The suggestion that XSJ2 was not monitoring the entire deformation zone prior to Loma Prieta, but has been since, requires a change over time in the location, or the nature, of principal shear deformation at that lati- tude. Additional alternatives may be assessed by examining transient strain and creep data at the time of episodic creep events to determine the regions where such events are generated. The very low strain rates recorded by the borehole strainmeter prior to 1988 require a different depth of slip to be applicable at that time. Strain rates recorded by SJT are lower than the 0.3 pustrain/year measured geo- detically by the USGS in the region northwest of San Juan Bautista (Lisowski and Prescott, 1981; Prescott and others, 1981). If we assume that the applied strain rate from 1986 to 1988 was 0.3 pustrain/year, then an esti- mated maximum observed borehole strain rate of 0.1 pustrain/year requires the fault at Nyland Ranch to slip above 4.8 km and below 5.1 km, whereas under the same conditions, the slip rate of 3.3 mm/year at XSJ2 implies that the fault is locked below 4.8 km. These large depths for surface creep appear to us implausible in the extreme northwestern end of the creeping section because the sur- face fault is completely locked less than 3 km to the north- west of Nyland Ranch. We conclude that the borehole strain rates reported by Gwyther and others (1992) may signify changes in the fault zone which are more complex than those modeled here. In the 6 months prior to the Loma Prieta event, slip was apparently confined to shal- low depths. The transient responses of the creepmeters in the months following the main shock confirm that the fault is responding differently along strike to strain changes induced by the Loma Prieta earthquake. D191 CONCLUSIONS Aseismic slip rates on the northwestern part of the San Andreas fault creeping section in central California in- creased substantially following the Loma Prieta earthquake and continue, three years after the main shock, to be above their mean pre-earthquake values. Farther southeast in cen- tral California, creep rates are substantially unaltered by the earthquake. Our two-dimensional elastic models indi- cate that maximum displacements of 20 cm on the north- western creeping section will result from strain changes accompanying the Loma Prieta main shock. Although these models differ from the creeping San Andreas and Cala- veras fault systems in that they are simulated as friction- less, vertical planes, they suggest that southeast of the San Andreas fault's intersection with the Calaveras fault, displacements induced on the creeping section should be significantly reduced in comparison to those northwest of the intersection. This finding qualitatively agrees with ob- served creep-rate increases. At current surface creep rates, the induced fault displacements predicted by two-dimen- sional models of the Loma Prieta earthquake will take a minimum of two decades to eliminate by creep processes alone, although three-dimensional models indicate that most, if not all, of the induced slip will occur within the next few years. Another possibility is that slip predicted for the northwestern creeping section, which has not yet occurred aseismically, may occur during future seismic activity. Our two-dimensional analysis of strain and creep data suggests that slip was apparently confined to shallow depths beneath creepmeter XSJ2 during the 6 months prior to the Loma Prieta earthquake. High strain rates reported near the northwestern end of the creeping section by Gwyther and others (1992) can be reconciled with high creep rates by permitting a substantial fraction of the San Andreas fault to slip below a locked zone from 1 to 2 km depth. However, we consider that the substantial aseismic slip at depth that is implied by these models is inconsis- tent with the setting at the northwestern end of the creep- ing section in that surface creep is not observed more than a few kilometers northwest of Nyland Ranch. We con- clude that the observed high strain rates recorded near San Juan Bautista must be related to a local strain adjust- ment process, perhaps caused by the concentration of strain near the creeping fault as it adjusts to the new strain con- ditions resulting from the Loma Prieta main shock. ACKNOWLEDGMENTS This work was funded by the U.S. Geological Survey grant 14-08-001-G1876. We wish to thank three anony- mous reviewers for their perceptive, insightful, and help- D192 ful comments. We thank Mr. Frank Avilla and his family for the use of the pasture and barn at Nyland Ranch. REFERENCES CITED Behr, J., Bilham, R., Bodin, P., Burford, R.O., and Burgmann, R., 1990, Aseismic slip on the San Andreas fault south of Loma Prieta: Geophysical Research Letters, v. 17, no. 9, p. 1445-1448. Bilham, R., 1989, Surface slip subsequent to the 24 November 1987 Superstition Hills, California, earthquake monitored by digital creepmeters: Bulletin of the Seismological Society of America, v. 79, no. 2, p. 424-450. Bilham, R., and King, G., 1989a, The morphology of strike slip faults: examples from the San Andreas fault, California: Journal of Geo- physical Research, v. 94, no. B8, p. 10204-10216. 1989b, Slip distribution on oblique segments of the San An- dreas Fault, California: observations and theory: U.S. Geological Survey Open-File Report 89-315, 447 p. Bilham, R., and Behr, J., 1992, A two-layer model for aseismic slip on the Superstition Hills fault, California: Bulletin of the Seismologi- cal Society of America, v. 82, no. 3, p. 1223-1235. Bilham, R., and Bodin, P., 1992, Fault zone connectivity: slip rates on faults in the San Francisco Bay area, California: Science, v. 258, p. 281-284. Burford, R.O., 1972, Continued slip on the Coyote Creek fault after the Borrego Mountain earthquake: U.S. Geological Survey Profes- sional Paper 787, p. 105-111. 1988, Retardations in fault creep rates before local moderate earthquakes along the San Andreas fault system, central Califor- nia: Pure Applied Geophysics, v. 126, nos. 2-4, p. 499-529. Burford, R.O., Allen, S.S., Lamson, R.J., and Goodreau, D.D., 1973, Accelerated fault creep along the central San Andreas fault after moderate earthquakes during 1971-1973, in Kovach, RL., and Nur, Amos, eds., Proceedings of the conference on tectonic prob- lems of the San Andreas fault system: Stanford University Publi- cations in the Geological Sciences, v. 13, p. 268-274. Burford, R.O., and Harsh, P.W., 1980, Slip on the San Andreas fault in central California from alignment array surveys: Bulletin of the Seismological Society of America, v. 70, no. 4, p. 1233-1261. Burford, R.O. and Schulz, S.S., 1988, A current retardation rate of aseismic slip on the San Andreas fault at San Juan Bautista, Cali- fornia: E O S, Transactions of the American Geophysical Union, v. 69, no. 44, p. 1424. Cohn, S.N., Allen, C.R., Gilman, R., and Goulty, N.R., 1982, Preearthquake and postearthquake creep on the Imperial fault and the Brawley fault zone: U.S. Geological Survey Professional Pa- per 1254, p. 161-168. Gladwin, M.T., Gwyther, RL., Higbie, J.W., and Hart, R.G., 1991, A medium term precursor to the Loma Prieta earthquake?: Geophysi- cal Research Letters, v. 18, no. 8, p. 1377-1380. Gwyther, RL., Gladwin, MT., and Hart, R.G., 1992, A shear strain anomaly following the Loma Prieta earthquake: Letters to Nature, v. 356, no. 6365, p. 142-144. Lisowski, M. and Prescott, W.H., 1981, Short-range distance measure- ments along the San Andreas fault system in central California, 1975-1979: Bulletin of the Seismological Society of America, v. 71, no. 5, p. 1607-1624. Lisowski, M., Prescott, W.H., Savage, J.C., and Johnston, M.J., 1990, Geodetic estimate of coseismic slip during the 1989 Loma Prieta, California, earthquake: Geophysical Research Letters, v. 17, no. 9, p. 1437-1440. AFTERSHOCKS AND POSTSEISMIC EFFECTS Marone, C.J., Scholz, C.H., and Bilham, R., 1991, On the mechanics of earthquake afterslip: Journal of Geophysical Research, v. 96, no. B5, p. 8441-8452. Mavko, G.M., 1982, Fault interaction near Hollister, California: Jour- nal of Geophysical Research, v. 87, no. B9, p. 7807-7816. Mavko, G.M., Schulz, S., and Brown, B.D., 1985, Effects of the 1983 Coalinga, California, earthquake on creep along the San Andreas fault: Bulletin of the Seismological Society of America, v. 75, no. 2, p. 475-489. Nason, R.D., and Tocher, D., 1970, Measurement of movement on the San Andreas fault, in Earthquake Displacement Fields and Rota- tion of the Earth, L. Mansinha and others, eds., Riedal, Dordrecht, Holland. Prentice, C.S., and Schwartz, D.P., 1991, Re-evaluation of 1906 sur- face faulting, geomorphic expression, and seismic hazard along the San Andreas fault in the southern Santa Cruz Mountains: Bul- letin of the Seismological Society of America, v. 81, no. 5, p. 1424-1479. Prescott, W.H., Lisowski, M., and Savage, J.C., 1981, Geodetic mea- surement of crustal deformation across the San Andreas, Hayward, and Calaveras faults near San Francisco, California: Journal of Geophysical Research, v. 86, no. B11, p. 10853-10869. Reasenberg, P.A., and Simpson, R.W., 1992, Response of regional seismicity to the static stress change produced by the Loma Prieta earthquake: Science, v. 255, p. 1687-1690. Rymer, M., 1990, Near-fault measurement of postseismic slip associ- ated with the 1989 Loma Prieta, California, earthquake: Geophysi- cal Research Letters, v. 17, no. 10, p. 1789-1792. Schulz, $.S., Mavko, G.M., Burford, R.O., and Stuart, W.D., 1982, Long-term fault creep observations in central California: Journal of Geophysical Research, v. 87, no. B8, p. 6977-6982. Schulz, S.S., 1989, Catalog of creepmeter measurements in California from 1966 through 1988: U. S. Geological Survey Open-File Re- port 89-650, 193 p. Sharp, R.V., K.E. Budding, J. Boatwright, M.J. Ader, M.G. Bonilla, MM. Clark, TE. Fumal, KK. Harms, J.J. Lienkaemper, D.M. Morton, B.J. O'Neill, C.L. Ostergren, D.J. Ponti, M.J. Rymer, J.L. Saxton and J.D. Sims, 1989, Surface faulting along the Supersti- tion Hills fault zone and nearby faults associated with the earth- quakes of 24 November 1987: Bulletin of the Seismological Soci- ety of America, v. 79, no. 2, p. 252-281. Simpson, RW., Schulz, S.S., Dietz, LD., and Burford, R.O., 1988, The response of creeping parts of the San Andreas fault to earth- quakes on nearby faults: two examples: Pure Applied Geophysics, v. 126, nos. 2-4, p. 665-685. Smith, S.W., and Wyss, M., 1968, Displacement on the San Andreas fault subsequent to the 1966 Parkfield earthquake: Bulletin of the Seismological Society of America, v. 58, no. 6, p. 1955-1973. Stein, R.S., King, G.C.P., and Lin, J., 1992, Change in failure stress on the Southern San Andreas fault system caused by the 1992 magni- tude = 7.4 Landers earthquake: Science, v. 258, p. 1328-1332. Sylvester, A.G., 1986, Near-field tectonic geodesy, in Wallace, K.E., Active Tectonics, ed.: Washington, D.C., National Academy Press, p. 164-180. Sylvester, A.G., Burford, R.O., and Schulz, $.8., 1990, Almost no sur- face displacement occurred at San Juan Bautista as a result of the Loma Prieta earthquake: E O S, Transactions of the American Geophysical Union, v. 71, no. 8, p. 290. Tse, S.T., Dmowska, R., and Rice, JR., 1985, Stressing of locked patches along a creeping fault: Bulletin of the Seismological Soci- ety of America, v. 75, no. 3, p. 709-736. U.S. Geological Survey Staff, 1990, The Loma Prieta, California, earthquake: an anticipated event: Science, v. 247, p. 286-293. THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989; EARTHQUAKE OCCURRENCE AFTERSHOCKS AND POSTSEISMIC EFFECTS EFFECT OF THE LOMA PRIETA EARTHQUAKE ON FAULT CREEP RATES IN THE SAN FRANCISCO BAY REGION By Jon S. Galehouse, San Francisco State University CONTENTS Page Abstract D193 Introduction 193 San Andreas fault 195 Hayward fault 195 Calaveras fault 199 Concord fault 202 Green Valley fault 204 Other faults 204 SUMMAry ANd CONCIUSIONS ence noen 204 Acknowledgments 206 References 206 ABSTRACT We have been measuring creep on San Francisco Bay region faults since 1979. Virtually no creep preceded the Loma Prieta earthquake (LPEQ) on the San Andreas fault northwest of the aftershock zone and virtually none has occurred since. In contrast, the post-LPEQ creep rate of 13.5 mm/yr near the northwestern end of the creeping segment of the San Andreas fault near San Juan Bautista is about twice the pre-LPEQ average. About 5 mm of coseismic right slip was triggered here by the LPEQ. Most of the Hayward fault creeps at a long-term rate of about 5 mm/yr. The median rate at five sites along the Hayward fault that we had been measuring for the decade before the LPEQ was 4.9 mm/yr. The median rate at these same rate has been particularly low at a site near the northwest- ern end of the fault in San Pablo and at sites near the ‘southeastern end in Fremont. In fact, left-lateral slip of a | few centimeters associated with the LPEQ may have oc- | |_creep rates during the first 6 months after the quake is in _- Galehouse (1990). curred in southern Fremont along this previously rapid, right-lateral-creeping segment of the fault. Small amounts of left-lateral slip have continued there since the LPEQ (through March 1993). The average creep rate on the southern Calaveras fault in the Hollister area before the LPEQ was 6.4 mm/yr at one site and 12.2 mm/yr at an- sites for the 3.4 yr following the LPEQ is 3.6 mm/yr. The other. More than 10 mm of right-lateral slip was triggered here by the M=6.3 Morgan Hill earthquake (MHEQ) in 1984 and again in 1989 by the M=7.1 LPEQ. Since the LPEQ-triggered slip, one site has slowed to 4.2 mm/yr and the other site has stopped creeping (the one that was creeping at 12.2 mm/yr), which has resulted in a slip defi- cit of more than 2 cm. The Concord fault creeps at an overall long-term average rate of about 3 mm/yr and the Green Valley fault at about 5 mm/yr. Both move episodi- cally and both have post-LPEQ rates of about 1-2 mm/yr slower than the longer-term averages. No creep has oc- curred at a site on the northern Calaveras fault in San Ramon either before or after the LPEQ. The same is true for sites on the noncreeping Seal Cove-San Gregorio, Rodgers Creek, West Napa, and Antioch faults. The changes in creep rates on San Francisco Bay area faults after the LPEQ are consistent with the static shear stress changes estimated for the LPEQ. INTRODUCTION My student research assistants and I have been measur- ing creep (aseismic slip) on active faults throughout the greater San Francisco Bay region since 1979. Over the past 13.5 yr (September 1979 to March 1993) we have made about 1,325 creep measurements at about two dozen sites (fig. 1). About 850 measurements were made in the decade preceding the Loma Prieta earthquake (LPEQ) and about 475 were made in the 3.4 yr following the quake. This data set enables us to determine the detailed creep rates and characteristics at each site and to compare mea- surements made after the LPEQ with those made previ- ously in order to determine the effects, if any, of the LPEQ. An earlier summary of the effects of the LPEQ on We use a theodolite triangulation method (Galehouse and others, 1982), which allows us to determine the amount of strike-slip surface creep by noting changes in angles between sets of measurements taken across a fault at D193 D194 different times. Most of the sites span a fault width of 50- 225 m, but a few must span a greater width because of site considerations. The precision of the measurement method is such that we can detect with confidence any movement more than 1-2 mm between successive mea- surement days. The M=7.1 LPEQ is the largest earthquake to have oc- curred in the San Francisco Bay region since 1906 (Plafker and Galloway, 1989). Before 17 October 1989, we had been remeasuring sites about once every 2-3 months. Be- cause of augmented funding following the LPEQ, we were able to increase the measurement frequency at sites on the San Andreas, Hayward, and southern Calaveras faults to AFTERSHOCKS AND POSTSEISMIC EFFECTS about once every 5-6 weeks through 1990. Since then we are again occupying all sites about once every 2-3 months. Results are presented in figures 2-5, which give the average creep rate at each site as determined by the slope of the least-squares line. These figures also indicate the fault width spanned (W) and the time of the LPEQ (shown as a vertical line). All creep rates presented in this paper should be considered minimum rates because undetected creep on additional active fault traces could be occurring outside of the fault width spanned. Most sites are located in low-relief areas, so creep due to mass movement is probably not significant. Even though our data include some of the area's wettest and driest years, the long-term | ] I 123° 00' 122° 00° 121° 30 380— [, - 37° 30 ® 37° 30- Seal Cove F. \ \_ 22\@ \p (90 \2 g \, |_ ¢, C _ 2. \ o - A , 0 10 20 30 _ \\ M 7.1% . - 87 Kilometers LPEQ 87° - 123° 00 122° 30 |__ 122e 00 Figure 1.-Map of San Francisco State University creep measurement sites (numbered dots). Epicenters and magnitudes are indicated for the 6 August 1979 Coyote Lake earthquake (CLEQ), the 24 April 1984 Morgan Hill earthquake (MHEQ), and the 17 October 1989 Loma Prieta earthquake (LPEQ). EFFECT OF THE LOMA PRIETA EARTHQUAKE ON FAULT CREEP RATES IN THE SAN FRANCISCO BAY REGION trends in creep rates do not seem to be significantly af- fected by these surficial weather conditions. We think that at most sites, the measured creep rate closely approxi- mates the right-lateral, strike-slip component of the tec- tonic creep rate. All creep and slip mentioned in this paper are strike slip with right-lateral slip defined as positive. SAN ANDREAS FAULT We presently measure six sites on the San Andreas fault, five of which are shown in figure 1. Measurements for the past 8.0 yr at Site 14 at the Point Reyes National Seashore Headquarters and for the past 12.9 yr at Site 10 in South San Francisco suggest that the segment of the San An- dreas fault is locked between these two sites that are about 135 km and 80 km northwest of the LPEQ epicenter. Virtually no creep was occurring prior to 17 October 1989 and virtually none has occurred thus far in the 3.4 yr since the quake (fig. 2). The same is true for our most northerly site (18) on the San Andreas fault in the Point Arena area (not shown in fig. 1). Langbein (1990) detected a few millimeters of postseismic slip within a small, multi-kilometer-long geo- detic network spanning the San Andreas fault at the north- western end of the LPEQ aftershock zone in the 7-200 day interval following the quake. In November 1989 in order to detect any post-LPEQ creep closer to the epicen- tral area, we began measuring a previously established U.S. Geological Survey (USGS) site in Woodside (Site 22) slightly northwest of the aftershock zone. Our results for the past 3.3 yr, when compared to unpublished USGS measurements in 1977 and 1980 (R. Burford, oral commun., 1989), show that little creep occurred at this site before, during, or after the LPEQ (through 13 Febru- ary 1993). The rate that we measure is <1 mm/yr (fig. 2). We also established a new site (23) on the San Andreas fault near the southeastern end of the LPEQ aftershock zone just northwest of San Juan Bautista. Virtually no post-LPEQ creep has occurred at this site either (through 20 February 1993). In July 1990, we established Site 25 on the creeping segment of the San Andreas fault just southeast of San Juan Bautista and the LPEQ aftershock zone. This site spans the location of USGS creepmeter XSJ2, which showed a slip of 5.2 mm triggered by the LPEQ shaking (Schulz, 1989). Our measurements give a creep rate of 13.5+1.0 mm/yr for the past 2.6 yr (through 20 February 1993). This rate is about the same as the creepmeter- determined rate of 12.4 mm/yr following the LPEQ and considerably faster than the pre-LPEQ longer-term creepmeter rate of about 7 mm/yr (Schulz, 1989; Breckenridge and Burford, 1990; Gladwin and others, 1991; Breckenridge and Simpson, 1992). Bilham (1992) D195 reports that the creepmeter-determined rate at Nyland Ranch, between our Site 23 and Site 25, was 20 mm/yr in the 2 years following the LPEQ which is approximately 3 times the mean creepmeter rate at this site from 1968-85 (Schulz, 1989). In summary, the San Andreas fault at five sites (18, 14, 10, 22, 23) along the locked portion of the fault both northwest and southeast of the LPEQ aftershock zone does not appear to have been affected by the LPEQ. Virtually no surface creep has been detected thus far. In contrast, the northwestern portion of the creeping segment of the fault at Site 25 near San Juan Bautista had about 5 mm of slip triggered by the LPEQ and has continued creeping at a post-LPEQ rate that is about twice the longer-term, pre- LPEQ creepmeter average. HAYWARD FAULT We have been measuring creep at five sites along the Hayward fault for about 13 years and have determined that the average long-term rate is slightly less than 5 mm/ yr (fig. 3). Although creep characteristics (steady or epi- sodic) differ from site to site, the average rates for all sites are similar. Table 1 gives the relationship between creep on the Hayward fault and the LPEQ and is a sum- mary of much of the following discussion. The Hayward fault at Site 17 at Contra Costa College in San Pablo was creeping at a rate of 4.7 mm/yr for the 9 years before the LPEQ. Since the LPEQ, however, the rate has been only 2.2 mm/yr, which has brought the overall average down to 4.2 mm/yr for the past 12.6 yr. The Hayward fault at Site 13 at Rose Street in Hayward has also moved episodically at an overall average rate of 4.9 mm/yr since June 1980; the same as the average in the 9.3 yr before the LPEQ. With J. Lienkaemper of the USGS, we remeasured curb offsets and old city of Hayward ar- rays at Rose Street in late 1992 and determined that the overall creep rate there since 1930 is 5 mm/yr. A least- squares line through our theodolite-measured data points collected after the LPEQ, however, indicates a rate of 6.8 mm/yr. This higher rate may be partly due to a statistical phenomena because of the episodic nature of creep events at this site. The overall rate of 4.9 mm/yr that includes all the theodolite data is the same as the pre-LPEQ rate. The total displacement since the LPEQ, however, is 20.1 mm in 3.37 yr, which gives an average of 6.0 mm/yr, closer to the long-term average rate but still about 1 mm/yr higher. The Hayward fault at Site 12 at D Street in Hayward has two active traces and moves fairly steadily and uniformly, unlike the episodic creep at Rose Street just 1.3 km to the northwest. At D Street the fault was creeping at a rate of 4.9 mm/yr for the 9.3 yr before the LPEQ. Since the LPEQ, however, the rate has dropped to 3.6 mm/yr, which has D196 AFTERSHOCKS AND POSTSEISMIC EFFECTS SF-18 POINT ARENA AREA (Alder Creek) LPEQ 1.0 + 0.1 mm/yr for 12.0 yrs W = 267.4 meters ne *" - 0 T T T T T T T T T T T T T -10 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 SF-14 POINT REYES NATIONAL SEASHORE HEADQUARTERS 0.6 + 0.1 mm/yr for 8.0 yrs LPEQ - 10 W = 70.6 meters 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 SF-10 SOUTH SAN FRANCISCO (Duhallow Way) -0.1 + 0.0 mm/yr for 12.9 yrs LPEQ 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 SF-22 WOODSIDE (Roberta Drive) . 0.9 + 0.3 mm/yr for 3.3 yrs LPEQ - 10 W = 91.2 meters - Virtually no movement between 16 Feb 77 and 4 Nov 89 I 1 U I I I U I I I I I I -10 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 SF-23 CANNON ROAD (San Juan Bautista area) -0.2 + 0.3 mm/yr for 3.3 yrs LPEQ - 10 W = 88.0 meters - CUMULATIVE RIGHT-LATERAL DISPLACEMENT (MM) I I I I I I I I I I I I I -10 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 40 SF-25 MISSION VINEYARD ROAD (San Juan Bautista area) f 13.5 + 1.0 mm/yr for 2.6 yrs Note: Different vertical [ 80 W = 134.2 meters scale on this graph. LPEQ - 20 - 10 I I I I I I I I I I ** 4 I I 0 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 Figure 2.-San Andreas fault displacement measurements at six sites square. A straight line represents the least-squares fit to all the pre- and between 1980 and 1993. LPEQ = Loma Prieta earthquake of 17 October post-LPEQ measurements at each site and its slope indicates the overall 1989. Squares represent individual measurements at each site. The *1 average creep rate. W is the width of the fault zone spanned by the standard deviation for each measurement almost always falls within its theodolite measurements. EFFECT OF THE LOMA PRIETA EARTHQUAKE ON FAULT CREEP RATES IN THE SAN FRANCISCO BAY REGION D197 80 SF-17 SAN PABLO (Contra Costa College) - so 4.2 + 0.1 mm/yr for 12.6 yrs '_4O W = 106.8 meters F 20 T0 I I I I I I I I I I I I I I ‘20 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 80 SF-13 HAYWARD (Rose Street) - 4.9 + 0.1 mm/yr for 12.7 yrs f 60 W = 153.9 meters h f- 20 [~ 0 T T T T T T T T T T T T T T -20 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 80 SF-12 HAYWARD (D Street) [ 60 4.6 + 0.1 mm/yr for 12.7 yrs . m: 40 W = 136.2 meters f CUMULATIVE RIGHT-LATERAL DISPLACEMENT (MM) LPEQ . f- 0 y- T T T T T T T T T T T T T -20 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 80 SF-02 UNION CITY (Appian Way) f e 4.7 + 0.1 mm/yr for 13.5 yrs F W = 125.2 meters - 40 LPEQ - 20 f- 0 U I I I 1 I I U I I I U I I “20 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 80 SF-01 FREMONT (Rockett Drive) f so 4.6 + 0.1 mm/yr for 13.5 yrs - W = 180.0 meters - 40 - 20 - 0 T T T T T T T - T T T T T T T -20 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 199480 SF-24 FREMONT (Camellia Drive) f so -0.9 + 0.2 mm/yr for 3.0 yrs - 40 W = 88.6 meters LPEQ R I - 20 | 0 U I I I I T I I I I I I I I "20 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 Figure 3.-Hayward fault displacement measurements at six sites between 1979 and 1993. See fig. 2 for further explanation. D198 AFTERSHOCKS AND POSTSEISMIC EFFECTS Table 1.-Hayward fault creep rates measured by theodolite between 1979 and 1993 [See fig. 1 for site locations.] All Pre-Loma Prieta Post-Loma Prieta Measurements Measurement Average Measurement Average Total - Average First Lasts Rate Firsta Last Rate Years _ Rate Site (yr) (yr) __ (mm/yr) (yr) (yr) _ (mm/yr) (mm/yr) 17 1980.609 1989.597 4.7+0.3 1989.866 1993.200 2.2+0.4 12.6 4.2+0.1 San Pablo 13 1980.481 1989.748 4.9+0.1 1989.847 1993.216 6.8+0.6 12.7 4.9+0.1 Hayward Rose St. 12 1980.478 1989.748 4.9+0.1 1989.847 1993.216 3.6+0.3 12.7 4.6+0.1 Hayward D St. 2 1979.729 1989.595 4.7+0.1 1989.808 1993.178 5.0+0.4 13.5 4.7+0.1 Union City 1 1979.726 1989.595 5.4+0.1 1989.808 1993.178 1.1+0.5 13.5 4.6+0.1 Fremont Rockett Dr. 24 1967 1987 9.5+0.3 ° 1990.115 1993.123 -0.9+0.2 3.0 -0.9+0.2 Fremont Camellia Dr. 27 1968 1982 8.2+0.2° 1992.262 1993.123 -4.0+3.8 0.9 -4.0+3.8 Fremont Parkmead- ow Dr. aThe Loma Prieta earthquake occurred at 1989.795. b Pre-Loma Prieta rate from Lienkaemper and others (1991) from curb offsets for the 20 years from 1967-1987. © Pre-Loma Prieta rate from Harsh and Burford (1982) from alignment array data from 1968-1982 at a site 0.3 km southeast of Parkmeadow Drive in Fremont. brought the overall average down to 4.6 mm/yr for the past 12.7 yr. It is interesting to note but difficult to ex- plain why the Rose Street and D Street sites that are so close together and had the same pre-LPEQ creep rate should be so different in the 3.4 yr following the quake, with Rose Street apparently speeding up and D Street slowing down. The Hayward fault at Site 2 at Appian Way in Union City also moves fairly steadily, though not as uniformly as at D Street in Hayward. The LPEQ appears to have had little effect on the fault at Site 2. Even though the post- LPEQ rate is 0.3+0.4 mm/yr higher than the average be- fore the LPEQ, the overall average rate has remained at 4.7 mm/yr since September 1979. The Hayward fault at EFFECT OF THE LOMA PRIETA EARTHQUAKE ON FAULT CREEP RATES IN THE SAN FRANCISCO BAY REGION Site 1 at Rockett Drive in Fremont moves episodically and was creeping at a rate of 5.4 mm/yr for the decade prior to the LPEQ. With J. Lienkaemper of the USGS, we measured the amount of curb offset along Rockett Drive in January 1993 and determined that the overall creep rate since 1964 is 5.3 mm/yr. This projects to a rate of about 5.8 mm/yr for the 25 years prior to the LPEQ, because the extremely low creep rate at this site since the LPEQ has brought the overall average down. The creep rate at Rockett Drive has been only 1.1 mm/yr since the LPEQ, which has reduced the average rate for the past 13.5 yr to 4.6 mm/yr. Although the creep data are equivocal because of their scatter and the normal episodic nature of movement at most sites, there is a possibility that a slowdown in creep on the Hayward fault began before the LPEQ. A crude estimate of the time when a slowdown may have begun can be made by noting the time of the last previous (be- fore the LPEQ) significant creep event at each site. Even though the results are highly subjective and other investi- gators could pick different times for the onset of the pos- sible slowdown, the following times were picked by "eyeball" from figure 3. The creep rate may have slowed about 0.3 yr before the LPEQ at Site 17, about 3.6 yr before at Site 13, about 1.4 yr before at Site 12, about 0.7 yr before at Site 2, and about 1.4 yr before at Site 1. Although these eyeball estimates are suggestive that a slowdown in creep on the Hayward fault may have begun a few months to a few years before the LPEQ, more ob- jective methods of evaluating the onset times of rate changes and a statistical measure of the confidence level of the proposed changes need to be developed before any pre-LPEQ slowdown can be considered likely. At the more recently installed Site 24 at Camellia Drive about 4 km southeast of Rockett Drive (Site 1) in Fre- mont, creep has only been measured subsequent to the LPEQ (first measurement on 11 February 1990). For the past 3 years, the fault has shown 0.9 mm/yr of left-lateral creep. Measurements for the past year at Parkmeadow Drive in Fremont (Site 27) only 0.4 km southeast of Ca- mellia Drive also indicate a small amount of left-lateral creep (not shown on fig. 3). Sites 24 and 27 are along a 4- km-long segment in Fremont near the southeastern end of the fault that before the LPEQ had been creeping right laterally at about 8-11 mm/yr since at least the 1920's YHarsh and Burford, 1982; Burford and Sharp, 1982; Lienkaemper and others, 1991; Lienkaemper and Borchardt, 1992). Lienkaemper and Borchardt (1992) con- sider this rate significant because they think it may reflect the long-term surficial slip rate and the deep slip rate that controls the recurrence between large earthquakes on the Hayward fault. Recent measurements of an offset fence along Union Street in Fremont (about 0.9 km southeast of Rockett Drive and 3.2 km northwest of Camellia Drive) by J. Lienkaemper and us suggest that a few centimeters D199 of left-lateral slip may have occurred here in conjunction with the LPEQ. In summary, our measurements at five sites (17, 13, 12, 2, 1) along the Hayward fault show that the right-lateral creep rate ranged from 4.7 to 5.4 mm/yr for the decade preceding the LPEQ. Equivocal data suggest the possibil- ity that a slowdown in creep may have occurred at these sites a few months to a few years before the LPEQ. No slip events appear to have been triggered at any of these sites by the quake. Since the LPEQ, however, rates have changed significantly at four of these five sites (see table 1). The pre- to post-LPEQ rate decreased in San Pablo from 4.7 to 2.2 mm/yr, at D Street in Hayward from 4.9 to 3.6 mm/yr, and at Rockett Drive in Fremont from 5.4 to 1.1 mm/yr. The rate increased at Rose Street in Hayward from 4.9 to 6.8 mm/yr and stayed nearly constant at 4.7 mm/yr in Union City. In addition, it appears that a few centimeters of left-lateral slip associated with the LPEQ in southern Fremont may have occurred along the previ- ously rapid, right-lateral-creeping southeastern segment of the fault. Slow, left-lateral creep at 0.9 mm/yr has contin- ued since the LPEQ along this segment of the fault at Site 24, and Site 27 has also shown left-lateral creep since we began measurements in April 1992. Historic creep rates for the Hayward fault determined over the years by various investigators from alignment arrays and offset cultural features have been summarized recently by Lienkaemper and others (1991). The rate of right-lateral creep for the 40-50 years before the LPEQ has been between 3.5 and 6.5 mm/yr for most of the Hayward fault, with a higher rate of about 9 mm/yr along a 4-km-long segment in southern Fremont. All of our pre- LPEQ measurements fall within the 3.5-6.5 mm/yr range. As pointed out above, however, several sites along the fault are now creeping at rates below this range, including \ slow left-lateral creep in southern Fremont. This indicates that the LPEQ has had a significant effect on creep rates along the Hayward fault. CALAVERAS FAULT San Francisco State University began measuring creep at Sites 4 and 6 on the southern Calaveras fault in the Hollister area (fig. 1) in 1979. During the 10 years of measurements before the LPEQ, creep at both sites was quite episodic, with intervals of fast creep typically last- ing a couple months or less, alternating with longer peri- ods when little creep occurs (fig. 4). More specifically, in the decade before the LPEQ, Site 4 along Seventh Street in the city of Hollister had nine episodes of fast creep of about 5 mm or more. This mode of movement had been noted previously by Langbein (1981), based on data from a precision multi-wavelength distance-measuring instrument. He reported that four strain D200 episodes of creep of about 4-6 mm per episode occurred over time intervals of about 2 months or less during 1975 and 1976 along the Calaveras fault in the Hollister area. He concluded that aseismic slip was the dominant mecha- nism for strain release, rather than slip associated with AFTERSHOCKS AND POSTSEISMIC EFFECTS earthquakes. At Site 4, times of relatively rapid move- ment since at least 1979 alternate with intervals of little net movement typically lasting about 8-12 months, with one lasting 2 years between January 1986 and January 1988. The LPEQ apparently triggered 14 mm of right- SF-19 SAN RAMON (Corey Place) 0.2 + 0.1 mm/yr for 12.2 yrs W = 111.1 meters 160 L140 L120 L100 LBO - so - 40 LZO T T T T T T -20 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 SF-06 HOLLISTER (Wright Road) 10.7 + 0.3 mm/yr for 13.4 yrs W = 51.7 meters 160 - 140 L120 L100 - so L60 - 40 - 20 - 0 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 T T T T T T -20 SF-04 HOLLISTER (Seventh Street) 7.1 + 0.1 mm/yr for 13.5 yrs W = 89.7 meters MHEQ 160 - 140 CUMULATIVE RIGHT-LATERAL DISPLACEMENT (MM) T T T T T T -20 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 Figure 4.-Calaveras fault displacement at three sites between 1979 and 1993. MHEQ = Morgan Hill earthquake of 24 April 1984. See fig. 2 for further explanation. EFFECT OF THE LOMA PRIETA EARTHQUAKE ON FAULT CREEP RATES IN THE SAN FRANCISCO BAY REGION lateral slip at Site 4 that marked the abrupt end of an interval of slow creep that had persisted for about a year. Total cumulative displacement from the time of the first measurement in 1979 went from 67 to 81 mm. Following this triggered slip, Site 4 has been creeping at a rate of 4.2 mm/yr for the past 3.4 yr (through 13 March 1993). This is considerably lower than the pre-LPEQ rate of 6.4 mm/yr. Creep at Site 6 along Wright Road, located 2.3 km northwest of Site 4, included 11 episodes of fast creep of about 5 mm or more in the 10 years before the LPEQ. This faster creep alternates with intervals of little net move- ment that typically last about 3-12 months at Site 6, with one lasting 18 months between June 1985 and December 1986. The LPEQ marked the end of an interval of slow creep that had persisted for about a year at Site 6, similar to the situation at Site 4. The earthquake apparently trig- gered 12 mm of right-lateral slip, and cumulative dis- placement increased from 119 to 131 mm. Following this triggered slip, the fault at Site 6 has virtually stopped creeping for the past 3.4 yr (through 13 March 1993). This has brought the pre-LPEQ average rate of 12.2 mm/ yr down to 10.7 mm/yr for the past 13.4 yr. As shown in figure 4, the fault at Site 6 now has a "slip deficit" of more than 2 cm. Creepmeter data also show that relatively long periods of time with little net movement have occurred along the Calaveras fault in the Hollister area. Evans and others (1981) note that the USGS creepmeter across the Cala- veras fault at Shore Road 10 km north of Hollister showed virtually no net creep for a 3-year period from July 1976 to July 1979. They point out that this period of no move- ment ended with a small slip event coincident with the 6 August 1979 M=5.9 Coyote Lake earthquake (CLEQ) on the Calaveras fault about 12 km northwest of the creepmeter site. Accelerated afterslip in the form of a cluster of creep events followed the small slip event (Ra- leigh and others, 1979). Our theodolite data alone cannot prove that the LPEQ on the nearby San Andreas fault coseismically triggered aseismic slip on the Calaveras fault. Our last measure- ment before the earthquake was on 19 August 1989 and our first measurement after the quake was on 21 October 1989. Strictly speaking, we can only say that 12-14 mm of displacement in the Hollister area occurred between these two dates. Other evidence, however, suggests that the timing of the slip was at least partially coseismic. We detected fresh en echelon cracks on Highway 25 near Hollister on the morning of 21 October 1989. These cracks extended from the asphalt into the dirt shoulder of the road. Fortunately, we were able to photograph them be- fore they were obliterated by rainfall that occurred that afternoon. The cracks had to have formed sometime after the last previous rainfall on 29 September 1989 (McClellan D201 and Hay, 1990) and probably formed shortly after the LPEQ (see the following discussion). The strongest evidence for coseismic slip on the Cala- veras fault due to the LPEQ comes from the USGS creepmeter at Shore Road, 8.5 km northwest of Wright Road (Site 6). The LPEQ occurred at 1704 hours, Pacific Daylight Time (Plafker and Galloway, 1989). The Shore Road creepmeter recorded slip of 5.0 mm within a 10- minute interval from 1700 to 1710 hours (Schulz, 1989). It is unlikely that slip on the Calaveras fault preceded the LPEQ. It is more likely that shaking from the LPEQ trig- gered slip on the Calaveras fault at the exact time of the earthquake or within a few minutes after it. Additional evidence that slip on the Calaveras fault was triggered by the LPEQ comes from field observations made by McClellan and Hay (1990), who observed fresh en ech- elon cracks and offset cultural features that indicated at least 5 mm of movement along 17 km of the Calaveras fault from the city of Hollister northwest to Highway 152. Table 2 shows a comparison of Calaveras fault creep rates in the Hollister area determined by USGS creepmeters and San Francisco State University (SFSU) theodolite mea- surements. Both data sets span 10 years or more and over- lap from the late 1970's to mid 1980's. The SFSU data exclude the LPEQ coseismic slip. Even though the mea- surement methods and time intervals are different, the creep rates determined by the USGS and SFSU are quite similar. Both indicate that the long-term creep rate at Wright Road (Site 6) is about 6 mm/yr faster than the rate at nearby Seventh Street (Site 4). Either the creep rate on the Calaveras fault decreases significantly from Wright Road southeast to Seventh Street or undetected surface movement is occurring outside our 89.7-m-long survey line at Seventh Street, which encompasses the former USGS creepmeter site. Since the slip triggered by the LPEQ, the creep rate at both sites has decreased and the sites with faster and slower creep have reversed. Site 4 is now creeping at 4.2 mm/yr, whereas Site 6 has stopped creeping. In contrast to the evidence for creep and triggered slip along the southern segment of the Calaveras fault near Hollister, Site 19 in San Ramon near the northwestern terminus of the Calaveras fault has shown virtually no creep since measurements began in 1980 (fig. 4). Previous investigators have determined that significant earthquakes occurring on particular faults in California have triggered coseismic slip or afterslip on other nearby faults. Examples of such earthquakes include the Borrego Mountain earthquake of 9 April 1968 (Allen and others, 1972), the Imperial Valley earthquake of 15 October 1979 (Fuis, 1982; Sieh, 1982), the Livermore Valley earthquakes of late January 1980 (Harsh and Burford, 1982), the Coalinga earthquake of 2 May 1983 (Mavko and others, 1985), the Tres Pinos earthquake of 26 January 1986 D202 Table 2.-Calaveras fault creep rates in the Hollister area measured by creepmeter and theodolite between 1970 and 1993 [See fig. 1 for site locations.] Average creep rate Site (mm/yr) Wright Road (SF-6) Usas (1971 - 1983) 13.41 SFSU (1979 - LPEQ)® 12.2 SFSU (LPEQ - 1993)» 0.0 Seventh Street (SF-4) USGS (1970 - 1987)2 6.8 SFSU (1979 - LPEQ)® 6.4 SFSU (LPEQ - 1993)b 4.2 a U. S. Geological Survey (USGS) creepmeter data from Schulz and others (1982) and Schulz (1989). b San Francisco State University (SFSU) theodolite data before and after the Loma Prieta earthquake (LPEQ) from this paper. (Simpson and others, 1988), the North Palm Springs earth- quake of 8 July 1986 (Williams and others, 1988), and the Superstition Hills earthquake of 24 November 1987 (Sharp, 1989). In central California, USGS creepmeters on the San Andreas, Hayward, and Calaveras faults have recorded small, abrupt movements during or shortly after nearby moderate earthquakes as well as increased or decreased creep rates either before or after some of these quakes (Schulz and others, 1982). King and others (1977) point out that during 1971-73, creepmeters along the creeping portion of the San Andreas fault in central California of- ten recorded minor coseismic steps (<1 mm) at times of local earthquakes of magnitudes 4-5. Although most of the nine moderate earthquakes that occurred locally dur- ing this time interval were on the San Andreas fault, some were not. In contrast, many larger creep events ( 21 mm) occurred during this same interval at times of no local seismic events. King and others (1977) concluded that during 1971-73, coseismic slip was negligible compared to slip by episodic and continuous creep processes. They also concluded that creep episodes often follow seismic events at depth, suggesting that some creep may be afterslip delayed by different fault rheology near the surface. In our observations, some phases of rapid creep on the Calaveras fault in the Hollister area are apparently related to nearby seismic events, but others are not. Our data suggest that some significant nearby earthquakes trigger slip on the Calaveras fault, but others do not. In addition AFTERSHOCKS AND POSTSEISMIC EFFECTS to the relationship between the LPEQ and triggered slip on the Calaveras fault discussed above, there also appears to have been a period of rapid creep on the fault near Hollister that was related to the 1984 M=6.3 Morgan Hill earthquake on the Calaveras fault. However, no immedi- ate surface displacement was observed at either of the Hollister sites when they were remeasured on 25 April 1984, the day after the MHEQ (Galehouse, 1987), even though surface displacement had occurred 2.7 km north- west of Site 6 (Galehouse and Brown, 1987). However, within the following 2.5 months, Site 4 showed 11 mm of creep and Site 6 showed 10 mm of creep. This relatively large amount of post-seismic creep was about the same as the slip (13 mm) recorded in the 18 hours following the earthquake at the USGS creepmeter at Shore Road 8.5 km northwest of Site 6 (Schulz, 1987). Since detailed monitoring began more than 20 years ago, the southern Calaveras fault has had several intervals with relatively large amounts of creep (2 1 ecm) occurring within a relatively short time interval (months), but these intervals were apparently unrelated to any nearby signifi- cant earthquakes. Fast creep was recorded by two USGS creepmeters in Hollister in May-August 1977 (Evans and others, 1981). This rapid creep occurred during a 3-year interval when the Calaveras fault at Shore Road, 10 km to the northwest, was in a period of no movement. It also occurred with no apparent relationship to any significant earthquake(s) in the area. Another example of distinctly aseismic rapid creep oc- curred in the 14-month interval between June 1980 and August 1981 (fig. 4). The Calaveras fault at Site 4 showed creep of 19 mm during this interval, more than twice the predicted slip of about 8 mm obtained for this 14-month interval by extrapolating the pre-LPEQ least-squares av- erage rate (6.4 mm/yr). The fault at Site 6 showed creep of 23 mm, also considerably larger than the expected slip of 14 mm, based on the average rate (12.2 mm/yr) for this interval. This rapid creep began 10 months after the M=5.9 CLEQ on the Calaveras fault (epicenter about 30 km north- west of Hollister) and ended about 32 months before the M=6.3 MHEQ on the Calaveras fault (epicenter about 55 km northwest of Hollister). Thus, this interval of rapid creep in the Hollister area appears to be unrelated to ei- ther large regional earthquakes (CLEQ, or MHEQ, or both), or any significant local earthquakes. CONCORD FAULT Sites 3 and 5 are in the city of Concord on what is known as the Concord segment of the Concord fault (Sharp, 1973). A detailed discussion of creep characteris- tics on the Concord fault is in Galehouse (1992). In gen- eral, the Concord fault creeps episodically, with intervals EFFECT OF THE LOMA PRIETA EARTHQUAKE ON FAULT CREEP RATES IN THE SAN FRANCISCO BAY REGION D203 of fast creep of about 7-10 mm over a period of a few months alternating with intervals of slow creep of about 1-2 mm/yr over a period of several years (fig. 5). This pattern of episodic creep was slightly more pronounced during the first eight to nine years of measurements. The average creep rate on the Concord fault for at least the past 13.4 yr is about 3 mm/yr (3.4 at Site 3 and 2.6 at Site 5). Because of the episodic nature of creep on the Con- cord fault and the lack of any triggered slip, it is difficult to assess any effect(s) of the LPEQ. Both sites completed 60 SF-20 GREEN VALLEY FAULT (near Cordelia) F 5.2 + 0.3 mm/yr for 8.6 yrs [ 29 W = 335.8 meters - 40 - 30 - 20 - 10 - 0 I I I I I I mJ I I I I I I I ‘1 o 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 *% => == eme | m p- LJ a> LL 60 _ Q SF-05 CONCORD (Salvio Street) | < 2.6 + 0.1 mm/yr for 13.4 yrs - 50 d W = 57.1 meters [ U - 40 mm - [@] f- 30 wel woe - 20 E - 10 _ _< [ = - 0 | m a m = I I I I I I I I I I I I I I “1 O g 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 CC 60 LA ' ->» SF-03 CONCORD (Ashbury Drive) m 3.4 + 0.1 mm/yr for 13.4 yrs [ 50 5 W = 130.0 meters bao - 3 f a= - 30 _ 22 - O - 20 f- 10 f- 0 I I I I I I -10 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988s 1989 1990 1991 1992 1993s 1994 Figure 5.-Concord fault displacement at two sites between 1979 and 1993 and Green Valley fault displacement at one site between 1984 and 1993 . See fig. 2 for further explanation. D204 their latest phase of fast creep in February 1988 and over- all have been creeping at about their longer-term average rate since then (through 7 March 1993). A least-squares line through only the post-LPEQ data points, however, shows a rate at both sites that is about 1-1.5 mm/yr lower than the long-term average. However, given the episodic nature of the fault movement recorded over the years at Sites 3 and 5, we cannot definitely state that the rate has slowed due to the LPEQ. GREEN VALLEY FAULT Frizzell and Brown (1976) and Helley and Herd (1977) suggest that the southern extension of the Green Valley fault connects to the northern extension of the Concord fault by a relatively short (<3 km) right step (fig. 1). The overall average rate of creep for the past 8.6 yr on the Green Valley fault is 5.2 + 0.3 mm/yr. This is virtually the same rate (5.4 mm/yr between 1922 and 1974) that Frizzell and Brown (1976) determined from offset power transmission lines about 3 km south of Site 20. Although the creep rates are different, creep characteristics of the Green Valley fault are crudely similar to those of the Concord fault (fig. 5), although there is more "noise" in the Green Valley data, which is at least partly due to the relatively large fault width being measured (335.8 m). A detailed comparison is in Galehouse (1991 and 1992). In general, the Green Valley fault also creeps episodically, with intervals of fast creep of 10 mm or more over a few months alternating with intervals of slower creep up to about 3 mm/yr over a few years. In the 8.6 yr since we began our measurements, there have been two intervals of fast creep and three of slow creep. The most recent fast creep interval began sometime between 6 August 1989 and 2 December 1989. Although there is a possibility that shaking from the 17 October 1989 LPEQ triggered the onset of this fast creep as it apparently did in Hollister, it is probably only a coincidence that the Green Valley fault began a period of faster creep at a time that can be crudely related to the quake that was centered more than 125 km away. Our measurement intervals on the Green Valley fault were too far apart to pin down the exact time when rapid creep started, and we know of no other evidence suggesting a relationship. The latest phase of slow creep is continuing, and in fact there has been no net slip on the Green Valley fault for the past 2.5 yr (through 30 January 1993), giving a post-LPEQ rate of only 2.9 mm/yr. OTHER FAULTS We also have measurement sites on the Seal Cove-San Gregorio, Rodgers Creek, West Napa, and Antioch faults AFTERSHOCKS AND POSTSEISMIC EFFECTS (fig. 1). None of these faults show unequivocal evidence of fault creep and none have shown any effects from the LPEQ thus far. SUMMARY AND CONCLUSIONS Reasenberg and Simpson (1992) point out that because the major San Francisco Bay area faults are probably loaded by right-lateral shear, the stress changes due to the LPEQ would either increase the shear load if the change was right-lateral or decrease the shear load if the change was left-lateral. Consequently, a right-lateral static stress change would tend to be associated with a post-LPEQ increase in seismicity rate and an increase in creep rate. A ; left-lateral change would tend to be associated with a de- crease in seismicity rate and either a decrease in creep rate or even a reversal in creep direction if the static stress lchange was particularly significant. Reasenberg and impson (1992) found that their calculations of stress \ changes agreed closely with changes in the regional seis- micity rate after the LPEQ. The data we have been col- lecting since the LPEQ are exactly what are needed to evaluate further the Reasenberg and Simpson (1992) cal- culations for static stress changes. For sites which clearly showed proposed stress changes, we compared our data with the Reasenberg and Simpson (1992) model. None of the stress changes were evidently large enough to cause noncreeping segments of faults to begin creeping. Whether the stress change was right-lateral or left-lateral, none of our measurement sites with pre-LPEQ rates less than 1 mm/yr increased in rate after the LPEQ. In other words, fault segments that were virtually locked before the LPEQ remained locked after the quake. This suggests that none of the noncreeping areas that experienced an increase in right-lateral stress were loaded to their breaking (fault slip) stress level. Those that received a left-lateral change probably had their next time of surface rupture delayed. The amount of static stress change and degree of shaking from the earthquake were not enough to induce move- ment at any of these noncreeping fault sites. In areas that were already creeping, we can compare the creep rates before and after the LPEQ to see if the changes corre- spond to the Reasenberg and Simpson (1992) model. In 11 of 13 cases, the changes in creep rates are consistent in sign with their model. In one additional case the data are consistent with the model, and in only one case is the change inconsistent. The San Andreas fault northwest of the LPEQ after- shock zone was not creeping before the quake, did not have any surface slip during the quake, and is still not creeping. There also has been no post-LPEQ creep at a site (23) just southeast of the aftershock zone. In contrast, about 42 km southeast of the epicenter, approximately 5 EFFECT OF THE LOMA PRIETA EARTHQUAKE ON FAULT CREEP RATES IN THE SAN FRANCISCO BAY REGION mm of slip was triggered by LPEQ shaking near the north- western end of the creeping segment of the fault near San Juan Bautista and the post-LPEQ creep rate at this site (25) is about twice the pre-LPEQ average. It appears that right-lateral shear stress from the Loma Prieta region was transferred to the region southeast of the rupture follow- ing the LPEQ. Most of the Hayward fault creeps at a long-term rate of about 5 mm/yr. Five sites along the fault that we had been measuring for the decade before the LPEQ had rates be- tween 4.7 and 5.4 mm/yr, with a median rate of 4.9 mm/ yr. These same sites have a median rate of only 3.6 mm/ yr during the 3.4 yr since the quake. The post-LPEQ rate is particularly low in San Pablo and Fremont, at either en the fault. A few centimeters of left-lateral slip assoc1ated with the LPEQ may have occurred in southern Fremont, and small amounts of left-lateral creep are con- tinuing. This left-lateral displacement is along the seg- ment of the Hayward fault that was creeping right-laterally at about 9 mm/yr for at least 50 yr before the LPEQ. Simpson and others (1988) point out two possible expla- nations for earthquake-related steps in creep (slip) along faults. Either the slip was triggered by earthquake shaking or the slip occurred in response to changes in static stress fields accompanying the earthquake. This left- lateral slip of a few centimeters in southern Fremont almost certamly that the s segment of the Hayward fault that had been creep- ing right-laterally at about twice the rate on the rest of the fault would have been loaded left-laterally and been primed to move in that direction. The left-lateral slip was almost certamly due to the nature of the static stress changes changes due to unknown factors unrelated to the LPEQ cannot be completely ruled out. A decrease in right-lat- eral static stress on the fault following the LPEQ was suggested by Reasenberg and Simpson (1992) based on their computer simulation of the earthquake and the post- LPEQ decrease in microseismicity along the Hayward fault. Their model assumes low values for the apparent coefficient of friction along the fault and agrees well with our measurements of low creep rates following the LPEQ. The segment of the Hayward fault in Fremont that moved and _is continuing to move left-laterally, however, prob- ably had a larger left-lateral stress change than indicated in the Reasenberg and Simpson (1992) model. The post- LPEQ slowdown in right-lateral creep over much of the rest of the Hayward fault is fully consistent with the sign of the predicted LPEQ stress change in their model. The model, however, does not help explain the puzzling post- LPEQ increase in creep ratg at Rose Street in Hayward or a pre-LPEQ slowdown along the Hayward fault, if that indeed occurred. The average creep rate on the southern Calaveras fault in the Hollister area before the LPEQ was 6.4 mm/yr at D205 Site 4 and 12.2 mm/yr at Site 6. The LPEQ triggered 12- 14 mm of slip here, less than 50 km from the epicenter. A similar amount of slip occurred at these sites after the MHEQ in 1984. It appears that the rapid slip that oc- curred in the Hollister area on the Calaveras fault in con- junction with the MHEQ in 1984 and the LPEQ in 1989 were of different character. The rapid slip associated with the MHEQ occurred during the 2.5 months following the quake, but no slip had occurred when measurements were taken the first day after the quake. This rapid slip was most probably in response to a change in the static stress field. In contrast, the rapid slip associated with the LPEQ almost certainly was triggered by earthquake shaking. For decades, the typical mode of movement has been episodic in the Hollister area, with longer periods of slower creep alternating with shorter periods of faster creep. When no significant earthquakes occur in the Hollister area over a relatively long period, a rapid phase of movement will occur with no association with any seismic event(s). When a significant earthquake occurs on either the Calaveras or another nearby fault and the Hollister area segment of the Calaveras fault is primed to move anyway, the earthquake will probably trigger a relatively rapid phase of move- ment. This may occur as coseismic slip triggered by shak- ing (as with the LPEQ) or as accelerated creep when a change in the static stress field caused by the earthquake is right-lateral (as with the MHEQ). On the other hand, when the fault is not primed, a nearby earthquake will not trigger any slip. Because the lengths of time of the slower and faster phases are so variable, predicting whether or not the southern Calaveras fault is "primed and ready to go" is extremely difficult. At the present time (March 1993), there appears to be a "slip deficit" of more than 2 cm at Site 6 in the Hollister area (fig. 4). A post-LPEQ decrease in the rate of right-lateral creep on the Calaveras fault in the Hollister area is qualitatively consistent with the static stress change following the LPEQ proposed by Reasenberg and Simpson (1992). The observed decrease in creep rates on the Concord and Green Valley faults is also qualitatively consistent with the stress changes estimated for these sites with the Reasenberg and Simpson (1992) model. The noncreeping northern Calaveras, Seal Cove-San Gregorio, Rodgers Creek, West Napa, and Antioch faults were apparently not affected by the LPEQ. In summary, changes in creep rates on San Francisco Bay region faults after the LPEQ are consistent with stress changes estimated for the LPEQ by Reasenberg and Simpson (1992). At one measurement site (25) on a creep- ing fault where Reasenberg and Simpson (1992) show a right-lateral static shear stress increase, the creep rate has doubled since the LPEQ. Among 12 sites on creeping faults where Reasenberg and Simpson (1992) calculate a left-lateral change (a decrease in right-lateral shear load), the post-LPEQ creep rate slowed significantly (>1.2 mm/ D206 yr) at ten of these sites (1, 3, 4, 5, 6, 12, 17, 20, 24, 27), stayed virtually the same at one (2), and increased at only one (13). As Simpson and others (1988) conclude, it is not yet possible to explain unambiguously all the observed creep on certain faults during and after earthquakes on other faults. We do, however, seem to be making some progress in understanding and predicting these relation- ships. ACKNOWLEDGMENTS I thank the many geology students at San Francisco State University who have been instrumental in helping collect this theodolite data since 1979. I am particularly grateful to Carolyn Garrison, Oliver Graves, Theresa Hoyt, and Carl Schaefer, who are presently working with me on this ongoing research effort. I also thank J. Lienkaemper, P. Reasenberg, and R. Simpson of the USGS for helpful suggestions and thoughtful reviews of the manuscript. The present work is supported by the USGS, Department of the Interior, under award number 14-08-0001-G1992. REFERENCES Allen, C.R., Wyss, M., Brune, J.N., Grantz, A., and Wallace, RE., 1972, Displacements on the Imperial, Superstition Hills and the San Andreas faults triggered by the Borrego Mountain earthquake, in The Borrego Mountain earthquake of April 9, 1968: U.S. Geo- logical Survey Professional Paper 787, p. 87-104. Bilham, R., 1992, Creepmeters on California faults: U.S. Geological Survey Open-File Report 92-258, p. 153-156. Breckenridge, K.S., and Burford, R.O., 1990, Changes in fault slip near San Juan Bautista, California, before the October 17, 1989, Loma Prieta earthquake-a possible precursor? [abs.]: Eos (Ameri- can Geophysical Union Transactions), v. 71, no. 43, p. 1461. Burford, R.O., and Sharp, R.V., 1982, Slip on the Hayward and Cala- veras faults determined from offset powerlines, in Hart, EW., Hirschfeld, S.E., and Schulz, S.S., eds., Proceedings - Conference on Earthquake Hazards in the Eastern San Francisco Bay Area: California Division of Mines and Geology, Special Publication 62, p. 261-269. Evans, K.F., Burford, R.O., and King, G.C.P., 1981, Propagating epi- sodic creep and the aseismic slip behavior of the Calaveras fault north of Hollister, California: Journal of Geophysical Research, v. 86, p. 3721-3735. Frizzell, V.A., Jr., and Brown, RD., Jr., 1976, Map showing recently active breaks along the Green Valley fault, Napa and Solano Counties, California: U.S. Geological Survey Miscellaneous Field Studies Map MF-743, scale 1:24,000. Fuis, G.S., 1982, Displacement on the Superstition Hills fault triggered by the earthquake, in The Imperial Valley, California, earthquake of October 15, 1979; U.S. Geological Survey Professional Paper 1254, p. 145-154. Galehouse, J.S., 1987, Theodolite measurements, in Hoose, S.N., ed., The Morgan Hill, California, earthquake of April 24, 1984: U.S. Geological Survey Bulletin 1639, p. 121-123. AFTERSHOCKS AND POSTSEISMIC EFFECTS 1990, Effect of the Loma Prieta earthquake on surface slip along the Calaveras fault in the Hollister area: Geophysical Re- search Letters, v. 17. no. 8, p. 1219-1222. 1991, Present-day creep on the Green Valley fault, in Figuers, S.H., ed., Field Guide to the Geology of Western Solano County: Northern California Geological Society and Association of Engi- neering Geologists Guidebook, p. 12-16. 1992, Creep rates and creep characteristics of eastern San Fran- cisco Bay area faults: 1979-1992, in Borchardt, G. and others, eds., Proceedings of the Second Conference on Earthquake Haz- ards in the Eastern San Francisco Bay Area: California Division of Mines and Geology, Special Publication 113, p. 45-53. Galehouse, J.S., and Brown, B.D., 1987, Surface displacement near Hollister, California, in Hoose, S.N., ed., The Morgan Hill, Cali- fornia, earthquake of April 24, 1984: U.S. Geological Survey Bul- letin 1639, p. 69-72. Galehouse, J.S., Brown, B.D., Pierce, B., and Thordsen, J.J., 1982, Changes in movement rates on certain East Bay faults, in Hart, EW., Hirschfeld, S.E., and Schulz, S.S., eds., Proceedings - Con- ference on Earthquake Hazards in the Eastern San Francisco Bay Area: California Division of Mines and Geology, Special Publica- tion 62, p. 239-250. Gladwin, M.T., Breckenridge, K.S., Hart, R.H.G., and Gwyther, RL., 1991, Recent acceleration of characteristic creep-strain events at San Juan Bautista [abs.]: Eos (American Geophysical Union Transactions), v. 72, no. 44, p. 484. Harsh, P.W., and Burford, R.O., 1982, Alinement-array measurements of fault slip in the eastern San Francisco Bay area, California, in Hart, EW., Hirschfeld, S.E., and Schulz, $.S., eds., Proceedings - Conference on Earthquake Hazards in the Eastern San Francisco Bay Area: California Division of Mines and Geology, Special Publication 62, p. 251-260. Helley, E.J., and Herd, D.G., 1977, Map showing faults with Quater- nary displacement, northeastern San Francisco Bay region, Cali- fornia: U.S. Geological Survey Miscellaneous Field Studies Map MF-881, scale 1:125,000. King, C.Y., Nason, R.D., and Burford, R.O., 1977, Coseismic steps recorded on creep meters along the San Andreas fault: Journal of Geophysical Research, v. 82, p. 1655-1662. Langbein, J.0., 1981, An interpretation of episodic slip on the Cala- veras fault near Hollister, California: Journal of Geophysical Re- search, v. 86, p. 4941-4948. 1990, Post-seismic slip on the San Andreas fault at the north- western end of the 1989 Loma Prieta earthquake rupture zone: Geophysical Research Letters, v. 17, no. 8, p. 1223-1226. Lienkaemper, J.J., and Borchardt, G., 1992, Hayward fault: Large earthquakes versus surface creep, in Borchardt, G. and others, eds., Proceedings of the Second Conference on Earthquake Haz- ards in the Eastern San Francisco Bay Area: California Division of Mines and Geology, Special Publication 113, p. 101-110. Lienkaemper, J.J., Borchardt, G., and Lisowski, M., 1991, Historic creep rate and potential for seismic slip along the Hayward fault, California: Journal of Geophysical Research, v. 96, no. B11, p. 18, 261-18, 283. Mavko, G.M., Schulz, S$.S., and Brown, B.D., 1985, Effects of the 1983 Coalinga, California earthquake on creep along the San An- dreas fault: Seismological Society of America Bulletin, v. 75, p. 475-489. McClellan, P.H., and Hay, E.A., 1990, Triggered slip on the Calaveras fault during the magnitude 7.1 Loma Prieta, California, earth- quake: Geophysical Research Letters, v. 17, no. 8, p. 1227-1230. Plafker, G., and Galloway, J.P., eds., 1989, Lessons learned from the Loma Prieta, California, earthquake of October 17, 1989: U.S. Geological Survey Circular 1045, 48 p. Raleigh, C.B., Stuart, W., and Harsh, P., 1979, Creep on the Calaveras EFFECT OF THE LOMA PRIETA EARTHQUAKE ON FAULT CREEP RATES IN THE SAN FRANCISCO BAY REGION fault near Coyote Lake [abs.]: Eos (American Geophysical Union Transactions), v. 60, p. 890. Reasenberg, P.A., and Simpson, R.W., 1992, Response of regional seismicity to the static stress change produced by the Loma Prieta earthquake: Science, v. 255, p. 1687-1690. Schulz, $.S., 1987, Response of U.S. Geological Survey creepmeters near Hollister, California, in Hoose, S.N., ed., The Morgan Hill, California, earthquake of April 24, 1984: U.S. Geological Survey Bulletin 1639, p. 111-115. 1989, Catalog of creepmeter measurements in California from 1966 through 1988: U.S. Geological Survey Open-File Report 89- 650, 193 p. Schulz, S.S., Mavko, G.M., Burford, R.O., and Stuart, W.D., 1982, Long-term fault creep observations in central California: Journal of Geophysical Research, v. 87, p. 6977-6982. Sharp, R.V., 1973, Map showing recent tectonic movement on the Concord fault, Contra Costa and Solano Counties, California: U.S. D207 Geological Survey Miscellaneous Field Studies Map MF-505, scale 1:24,000. 1989, Pre-earthquake displacement and triggered displacement on the Imperial fault associated with the Superstition Hills earth- quake of 24 November 1987: Seismological Society of America Bulletin, v. 79, p. 466-479. Sieh, KE., 1982, Slip along the San Andreas fault associated with the earthquake, in The Imperial Valley, California, earthquake of Oc- tober 15, 1979; U.S. Geological Survey Professional Paper 1254, p. 155-159. Simpson, R.W., Schulz, S.S., Dietz, L.D., and Burford, R.O., 1988, The response of creeping parts of the San Andreas fault to earth- quakes on nearby faults: two examples: Pageoph, v. 126, nos. 2-4, p. 665-685. Williams, P.L. and others, 1988, Triggered slip along the San Andreas fault after the 8 July 1986 North Palm Springs earthquake: Seis- mological Society of America Bulletin, v. 78, p. 1112-1122. THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989: EARTHQUAKE OCCURRENCE AFTERSHOCKS AND POSTSEISMIC EFFECTS POSTSEISMIC STRAIN FOLLOWING THE LOMA PRIETA EARTHQUAKE FROM REPEATED GPS MEASUREMENTS By Roland Biirgmann, University of California, Davis; Paul Segall, Stanford University; and Mike Lisowski and Jerry L. Svarc, U.S. Geological Survey CONTENTS Page Abstract D209 Introduction 209 Methods: GPS data collection, processing, and error analysis - 211 Data collection 211 Data analysis 211 Data precisiOn ANd ACCUTACY 225 Results 227 Pre-Loma Prieta earthquake displacement field ------------- 227 Post-Loma Prieta earthquake displacement field ------------ 230 Data interpretation And MOGel 233 Discussion 234 Alternative models 234 Thrusting northeast of the San Andreas fault ---------------- 236 Implications for seismic hazard in the San Francisco Bay region 237 Acknowledgments 241 References cited 241 ABSTRACT The postseismic velocity field in the epicentral region of the Loma Prieta earthquake differs significantly from displacement rates measured in the two decades preced- ing the event. The post-earthquake displacement rates along the Black Mountain profile, which crosses the San An- dreas fault 44 km northwest of the Loma Prieta epicenter, do not differ significantly from those determined from 20 years of trilateration measurements. However, station ve- locities along the Loma Prieta profile, which passes through the epicentral region, significantly exceed pre- earthquake rates within 20 km of the Loma Prieta rupture. There is also significant San Andreas normal shortening centered near Loma Prieta. Aseismic right-lateral strike slip at a rate of 0.10 m/yr on the Loma Prieta rupture and reverse slip at a rate of 0.12 m/yr on a reverse fault of the Foothills thrust belt can explain the observed postseismic deformation. Slip within the Foothills thrust belt appears to have been triggered by the stressing during the Loma Prieta main shock. INTRODUCTION Geodetic observations have shown accelerated strain rates adjacent to faults in the years to decades following some large earthquakes (for example, Thatcher, 1986). Following the 1906 San Francisco earthquake, which rup- tured a 400 km segment of the San Andreas fault (Lawson, 1908), shear strain rates near Point Reyes and Point Arena were 2 to 3 times greater than rates measured in the 1970's (Thatcher, 1974). Postseismic transients have also been observed following some subduction-zone earthquakes in Japan (Kasahara, 1981; Thatcher, 1986). Transient postseismic deformation has generally been explained either by viscous relaxation of a ductile (asthenospheric) layer underlying an elastic (lithospheric) plate, or by downward propagation of aseismic slip along an extension of the fault zone into the lower crust (Bott and Dean, 1973; Nur and Mavko, 1974; Thatcher, 1974; Anderson, 1975; Budianski and Amazigo, 1976; Rundle and Jackson, 1977; Savage and Prescott, 1978; Cohen, 1979; Lehner and others, 1981; Thatcher, 1983; Li and Rice, 1987). Unfortunately, there is a lack of adequate data to test and differentiate among the various kinematic models. Earthquakes have been observed to propagate along fault zones, rupturing adjacent fault segments over the time span of several years. Examples involving strike-slip faults have been reported in northeast China preceding the 1975 Haicheng earthquake (Scholz, 1977), and along the North Anatolian (Toksoz and others, 1979), the Calaveras (Oppenheimer and others, 1990), and the San Andreas (Savage, 1971; Wood and Allen, 1973) fault zones. In the 19th century, earthquakes along the San Andreas fault and the Hayward fault in the San Francisco Bay area oc- curred in clusters spread over several years (Ellsworth, 1990). These observations suggest the existence of tran- sient loading processes that occur with characteristic time scales of several years. However, without more data on time-dependent deformation following large earthquakes, we can only speculate on the mechanics of postseismic D209 D210 deformation and the relation between strain transients and the clustering of earthquakes. The Loma Prieta earthquake was the largest earthquake in California in 40 years and the largest earthquake in the San Francisco Bay area since 1906. The horizontal defor- mation field in the epicentral region prior to the 1989 earthquake was already well established on the basis of 20 years of Geodolite measurements (Lisowski and oth- ers, 1991a) and several years of Global Positioning Sys- tem (GPS) measurements (Davis and others, 1989; Lisowski and others, 1990b). Because of the extensive pre-earthquake monitoring, it is possible to identify fea- tures in the present-day deformation field that are attrib- utable to the earthquake. Furthermore, the source characteristics of the Loma Prieta earthquake itself are well established by both seismic and geodetic measure- ments (Beroza, 1991; Hartzell and others, 1991; Steidl AFTERSHOCKS AND POSTSEISMIC EFFECTS and others, 1991; Wald and others, 1991; Lisowski and others, 1990a; Marshall and others, 1991; Williams and others, 1993; Arnadottir and Segall, 1994). The magni- tude of the earthquake, together with the extensive pre- and co-seismic geodetic monitoring, make the Loma Prieta earthquake an important source of data for the study of postseismic processes. The possibility that accelerated postseismic straining could advance the occurrence of other earthquakes in the San Francisco Bay area adds additional importance to the study. Postseismic slip at shallow depths along the trace of the San Andreas fault was generally less than 1 cm in the first year following the Loma Prieta earthquake (Behr and oth- ers, 1990; Langbein, 1990; Rymer, 1990). The cumulative slip of aftershocks on the Loma Prieta rupture and adja- cent faults is about 5 mm (King and others, 1990), which is too small to be detected geodetically. Behr and others y. . 'San Jose,. *%, ‘Black Mtn. A Santa Cruz Mountains Figure 1.- Location map of San Francisco Bay area showing major fault zones and localities discussed in text. BeF = Berrocal fault, SgtFZ = Sargent fault zone, ShF = Shannon fault, ZF = Zayante fault. POSTSEISMIC STRAIN FOLLOWING THE LOMA PRIETA EARTHQUAKE FROM REPEATED GPS MEASUREMENTS (this chapter) find that slip rates on the creeping section of the San Andreas fault near San Juan Bautista acceler- ated from a pre-earthquake rate of 7 to 8 mm/yr to a rate of about 13 mm/yr following the earthquake. About 2 to 3 cm of excess slip occurred on the northernmost 15 km of the creeping section in the 3 years following the earth- quake, probably down to a depth of 1 to 3 km (Behr and others, this chapter; Gwyther and others, 1992). Surficial creep rates on the southern Hayward fault have decreased to about half the pre-earthquake rates of 3 to 9 mm/yr. Near the southeast end of the Hayward fault, where the pre-earthquake-creep rate was highest, the fault has been slipping left-laterally since the earthquake (Lienkaemper and others, 1992) with a possible reversal in early 1993 (Lienkaemper, oral commun.,1993). In this study we report on post-earthquake GPS mea- surements from the Black Mountain profile, which crosses the San Andreas fault 44 km northwest of the Loma Prieta epicenter. We analyze these data together with GPS data from a similar profile (the Loma Prieta profile) which crosses the SAF through the epicentral region (Savage and others, 1994), as well as from the San Francisco Bay area monitor network (fig. 2). The post-earthquake net- work geometry was chosen to measure postseismic tran- sients away from and along the San Andreas fault trace. Twenty years of trilateration measurements prior to the earthquake constrain the preseismic displacement field (Lisowski and others, 1991a). The five-station San Fran- cisco Bay area monitor network and two VLBI (Very Long Baseline Interferometry) sites had also been repeatedly measured for several years prior to the earthquake (Lisowski and others, 1990b; Clark and others, 1990). Our objectives are to characterize the deformation of the crust following the Loma Prieta earthquake, to deter- mine the sources of the postseismic deformation, and to assess the implications of these results for earthquake haz- ards in the San Francisco Bay region. METHODS: GPS DATA COLLECTION, PROCESSING, AND ERROR ANALYSIS DATA COLLECTION A total of 17 occupations during 11 campaigns of the Black Mountain profile (fig. 2) were carried out with dual frequency C/A-code Trimble 4000, P-code TI 4100, and C/A-code Ashtech LM-XII receivers. The results we present here are derived from 9 campaigns between Janu- ary 1990 and January 1993 (fig. 3). We also interpret data from 9 occupations of the Loma Prieta profile during the first 2 years following the Loma Prieta earthquake (Lisowski and others, 1991b; Savage and others, 1994), and 12 to 14 occupations of the San Francisco Bay area D211 monitor network baselines (fig. 4). Table 1 shows the names and site coordinates of the postseismic GPS sta- tions. Table 2 details the occupation dates and receiver type used in the Black Mountain profile. The Loma Prieta profile and the monitor network were surveyed by the U.S. Geological Survey with TI 4100 receivers until Au- gust 1991 and Ashtech receivers since that time. TI 4100 and Ashtech carrier phase and pseudorange data were collected at 30-s intervals for approximately 6 hours, whereas the Trimble receivers recorded data at 15- s intervals for 6 to 14 hours. Occupations exceeding 8 hours in the May 1992 campaign appear to have contrib- uted significantly to the precision (see below) of this sur- vey. Survey times were chosen to maximize the number of satellites visible during an experiment. DATA ANALYSIS Our analysis of the GPS data follows the methods de- scribed by Davis and others (1989). Doubly differenced, ionosphere-free carrier phase observations were processed to estimate station coordinates, tropospheric zenith de- lays, satellite orbit parameters, and integer phase ambigu- ities. We used the Bernese GPS analysis software (Versions 3.2 and 3.3) for parameter estimation (Beutler and others, 1987; Rothacher and others, 1990). Errors in the orbit information broadcast by the GPS satellites can be significantly reduced by monitoring the satellites from global tracking networks with well known station coordinates (for example, Larson and others, 1991). We improved GPS satellite orbits using data from fiducial stations of the CIGNET network; usually Mojave, Cali- fornia, Westford, Massachusetts, and Richmond, Florida. The fiducial sites were constrained to their SV6 coordi- nates (M. Murray, written commun., 1992). During the June 1991 campaign we added data from tracking stations at Penticton and Yellowknife in Canada, and from Kokee in Hawaii to compensate for the lack of data from the three North American stations. In three campaigns (Janu- ary and September, 1990 and February, 1991) we have not been able to produce reliable improved orbits with the tracking data from the CIGNET stations. In these cam- paigns, broadcast orbits with inferred precision of about 5 parts in 10" (Hager and others, 1991) were used. For the Trimble and Ashtech data, we determined re- ceiver clock corrections at each measurement point using C/A-code pseudorange data. For the TI 4100 data we esti- mated corrections with a polynomial in time using P-code pseudorange data (Davis and others, 1989). The data were processed with an automatic program to fix cycle slips, and then all doubly differenced carrier-phase residuals were visually inspected to remove any remaining cycle slips and outliers. D212 All data were processed with a satellite elevation cut- off angle of 15°, because higher cut-off angles appear to reduce the short-term repeatability of the station coordi- nates in the June 1991 survey. Carrier phase data from GPS satellites are known only up to an integer number of wavelengths. We resolve these so-called integer ambigu- ities using phase and pseudorange measurements (Blewitt, 1989). We initially resolved "wide-lane" ambiguities-a AFTERSHOCKS AND POSTSEISMIC EFFECTS linear combination of the L1 and L2 carrier frequencies- with a wavelength of 86 cm. Using the resolved wide- lane ambiguities, we fixed the remaining linearly independent integer-cycle ambiguities using the iono- sphere-free linear combination. This combination of the two L-band frequencies (L3) eliminates the first-order ef- fects of the dispersive ionospheric delays. Practically all integer-cycle ambiguities in the doubly differenced phase 38° 4 o tp z Te a \ ~ g h A - BQ Presidio (VLBI). 0 A residio ( ) * § $6 .- A A f A & \ G* & co f Antelynx Mt, Oso 30 | (£90 Bayrs‘hore A A Mocho _ 0& /g Bend @,Allison Q0 Q/ \® (06k Foothill flag Hamilton | © @&cHaul (bQ "‘ % | OT’UG Bmio gC Halmaur " |- Pigeon Pt. p i 37° |~ i @ GPS site |. A Trilateration site /D RPN | O VLBI site JB; *e |- -! Fort Ord (VLBI) ll 7 30 km "" Brush 2/Fort Ord S (VLBI) | I | 1 l 1 1 | m 30° 122° 30° Figure 2.- Trilateration network, GPS, and VLBI station locations in the south San Francisco Bay-Monterey Bay area. Pre-earthquake data are available from the trilateration network, VLBI measurements at Presidio and Ft. Ord, and GPS occupations of the Bay area monitor network (underlined station names). Postseismic measurements are available from the VLBI sites, the Bay area GPS monitor network, the Black Mountain profile, and the Loma Prieta profile. POSTSEISMIC STRAIN FOLLOWING THE LOMA PRIETA EARTHQUAKE FROM REPEATED GPS MEASUREMENTS D213 m Bend - Pigeon Pt. m -27064.33 ; ; pre ; ; ; 143.5 -27064.34 |- + -27064.35 | 149.4 -27064.36 } -27064.37 | ’ 14 143.3 -27064.38 -27064.39 Up t L -31835.07 |- f 1 41784.87 -31835.08 F -31835.09 | I] 41784.86 -31835.1 | 41784.85 -31835.11 | f ] 41784.84 -31835.12 |- -31835.13. h rhc] 41784,88 & o o r- r- ad ad O o r r O l c & & & & & & & & & & & & som E E 5 Eos o s Hof § Sof 6 & 5 24 56 4 68 3 5 4 58 4 58 m Bend - Halmaur m -22018.63 --- r--p pe- , ; --- 14841 -22018.64 |- -22018.65 148 -22018.66 | -22018.67 |- 147.9 -22018.68 | ~28731.4 t 36198.6 c2g731.41 | Figure 3.- Relative changes in the Cook 36198.59 _ north, east, and vertical baseline com- -28731.42 | ponents and line length changes in the 36198.58 BlackiMountmn profile VIN/1th respect -28731.43 to station Bend as a function of time. The line is a least-squares fit to the -28731.44 36198.57 _ data. Error bars are + 1 sigma based | on the observed residuals to the linear -28731.45 C f P ' 1 36198.56 _ fit. Standard errors are 6 mm in the north and 7.1 mm in the east compo- nent. The line fit in fig. 31 is based on the data collected after a benchmark disturbance at Bayshore in 1990. Oct 17, 89 Apr 17, 90 | Oct 17, 90 | Apr 17, 91 Oct 17, 91 Apr 17, 90 | Oct 17, 90 | Apr 17, 91 Oct 17, 91 Apr 17, 92 |- D214 m -16967.22 -16967.23 -16967.24 -16967.25 -16967.26 -16967.27 -16967.28 -16967.29 -22065.27 -22065.28 -22065.29 -22065.3 -22065.31 -22065.32 -12675.8 ; -12675.81 -12675.82 -12675.83 -12675.84 -12675.85 -16061.53 -16061.54 -16061.55 -16061.56 -16061.57 -16061.58 AFTERSHOCKS AND POSTSEISMIC EFFECTS sould 1 1 h- | Tn | & 0 («] r- v- (l [ad 0 co & & & & & & & N N N N N N N N No ONR ONR ONR OR or ~ 6 a. 6 a. 6 a. 6 a. O < O < o] < O < Bend - True Oct 17, 90 |- Apr 17, 91 Oct 17, 91 Apr 17, 92 |- Oct 17, 92 |- -W- +- B- +--- bo- fifo w a C | C © 1 1 1 im-f-- Oct 17, 89 Apr 17, 90 - Oct 17, 90 |- Apr 17, 91 Oct 17, 91 Apr 17, 92 |- Oct 17, 92 |- Apr 17, 90 - Figure 3.- Continued Oct 17, 90 |- Apr 17, 91 Oct 17, 91 Apr 17, 92 - Oct 17, 92 |- 142.85 142.75 142.65 27834.98 27834.97 27834.96 27834.95 27834.94 569.05 568.95 568.85 20468.86 20468.85 20468.84 20468.83 20468.82 POSTSEISMIC STRAIN FOLLOWING THE LOMA PRIETA EARTHQUAKE FROM REPEATED GPS MEASUREMENTS m Bend - Pawt -11312.23 * ~~ T T T T ~T T T "T T" T -11312.24 -11312.25 | { -11312.26 + 11312.27 - -11312.28 l— L Up L 1171041 |E 6 41171011 1171012 1171013 1171014 E 1171015 r + + + ' ' | o [«] («] vo r- al [sQ o o x r- O [ad c & & & & & & & & & & & & €os 5 § § § § fo § § oR §o5 6 & 58 £ 8 £ 8 4 06 4 6 4 58 Bend - BM10 ~15118.8 e -15118.81 -15118.82 + - -15118.83 | -15118.84 l - -15118.85 |- Up L -10491.95 7 -10491.96 1 -10491.97 -10491.98 -10491.99 4 _10492 1 1 1 1 1 1 1 1 T idl . 1 o o [«] <- r- o [SQ o o r- r- ad sd co & & & & & & & & & & & & sos 5 § § § s § E § 5 fof 6 & 6 £ 58 £ 8 4 5 4 86 4 58 Figure 3.- Continued m 760.75 760.65 760.55 16299.48 16299.47 16299.46 16299.45 16299.44 735.3 735.2 735.1 18417.42 18417.41 18417.4 18417.39 18417.38 D215 D216 m AFTERSHOCKS AND POSTSEISMIC EFFECTS Bend - Foothill 7068.26 ¢ 7068.27 | 7068.28 | 7068.29 | -zo0ss.3 | -7068.31 | +--- 1 ' 7826.91 |- 7826.9 |- 7826.89 |- -z826.88 |- -7826.87 --r -7826.86 - +B +B ~ | C kel I 1 I L Oct 17, 89 Apr 17, 90 |- m Oct 17, 90 |- Apr 17, 91 F Oct 17, 91 Apr 17, 92 |- Oct 17, 92 | Apr 17, 90 p- Oct 17, 90 |- Apr 17, 91 Oct 17, 91 Apr 17, 92 p Oct 17, 92 - Bend - Antelynx 8913.69 --- sois3.6s | sois.67 | sais.66 | sais.65 | 8913.64 ~pr- 10268.25 - 10268.24 | 10268.23 102e8.22 | 10268.21 10268.2 Oct 17, 89 Apr 17, 90 |- Figure 3.- Continued Oct 17, 90 |- Oct 17, 91 Apr 17, 92 p Oct 17, 92 | Apr 17, 90 | Oct 17, 90 p- Apr 17, 91 Oct 17, 91 Apr 17, 92 F- Oct 17, 92 | 96.65 96.55 96.45 10546.59 10546.58 10546.57 10546.56 10546.55 78.1 78 77.9 13597.65 13597.64 13597.63 13597.62 13597.61 POSTSEISMIC STRAIN FOLLOWING THE LOMA PRIETA EARTHQUAKE FROM REPEATED GPS MEASUREMENTS m Bend - Bayshore/Bayshore RM 1 m 5966.11 n T T T T T T T T T T T T 10 s966.1 | I + 5966.09 |- f | 9.9 5966.08 |- 5966.07 |- + 5966.06 |- + + 4 98 5966.05 |- f Up L 7eai.29 | 7 ssse.62 7631.28 |- | 9686.61 7631.27 |- | 9686.6 7631.26 |- 7631.25 |- 1 ge8e.59 7631.24 bocca Placa" 0686.58 & O o r- = O O O =] pref a l (l co & & & & & & & & & & & & No ER OR OR OK OR OK No R OR OR OR oR 6 & 58 £ 8 £ 8 4 6 4 8 4 8 m Bend -LP1 m ~85142.26 - r--p por-, 1032.9 -35142.27 |- J -35142.28 |- 35142.28 1 1032.8 -35142.29 |- -35142.3 + 4—1 1 . -35142.31 |- T 032.7 -35142.32 |- Up E L 16910.62 |- 1 -39013.01 16910.61 |- -39013.02 16910.6 |- -39013.03 16910.59 |- 16910.58 |- 1 -39013.04 16910.57 pocco pocco clau uud) ©39013.05 & 0 o = r O O o o r- r sQ ad co O & & & & & O O & & & & No R ONR OK ONR OR ox OR ONR OR on o & 585 4 8 £ 8 4 6 4 8 £ 8 Figure 3.- Continued D217 D218 m -17761.86 ; -17761.87 | -17761.88 | -17761.89 -17761.9 -17761.91 -17761.92 | -17761.93 | -18398.66 -18398.67 | -18398.68 | -18398.69 | -18398.7 | -18398.71 -18398.72 m ~14135.94 --- -14135.95 -14135.96 t -14135.97 -14135.98 -14135.99 -14136 -14136.01 -7024.04 -7024.05 | -7024.06 | -7024.07 | -7024.08 | -7024.09 | -7024.1 Lp1 - Cliff oy +4 Normal Parallfilé & i 1 1 1 Apr 17, 89 Oct 17, 89 |- Apr 17, 90 |- Oct 17, 90 |- Oct 17, 91 Oct 17, 89 |- Apr 17, 90 - Oct 17, 90 |- Apr 17, 91 Normal Apr 17, 89 p- Oct 17, 89 |- Apr 17, 90 |- Oct 17, 90 |-- Apr 17, 91 Oct 17, 91 Oct 17, 89 |- Apr 17, 90 - Oct 17, 90 | Apr 17, 91 AFTERSHOCKS AND POSTSEISMIC EFFECTS m -25569.37 -25569.38 -25569.39 -25569.4 -25569.41 -25569.42 450.32 450.31 450.3 450.29 450.28 450.27 450.26 450.25 450.24 Oct 17, 91 Apr 17, 92 m -14962.38 -14962.39 -14962.4 -14962.41 -14962.42 -14962.43 -5028.845 -5028.855 -5028.865 -5028.875 -5028.885 -5028.895 -5028.905 -5028.915 -5028.925 Oct 17, 91 Apr 17, 92 Figure 4.- Relative changes in the north, east, San Andreas fault-normal and fault-perpendicular baseline com- ponents in the Loma Prieta profile with respect to station LP1 as a function of time. The line is a least-squares fit to the data. Error bars are + 1 sigma based on Bernese formal errors scaled by a factor of 5. Significant fault-nor- mal motions occur at sites within about 20 km of the Loma Prieta rupture. POSTSEISMIC STRAIN FOLLOWING THE LOMA PRIETA EARTHQUAKE FROM REPEATED GPS MEASUREMENTS m Lp1 - Fire m 6958.03 p ; pope -- , , , - ; , -10773.76 6958.04 |- C 10773.77 -6958.05 | 6958.06 -10773.78 -6958.07 -10773.79 -6958.08 | -10773.8 -6958.09 | F -10773.81 -6958.1 | N Normal Parallel -8278.34 |- 3 933.63 sores : 933.62 -8278.35 | 5 933.61 -8278.36 | 933.6 -8278.37 | j 933.59 -8278.38 | 933.58 R 933.57 8278.39 | 933.56 -8278.4 1 1 1 1 1 1 1 1 1 1 933.55 & & l=] o r- r- & o O r- r- l co co & & & & co & & & & & §oo s oR § Roof og of § of 4 6 £ 85 4 8 6 & 6 4 8 £ m Lp1 - Lp4 m ; ; ; ; --- ~4305.63 -6608.06 1 -4305.64 -6608.07 * 6608.08 1 -4305.65 -6608.09 i 1 -4305.66 6608.1 ¥ + 1 -4305.67 -6608.11 6608.12 -4305.68 F Normal Parallel 519 [ 1 -5039.56 o -5039.57 518.99 -5039.58 518.98 -5039.59 518.97 -5039.6 - -5039.61 518.96 039.6 -5039.62 518.95 i -5039.63 518.94 1 1 1 1 -5039.64 Apr 17, 89 Oct 17, 89 | Figure 4.- Continued Apr 17, 90 F Oct 17, 90 F Apr 17, 91 Oct 17, 91 Oct 17, 89 |- Apr 17, 90 |- Oct 17, 90 F Apr 17, 91 Oct 17, 91 Apr 17, 92 D219 D220 AFTERSHOCKS AND POSTSEISMIC EFFECTS m Lp1 - Lp2 m -629.29 --- , ro- - pr nra al -4472.94 -629.3 | 629.3 E 1 -4472.95 -629.31 |- 2629.32 -4472.96 -629.33 4472.97 -629.34 -4472.98 -629.35 -4472.99 -629.36 @ Normal Parallel 5696.37 | 4 3582.98 3582.97 -5696.38 3582.96 -5696.39 3582.95 -5696.4 3582.94 -5696.41 258293 f 582.92 5696.42 |- + 3 3582.91 -5696.43 poclcc 1 1 1 icc 1 1 1 1 3582.9 & O O O x- x= O o O r- r- l c co & & & & co & & & & & 15 km) slip on the Peninsular segment of the San An- dreas fault at about 20-25 mm/yr (prior was 20+3 mm/yr), (2) deep slip (>11 km) on the Hayward fault at about 14 mm/yr (prior was 9+2 mm/yr), (3) deep slip (>10 km) on the northern Calaveras fault at about 10-12 mm/yr (prior was 6+4 mm/yr) and at up to 28 mm/yr (prior was 17+5 m Fort Ord Fort Ord S Presidio m ‘2'2_"'I"'I"'1"‘I"“"'l".l"'l"' I Cu C 1 C OT C ‘I"‘|"_I"'I"'0'2 -2.25 — : ; Station : 0.15 -2.3 |- offset : F removed 04 -2.35 |- M : - - I R J : I 0.05 ; Earthquake 'Earthquake] p : offset : offset : removed 'removed : : -| -0.05 I : North : | North F East East 0.55 | -| -0.05 0.5 |- | -| -04 0.45 |- I ] | I ; I -015 0.4 0.35 | -| -0.2 0.3 L 1 1 1 1 pal ca pala t pul " fran ao 1 9.25 1983 1984 1985 1986 1987 1988 1989 1990 1991 1984 1985 1986 1987 i988 i989 1990 1991 1992 Figure 6.- Time-displacement plots of pre- and post-Loma Prieta earth- quake VLBI measurements in the San Francisco Bay area (Presidio, Fort Ord, and Fort Ord S), relative to a North American reference frame. The displacements of Fort Ord and Fort Ord S located 8.9 km to the south are shown on the same plot. Loma Prieta earthquake offsets and the offset introduced in the 1988 change of stations (Fort Ord to Fort Ord S) have been removed from the data (Clark and others, 1990). The line fits are based on the pre-earthquake data at Presidio and on the Fort Ord data between 1983 and early 1988. The preseismic and postseismic displace- ment rates at Presidio and at Fort Ord/Fort Ord S$ are not significantly different. D230 mm/yr) on the southern Calaveras fault, and (4) shallow creep on the East Bay and southern San Andreas faults at rates comparable to the a priori values. The deep slip rate on the southern Calaveras fault appears too high and may be related to effects of the Coyote Lake and Morgan Hill earthquakes that occurred during the measurement period. Figure 5 shows a comparison of the measured and mod- eled station velocities. To facilitate the comparison, the model velocities are adjusted so that Loma Prieta moves at the observed velocity. We note that most velocities are fit well within their 95-percent uncertainties except for several of the sites near the southern Calaveras fault. The model also slightly underpredicts the velocities southwest of the San Andreas fault south of Loma Prieta. The kinematic model shown in figure 7 can be used to compute the predicted preseismic velocities of all the sta- tions monitored after the earthquake (table 3). Figure 8 shows the measured postseismic velocities in comparison with pre-earthquake velocities, either measured at these sites or derived from the kinematic model. POST-LOMA PRIETA EARTHQUAKE DISPLACEMENT FIELD Figure 3 shows the changes with time of the coordi- nates of the stations in the Black Mountain profile rela- tive to station Bend. The least-squares estimates of the relative station velocities are shown as solid lines. Within the measurement errors we see no evidence for changes in velocity during the surveyed 3-year period following the AFTERSHOCKS AND POSTSEISMIC EFFECTS earthquake. The average root-mean-square (rms) residu- als about the best fit line are 6.0 mm and 7.1 mm in the north and east, respectively. These residuals are slightly larger than the short-term precision estimates, presumably because of the more poorly determined broadcast orbit solutions and the fact that some error sources (for ex- ample, atmospheric and orbital) are correlated in the day- to-day comparisons (Davis and others, 1989). The estimates of the vertical displacement component do not show a trend above the data noise; the average rms re- sidual about the mean is 3.3 cm. A tie between the two profiles is provided by four mea- surements of the Bend-to-LPI1 baseline (fig. 3J). Time- displacement plots of stations in the Loma Prieta profile relative to station LPI are shown in figure 4. We note that several sites show a significant convergence normal to the San Andreas fault. There is some indication that fault- normal rates decayed during the 2-year period (Savage and others, 1994) at some sites (for example, figs. 4G and H). However, we assume constant postseismic velocities in our analysis. The solution for the baseline to Brush 2 shows a large scatter in the east component (fig. 40), resulting in a weak tie to the VLBI reference frame. Figure 8 shows the postseismic velocity field in the region based on (1) nine occupations of the Black Moun- tain profile between January 1990 and January 1993, (2) nine occupations of the Loma Prieta profile between Oc- tober 1989 and December 1991 (Savage and others, 1994), (3) 12-14 occupations of the San Francisco Bay area moni- tor network between October 1989 and December 1991, (4) four occupations of ties between LP1 and Bend, and ‘ depth (km) | o | 10 | 20 30 CZ 3 | locked 4 12 20 28 36 mm/yr Figure 7.- Slip rates on segments of the San Francisco Bay area fault system used to model the preseismic velocity field. The fault elements that end at 30 km depth in the plot continue to 1000 km depth in the model. POSTSEISMIC STRAIN FOLLOWING THE LOMA PRIETA EARTHQUAKE FROM REPEATED GPS MEASUREMENTS Table 3.- Preseismic and postseismic relative station velocities D231 [The pre-earthquake horizontal velocities were computed from a kinematic model of the preseismic deformation (fig. 7). The postseismic velocities and their uncertainties (1 sigma) were determined from a least-squares fit to the GPS data and their full-position covariances] Baseline Preseismic (cm/yr) Postseismic (cm/yr) From __To North East North +/- __ East +/- Up +/- GPS Bend - - Pigeon Point 0.8 -0.6 0.77 - O.11 -1.41 _ 0.12 1.5 0.9 Bend - - Halmaur 0.8 -0.6 0.75 _ 0.14 -1.69 - 0.14 1.3 1.0 Bend - - Haul 0.7 -0.5 0.51 _ 0.11 -1.28 _ 0.13 1.3 1.0 Bend - - True 0.5 -0.4 0.67 0.09 -0.93 _ 0.09 0.1 0.7 Bend- - BM1O0 0.5 -0.3 0.30 - 0.10 -0.62 - 0.10 _ -0.2 0.8 Bend - - PAWT 0.4 -0.3 0.13 0.10 -0.76 _ 0.10 _ -0.2 0.7 Bend - - Foothill 0.2 -0.1 0.69 0.09 -0.96 0.09 _ -0.2 0.7 Bend - - Bayshore RM 1 -0.2 0.2 -0.38 - 0.29 _ 0.66 _ 0.30 0.6 1.1 Bend - - Antelynx -0.8 0.6 -0.43 _ 0.08 _ 0.38 _ 0.08 0.2 0.6 Bend - - LPI -0.4 0.1 -0.18 - 0.14 -0.04 _ 0.14 _ -0.7 1.1 LPI - _ Cliff 0.5 -0.5 1.37 - 0.26 -2.18 - 0.25 0.7 1.8 LP1 - - Gregor 0.4 -0.4 1.63 - 0.25 -2.32 - 0.25 1.0 1.8 LPI - - Fire 0.3 -0.3 0.52 - 0.19 -2.24 - 0.19 0.7 1.3 LPI - LP2 0.1 -0.1 -0.04 _ 0.24 -1.69 _ 0.24 3.6 1.7 LPI - _ LP4 0.1 -0.1 0.34 - 0.23 -1.56 - 0.23 0.6 1.6 LP1 - _ End 0.0 0.0 0.23 0.25 -0.61 _ 0.23 -1.8 1.7 LP1 - _ Mazzone -0.2 0.2 0.31 _ 0.21 0.89 _ 0.21 2.2 1.7 LP1 - - Calero -0.3 0.3 0.46 _ 0.15 1.51 _ 0.15 0.5 1.2 LP1 - - Coy -0.4 0.4 0.79 0.14 0.86 _ 0.14 2.6 1.1 LP1 - - Hamilton -1.7 1.1 -0.13 - 0.14 _ 1.77 - 0.13 1.4 1.0 LP1 - _ Mocho -2.0 1.1 -0.12 - 0.24 _ 1.41 _ 0.23 0.7 1.6 LP1 - - Mt. Oso -2.0 1.1 -0.32 - 0.27 - 1.48 0.25 1.4 1.7 LPI - - Eagle Rock 0.3 -0.4 1.51 _ 0.21 -1.86 _ 0.17 _ -0.9 1.2 LP1 - _ Allison -1.2 0.7 0.10 - 0.20 _ 0.94 0.16 _ -0.5 1.1 LP1 - - Brush 2 0.8 -1.3 1.08 _ 0.51 -1.84 _ 0.21 2.8 1.4 VLBI Westford Presidio 2.6 -2.1 2.09 0.3 -1.97 0.16 -1.42 - 0.48 Westford Fort Ord 3.8 -3.0 Westford Fort Ord S 3.0 -4.5 4.08 ___ 0.25 __-3.16 _ 0.18 1 __ 1.36 (5) Very Long Baseline Interferometry (VLBI) data col- lected at Fort Ord S and at Presidio between October 1989 and July 1991 (fig. 6, D. Caprette, written commun., 1992). The GPS station velocities were determined using a least squares algorithm that computes station velocities and covariances relative to a local reference station from the station coordinates and the associated covariance ma- trices of the individual campaigns (M. Murray, written commun., 1991). Formal errors from the Bernese GPS solutions were scaled by a factor of five, as they appear to underestimate the observed scatter of the residuals about the solution by about that factor (M. Murray, oral commun., 1992). Table 3 lists the computed relative GPS station velocities and their 1-sigma uncertainties. The velocities of the VLBI stations Fort Ord S and Presidio are shown relative to fixed North American sites (fig. 8, Clark and others, 1990). Because the tie to the VLBI network through Brush 2 is noisy (fig. 40), we present the velocities of the GPS sites holding the veloc- ity vector at Mt. Oso, located about 60 km northeast of Loma Prieta, at its pre-earthquake rate (fig. 8). Several features in the postseismic velocity field (fig. 8) are worthy of note. Stations east of the Calaveras fault as well as those in the Black Mountain profile have ve- locities consistent with their pre-earthquake velocities. Many of the Black Mountain profile sites have moved slightly more northeasterly since the earthquake (fig. 8). This may be caused by the weak tie between Bend and LP1, which is based on only four surveys. Over the 60- km-long Black Mountain profile, from the Pacific coast (Pigeon Pt.) to just east of the Hayward fault (Antelynx), right-lateral shear at 22+5 mm/yr occurred since the earth- D232 quake. The postseismic displacement rates do not differ significantly from the pre-earthquake rate of 20+4 mm over approximately the same region. In contrast, some stations in the Loma Prieta profile show dramatic changes in velocity after the earthquake (Lisowski and others, 1991b; Savage and others, 1994). Stations near Santa Cruz and at Eagle Rock moved ob- AFTERSHOCKS AND POSTSEISMIC EFFECTS liquely towards the fault at an accelerated rate. The most significant difference between the pre- and post-earthquake velocity fields is the fault-normal contraction centered northeast of Loma Prieta mountain. Stations Fire to Coy- ote show significant fault-normal motion (see also fig. 4) that would not have been anticipated from the pre-earth- quake data. {coe GoGo oa I|ll|||ll||lllll||| coon oo oooo foco aao aaae f T I|IIIIlIIIT]|IT1|11¥||III|III|I[\llrlllll]IIIIIIIII|I|II|Ifi1jllllxlll l 50° 40° 30° - 20" -< 10° < 7 37 °-] so: < 1970-1989 \ 1 - velocities \ 7 \ 7 \ 1 __ 5 cmlyr \ sla - a R - 40 - \\ *» Fort Ord \\__ 1 20 km \\\ v Brush 2 x; _ITl'llllllI|IIITIT]IIIIXIIII|I>IIl\\!ll‘!||lli!|l]llll|l‘lll|IlrllllIII]I[I|IIT\]TTI\_ 30° 20° 10' 122° - 50° 40° 30' 20° Figure 8.- Postseismic velocity field determined from GPS measurements. The velocities (solid arrows) and associated errors were computed relative to LP1. A rigid body translation is then added so that Mt. Oso displaces at its pre-Loma Prieta velocity. Postseismic velocities relative to stable North America are shown for the two VLBI stations in the area. The discrepancy between the GPS-derived vector at Brush 2 and the VLBI vector at Fort Ord S, which is not significant at 95- percent confidence, is likely due to the poor results in the Brush 2 - LP1 baseline (fig. 40). Dashed arrows show the preseismic velocities, either measured with error ellipses or computed from the kinematic model (fig. 7), without error ellipses. POSTSEISMIC STRAIN FOLLOWING THE LOMA PRIETA EARTHQUAKE FROM REPEATED GPS MEASUREMENTS There is a large difference between the postseismic ve- locity vectors at the GPS station Brush 2 and the VLBI station at Fort Ord S, less than 100 m away. The discrep- ancy, which is not significant at the 95-percent confi- dence level, is almost certainly associated with errors in the GPS baseline determinations of LP1 to Brush 2. The large scatter in the east component of Brush 2 in the time- displacement plots (fig. 40) may be caused by incorrectly resolved integer phase ambiguities. Whereas only minor changes have been observed in the Black Mountain profile 44 km northwest of the 1989 epi- center, measurements along the Loma Prieta profile show significant changes in displacement rates from pre-earth- quake values. The difference between the post- and pre- earthquake velocities (fig. 9) represents the component of the measured signal that exceeds the secular deformation gol ou IllI|1|IIllll|||l||‘llll|lll|[lIILIIIII «[ I [ --- 30° 20 10° 37 ° IIIIIII‘lIl I]IllllIIIIIIIIIIlllllll‘lllllllll D233 field. The displacement rates predicted by the preseismic model (fig. 7) were subtracted from the postseismic rates, assuming that the motion of Mt. Oso did not change after the Loma Prieta earthquake. The error ellipses were de- termined under the assumption that the errors in the pre- and post-earthquake velocities are equal. DATA INTERPRETATION AND MODEL We model the residual site velocities with rectangular uniform-slip displacement discontinuities embedded in a homogeneous elastic half-space (Okada, 1985). Using a forward-modeling approach, we conclude that two dis- tinct fault planes are required to fit the data. A single fault plane, at any depth, does not produce the observed fault- oc ooc (Goofs ~ Mt. Oso # ----I» Model velocities ---> Post-seismic velocities reduced by pre- seismic rates \\ \\ W. \ C \\ \ \ a TllllljllllljlllIIIllllllllllllllllllllTlllll III]IIIIIXIIIIIIIIIIIIIIIIIIII||I| 20' 10' 122 ° Figure 9.- Excess postseismic station velocities (secular displacement rates subtracted from postseismic rates) assuming the velocity of Mt. Oso was unaffected by the earthquake. Error ellipses for the sites near Loma Prieta are not shown for clarity. They are the same order of magnitude as the error ellipses at other sites. Also shown are the veloci- ties (filled arrows) computed from a dislocation model involving slip on two faults; strike slip on the Loma Prieta fault and thrusting on a reverse IIII|IIIllllllfllllllIIITIIIIIIIW 50' 40" 30° 20' fault in the Foothills thrust system. The two shaded rectangles show the projections of the faults used to compute the modeled station displace- ments. The first fault lies in the Loma Prieta aftershock zone, from 6 to 17 km depth, dipping 70° SW, with a strike-slip displacement rate of 0.10 m/yr. The second fault slips 0.12 m/yr on a reverse fault dipping 47° SW, striking subparallel to the San Andreas fault, from 1.4 to 5.8 km depth. D234 normal and fault-parallel displacements at the surface. As an example, we show the predicted displacement fields from uniform slip on a 5-km-wide, down-dip extension of the Loma Prieta rupture. Figure 10 compares the mea- sured and predicted excess velocities computed from mod- els involving strike slip, dip slip, and oblique slip. The magnitude of the slip is adjusted to provide the best fit (in a least-squares sense) to the residual displacement rates. The models involving a deep source do not fit the data along the Loma Prieta profile well and predict a signifi- cant acceleration along the Black Mountain profile that is not observed in the data. Table 4 lists the model param- eters and misfits to these models. We use a quasi-Newton algorithm (Arnadottir, 1993; Arnadottir and Segall, 1994) to estimate the geometry and position of the dislocation that best fits the data. The al- gorithm allows upper and lower bounds on all of the dis- location parameters. The best-fitting one-fault model with the strike and dip of the Loma Prieta rupture is located at a depth of 13 to 15 km (table 4). This suggests that tran- sient slip may have occurred within rather than below the rupture plane. However, the misfit, , table 4), is 2.3, which suggests that the model is inadequate. More importantly, the spatial distribution of residuals demon- strates that even the best-fitting one-fault model does not provide an adequate fit to the data. We used the quasi-Newton method to investigate the possibility that the data could be satisfactorily fit with multiple faults. In this calculation, one dislocation is con- strained to have the strike and dip of the Loma Prieta rupture, while the other dislocation is not constrained. We found a better fit to the data with two slipping faults (table 4). One dislocation is roughly coincident with the Loma Prieta coseismic rupture, the second is located north- east of the San Andreas fault near the Foothills thrust belt. The surface projections of the two model slip planes and the predicted velocity vectors are shown in figure 9. The best-fit model, (,/RSS/df =1.8), includes a combina- tion of strike-slip at a rate of 0.10 m/yr, from 6 to 17 km depth on the Loma Prieta fault and reverse slip at a rate of 0.12 m/yr on a 47° southwest-dipping thrust plane at shal- low depth (1.4 to 5.8 km) northeast of the San Andreas. The velocities predicted by the model are shown in figure 9. While the model by no means provides a perfect fit, it does explain both the accelerated fault-parallel motion and the fault-normal convergence. The depth of postseismic slip on the Loma Prieta fault is primarily constrained by the narrowness of the zone of accelerated deformation and the absence of increased dis- placement rates on the Black Mountain profile. There is some trade-off between the down-dip fault width and the amount of slip on the faults. For example, models with somewhat wider faults and lower slip rates would pro- duce similar surface displacements. It is possible that lim- ited aseismic slip or ductile flow occurred below the 1989 AFTERSHOCKS AND POSTSEISMIC EFFECTS rupture zone, but that this was masked by shallower de- formation sources. Our inversion results indicate pure strike-slip motion on the Loma Prieta rupture plane since the earthquake. We believe that significant dip slip has not occurred on the Loma Prieta fault as any reverse com- ponent would cause the sites southwest of the San An- dreas to move oceanward (fig. 10B and C), opposite to the observed trend. Finally, the location, strike, and dip- parallel rake of the thrust fault is constrained by the nearly San Andreas fault-perpendicular contraction northeast of the fault. DISCUSSION ALTERNATIVE MODELS The postseismic displacement pattern following the Loma Prieta earthquake has also been interpreted by Linker and Rice (1991, and this chapter) and by Lisowski and others (1991b). Linker and Rice (this chapter) model the displacement field resulting from the interaction of the earthquake rupture with a linear viscoelastic zone that represents the deep aseismic portion of the fault. They also developed a non-linear model that assumes rate- and state-dependent friction on the down-dip aseismic region of the fault zone. Their models, involving a relatively deep deformation source below the coseismic rupture, pre- dict the fault-parallel velocities of the Loma Prieta profile well, but do not predict the observed fault-normal con- traction and appear to overpredict the rates in the Black Mountain profile. Figure 10D shows the displacements predicted by model 1 of Linker and Rice (this chapter, figure 6A) for comparison. As with the deep-fault ob- lique-slip model in figure 10C, their relaxation model is not capable of producing fault-trace normal contraction. Both deep models also overpredict displacement rates in the Black Mountain profile. On the other hand, some of the increased fault-parallel shear observed in the Loma Prieta profile may be explained by the transfer of stress to a ductile region below the fault zone, as proposed by Linker and Rice (this chapter). The model proposed by Lisowski and others (1991b) involves strike-slip shear on the Loma Prieta rupture plane (0.83 m/yr at 4 to 11 km depth), as well as fault-normal compaction of the fault zone at a rate of 0.11 m/yr (3 to 7 km depth). The compaction is required to explain the fault- normal site-velocity vectors. This model successfully pre- dicts the horizontal displacement measurements. Lisowski and others (1991b), however, offer no satisfactory physi- cal model of the proposed fault-zone collapse at decime- ter rates. Our model involves contemporaneous pure strike-slip and thrust faulting at high rates on two faults with similar strikes. The inferred slip on the Loma Prieta fault may D235 POSTSEISMIC STRAIN FOLLOWING THE LOMA PRIETA EARTHQUAKE FROM REPEATED GPS MEASUREMENTS 't 91qt) ut are y-y oJ siopgurered Jopow ay, siy}) sory pur soyur7 jo jopour uonexejar onsejaoosta v 4q porotpaid (() sores pue '(7) dis aszaao1 onbrgo (g) 4juo dis as1oaa1 '(y) Kuo dis ays somof ay) ut moj 10 digs anustasisod Sunuasaidar amdnu ewo7 ay; mojaq aue;d jing} poungq e uo djs 4q poarorpoud pjoy -' amity D236 AFTERSHOCKS AND POSTSEISMIC EFFECTS represent creep around high-slip asperities that released most of the earthquake moment (Beroza, 1991; Hartzell and others, 1991; Steidl and others, 1991; Wald and oth- ers, 1991; Arnadottir and Segall, 1994). We do not have a i - sly e gig-E; S i E g E ® unique explanation for the dominance of the strike-slip ‘§“5§ A component on the Loma Prieta fault. One possibility is i fig E & that postseismic fault creep may be concentrated around g a 2 a § 5 g E E the southern half of the rupture that underwent mostly a £ E m strike slip during the earthquake (Beroza, 1991; Arnadottir 55 i a - |a w in o o and Segall, 1994). A large postseismic strain anomaly, “TE: E 3 C c g S 2 l; < g measured with a borehole tensor strain meter near the San é}: 8 A2 E o Andreas fault about 30 km southeast of the Loma Prieta zo E & - - rays € po ca eplcfzntcrj may: be further evidence of continued aseismic é s é g- cls 5 |I |8 Ela slip in this region (Gwyther and others, 1992). ii | is | |f |F (°° BH ag o joe |e je je THRUSTING NORTHEAST OF THE SAN § A s s el- a f $: |? |? |? |°= ANDREAS FAULT 3 & 5 3 € 338 A 2 Northeast of the San Andreas fault lie several sub-par- fig E sfc |G |3 |2 C 2 allel southwest-dipping faults of the Foothills thrust belt J § £ g # ' ' ' (fig. 1, and Aydin and Page, 1984), with a dominant com- Egg é E ponent of thrust motion (McLaughlin, 1990). Some of .e 3 these faults appear to have been active in the Quaternary, 6 22 5 -la |a |a |a |a 0 as evidenced by offset young alluvial deposits and soil g s E E ° 2 |2 |R |R |R E2 horizons (Haugerud and Ellen, 1990; McLaughlin, 1990). < é g < - Broad zones of contractional surface deformation during 5.2 $ 2 o Z s ls le |R |R t the Loma Prieta earthquake follow the trend of these fault $85 A P| . * f | ' zones intermittently over a total distance of about 20 km £ 3 i? o 2 [s ls |e |8 |8 2 (Haugerud and Ellen, 1990). Our model thrust fault ap- § g g €5 |f I Tle | | a - 18 °° pears to coincide with the Foothills thrust system. §0 sg g 2 > EE # A cross section perpendicular to the San Andreas fault 5 3 E Jig id shows the mapped faults, the post-earthquake seismicity, 8 E $854 |° $, _ s |s (2 12 2 and the model faults found from the inversion of the ex- § s P2 f €. € Ele » | [~ - cess postseismic velocities (fig. 11). The modeled reverse § E 3g § 8 A ? 7 fault lies close to the Berrocal fault zone and has a com- a s 23 f ala |o |om |a |» o parable dip (McLaughlin, 1990). This zone has experi- § £233? a 2°56 ® ° 8 ° °C enced strongly increased earthquake activity since the § 5 is 24 |§ 3 s Loma Prieta event (Reasenberg and Simpson, 1992). Fo- {g E w g 2 - 'B cal mechanisms of all aftershocks greater than M=2.0 along pec the Foothills thrust belt indicate that these events occurred § I; § E “25: - |» |» a é predominantly on reverse faults (fig. 11B). The seismicity i G g $ 2 3 |3 |8 E S near the Loma Prieta rupture, on the other hand, is com- “5 EE $5 $ f $ E 2 plex and includes all types of fault plane solutions § f if? £ E I& |£ 3 € (Oppenheimer, 1990). Note that the fault slip we model § 3 E g r 3 3 e é exceeds by several orders of magnitude the estimated slip fs gg‘cég a 44 J4 |.. accumulated in aftershocks and must be dominantly B ggég _“5°“6~:”‘5§~g 3 aseismic. ‘T pes § 3 3 g 8 25 2|£ E Stress redistribution accompanying the Loma Prieta main 30 § s $ £ pS g 2 g "a g E 3 |3 shock increased the loading on thrust faults northeast and 2 e cEle H&R elLG 12 updip of the rupture (Reasenberg and Simpson, 1992). 2 agi | A SJA "el 1 ca Figure 12 shows a cross section perpendicular to the San Andreas fault, with the contoured coseismic stress changes resolved on planes striking N6O°W and dipping 45° SW overlaid on a plot of the aftershocks. Shear and normal POSTSEISMIC STRAIN FOLLOWING THE LOMA PRIETA EARTHQUAKE FROM REPEATED GPS MEASUREMENTS stress changes both affect the tendency for slip through a Coulomb Failure Function (Reasenberg and Simpson, 1992) CFF = t, + LG,, where t, is the shear traction in the slip direction, 0, the fault-normal traction, and p is the coefficient of friction. A coefficient of friction of 0.6 has been chosen for the calculation in figure 12. The slip distribution during the earthquake is important in determining the induced stress field in the near field of the earthquake rupture. In this calculation we used the heterogeneous fault slip distribu- tion of Beroza (1991). We found that co-seismically in- duced static stress changes of 1 to 2 MPa that enhance thrust loading occurred northeast of the San Andreas fault. The stress changes on fault planes of somewhat different strikes and dips are comparable to those shown in figure 12. We conclude that the increased micro-seismicity and the aseismic thrusting inferred from geodetic data reflect deformation in response to coseismic stress changes. Af- tershock seismicity near the Foothills thrust belt is re- stricted to an approximately 15-km-wide zone along the northern half of the Loma Prieta rupture. This may be related to the observation that most of the thrust compo- nent of the earthquake slip is concentrated northwest of the Loma Prieta epicenter (Beroza, 1991; Steidl and oth- ers, 1991; Marshall and others, 1991; Arnadottir and Segall, 1993). IMPLICATIONS FOR SEISMIC HAZARD IN THE SAN FRANCISCO BAY REGION Continued slip at decimeter rates on faults in the San Francisco Bay region will change the loading of nearby segments of the San Andreas fault and the Hayward fault. We modeled the change in loading on northwest-striking fault planes due to the Loma Prieta earthquake and mod- eled postseismic fault slip with the method outlined above, assuming that fault-parallel shear stress as well as fault- normal stresses determine the fault response through a Coulomb failure criterion. A detailed analysis of coseismic static stress changes on San Francisco Bay area faults shows that micro-seismicity rates increased or decreased in accordance with the computed stress changes (Reasenberg and Simpson, 1992). Reduced creep rates, observed along the southern Hayward fault following the earthquake, are evidence of induced left-lateral shear stress across that segment of the fault (Lienkaemper and others, 1992). Figure 13 shows the contoured changes in frictional failure conditions for right-lateral strike-slip shear on north- west-striking fault planes at a depth of 10 km due to the D237 coseismic rupture (fig. 13A), and from the predicted yearly postseismic fault slip (fig. 13B). Secular tectonic loading is not included, and the modeled postseismic stresses in- dicate a relaxation or enhancement of fault loading above the background. The results are only directly applicable to N45°W-striking fault planes. However, deviations of fault strike by up to 10° do not change the pattern signifi- cantly. We find that the postseismic fault slip further in- creases the load on the San Andreas fault immediately northwest of the Loma Prieta rupture, while the combined coseismic and postseismic effects have a retarding effect on the Hayward fault zone and on most of the Calaveras and San Gregorio-Hosgri fault zones. The yearly postseismic stress changes are about an order of magni- tude less than those induced coseismically. If the Loma Prieta earthquake stress changes advanced the occurrence of the next earthquake north of the rupture by 2-25 years (Reasenberg and Simpson, 1992), then the postseismic changes enhance this effect by about 10 percent per year. Similar postseismic adjustments occurred following the 1979 Coyote Lake and the 1984 Morgan Hill earthquakes along the Calaveras fault (Oppenheimer and others, 1990). While the coseismic stress perturbations on neighboring fault segments induced by these moderate earthquakes were small (Du and Aydin, 1990), postseismic fault creep adja- cent to the coseismic ruptures may have caused the ob- served south-to-north propagation of earthquakes along the Calaveras fault (Oppenheimer and others, 1990). The stress changes caused by coseismic and postseismic fault slip only add to stresses accumulated by background tectonic loading of the San Francisco Bay area faults and to stress changes produced by previous earthquakes (for example, the 1906 San Francisco earthquake). If a fault segment is near instability, the stress release associated with the Loma Prieta earthquake may only slightly retard the imminent failure. The January 15, 1993, Gilroy earth- quake that occurred in a region of induced left-lateral shear east of the southern termination of the Loma Prieta rupture is a timely reminder of this. The Peninsular seg- ment of the San Andreas fault, the updip extension of the Loma Prieta rupture (from about 8 km to the surface), and several thrust faults that parallel the San Andreas fault through most of the San Francisco Bay region appear to represent significant seismic hazards in the wake of the Loma Prieta earthquake. Continued reverse faulting may also relieve normal traction along the southern Hayward fault and increase the likelihood of faulting there, despite a decrease in right-lateral stress. Rapid fault slip along a fault of the Foothills thrust belt since the Loma Prieta earthquake emphasizes the poten- tial earthquake hazards associated with these faults. While records of historic seismicity in the area reach only as far back as the early 19th century, they show that the last 80 years have been marked by unusual seismic quiescence. In 1865, a M=6.5 earthquake may have ruptured a thrust AFTERSHOCKS AND POSTSEISMIC EFFECTS D238 'searpuy ueg ay} JO 1sea -you 7 = J wey} Jojea183 soxyenbyirea Jo suonnjos peoo; pue Ajtorustas -o11u ayenbyprea-jsod jo mata dep; 'g 'z66l 1oqwo0ag nun ou ewo7 ay) Joye yjuow auo sjuaaa pajeoof-[JoM [Je opnjout put O . 1 . 1 1 1 1 1 1 OT a i 1 ___ 1 ___ (g ut xoq) uonoas Jo aur ay} uo palrajuad auoz e wos; are sayenb -yirea ponord oy, jng} reoouag oy} uo dis dip ay; pue jng} vjoug ewo7 ay} uo difs ayLns ay} ajo 'sing} jopour om) ay} are umoys osfy toug ewo7 pue (066] wos payipow) 430j0a83 ay} Jo (z am ut {07 01 108ain) uondas ssord iseaypiou-0}-4samyJnog 'y -J J aindlq OI 0 ~ + 0S _ o | + O'FH _ o + 0°C _ a I + 0°O = SAHAULLLINDVW guoz Jne} .< uouuereys oJ2AOIq UG guoz ne} guoz ne} jeoousg - juabueg nCJ 1 1 1 1 _ 1 1 1 1 1 1 1 1 1 O N 01 - guoz yng} seaipuy ueg POSTSEISMIC STRAIN FOLLOWING THE LOMA PRIETA EARTHQUAKE FROM REPEATED GPS MEASUREMENTS - D239 MAGNITUDES } Figure 11.- Continued AFTERSHOCKS AND POSTSEISMIC EFFECTS D240 pue 'ordonost 'ouseja ue ut sjuawraja j[ng] difs 7/7 SEM (1661) 'g'p = t uonaty Jo juato __ ezorag 4q poutungiap uonngLysip orustasoo ay J, 'sastarout sinoju03 pIJOS -1JJaoa e Surunsse ain{ie} quo[no; t ySnouy} jjng] ay} Jo Surpeof ay) joaJJ3 _ 'q4o UI Saseaioop ajeotput sinojuod payseq 'sour}d jng} Surddip mS uo ayenb sossans seoys jojfered-dip pue sassans jeurou-j[ng} yjog snoauaSowoy _ joing ewo7 ay} Aq poonput (eq)Q UI) saSueyo ssans painoju0) -J amsly (WNY) HDNVILSIG 0€ OC 01 0 | +0'° € - o | +0 C = - +0 Jp => +0° 0 SHMUMLINDVW I OT- am] (WY) HLd4U POSTSEISMIC STRAIN FOLLOWING THE LOMA PRIETA EARTHQUAKE FROM REPEATED GPS MEASUREMENTS fault northeast of the San Andreas fault with two M=5 foreshocks in the preceding year (Tuttle and Sykes, 1992). This earthquake was preceded by a M=7 break in 1838 that involved the Peninsular San Andreas fault segment and maybe the Loma Prieta fault segment and followed in 1868 by a large earthquake on the southern Hayward fault (for example, Tuttle and Sykes, 1992). An earthquake com- parable to the 1865 event in the southern Santa Cruz Moun- tains constitutes a significant hazard to the south San Francisco Bay region. ACKNOWLEDGMENTS We thank Dave Oppenheimer and Greg Beroza for af- tershock data and Doug Caprette for providing the VLBI data. Thora Arnadottir, Yijun Du, Jeff Freymueller, and Mark Murray wrote very useful programs used in our analyses. A large number of Stanford students and faculty helped with the GPS surveys of the Black Mountain pro- file. We benefited from discussions with Mark Linker. Ruth Harris and Mark Murray provided very helpful re- views. We acknowledge support from NSF grants EAR- 9003575, EAR-9002164, and EAR-9116117, and from the U.S. Geological Survey. REFERENCES CITED Anderson, D.L., 1975, Accelerated plate tectonics: Science, v. 187, p. 1077-1079. Arnadottir, T., 1993, Earthquake dislocation models derived from in- version of geodetic data: Stanford, CA, Stanford University, Ph.D. dissertation, 127 p. Arnadottir, T., and Segall, P., 1994, The 1989 Loma Prieta earthquake imaged from inversion of geodetic data: Journal of Geophysical Research, v. 99, p. 21,835-21,855. Aydin, A., and Page, B.M., 1984, Diverse Pliocene-Quaternary tecton- ics in a transform environment, San Francisco Bay region, Cali- fornia: Geological Society of America Bulletin, v. 95, p. 1303- 1317. Behr, J., Bilham, R., Bodin, P., Burford, R.O., and Biirgmann, R., 1990, Aseismic slip on the San Andreas fault south of Loma Prieta: Geophysical Research Letters, v. 17, p. 1445-1448. Beroza, G.C., 1991, Near-source modeling of the Loma Prieta earth- quake: Evidence for heterogeneous slip and implications for earth- quake hazard: Bulletin of the Seismological Society of America, v. 81, p. 1603-1621. Beutler, G., Bauersima, I., Gurtner, W., Rothacher, M., Schildknecht, T., Mader, G.L., and Abell, M.D., 1987, Evaluation of the 1984 Alaska Global Positioning System campaign with the Bernese GPS software: Journal of Geophysical Research, v. 92, p. 1295- 1303. Blewitt, G., 1989, Carrier phase ambiguity resolution for the Global Positioning System applied to geodetic baselines up to 2000 km: Journal of Geophysical Research, v. 94, p. 10187-10203. Bott, N.H.P., and Dean, D.S., 1973, Stress diffusion from plate bound- aries: Nature, v. 243, p. 339-341. Budianski, B., and Amazigo, C., 1976, Interaction of fault slip and D241 lithosphere creep: Journal of Geophysical Research, v. 81, p. 4897-4900. Clark, TA., Ma, C., Sauber, J.M., Ryan, J.W., Gordon, D., Shaffer, D.B., Caprette, D.S., and Vandenberg, N.R., 1990, Geodetic mea- surements of deformation in the Loma Prieta, California earth- quake with Very Long Baseline Interferometry: Geophysical Re- search Letters, v. 17, p. 1215-1218. Cohen, S.C., 1979, Postseismic surface deformations due to lithos- pheric and asthenospheric viscoelasticity: Geophysical Research Letters, v. 6, p. 129-131. Davis, J.L., Prescott, W.H., Svarc, J.L., and Wendt, K.J., 1989, Assesment of Global Positioning System measurements for studies of crustal deformation: Journal of Geophysical Research, v. 94, p. 13635-13650. DeMets, C., Gordon, R.G., Stein, S., and Argus, D.F., 1987, A revised estimate of Pacific-North America motion and implications for western North America plate boundary zone tectonics: Geophysi- cal Research Letters, v. 14, p. 911-914. Du, Y., and Aydin, A., 1990, Stress transfer during three sequential moderate earthquakes along the central Calaveras fault: EOS Transactions American Geophysical Union, v. 71, p. 1652. Du, Y., Aydin, A., and Segall, P., 1992, Comparison of various inver- sion techniques as applied to the determination of a geophysical deformation model for the 1983 Borah Peak earthquake: Bulletin of the Seismological Society of America, v. 82, p. 1840-1866. Ellsworth, W.L., 1990, Earthquake history, 1769-1989; U.S. Geologi- cal Survey Professional Paper 1515, p. 153-181. Erickson, LL., 1987, User's Manual for DIS3D: A three dimensional dislocation program with applications to faulting in the earth: Stanford, CA, Stanford University, M.S. thesis, 167 p. Gwyther, RL., Gladwin, M.T., and Hart, R.H.G., 1992, A shear strain anomaly following the Loma Prieta earthquake: Nature, v. 356, p. 142-144. Hager, B.H., King, RW., and Murray, M.H., 1991, Measurements of crustal deformation using the Global Positioning System: Annual Review of Earth and Planetary Sciences, v. 19, p. 351-382. Hartzell, SH., Stewart, G.S., and Mendoza, C., 1991, Comparison of L, and L, norms in a teleseismic waveform inversion for the slip history of the Loma Prieta, California, earthquake: Bulletin of the Seismological Society of America, v. 81, p. 1518-1539. Haugerud, R.A., and Ellen, S.D., 1990, Coseismic ground deformation along the northeast margin of the Santa Cruz Mountains: U.S. Geological Survey Open File Report 90-274, p. 32-37. Kasahara, K., 1981, Earthquake mechanics: Cambridge, Cambridge University Press, 248 p. King, G.C.P., Lindh, A.G., and Oppenheimer, D.H., 1990, Seismic slip, segmentation, and the Loma Prieta earthquake: Geophysical Research Letters, v. 17, p. 1449-1452. Kovach, RL., and Beroza, G.C., 1993, Seismic potential from reverse faulting on the San Francisco Peninsula: Bulletin of the Seismo- logical Society of America, v. 83, p. 597-602. Langbein, J.O., 1990, Postseismic slip on the San Andreas fault at the northwestern end of the 1989 Loma Prieta earthquake rupture zone: Geophysical Research Letters, v. 17, p. 1223-1226. Larson, K.M., and Agnew, D.C., 1991, Application of the Global Postioning System to crustal deformation measurements 1. Preci- sion and accuracy: Journal of Geophysical Research, v. 96, p. 16547-16565. Larson, K.M., Webb, FH., and Agnew, D.C., 1991, Application of the Global Postioning System to crustal deformation measurements 2. The influence of errors in orbit determination networks: Journal of Geophysical Research, v. 96, p. 16567-16584. Lawson, J.J., 1908, The California earthquake of April 18, 1906: Re- port of the State Earthquake Investigation Commission, Carnegie Institution of Washington, Washington, D.C. D242 Lehner, FK., Li, V.C., and Rice, JR., 1981, Stress diffusion along rupturing plate boundaries: Journal of Geophysical Research, v. 86, p. 6155-6169. Li, V.C., and Rice, J.R., 1987, Crustal deformation in great California earthquake cycles: Journal of Geophysical Research, v. 92, p. 11533-11551. Lienkaemper, J.J., Borchardt, G., and Lisowski, M., 1991, Historic creeprate and potential for seismic slip along the Hayward Fault, California: Journal of Geophysical Research, v. 96, p. 18261- 18283. Lienkaemper, J.J., Galehouse, J.S., and Simpson, R.W., 1992, Hayward fault, California: Longterm creep rates versus slower AFTERSHOCKS AND POSTSEISMIC EFFECTS physical Union Transactions), v. 73, p. 119. Linker, M.F., and Rice, J.R., 1991, The effective viscosity of the San Andreas fault beneath the seismogenic zone: Constraints from the response to the 1989 Loma Prieta earthquake: Eos (American Geophysical Union Transactions), v. 72, p. 310-311. Lisowski, M., Prescott, W.H., Savage, J.C., and Johnston, M.J., 19902, Geodetic estimate of coseismic slip during the 1989 Loma Prieta, California earthquake: Geophysical Research Letters, v. 17, p. 1437-1440. Lisowski, M., Prescott, W.H., Savage, J.C., and Svarc, J.L., 1990b, A possible geodetic anomaly observed prior to the Loma Prieta, California, earthquake: Geophysical Research Letters, v. 17, p. creep since 1989 Loma Prieta earthquake: Eos (American Geo- 1211-1214. 1 ltllllillil“1111|||111|l||11111||l||||||||||||||I|||111||1|||1||||||f 40 tI k \ 10 : -A - I 20-~ ~ _ 40 [_ - 50 [- 20° - E a [L - - 10° 7 - 37 °- - 50' 7 - 30° 20' 10' Figure 13.- A, Horizontal stress changes (in kPa) at 10 km depth on N45°W-striking strike-slip faults caused by the Loma Prieta earthquake. Both fault-normal stresses and strike-parallel shear stresses effect the loading of the faults through a Coulomb failure criterion assuming a coefficient of friction u = 0.6. The peninsular segment of the San An- dreas fault, the northernmost segment of the creeping section of the San \ & - IlllllIIIIIIIIIINIlIIIIIIlllllllll|llll])1||l||ll|IllYIIIII|Illllllll 122 ° 50° 40" 30° Andreas fault, and the central Calaveras fault experienced increased right-lateral stress (positive), whereas the San Gregorio-Hosgri and the southern Hayward fault show a stress decrease. B, Stress changes (in kPa) induced by postseismic slip inferred from geodetic data. The pat- tern of postseismic stress changes is similar, but of lower magnitude, to the coseismically induced static stress changes. POSTSEISMIC STRAIN FOLLOWING THE LOMA PRIETA EARTHQUAKE FROM REPEATED GPS MEASUREMENTS Lisowski, M., Savage, J.C., and Prescott, W.H., 1991a, The velocity field along the San Andreas fault in central and southern California: Journal of Geophysical Research, v. 96, p. 8369-8389. 1991b, Surface deformation after the Loma Prieta, California, earthquake: Eos (American Geophysical Union Transactions), v. 72, p. 119. Marshall, G.A., Stein, R.S., and Thatcher, W., 1991, Faulting geom- etry and slip from co-seismic elevation changes: The October 17, 1989 Loma Prieta, California, earthquake: Bulletin of the Seismo- logical Society of America, v. 81, p. 1660-1693. Matsu'ura, M., Jackson, D.D., and Cheng, A., 1986, Dislocation model for aseismic crustal deformation at Hollister, California: Journal of D243 Geophysical Research, v. 91, p. 12661-12674. McLaughlin, R., 1990, Sargent fault zone at Loma Prieta: U.S. Geo- logical Survey Open File Report 90-274, p. 19-22. Nur, A., and Mavko, G., 1974, Postseismic viscoelastic rebound: Sci- ence, v. 183, p. 204-206. Okada, Y., 1985, Surface deformation due to shear and tensile faults in a half-space: Bulletin of the Seismological Society of America, v. 75, p. 1135-1154. Oppenheimer, D.H., 1990, Aftershock slip behavior of the 1989 Loma Prieta, California earthquake: Geophysical Research Letters, v. 17, p. 1199-1202. Oppenheimer, D.H., Bakun, W.H., and Lindh, A.G., 1990, Slip parti- tioning of the Calaveras fault, California, and prospects for future 40" \l||||11||| o luo Illllllll‘llllllllllIJ!_I|IIII1IIllllllllllll| 30' 20° 1 au oa fog 10 37 ° 50° lllllllllllllllll|lll \ IIIITIIIIIIIIIIII|11_TIIIIll|llfillllllllllllIIIII 30° 20' 10' Figure 13.- Continued 122 ° 50° 40° 30° D244 earthquakes: Journal of Geophysical Research, v. 95, p. 8483- 8498. Reasenberg, P.A., and Simpson, R.W., 1992, Response of regional seismicity to the static stress change produced by the Loma Prieta earthquake: Science, v. 255, p. 1687-1690. Rothacher, M., Beutler, G., Gurtner, W., Schildknecht, T., and Wild, U., 1990, Documentation for Bernese GPS Software Version 3.2: Bern, University of Bern. Rundle, J.B., and Jackson, D.D., 1977, A three-dimensional viscoelas- tic model of a strike-slip fault: Geophysical Journal of the Royal Astronomical Society, v. 49, p. 575-591. Rymer, M.J., 1990, Near-fault measurement of postseismic slip associ- ated with the 1989 Loma Prieta, California, earthquake: Geophysi- cal Research Letters, v. 17, p. 1789-1792. Savage, J.C., 1971, A theory of creep waves propagating along a trans- form fault: Journal of Geophysical Research, v. 76, p. 1954-1966. Savage, J.C., and Prescott, W.H., 1978, Asthenosphere readjustment and the earthquake cycle: Journal of Geophysical Research, v. 83, p. 3369-3376. Savage, J. C., Lisowski, M., and Svarc, J. L., 1994, Postseismic defor- mation following the 1989 (M = 7.1) Loma Prieta, California, earthquake: Journal of Geophysical Research, v. 99, p. 13757- 13765. Scholz, C.H., 1977, A physical interpretation of the Haicheng earth- quake prediction: Nature, v. 267, p. 121-124. Steid1, J.H., Archuleta, R.J., and Hartzell, S.H., 1991, Rupture history of the 1989 Loma Prieta, California, earthquake: Bulletin of the AFTERSHOCKS AND POSTSEISMIC EFFECTS Seismological Society of America, v. 81, p. 1573-1602. Thatcher, W., 1974, Strain release mechanism of the 1906 San Fran- cisco earthquake: Science, v. 184, p. 1283-1285. 1983, Nonlinear strain buildup and the earthquake cycle on the San Andreas fault: Journal of Geophysical Reseach, v. 88, p. 5893-5902. 1986, Cyclic deformation related to great earthquakes at plate boundaries: Royal Society of New Zealand Bulletin, v. 24, p. 245- 272. Toksoz, M.N., Shakal, A.F., and Michael, A.J., 1979, Space-time mi- gration of earthquakes along the North Anatolian fault zone and seismic gaps: Pure and Applied Geophysics, v. 117, p. 1258-1270. Tuttle, M.P., and Sykes, L.R., 1992, Re-evaluation of several large his- toric earthquakes in the vicinity of the Loma Prieta and Peninsular segments of the San Andreas fault, California: Bulletin of the Seismological Society of America, v. 82, p. 1802-1820. Wald, D.J., Helmberger, D.V., and Heaton, TH., 1991, Rupture model of the 1989 Loma Prieta earthquake from the inversion of strong- motion and broadband teleseismic data: Bulletin of the Seismo- logical Society of America, v. 81, p. 1540-1572. Williams, C.R., Arnadottir, T., and Segall, P., 1993, Coseismic defor- mation and dislocation models of the 1989 Loma Prieta earth- quake derived from Global Positioning System measurements: Journal of Geophysical Research, v. 98, p. 4567-4578. Wood, M.D., and Allen, S.S., 1973, Recurrence of seismic migrations along the central California segment of the San Andreas fault sys- tem: Nature, v. 244, p. 213-215. THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989; EARTHQUAKE OCCURRENCE AFTERSHOCKS AND POSTSEISMIC EFFECTS SHALLOW, POSTSEISMIC SLIP ON THE SAN ANDREAS FAULT AT THE NORTHWESTERN END OF THE LOMA PRIETA EARTHQUAKE RUPTURE ZONE By John Langbein, U.S. Geological Survey CONTENTS Page Abstract D245 D3 Introduction 245 Data 247 Discussion 248 References 251 ABSTRACT A small, 10 km by 10 km geodetic network spanning the San Andreas fault was measured 7, 77, 157, 194, 422, 547, and 756 days following the Loma Prieta earthquake. This network is located at the northwestern end of the rupture plane defined by the locations of numerous after- shocks. In the initial 70-day interval, the measured line- length changes revealed that 4.74+0.6 of right-lateral slip occurred within the network. However, during the later 2- year interval only 2.1+0.5 mm/yr of right-lateral slip could be detected. Thus, it appears that the measured slip is a typical response of a fault following a major shock in that the rate of slip decreases rapidly with time. However, the magnitude of the postseismic slip is less than 0.5 percent of the inferred co-seismic slip at depth. In addition, we estimate that secular strain is accumulating at 0.19+0.05 ppm/yr of fault-parallel strain, -0.37+0.06 ppm/yr of fault- normal strain, and 0.20+0.05 ppm/yr of fault-tensor-shear strain which appears to be caused by postseismic, dextral slip on the rupture plane of the Loma Prieta earthquake. INTRODUCTION Immediately following the Loma Prieta earthquake, U.S. Geological Survey personnel installed and measured a small geodetic network on the northwestern end of a zone defined by the epicenters of aftershocks that occurred within 2 days of the main shock (fig. 1). The intent of these observations is to detect possible fault slip follow- ing the main shock and to monitor the deformation rate for anomalous changes that might precede a large and potentially damaging earthquake on the San Francisco pen- insula segment of the San Andreas fault northwest of the Loma Prieta shock. The network was set up to the north- west rather than to the southeast for several reasons. With only a short segment between the southern end of the rupture zone and the creeping section of the San Andreas fault south of San Juan Bautista (Burford and Harsh, 1980), the potential for an earthquake larger than M=6 is low. Furthermore, there already were several geodimeter baselines (Lisowski and Prescott, 1981; King and others, 1981; Savage and others, 1979) and volumetric strain meters (Johnston and others, 1986; Johnston and others, 1987) for detecting deformation on the segment to the southeast. Within the zone defined by the aftershocks, there was already geodetic coverage because several baselines are monitored using a central point at Loma Prieta (Savage and others, 1987). However, geodetic cov- erage just to the northwest of the aftershock zone has been poor because of dense tree growth and steep topog- raphy. Nonetheless, we were able to install a few monu- ments that could be measured with a two-color geodimeter. The advantage of this instrument over a global position- ing system (GPS) (Prescott and others, 1989a) and the more traditional terrestrial distance measuring instrument (Savage and Prescott, 1973) is its higher precision of 0.10 to 0.15 part per million for lengths between 2 and 10 km (Langbein, 1989), which makes it possible to detect small displacements and possible changes in rates. The data to be discussed here indicate displacements of 4 mm in the 2.5-month period following the main shock, and these displacements can be interpreted as 5 to 6 mm of shallow, cumulative fault slip on this segment of the San Andreas fault between Black Mountain and Loma Prieta. How- ever, the measurements during the next 2 years detected a similar amount of displacements, indicating a substantial reduction in rate. Although other geodetic measurements D245 AFTERSHOCKS AND POSTSEISMIC EFFECTS D246 e Sursn (46861) s1oy10 pue noosoaig 4q 0/61 s0uts painseaur y1ompou urejunojy yoefg ay) st ouarag [q Jo you y10M Jou anapoos oy J, '(wojdeyo sy) 'yuoms|[q pue z191g) sunojorarm pop10031 [Jom Sursn jopow Ajtoofaa pamoidut ue itm sjuaaa ay} Sunesojar Jo jnsa1 e are astoord O6 LoL o +0 't +09 D (] Q +06 +0 ° 2 o s30nLINDVN +002 V +0°€ L <> +0°'g O +00 + SHLd43G0 oUL yooys urewr oy} soupe sXep / jsiy oy} urgiim Sununooo pue 10 z=py aso are syooysroye oy J, yosut ay) ur umoys st erwojije; uriim dew si) jo uonesof areur -rxoidde ay}, 'efnsuruad uey ay} uo aSe1aa03 peuonippe ay} Jo auros pue 'sing} paddew 'auoz yooysroye ayenbyirea vioug euo7 oy 01 itm Spry quarag |q uo paroju0d y1omjou 10fo2-0m) ay} jo uoneso1 ay) Sumoys dep- 1 .ON0NNF ob Lp ppp [_ .mvo 98 oualas |3 POSTSEISMIC SLIP ON THE SAN ANDREAS FAULT AT THE NORTHWESTERN END OF RUPTURE ZONE have detected significant deformation following the main shock (Lisowski and others, 1991; Burgmann, and others, 1994), these measurements have not been able to unam- biguously detect any decrease in rate that is usually char- acteristic of postseismic slip because of the lower precision of these techniques (Savage and Prescott, 1973; Prescott and others, 1989b). It also is possible that these large, 100-km networks are more sensitive to slip at depth caused by visco-elastic relaxation of the lower crust (Savage and 1 | 1 1 1 1 | 1 D247 Prescott, 1978), which may have a time scale of years rather than days, as is often found with shallow slip. THE DATA Within one week following the Loma Prieta earthquake, we installed and measured the lengths of the four baselines shown in figures 1 and 2. During the first week of Janu- 1 SARA / \ / A \ / * DEPTHS + 0.0+ > O 5.0+ [ 13.0+ A 20.0+ MAGNITUDE = 2.0+ CJ 3.0+ 37°10 7 t T T T T T T T | 122 00° Figure 2.-Detailed map showing the locations of the six baselines of the El Sereno network. The four baselines shown with solid lines were measured starting 7 days after the Loma Prieta earthquake. The other two baselines, shown with dashed lines, have measurements starting 7 days after the main shock. Also plotted are the locations of the after- [j 4.0+ shocks (Dietz and Ellsworth, this chapter) near and within the network for the period spanning the surveys of the network. Most of the shocks occurred within a month of the main shock. Again, only earthquakes with magnitudes greater than 2 are shown. D248 ary 1990, these baselines were surveyed again, and two additional baselines were measured for the first time. In mid-March 1990, three out of the original four baselines were measured for the third time. Finally, in late April 1990, all six baselines were measured again. The three most recent surveys included measurements on all six baselines with the exception of the December 1990 sur- vey, when the monument at GAP was not recovered. Be- cause of difficult line-of-sight conditions, we needed to use two central monuments, BOHL and BAY, on El Sereno Ridge so that we could have a 360° view of the surround- ings. Two of the six baselines cross the surface trace of the San Andreas fault at about 45° angles, and their length changes are sensitive to fault slip. A third baseline, from BAY to VASONA, does not cross the San Andreas fault but is oriented at 45° to the fault strike and provides a sensitive measure of off-fault shear strain. A fourth baseline crosses the San Andreas fault at a steep angle and is useful for detecting displacements perpendicular to the fault. The last two baselines, from BAY to SARA and TABLE, are useful for detecting off-fault deformation. The instrument used to measure the baseline lengths is a portable, two-color geodimeter (Slater and Huggett, 1976), with a nominal precision between 0.4 mm and 0.9 mm for lengths between 2 and 7 km (Langbein, 1989). For distances of less than 10 km, this instrument has at least a factor of three better precision than any other geo- detic techniques (Savage and Prescott, 1973; Prescott and others, 1989a). Because the instrument ranges on two wavelengths, it is able to detect differences in travel time due to dispersion in the atmosphere. Whereas the use of two optical wavelengths gives the instrument its high pre- cision, it also limits the maximum range to that obtained with its long wavelength, blue. The scatter of the blue light during hazy conditions limits the use of the instru- ment to clear nights. The March 1990 survey was incom- plete due to foggy conditions. The results of measuring the six baselines are shown in figure 3 and clearly show that two of the fault-crossing baselines changed length by 4 mm between the first two surveys and show some changes in length during the next 2 years. The sense of the initial set of displacements is consistent with approximately 5 to 6 mm of right-lateral slip on the San Andreas fault, or localized displacement of the monument at BOHL to the southeast by the same amount. While the baselines to SUMMIT and GAP showed significant displacements during the initial 2.5 months, the other two baselines did not show any length changes within their one standard deviation error bars. With any instrument that measures distance, one needs to be careful about possible drift in the instrumentation which could manifest itself as a length change. Typically, the drift would cause a change in the length scale, thus causing all of the baselines to either extend or contract in length. However, since the lengths of two baselines remain un- AFTERSHOCKS AND POSTSEISMIC EFFECTS changed and the other two baselines show displacements of equal magnitude but in the opposite sense, a change of the length scale in the instrumentation is unlikely. The 2- to 4-mm amplitude of displacement on the baseline BOHL-CENTRAL has a suspicious seasonal pe- riodicity that has been seen elsewhere in central Califor- nia (Langbein and others, 1990). The monument at CENTRAL is placed at the edge of a steep embankment next to a highway. Any soil dilation corresponding to the seasonal rainfall would contaminate the data from this baseline. Accordingly, the data from this baseline will be ignored in the following discussion. DISCUSSION Although a localized displacement vector to the south- east of the monument BOHL is consistent with the obser- vations, the simplest tectonic model is 4.7+0.6 mm of shallow, right-lateral slip on the segment of the San An- dreas fault located within the network during the 2.5 months following the main shock (fig. 4). Because the network is limited spatially, we cannot place any defini- tive bounds on the extent of the slip plane. Accordingly, we use the simplest model of a throughgoing fault plane that splits the crust into two rigid blocks. The results of estimating right-lateral slip as a function of time on the basis of this simple model are shown in figure 4. How- ever, in a few tests of varying the top and bottom edges of the dislocation surface, we could reject with a 99 percent confidence level those models that specify zero slip be- tween the surface and 3 km depth. All other models with slip on the San Andreas fault provide an adequate fit to the observations, and it seems likely that slip did occur at shallow depths. The inferred value of slip for late 1990 appears anoma- lous. However, confidence for that particular value of slip is low because only one out of the two fault-crossing baselines was measured during that survey. The baseline to GAP was not measured. Accordingly, in the following discussion, the slip value from late 1990 is ignored. The observed 4.7 mm of fault slip should be interpreted as postseismic slip, which often occurs after large earth- quakes (Smith and Wyss, 1968; Langbein and others, 1983). As shown in figure 2, there are a number of after- shocks located within the El Sereno network. Just to the north of the network, the number of aftershocks decreases significantly. The cumulative moment of these tiny after- shocks would translate into slip much smaller than the inferred 4.7 mm. The presence of seismicity is evidence for deformation, but most of the deformation could be happening aseismically. While estimating the slip as a function of time, the three components of tensor strain rate are fit simultaneously to the data using the method discussed by Langbein (1989). POSTSEISMIC SLIP ON THE SAN ANDREAS FAULT AT THE NORTHWESTERN END OF RUPTURE ZONE The results of this simultaneous adjustment for both slip and strain rate indicate 0.19+0.05 ppm/yr of fault-parallel strain, -0.37+0.06 ppm/yr of fault-normal strain, and 0.20+0.05 ppm/yr of tensor shear. The local strike of the fault is taken to be N54°W. Recall, however, that data from the five baselines are used to estimate four param- eters, so these estimates are statistically co-dependent. There are three sets of geodetic measurements that have detected postseismic deformation. Two sets of these mea- surements use GPS to re-occupy a profile of monuments oriented perpendicular to the San Andreas fault and span- ning a 50- to 80-km-wide zone. The profile of Lisowski and others (1991) spans the epicentral area of the Loma Prieta earthquake and the second profile of Burgmann and Segall (1991) is located approximately 10 km north- west of the two-color network. Lisowski and others' (1991) measurements show 8 mm/yr near surface slip, an addi- D249 tional fault-parallel 20-mm/yr displacement distributed over a 20-km-wide zone centered on the San Andreas fault, and 18 mm/yr contraction over 15-km-wide zone for which the San Andreas forms the southwestern bound- ary. These displacement rates translate into a 0.5 ppm/yr shear strain and 1.2 ppm/yr fault-normal contraction. By comparison, the two-color network has detected lower rates by a factor of two to three than the GPS observations spanning the epicentral area. However, to the northwest of the two-color network the data from the GPS network of Burgmann and Segall (1991) show rates that can be explained by secular strain accumulation. Thus, the strain rates derived from the two-color data fit between the high rate within the epicentral area and the low rate to the northwest. Burgmann and others (this volume) have modeled the GPS data measurements from both the epicentral area and DISTANCE CHANGES 3 BAY-TAB 50 a I______£LE :» e e C - $ --- p__ _ T _ F BAY-SARA - IC p O BAY-VASONA (Wo) O DD BOHL-GAP u |a >--- € DISTANCE CHANGES, mm L LU OLL L \ ~ oil 7 E\_h_\ : 2 7 ::‘*§\ _______ § c. 9° 3 BOHL-CENTRAL _ _- -- *~ ~ + F —I_—:E‘ Pikes \ é |- I| =E ~- -- --- _ 10 4 BOHL-SUMMIT _ _ - ~ ~- ap 1 c c --- F 1 T _ = - A ~-; I- Figure 3.-Line-length changes observed on the six baselines of the El Sereno network using a two-color geodimeter. The error bars represent the one standard deviation level derived by Langbein (1989). Typically, each baseline is measured twice during a single evening using different II|IIIIIIIIIIT| 1991 TIME reflectors. Hence, the observations of length from a single evening are plotted with overlapping error bars. The data for the baseline BOHL- CENTRAL were not used in the modeling that is discussed in the text. 1992 D250 the Black Mountain profile in terms of deep, dextral slip on the Loma Prieta rupture plane and thrust slip on a fault plane northeast and paralleling the San Andreas fault. Al- though the lateral extent of the thrust fault is poorly con- strained from the GPS data, Burgmann and others model this fault as extending beneath the baseline from BAY to VASONA. Given Burgmann and others' slip rate of 120 mm/yr of thrust on this foothills fault, it would predict approximately 3.6 mm/yr of extension on this baseline, which is clearly not the case. Instead, if thrust is indeed occurring, then it must be located southeast of the two- color network. However, the 100 mm/yr of dextral slip on the plane representing rupture of the Loma Prieta fault determined by Burgmann and others is broadly consistent with the two-color measurements. In fact, if we use the model of time-dependent slip on the San Andreas fault shown in figure 4, and slip on the two planes discussed in Burgmann and others, a satisfactory fit to the two-color data is ob- AFTERSHOCKS AND POSTSEISMIC EFFECTS tained. The inferred value of postseismic slip on the Loma Prieta plane using the two-color data is 157+23 mm/yr, which is within 50 percent of the value estimated by Burgmann and others of 100 mm/yr. However, the model derived from the two-color data implies -14+6 mm/yr of thrust on the foothills fault, which is not consis- tent with the estimate of Burgmann and others of 120 mm/yr. Finally, the third set of postseismic measurements, which have been described by Rymer (1990), are a series of taped distance measurements on small-aperture (0.5 to 7.7 m) quadrilaterals. One quadrilateral is located 3.4 km southeast of the station GAP straddling a subparallel trace of the San Andreas fault. At this site during the 6 months following the main shock, Rymer (1990) recorded a cu- mulative of 5+2 mm of right-lateral slip consistent with the observations described here. The unique feature of the two-color data is that we can infer a decrease in slip rate over the 2 years following the INFERRED SLIP 10 L __ d I 20 llAXS—f’ | e — /’_/'_’__’:____—_———:—‘- I ~ L . - I | - 0 — --- ___ 0.5 DAYS | 5 e --I r -] - I ol h c | / _ / 1 / _ e _ | £ i -10 1 [ PT rr rrr opt OTT 1990 1991 m Figure 4.-Values of inferred slip as a function of time following the Loma Prieta main shock. The error bars represent one standard devia- tion. Since the total amount of slip following the earthquake is un- known, only the relative slip is plotted. The two curves show the predicted TIME slip as a function of the logarithm of time for two differing decay times, 0.5 and 20 days. The value of slip for December 1990 was not used in estimating the parameters of the logarithmic curves. POSTSEISMIC SLIP ON THE SAN ANDREAS FAULT AT THE NORTHWESTERN END OF RUPTURE ZONE main shock. To demonstrate that slip has been occurring during the interval between January 1990 and November 1991, a different function of slip with time is fit to the observed line-length changes. For the interval between October 1989 and the January 1990 surveys, slip is as- sumed to occur as a single event at an unspecified time. However, for the observations from early January through November 1991, slip is assumed to occur at a constant rate. Estimates of these two parameters yield 4.7+0.6 mm of slip between early October 1989 and early January 1990, and a significant slip rate of 2.1+0.5 mm/yr for the period following January 1990. This rate appears to be marginally faster by 1.7 standard deviations than the ap- parent background rate of mm/yr 1.2+0.2 mm/yr estimated for the Black Mountain network (fig. 1) by Prescott and others (1989b). Although the slip rate recorded within days after the main shock is significant, the rate of postseismic slip ap- pears to be decreasing inversely with time. Along with the estimated slip, figure 4 shows two curves representing postseismic slip as Ulog(t/t+1)+B, where t is the time of observation since the time of the main shock (Langbein and others, 1983). Here we have specified two differing values for the decay time t as 0.5 days and 20 days and estimated corresponding values for U, and B by least- squares fitting. With the exception of the inferred slip in late 1990, the comparison in figure 4 shows reasonable agreement between the inferred values of slip as a func- tion of time and the postseismic slip predicted from the logarithmic functions. However, those functions having decay times of less than a month fit the observed slip data better than those functions with decay periods in excess of several months. Thus the short decay time (t) found here is consistent with that found in other cases of shal- low, postseismic slip following the 1966 Parkfield and 1979 Imperial Valley earthquakes (Smith and Wyss, 1968; Langbein and others, 1983). Breckenridge and Simpson (this volume) show data for changes in the rates of fault slip on other fault segments in response to the Loma Prieta earthquake, which they suggest is caused by the Loma Prieta earthquake perturb- ing the regional stress field. They demonstrate that the spatial pattern of stress changes predicted from their stress model was consistent with the changes in slip that they observed from creepmeters along the Hayward, San An- dreas, and Calaveras fault system. Furthermore, their model predicts an increase in rate of slip on the section of the San Andreas fault that bisects the two-color network immediately following the Loma Prieta earthquake. The results presented here support their model. However, their creep data show that the time constant is about a year for the postseismic slip rate to return to its secular rate. This observation is inconsistent with the time-constant for postseismic slip determined from the measurements dis- D251 cussed here. It could be that the time-constant of defor- mation detected by the creep data and the GPS observa- tions is due to a deep-seated process with a long time constant, but that there is a short time constant associated with near-surface fault-slip. REFERENCES Burford, R.O., and Harsh, P.W., 1980, Slip on the San Andreas fault in central California form alinement array surveys: Bulletin of the Seismological Society of America, v. 70, p. 1233-1261. Burgmann, R., and Segall P., 1991, Postseismic GPS monitoring NW of the 1989 Loma Prieta rupture zone: Eos, v. 72, no. 44, p. 119. Johnston, M.J.S., Linde A.T., Gladwin MT., and Borcherdt, R.D., 1987, Fault failure with moderate earthquakes: Tectonophysics, v. 144, p. 189-206. Johnston, M.J.S., Borcherdt R.D., and Linde A.T., 1986, Short period strain (0.1-105 s): Near source strain field for an earthquake (M, 3.2) near San Juan Bautista, California: Journal of Geophysical Research, v. 91, p. 11497-11502. King, N.E., Savage, J.C., Lisowski, M., and Prescott, W.H., 1981, Preseismic and co-seismic deformation associated with the Coyote Lake, California earthquake: Journal of Geophysical Research, v. 86, p. 892-898. Langbein, J., McGarr, A., Johnston, M.J.S, and Harsh, P.W., 1983, Geodetic measurements of postseismic crustal deformation follow- ing the 1979 Imperial Valley earthquake, California: Bulletin of the Seismological Society America, v. 73, p. 1203-1224. Langbein, J., 1989, The deformation of the Long Valley caldera, east- ern California, from mid-1983 to mid-1988; measurements using a two-color geodimeter: Journal of Geophysical Research, v. 94, p. 3833-3850. Langbein, J., Burford, R.O., and Slater, LE., 1990, Variations in fault slip and strain accumulation at Parkfield, California: Initial results using two-color geodimeter measurements, 1984-1988: Journal of Geophysical Research, v. 95, p. 2533-2552. Lisowski, M., and Prescott, W.H., 1981, Short-range distance measure- ments along the San Andreas fault system in Central California, 1975-1979: Bulletin of the Seismological Society America v. 71, p. 1607-1624. Lisowski, M., Prescott, W.H., Savage, J.C., and Johnston, M.J.S., 1990, Geodetic estimate of co-seismic slip during the 1989 Loma Prieta, California, earthquake: Geophysical Research Letters, v. 17. Lisowski, M., Savage, J.C. and Prescott, W.H, 1991, Surface defor- mation after the Loma Prieta, California Earthquake: Eos, v. 72, no. 44, p. 119. Prescott, W.H., Davis, J.L, and Svarc, J.L, 1989a, Global positioning system measurements for crustal deformation: Precision and accu- racy: Science, v. 244, p. 1337-1340. Prescott, W.H., Savage, J.C. and Lisowski, M., 1989b, Crustal strain, in National Earthquake Reduction Program, Summaries of Techni- cal Reports (v. 29): U.S. Geological Survey Open-File Report 90- 54. Rymer, M.J., 1990, Near-fault measurement of postseismic slip associ- ated with the 1989 Loma Prieta, California, Earthquake: Geo- physical Research Letters, v. 17, p. 1789-1792. Savage, J.C., and Prescott, W.H., 1973, Precision of geodolite dis- tance measurements for determining fault movements: Journal of Geophysical Research, v. 78, p. 6001-6008. Savage, J.C., and Prescott, W.H., 1978, Asthenosphere readjustment D252 AFTERSHOCKS AND POSTSEISMIC EFFECTS and the earthquake cycle: Journal of Geophysical Research, v. 83, p. 3369-3376. Savage, J.C, Prescott, W.H., Lisowski, M., and King, M, 1979, Geodolite measurements of deformation near Hollister, California, 1971-1978:, Journal of Geophysical Research, v. 84, p. 7599- 7615. Savage, J.C, Prescott, W.H, and Lisowski, M., 1987, Deformation along the San Andreas fault 1982-1986 as indicated by frequent Geodolite measurements: Journal of Geophysical Research, v. 92, p. 4785-4797. Slater, LE., and Huggett, G.R, 1976, A multi-wavelength distance- measuring instrument for geophysical experiments: Journal of Geophysical Research, v. 81, p. 6299-6306. Smith, S.W., and Wyss, M., 1968, Displacement on the San Andreas fault initiated by the 1966 Parkfield earthquake: Bulletin of the Seismological Society of America, v. 68, p. 1955-1974. THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989; EARTHQUAKE OCCURRENCE AFTERSHOCKS AND POSTSEISMIC EFFECTS MODELS OF POSTSEISMIC DEFORMATION AND STRESS TRANSFER ASSOCIATED WITH THE LOMA PRIETA EARTHQUAKE By MF. Linker, Harvard University and U.S. Geological Survey; and J.R. Rice, Harvard University CONTENTS Page Abstract D253 Introduction 253 Physical models of postseismic deformation ----------------------- 255 Three-dimensional finite element models --------------------------- 256 Magnitude of coseismic and long-time postseismic stress transfer 258 Analytical models 258 Coseismic and long-time fully relaxed postseismic stress States FOT LOMA PMI@t@ once snes ecco cee 260 A relative time-scale for postseismic relaxation ------------ 263 Time-dependent models 263 Linear viSCO@IAStiC MOG@IS 264 Non-linear viscoelastic models: Hot-friction ---------------- 266 DiSCUSSiON ANG non- no- 271 Alternative gEOMEtTIC MOGEIS ----------------------------.-.---- 271 Comparison to strain-meter observations at San Juan Bautista 272 Alternative PhySiCal PMOCES$ 272 Summary 272 Acknowledgments 273 References cited 273 ABSTRACT Recent geodetic observations indicate that the velocity field at the Earth's surface has been perturbed by the Loma Prieta earthquake. We interpret this change in terms of models in which transient postseismic slip occurs beneath the locked seismogenic portion of the San Andreas fault zone. In our three-dimensional finite element calculations, the deep aseismic region of the fault zone is either treated as linear Maxwell viscoelastic or is made to follow a steady-state version of the laboratory-derived rate- and state-dependent friction law, in which slip rate depends exponentially on the ratio of shear stress to effective nor- mal stress. We refer to this second rheology as our "hot- friction" model. Comparison of model predictions to observed postseismic displacements provides a constraint on the ratio of Maxwell relaxation time, t,, for material in the deep aseismic region of the fault zone to fault-zone thickness, /. Neither quantity can be constrained indepen- dently. Best fit to the initial 1.3 years of data is obtained with 1t,/h = 0.3 yr/km thickness. If we assume that postseismic slip has occurred in a fault zone of thickness h<1 km, then using a shear modulus of 30 GPa we obtain an estimate of the effective viscosity of material in the deep aseismic region of the fault zone, 17, <3x10'7 Pa-s. Previous estimates made for the material of the lower crust, based on laboratory measurements of steady-state dislocation or diffusion creep, exceed this value by at least an order of magnitude. The laboratory-derived hot- friction model can yield postseismic deformation of mag- nitude comparable to that observed, but only if dwd(In V) is of the order 0.5 MPa, where 7 is the resistance to slid- ing and V is the slip rate. Laboratory measurements indi- cate that dt/d(In V) can be written approximately as co, where G is the effective normal stress equal to total nor- mal stress minus pore pressure and c is on the order of 0.015 for mid-crustal conditions. To be consistent with our models, the effective normal stress in the deep aseismic portion of the fault zone must, therefore, be extremely low, perhaps on the order of 30 MPa. In the context of the laboratory-derived hot-friction model, the occurrence of postseismic deformation may be evidence that pore pres- sure in the aseismic portions of the fault zone and perhaps the lower crust is near lithostatic. INTRODUCTION The occurrence of the Loma Prieta earthquake brought immediate concern that another large and damaging earth- quake might occur in the San Francisco Bay region in the near future (Working Group, 1990). This concern rose from numerous examples of earthquake triggering and migration (see, for example, Richter, 1958; Mogi, 1968; Allen 1969; Yonekura, 1975; Scholz, 1977; Sykes and others, 1981; Doser, 1986; Stein and Ekstrom, 1992) in- D253 D254 cluding two earthquake pairs in the San Francisco Bay region (Toppazada and others, 1981). In 1836 an M=7 event occurred on the northern half of the Hayward fault and 2 years later an M=7 event occurred on the San Fran- cisco peninsula. Similarly, in 1865 an M=6.5 event oc- curred in the vicinity of Loma Prieta and 3 years later an M=7 event occurred on the southern segment of the Hayward fault. More recently, on the Calaveras fault three moderate events have occurred in spatial sequence, mi- grating towards the north (Oppenheimer and others, 1990). Previous workers have suggested that earthquake mi- gration and triggering might be the result of stress trans- fer between earthquake source regions. Numerous attempts to relate static coseismic stress changes to the occurrence of future events (Stein and Lisowski, 1983; Mavko and others, 1985; Oppenheimer and others, 1988; Poley and others, 1987; Hudnut and others, 1989; Du and Aydin, 1990; Seeber and Armbruster, 1990; Michael and others, 1990; Michael, 1991; Reasenberg and Simpson, 1992) demonstrate that the magnitude of the static coseismic stress change is likely to be small, except perhaps in the region immediately adjacent to the main rupture. Though postseismic relaxation associated with inelastic deforma- tion processes could act to increase the magnitude of the stress transfer relative to the static coseismic stress change, the total stress change in the far-field is still likely to be modest compared to typical earthquake stress drop (Rice and Gu, 1983). Furthermore, the rate of stress transfer resulting from postseismic relaxation may be small com- pared to the background loading rate, and so its contribu- tion to the far-field stress state is likely to be negligible, AFTERSHOCKS AND POSTSEISMIC EFFECTS unless postseismic relaxation occurs at a sufficiently high rate (Lehner and others, 1981). Nevertheless, the observa- tions of earthquake triggering and migration are numer- ous, and so we have undertaken a study of the stress transfer associated with the Loma Prieta earthquake. In the first 50 years following the great 1906 San Fran- cisco earthquake, the shear strain rate near the San An- dreas fault decreased by about a factor of three (fig. 1) (Thatcher, 1983), and so we might expect similar tran- sient postseismic deformation to follow the Loma Prieta earthquake. Indeed, geodetic data indicate that postseismic deformation has taken place at rates that exceed the pre- earthquake rate (Biirgmann and others, 1991, this chapter; Lisowski and others, 1991a, 1991b; Biirgmann and oth- ers, 1992; Savage and others, 1994). Therefore, the Loma Prieta earthquake provides us an opportunity not only to study the stress transfer that could potentially lead to a future major earthquake, but more generally, to obtain constraints on the rheological properties of the crust, fault zone, and upper mantle. We use the finite element method to examine three di- mensional models of postseismic deformation and stress transfer associated with the Loma Prieta earthquake. We use preseismic (Lisowski and others, 1991b) and prelimi- nary postseismic measurements made during the initial 1.3 years following the earthquake (Biirgmann and Segall, 1991; Lisowski and others, 1991a; Biirgmann and others, 1992) to constrain parameters that control the relaxation process. Our models are additionally constrained by con- sideration of heat flow data, laboratory-based rheologies and constitutive parameters, seismic observations of pre- /- 4 42N M |_ I I I I I I I I o - | E E [_ sC é g 3 |- | PA |- a FR 2 && [1- FR i PR PR 3 2 , ; SP 24 2 |- PA) | m laval | _ L m a |_ 126W ”SS/SN < p \- SC i H 0 1 |- +- 4 - T pd m PR - p3 |- | PR SF 7 w O | I I I | I | I I l I I I 0 20 40 60 80 TIME SINCE 1906 (YEARS) Figure 1.-Geodetically measured right-lateral shear strain rate near the San Andreas fault in northern Califor- nia following the great 1906 San Francisco earthquake (from Thatcher, 1983). Inset map indicates the location of the networks. MODELS OF POSTSEISMIC DEFORMATION AND STRESS TRANSFER ASSOCIATED WITH THE EARTHQUAKE earthquake microseismicity, and seismological studies of the main shock. Our paper is organized as follows. First, we study the magnitude of stress transfer by examining models of the static coseismic stress field and long-time limits of postseismic relaxation. We then examine time-dependent models and compare predicted displacements to geodetic data in attempts to constrain the parameters that control the relaxation time-scale. Our first set of time-dependent models incorporate linear Maxwell viscoelasticity in the inelastically deforming regions of the model. Our second set of time-dependent models employ a steady-state ver- sion of the laboratory-derived rate- and state-dependent friction model in which aseismic slip rate depends expo- nentially on the ratio of shear stress to effective normal stress (Blanpied and others, 1991), where the effective normal stress is the total normal stress minus the pore pressure. We refer to the nonlinear model as our "hot- friction" model. Our principle conclusions are as follows. (1) The larg- est stress changes associated with the earthquake occur adjacent to the main rupture, for example, along the San Francisco Peninsula segment of the San Andreas fault. (2) In that region, postseismic relaxation processes could lead to an increase in shear stress by an amount that exceeds the static coseismic stress increase. (3) Northeast to north of Loma Prieta, both coseismic slip and postseismic re- laxation reduce the right-lateral shear stress and the com- pressive stress along the Hayward and Calaveras faults. (4) Models in which deep aseismic fault slip occurs be- neath the seismogenic zone are capable of producing strike- slip displacements consistent with the initial 1.3 years of postseismic deformation observed along the profile of GPS stations that crosses through the epicenter (Lisowski and others, 1991a). These deep-slip models, however, appear to be incapable of producing fault-trace normal contrac- tion compatible with the observations there and an addi- tional source of deformation may be required. (5) Northwest of Loma Prieta, the postseismic displacements predicted by our deep-slip models may exceed the obser- vations, though a careful treatment of all of the pre-earth- quake and post-earthquake geodetic data is required to better establish a measure of the change in the velocity field associated with the earthquake. (6) The time-scale for postseismic relaxation appears to be sufficiently long so that the average background tectonic stress rate is likely to exceed the stress rate resulting from relaxation, except perhaps within about 20 km of the edge of the coseismic rupture. (7) In linear viscoelastic models that incorporate postseismic fault slip, the parameter controlling the relax- ation time is the ratio of fault-zone material relaxation time to the thickness of the fault zone, t,/k. This param- eter appears to have a value of approximately 0.3 years/ km thickness. If relaxation takes place in a shear zone that is less than 1 km thick and if the shear modulus is 30 D255 GPa, then the effective viscosity in the deforming region must be less than about 3x10" Pa-s. This value is at least one order of magnitude lower than any obvious interpre- tation of laboratory measurements of the steady-state creep of solid crustal rocks controlled by dislocation or diffu- sion creep. Transient creep response to a sudden change in stress has not been as well characterized in the labora- tory and so probably cannot be ruled out as a possible mechanism. (8) The laboratory-derived hot-friction model can yield postseismic deformation of magnitude compa- rable to that observed, but only if dt/d(In V) is of the order 0.5 MPa, where T is the resistance to sliding and V is the slip rate. Laboratory measurements indicate that dt/ d(ln V) can be written approximately as co (Stesky, 1975, 1978; Dieterich, 1981; Ruina, 1983), where 0 is the effec- tive normal stress and c is of order 0.015 for mid-crustal conditions (Blanpied and others, 1991). Therefore, to be consistent with our models, the effective normal stress must be extremely low, perhaps of order 30 MPa. (9) In the context of the laboratory-derived hot-friction model, the occurrence of postseismic deformation may be evi- dence that pore pressure in the aseismic portions of the fault zone and perhaps the nearby lower crust is near lithostatic. PHYSICAL MODELS OF POSTSEISMIC DEFORMATION Models of postseismic deformation have been motivated primarily by observations of deformation that followed large earthquakes in subduction zones and along the San Andreas fault. In these models deformation occurs as the result of fault slip on a down-dip projection of the seismic rupture or as distributed shear in a broad zone beneath the brittle seismic layer (Savage and Prescott, 1978; Savage, 1983; Thatcher, 1983; Thatcher and Rundle, 1984; Li and Rice, 1987; Lyzenga and others, 1991). Unfortunately, the displacement field at the earth's surface is rather in- sensitive to the location of the deformation source, and so generally it is not possible to reject either of these models on the grounds that they do not fit the deformation data (Savage and Prescott, 1978; Savage, 1990; Segall, 1991). Therefore, physical models of postseismic deformation must be defended as well as motivated on broad geo- physical and geological grounds. England and Molnar (1991), for example, propose that the upper continental crust is weak relative to the lower parts of the lithosphere and that crustal blocks passively follow the deformation of the substrate. Within this context, postseismic defor- mation could be caused by transient interaction between the individual crustal blocks as well as between the blocks and the substrate. This concept has been adopted either explicitly or implicitly in much of the modeling men- tioned above. D256 We examine three models of the geometrical distribu- tion of inelastic material that shears postseismically (fig. 2). In Model 1, postseismic fault slip is allowed to occur down-dip from the seismogenic zone but is confined to the crust. In Model 2, slip is allowed to extend from the base of the seismogenic zone downward through the crust and into the upper mantle. In Model 3, slip is allowed to extend from the base of the seismogenic zone downward through the crust to where it terminates at its intersection with a horizontal detachment zone in the lower crust. In the Earth, immediately after the earthquake, stress changes and presumably slip rates are highest near the edge of the coseismic rupture. Therefore, Model 1 can be thought of as a short-time version of either Model 2 or Model 3. In Model 2, crustal blocks are thought of as effectively welded to the substrate, at least on the time-scale of the repeat-time for large crustal earthquakes. Postseismic mo- tion is interpreted as the result of transient motion be- tween the crustal blocks whose boundaries extend into the upper mantle. We refer to Model 2 as our throughgoing fault model. Model 3 is based on previous studies of postseismic deformation (for example, Li and Rice, 1987) that were motivated by the recognition that elevated temperatures in the lower crust may give rise to inelastic flow in re- sponse to the stress changes associated with earthquakes (Brace and Kohlstedt, 1980; Meissner and Strehlau, 1982; Sibson, 1982; Chen and Molnar, 1983). This class of mod- els can be further motivated by compilations of seismic, geological, and geophysical data in central California (see, for example, Saleeby, 1986; Fuis and Mooney, 1990) and by electrical conductivity measurements, which indicate that horizontal seismic reflectors in the lower crust fre- quently coincide with horizontal layers of high conductiv- ity and, by inference, with regions of elevated fluid content (Hyndman and Shearer, 1989) and fluid pressure (Rice, 1992). All of these observations are in accord with the commonly held view that large-scale horizontal shearing or fault-slip occurs in the middle to lower crust. In Model 3, motion of the shallow portion of the crust is coupled through this horizontal shear zone to the motion of the substrate that drives the system. We refer to Model 3 as our detachment model. Tse and Rice (1986) studied models based on rate- and state-dependent friction (Dieterich, 1981; Ruina, 1983), in which the fault-zone rocks at depth are hot enough to be in the "velocity strengthening" range of frictional re- sponse. In this case, time-varying stress occurs naturally in the aseismic portion of the fault zone. In contrast, Li and Rice (1987), Fares and Rice (1988), and Ben-Zion and others (1993) treated the aseismic portion of the fault zone within the crust as freely slipping, so that only the shear stress in the basal shear zone varies with time. Their free-slip condition is consistent with an assumption that AFTERSHOCKS AND POSTSEISMIC EFFECTS the relaxation time for postseismic crustal fault slip is considerably shorter than for basal shear. In this paper, we include the effects of time-varying stress in the aseismic portion of the fault zone as well as in the basal shear zone. We treat the region outside the fault zone and basal shear zone as purely elastic, with the thought that stress changes associated with earthquakes are too small to cause any appreciable inelastic deforma- tion there. We additionally include the effects of finite rupture length and so our models are three dimensional. THREE-DIMENSIONAL FINITE ELEMENT MODELS We use the finite element method to model stress changes and deformation associated with the earthquake. Our modeling strategy is as follows. Coseismic slip is imposed on a segment of the fault zone that represents the coseismic rupture surface. This imposed slip produces a stress field throughout the body that represents the static coseismic stress change in the earth. We then use two approaches to model the postseismic deformation and stress transfer resulting from that coseismic stress field. First we examine coseismic and long-time fully-relaxed postseismic stress states. Then we examine time-dependent models in which stress and displacement are calculated while the system relaxes toward the fully-relaxed state. By compar- ing the computed time-dependent displacements to defor- mation data, we attempt to constrain the parameters that control the time-scale for relaxation. The solutions we present employ fixed-displacement boundary conditions along the outer surfaces of the model where it has been cut away from the surroundings. The remote boundaries of our model are located sufficiently far from the coseismic rupture so that fixed-displacement and stress-free boundary conditions give nearly the same solution in the regions of interest. We ignore the effects of the shallow creeping zone along the San Andreas fault that begins southeast of Loma Prieta near San Juan Bautista and only consider models that are geometrically symmetric along strike about the epicenter. We additionally ignore any deformation associated with slip on the Hayward and Calaveras faults. We assume a crustal thickness of 25 km, which is representative of the region near the San Andreas fault (Oppenheimer and Eaton, 1984), and represent the horizontal shear zone in the lower crust with a 5-km-thick layer of elements that extends from 20 to 25 km in depth. Our finite element mesh represents a volume with di- mensions 217 km along strike by 265 km perpendicular to strike by 62.5 km in depth (fig. 24). Minimum element- dimensions correspond to 3.5%x2.5x1.0 km. Approximately 10,000 nodes, or 30,000 degrees of freedom, comprise the D257 MODELS OF POSTSEISMIC DEFORMATION AND STRESS TRANSFER ASSOCIATED WITH THE EARTHQUAKE "snJo Jomof ay} ut yey} 'auoz juouwyoriap 10 'auoz stays [EjuOZHOY E SJoosJojuf nq oy} 0) paulyuoa st auoz jing] oy) jo uormod Surddr}s ay} yorym ut ¢ jopopy jo mata (7 'opueu saddn ay out pure oyopy ay} ySnouy} spuarxa auoz jing} ay} jo uonsod Ajreotwustase ay) yorym ur z jopopy Jo moalA (G 'IsSnuo oy Jo aseq ay} pug auoz moj[eys ay} usamjoq alf 01 pouty uo st jing} ay} yo uorod Ajreotwstase ouf, 'auoz jng} Surddip ay) jeaaot 01 {eme mmo uaaq sey japow Jo apis isamjnos ay} yom UI | Jopopy JO mol '> 'payoof are auoz Surddrs duoz eous + jng} '€# IOPOW - HA (1# wor) auoz reau [reap juowaja jiney - g ing; Ajpeonustose ay; mojag pue amoge Aen) 'dis onustosoo Aq paonput sasueyo ssans ay) 01 asuodsa1 ut 4fpeatustase sdrfs uor8a1 4e13 1s274317 'anydn1 ay} jo arooa ay} wroj; Aeme o13z 01 s1ode; difs arustasoa pasodut ay} uorSa1 418 -yrep Surpunouns ay} up 'wojtun st digs orwstasoo pasodut ay aroym amidnu orustasoo Jo aoa sajouap uor3a1 yoefg 'J Jopoj 'uonexgja1 Jo japou sefnonred e 10j auoz arndn1 arustasoo ay} punore pue ut sjuawajfa auoz-4]ng] o mota dn-as0fp 'g 'voug ewo7 seou apuwew soddn pue jsnu> ay; jopour 01 posn ysow q4 peuorsuawrp-so014 J, 'y onsejacosta SuroS1opun Jo suonnguysip aar juasoidar 01 pasn sapour (g) juawaja altuty-7 ng; jes 'I# POW J '# POW C 7T] '' Aim. 2 -OH N 11 \| -| --t aa.. ur 55 z9| E a ccs - -- G- toons Oger --- --- <_ ess ~- -ove c w- s <--- <- tory ewOo7 10J YSJ Gf y D258 mesh of eight-noded isoparametric-brick elements. We use the finite element code ABAQUS (Hibbitt and others, 1991). Elastic three-dimensional and viscoelastic two-di- mensional models were run on a SUN 4 workstation and require several hours of CPU time and 10 to 20 Mbytes of disk space for output. All computations for three-dimen- sional time-dependent models were performed on Cray 2 and Cray YMP supercomputers at the National Center for Supercomputing Applications (Urbana, Illinois). The lin- ear viscoelastic solutions typically require about 3 hours of Cray CPU time for 65 time increments using an ex- plicit time-integration method to reach about 5 years of model time. The nonlinear hot-friction models use appre- ciably more CPU time, since the creep rate varies dra- matically during the relaxation process. Typical run times for three dimensional nonlinear models range from 4 to 10 hours on the Cray 2 to reach 2 years of model time. It is likely that some computation time could be saved if an implicit integration procedure was used, but we did not take that approach. In our models, the fault zone is represented by a thin layer of finite elements (fig. 2). The seismic portion of the fault zone is treated as elastic, and coseismic slip is imposed via a shear transformation strain, yT, which changes the stress-free shape of the material in the rup- ture zone (Aki and Richards, 1980; Rice, 1980). The equivalent seismic slip is hy", where h is the thickness of the layer of fault elements. At distances greater than order h from the edge of the rupture, this technique produces results equivalent to those obtained by imposing slip on a surface. Any estimate of the stress field close to the edge of the coseismic rupture will be extremely sensitive to one's es- timate of the slip distribution. One might incorporate esti- mates of the slip distribution determined from seismic records (Beroza, 1991; Hartzell and others, 1991; Steidl and others, 1991; Wald and others, 1991) or measure- ments of the static displacement field (Marshall and oth- ers, 1991), but we have not taken this approach. Future work on this problem may be warranted since the largest stress changes occur in this near-rupture region. Farther from the edge of the rupture, the stress field is less sensi- tive to the slip distribution and can therefore be estimated with greater confidence. In this spirit, we assume that the transformation strain and hence coseismic slip are uniform except in the outer- most elements of the rupture zone (fig. 2B). In these pe- rimeter elements, we specify that the shear modulus is only 1 percent of the nominal shear modulus, and Poisson's ratio is chosen so that the bulk modulus remains uniform. In addition, the transformation strain tapers to zero here. This procedure roughly simulates a freely slipping zone and has the effect of suppressing large local stresses that are poorly modeled without extensive mesh refinement AFTERSHOCKS AND POSTSEISMIC EFFECTS and that have magnitudes which depend critically on the slip distribution near the edge of the rupture. The characteristics of our coseismic rupture were cho- sen to correspond to estimates obtained by other workers. Specifically, we used scalar moment = 3x10" Nm, dip=64°, and rake=145°. The rupture extends from 5 km to 17.5 km in depth (fig. 2B). The shape of the rupture perimeter was chosen to mimic the distribution of preseismic microseismicity (Olson, 1990) and initial af- tershock distribution (Dietz and Ellsworth, 1990). The cho- sen moment and rake is that from Lisowski and others' (1990) preferred model, in which they specify uniform slip on a rectangular dislocation in a homogeneous elastic half space to model horizontal geodetic data. Seismic and other geodetic estimates of the moment generally range from 1 to 4x10'° Nm, dips range from 55° to 80°, and rakes range from 115° to 155° (see summary by Marshall and others, 1991, table 7). MAGNITUDE OF COSEISMIC AND LONG-TIME POSTSEISMIC STRESS TRANSFER ANALYTICAL MODELS Before presenting results for the Loma Prieta earth- quake, we review simple analytical models to obtain an understanding of how stress fields associated with generic models of stress transfer vary in space. Our analysis fol- lows the work of Rice and Gu (1983). We consider two limiting states: (1) immediate coseismic and (2) fully-relaxed downward continuation of the crustal fault zone and underlying horizontal substrate. The latter corresponds to Model 3 (fig. 26) at times suffi- ciently large that relaxation has gone to completion. We compare the far-field solution for a finite, vertical, strike- slip dislocation in an elastic half-space to the correspond- ing solution in an elastic plate (fig. 3). The static coseismic change in right-lateral shear stress on the still locked segment of the fault zone as a function of distance, r, from the center of the rupture is 1,.(n ~ (1MA)(DPPL/P)At, and the long-time state when the deep aseismic portion of the fault zone and the horizontal detachment are fully- relaxed is T.(r) ~ (5/16)(HL/P)At, where D is rupture depth, L is rupture length, H is the thickness of the elastic portion of the crust, and AT = 2uAu/ID is the coseismic stress drop associated, via a MODELS OF POSTSEISMIC DEFORMATION AND STRESS TRANSFER ASSOCIATED WITH THE EARTHQUAKE crack model, with a long surface-breaking strike-slip rup- ture over depth D, with average slip Au, in a half-space with shear modulus u. These solutions are valid at dis- tances from the center of the rupture that are large com- pared to rupture half-length. The ratio of long-time to short-time stress is T/T,, =(5/4)(H/D)(r/D). At distances from the rupture r >> D relaxation of the horizontal sub- strate yields postseismic stress changes that greatly ex- ceed the coseismic change. D259 For example, given an earthquake with average stress drop At, rupture length equal to twice the elastic plate thickness, L=2H, and rupture depth D=H/2, then at a dis- tance from the center of the rupture equal to 2H the static coseismic stress change is approximately 0.01647 while the stress change after complete relaxation of basal shear stress is approximately 0.16At. In this example, postseismic relaxation increases the far-field stress by an order of magnitude relative to the static coseismic stress Average stress drop = At Figure 3.-Dislocation models comparing (A) short-time coseismic static shear stress change, t,,. to (B) long-time limit, T_, of Model 3 (fig. 2€) in response to coseismic stress drop, At, on the main rupture surface of depth D and length L. In the first case stress is carried by the elastic half space, while in the latter case stress is carried by an elastic plate, with thickness H, floating on a traction-free substrate meant to represent a fully-relaxed horizontal shear zone in the lower crust. The solutions are valid at distances from the center of the rupture that are large compared to rupture half-length. The transition from short-time to long-time stress state corresponds to a change from 1/r 3 to 1/r 2 decay of stress with distance, r, from the center of the rupture. D260 change. Note, however, that the total magnitude of the stress change in the far-field is still only about 16 percent of the earthquake stress drop. COSEISMIC AND LONG-TIME FULL-RELAXED POSTSEISMIC STRESS STATES We estimate the coseismic and long-time fully-relaxed postseismic static stress changes associated with our three geometrical models of the Loma Prieta region (fig. 2). Note that in inhomogeneous viscoelastic systems, the fully- relaxed-limit solution does not necessarily provide an up- per bound on changes of stress during the relaxation process. The models are now purely elastic and postseismic re- laxation is represented by introducing freely slipping re- gions that, in the earth, might deform in a time-dependent manner. This approach saves dramatically on computa- tion time since only one "time-step" is need to achieve "relaxation." The states obtained with these elastic mod- els should be interpreted as representing the long-time fully-relaxed limits of the corresponding viscoelastic mod- els that are presented in later sections of this paper. The calculations are done using the finite element method as discussed previously. The "free-slip" regions are incorporated by specifying that the shear modulus of the "relaxed" material is reduced relative to the surround- ings. To simulate free-slip with this procedure, one needs ”faul/h t.o be small compared to H.,,;,,,, dings/H' wher.e h is the thickness of the layer of fault elements and H is a scale-length for the surrounding region. We use a reduced shear modulus that is 1 percent of the nominal shear modulus, 30 GPa, and choose Poisson's ratio so that the bulk modulus is uniform. In the finite element mesh, the thickness of the fault zone and basal shear zone are 1 km and 5 km, respectively. Further reduction of the low shear modulus does not affect the stress distribution apprecia- bly. The mantle is made stiffer than the crust by a factor of two. Profiles of the changes in stress along profiles corre- sponding approximately with the San Francisco peninsula segment of the San Andreas fault and to the Hayward and Calaveras faults (fig. 4) are plotted in figure 5 as compo- nents of the traction vector resolved onto vertical planes. The values plotted in figure 5 are in general agreement with those obtained by Simpson and Reasenberg (this chap- ter), who used shear traction-free rectangular dislocations to represent relaxed, deep fault zones and horizontal de- tachment faults. In the San Andreas profiles (figs. 5A and 5B), the val- ues of stress represent averages over the bottom one-quar- ter of the locked seismogenic zone, which extends from 0 to 10 km in depth in our models (fig. 2). Very close to the AFTERSHOCKS AND POSTSEISMIC EFFECTS end of the rupture, computed stress changes are sensitive to the assumed spatial distribution of slip. We have not attempted to incorporate detailed estimates of the coseismic slip distribution in our models and have only plotted those stress changes that we consider to be relatively reliable in this context. Along the San Andreas profile, the coseismic change in static shear stress is largest close to the end of the rupture, as expected (fig. 5A). Here, relaxation can double or even quadruple the change in shear stress. Farther from the end of the rupture, at distances that are large compared to the 17.5 km rupture depth, the ratio of postseismic to coseismic shear stress reaches values exceeding 10:1, but the total change in shear stress is quite small. These three observa- tions are qualitatively consistent with the results obtained above with simple analytical models. Both coseismic slip and subsequent postseismic relaxation result in fault- normal compression relative to the pre-seismic state (fig. 5B). The enhancement of shear and normal stress on the San Andreas fault that results from postseismic relaxation can be rationalized if one considers the stress concentration in the vicinity of a shear crack in an elastic body. In our fully-relaxed models, the tip of this shear crack lies along the base of the seismogenic zone, adjacent to the points sampled in figures 5A and 5B. Here, the stress field will be dominated by near-field terms so that the state of stress along the up-dip projection of the aseismically slipped zone will be approximately pure shear. Recall that coseismic slip was oblique, reverse plus right-lateral, on a steeply SW-dipping fault segment so that postseismic slip on the aseismic portion of the fault zone will likewise be oblique, reverse plus right-lateral. By Mohr-circle analy- sis it is simple to demonstrate that the resolved shear and normal stress on a vertical plane extending upward from the tip of the aseismically slipped zone will therefore be right-lateral and compressive, in agreement with figures 5A and 5B, respectively. Stress changes along the Hay ward/Calaveras trend (fig. 4) are plotted in figures 5C and 5D. There we plot the change in traction that occurs on a vertical plane trending 21° clockwise from the strike of the modeled coseismic rupture plane versus position along strike measured from a point lying adjacent to the Loma Prieta epicenter and coinciding, approximately, with the southeast end of the M=6.2, 1984 Morgan Hill rupture (Bakun and others, 1984). Coseismic slip and postseismic relaxation lead to de- creases in both right-lateral shear stress and in compres- sive stress along the Hayward/Calaveras trend. At positions 0 to 7 km the coseismic decrease in right-lateral shear stress is exceeded by the decrease in compressive stress. The ratio of the two is about 0.10/0.23 = 0.4 (figs. 5C and 5D). Therefore, this portion of the fault would be moved toward failure if the effective coefficient of friction is MODELS OF POSTSEISMIC DEFORMATION AND STRESS TRANSFER ASSOCIATED WITH THE EARTHQUAKE D261 greater than about 0.4 (see extensive discussion by Calaveras trend. In the far-field, postseismic relaxation of Reasenberg and Simpson (1992) and by Simpson and _ the lower crust can result in changes in shear stress that Reasenberg (this chapter)). are large relative to the static coseismic change (fig. 5C, Postseismic relaxation appears to have only a moderate _ Model 3) but, as noted previously, the total magnitude of effect on the overall state of stress along the Hayward/ _ the stress change is rather small. EXPLAN ATION ”ALP Loma Price profile e BM Black Mountain profile |___ ___ x SA San Andreas profile |- |__| H/C Hayward/Caldveras profile C Loma Pricta hypocenter : A GPS stdtlon f 0 20 40 60 80 100 KM L I 1 1 | ] Figure 4.-Overlay of finite element mesh onto a map of the San Francisco Bay region. The shoreline and major faults are indicated by light and heavy sinuous lines, respectively. The Loma Prieta epicenter (U.S. Geological Survey Staff, 1990) is plotted as a square and geodetic stations occupied by the GPS are plotted as triangles. The perimeter of the finite element mesh is indicated by the dark bounding box and the profiles along which computed stresses and displacements are plotted in subsequent figures are indicated by solid lines. The relative position of the finite element mesh is chosen so that the strike of the model San Andreas fault is N44°W. The model center-line, normal to the fault trace, passes though the main shock epicenter and the model San Andreas fault overlies the mapped fault trace NW of Loma Prieta. AFTERSHOCKS AND POSTSEISMIC EFFECTS D262 7 ut se aryoid aures Suofe ssans aftsua}, 'C '(p 34 298) jng; prem{ep ay) jo puo jseaynos ay} seau e wou; poinseou st ayLns Suore uonisog 'wy ¢'/ 01 0'g feajojut yjdap ay) soa0 paStroate sta -efeqp pue ay) yiim Ajarewnxoidde Surpuodsanoo puan e Suofe uonoen seayg '> +y ut se ajyoud aures Suoje ssans aftsuay, 'g 'yidop ury qJ 01 5, 'auoz atuasowstas AN (SMALAIWNWOTIY) AMRLLS ONOTV NOLLISOd - 4S OOT 09 Of 08 I OT _ 0 --- 0° O - 7TO 1 I 1 0° O 1 TO 1 I 1 | UNAHL SVHMAVTVD/UYVMAVH poy20| 94) Jo y/] Wwonoq ay} ut jng] seaipuy ueg ay} Jo juowdsas ueg ay} Suope uonoen seaygy 'y '(7 '3y) ¢ pue 'z 'I SJopoW Jo suoisioa poxtjal owum-Suof 01 'Ajoanoadsar 'puodsanoo sammo poysep-3uo1 pue 'poysep-uoys 'panoG "aes orustasoo anes 01 puodsanooa saamumns 'aanisod pauoyoot UuoIs p angy ut poreotput spuan Suore Sur4| sour}d peomuioa uo uonoen pue uonotn Jseays fesojef-443U [ejuozuOy Uf sagueyo ay} Jo soyoig-'g amsty (SYAILAIWNOTY) MNALNYOIIH X0 MN HONVLSIG OOT 09 _ 0 t - OT 0 08 |. Noy \ NA \ L \ // |_ y \ I i I i | i I Do \ \\ C I POW \ T POW \ £ fpoW R | i LTIAVA SVAHUUNYV NVS 0° O T 0 0° O T 0 t O (STVSVIVOAW) SSHHLLS 4HTISNAL SSHRLLS YVAHS dI TIS YHMIMLS TV MALVT-LHOI MODELS OF POSTSEISMIC DEFORMATION AND STRESS TRANSFER ASSOCIATED WITH THE EARTHQUAKE A RELATIVE TIME-SCALE FOR POSTSEISMIC RELAXATION In later sections of this paper we analyze time-depen- dent viscoelastic models to constrain the absolute time- scale for the relaxation process. Here, we construct a relative time-scale by estimating the amount of time that background tectonic loading requires to contribute a change in shear stress equal to that resulting from long-time postseismic relaxation. This relative time-scale can later be compared to estimates of the system relaxation time. If, for example, the average tectonic contribution to the rate of shear-stress accumulation is the ratio of a typical earthquake stress drop to an earthquake cycle time, then this tectonic loading rate is likely to be about 5 MP2/200 yr = 0.025 MPa/yr. Along the Peninsula segment of the San Andreas fault, 80 km from the epicenter, the increase in shear stress resulting from complete relaxation of the lower crust is about 0.05 MPa (fig. 5A, Model 3). This increase is then equivalent to about 2 years of tectonic loading. Postseismic relaxation is, therefore, unlikely to contribute significantly to the stress rate at distances be- yond 80 km, unless the time-scale for relaxation of the lower crust is less than a few years. In contrast, 40 km from the epicenter the corresponding increase in shear stress is about 0.4 MPa (fig. 5A, Model 3) and conse- quently represents about 16 years at the tectonic loading rate. Therefore, relaxation of the lower crust will contrib- ute significantly to the stress rate in the region near the end of the Loma Prieta rupture, provided that the time- scale for relaxation is on the order of 16 years or less. TIME-DEPENDENT MODELS We analyze our three geometrical models of postseismic relaxation (fig. 2) while employing two different Maxwell viscoelastic rheologies. First, we use linear viscoelasticity to model the presumably nonlinear behavior of materials in the inelastically shearing portions of the fault zone and lower crust. With this approach, any estimate of relax- ation time, obtained by fitting our models to data, must be interpreted as an effective relaxation time that physically reflects the response of a nonlinear creep process to stress changes resulting from the earthquake. Our second set of time-dependent models employ a nonlinear rheology that represents a steady-state creep version of the laboratory- derived rate- and state-dependent friction law (Stesky, 1975, 1978; Dieterich 1981; Ruina, 1983; Blanpied and others, 1991). With this law the aseismic slip rate de- pends exponentially on the ratio of shear stress to effec- tive normal stress. Using either viscoelastic rheology, we treat the static coseismic stress change and resulting deformation as D263 changes to the state that would have existed in absence of the earthquake. It can be shown that the postseismic prob- lem can be treated exactly in this manner if one assumes that a preexisting steady-state of inelastic shearing existed in all aseismic regions of the model prior to the earth- quake. The more traditional approach to this problem is to precondition the model by imposing a cycle of repeated characteristic earthquakes until the system responds in a periodic manner (Savage, 1983; Thatcher and Rundle, 1984; Li and Rice, 1987; Dmowska and others, 1988; Fares and Rice, 1988; Lyzenga and others, 1991; Ben- Zion and others, 1993; Reches and others, 1994). In our situation, this strategy may be difficult to apply with any confidence since there is some question as to how the Loma Prieta earthquake relates to characteristic events for the region (Anderson, 1990; Segall and Lisowski, 1990). We represent the aseismically shearing regions (fig. 2) with Maxwellian viscoelasticity as 8-A—8 +8“, where e“ is the elastic strain, g;/ is the inelastic creep strain, and €,, is the total strain. We adopt the standard assumptions of elementary plasticity theory-that the in- elastic creep strain is not affected by hydrostatic pressure and that the volumetric creep strain is zero (see, for ex- ample, Malvern, 1969). The creep law can then be written in terms of deviatoric stress, S; =9, -(1/ 3)(5kk8v, where G,, is the stress tensor and we sum on repeated indices. In the creeping portion of the model 0; = (K‘§“)€kk5rj +21(E,, - eg). where K and u are the elastic bulk and shear moduli, respectively. We adopt the von Mises formulation and write the flow law in terms of the second invariant of the deviatoric stress, J, =S,,8,, /2. The Mises equivalent ten- sile stress, q, is defined so that, in a state of uniaxial tension, q is equal to the tensile stress: The creep law is then written as S,, UJ 3 8,4 =---£ 7 q €.. (q), Y where €.,(q) is the scalar Mises equivalent tensile creep rate, defined so that g.,(q)=&{" in response to a uniaxial load G;, =q. D264 LINEAR VISCOELASTIC MODELS In the case of linear viscoelasticity €" (q)=q/A, where A is a constant. In fluid mechanics, it is conventional to define a shear viscosity, 1, such that in a state of pure shear t=1y,. where T is the shear stress and yis the engineering-shear-strain rate. We adopt this convention so that A=37 and ¢ = -L, €..(4) m, The linear creep law as employed in our finite element calculations is, therefore, ner __ Sij U 2T] The viscous resistance to shear across a thin layer de- pends on the ratio of viscosity to layer thickness, /A. Dimensional analysis, therefore, tells us that the time- scale for relaxation of the system will be proportional to H h where H is the effective length-scale of the elastic sur- roundings and ¢, is the relaxation time of the viscoelastic material. In Models 1 and 2, H corresponds to the down- dip dimension of the coseismic rupture while in Model 3, H corresponds to the thickness of the elastic portion of the crust. In our finite element calculations we treat H as a known quantity. By comparing model predictions to data we can, therefore, hope to constrain the ratio t,/k but nei- ther 1, or h independently. We include a 1-km-thick fault zone in our finite element models, and in Model 3 addi- tionally include a 5-km-thick basal shear zone. We then make 1,/h uniform throughout the model by specifying that the viscosity of the material in the basal shear zone is five times that of the material in the aseismic portion of the fault zone. By this procedure, time is measured in units of 1, km/h and t, and h are the unknown material relaxation time and shear-zone thickness in the earth. By comparing computed displacements to those ob- served geodetically we can hope to equate model-time O, km/h with time ¢ and by doing so constrain 1,/h. Rewriting this expression, we obtain which demonstrates how the parameters used to constrain viscosity will trade off. Note that if (/@&)(h/km) is of or- AFTERSHOCKS AND POSTSEISMIC EFFECTS der one year and =30 GPa, then n~10'8 Pa-s. In what follows we examine solutions at ¢=5 and =20 and geo- detic observations at t=1.3 years. In figure 6 we plot calculated and observed postseismic displacements measured relative to extrapolation of the pre-earthquake velocity field. The displacements are plot- ted along two profiles that lie perpendicular to the strike of the model fault zone (fig. 4). The first profile crosses through the center of the rupture and is referred to as the Loma Prieta profile. The second lies 44 km to the north- west and is referred to as the Black Mountain profile. These two profiles correspond approximately with two profiles of GPS stations measured frequently since the earthquake (Biirgmann and others, 1991, and this chapter; Lisowski and others, 199 1a; Biirgmann and others, 1992; Savage and others, 1994). We plot three components of displacement from Models 1, 2, and 3 (fig. 2) at two model times, 1=5¢,. km/h in figure 6A and 1:=20¢, km// in figure 6B. At t=5t, km/h the computed displacements are nearly equal for the three models (fig. 6A). This indicates that at short-time all three models are dominated by aseismic fault slip in the down-dip portion of the fault zone con- fined above the Moho. At 1=20t,, km/h the computed dis- placements for the three models differ to a larger degree (fig. 6B). Therefore, if one of these models is correct for Loma Prieta at long-time, there is some hope of resolving which one it is. Also plotted in figure 6 are measures of the observed postseismic displacement field 1.3 years after the earth- quake, made relative to extrapolation of the pre-earth- quake velocity field. The bulk of the pre-earthquake geodetic data are from ground-based laser ranging mea- surements (EDM) and so do not constrain the rigid body component of the pre-earthquake velocity field. In con- trast, all of the post-earthquake data are from GPS mea- surements in which the rigid body motion is constrained by making simultaneous observations to remote stations. As of this writing, though there are some pre-earthquake GPS observations, there is no self-consistent estimate of the change in the velocity field-post-earthquake minus pre-earthquake-based on all of the geodetic data, and so we have plotted the data as we describe in the following paragraphs. The data plotted along the Loma Prieta profile repre- sent the difference between estimates of post- and pre- earthquake velocity obtained by assuming that the velocity field far to the northeast of the San Andreas fault did not change (Lisowski and others, 1991a, 1991b). This mea- sure of the change in velocity is then multiplied by 1.3 years to obtain an estimate of the postseismic displace- ment measured relative to extrapolation of the pre-earth- quake velocity. The error bars assume no error in the estimate of the pre-earthquake velocity field and thus un- derestimate the actual uncertainty. MODELS OF POSTSEISMIC DEFORMATION AND STRESS TRANSFER ASSOCIATED WITH THE EARTHQUAKE A DISPLACEMENT (MILLIMETERS) DISPLACEMENT (MILLIMETERS) Figure 6.-Profiles of observed and calculated postseismic displacements measured relative to extrapolation of the pre-earthquake velocity field. The Loma Prieta and Black Mountain profiles lie perpendicular to the strike of the San Andreas fault. The former passes through the center of the model rupture and also through the Loma Prieta epicenter, while the latter lies 44 km to the northwest along strike (see fig. 4). Observed displacements are based on geodetic measurements and correspond to 1.3 years after the earthquake (Lisowski and others, 1991a, 1991b; Biirgmann and others, 1992). Calculated displacements represent versions of Models 1, 2, and 3 (solid, dotted, and dashed curves, respectively) in which the relaxing portions of the models are made linear Maxwell viscoelastic (see fig. 2). Figures A and B correspond to model times of 5t, km/h and 20t, km/h respectively, where 1, is the Maxwell relaxation time NORTHEAST NORTHWEST NORTHEAST NORTHWEST 100 50 -50 100 50 -50 100 100 100 50 -50 100 LOMA PRIETA PROFILE T T T T T i il t. - 7 _ + H 1 1 1 1 1 -20 0 20 _ 40 _ 60 T T T T t - m . *+ * £) 1 i1__a ___ 1 i -20 0 20 40 60 BLACK MOUNTAIN PROFILE 100 T T T I I 50 | 7 _50- i 1 1 1 i 100 T * T T * T OT 50 E 4 -50 L _i __ 1 1 1 100 - t t t T _50 C 1 1 " 1 1 -20 0 20 _ 40 60 DISTANCE NORTHEAST OF FAULT TRACE (KILOMETERS) LOMA PRIETA PROFILE BLACK MOUNTAIN PROFILE 100 - t t t DISTANCE NORTHEAST OF FAULT TRACE (KILOMETERS) of the viscoelastic material, and / is the thickness of the viscously deforming region. D265 D266 Analysis of the Black Mountain data indicates that the change in velocity there-post- minus pre-earthquake- may not be measurably different from zero (Biirgmann and others, 1992). Lacking a better estimate of the rela- tive postseismic deformation, we plot zero displacement along the Black Mountain profile to indicate that the rela- tive postseismic displacement is likely to be small. Along with the null data, we also plot an estimate of the uncer- tainty in the displacement accumulated during the first 1.3 years after the earthquake. We estimate this uncer- tainty by taking the uncertainty in the velocity measured during the post-earthquake interval January 1990 to May 1992 (Biirgmann and Segall, written commun., 1993) and multiplying by 1.3 years. The resulting error bars place a lower bound on the detection threshold of postseismic deformation relative to the pre-earthquake rate. Relative postseismic deformation smaller than these error bars would go unnoticed in 1.3 years. At 1=5t, km/h the predicted fault-trace-normal compo- nents of displacement are inconsistent with the data along the Loma Prieta profile, but along the Black Mountain profile they may be small enough to satisfy the data. If we were to add about 0.3 ppm or 0.2 ppm/yr of regional fault-normal compression to our model, then we could probably match the data along the Loma Prieta profile but then probably violate the Black Mountain data. The pre- dicted strike-slip components of displacement appear to be remarkably consistent with the data along the Loma Prieta profile but may not be consistent with the observa- tions along the Black Mountain profile, where the com- puted displacements appear to exceed the observations. The predicted vertical displacements along the Loma Prieta profile may be sufficiently large to be detected by GPS measurements, whose precision in the vertical is about 20 to 30 mm (Davis and others, 1989; Biirgmann and others, this chapter). The predicted vertical displacements along the Black Mountain profile are extremely small. The pre- dicted tilts along the Loma Prieta profile exceed 1 ura- dian and so could potentially be detected by leveling measurements. At 1=20t,, km/h, the predicted displacements far exceed all of the observations, and so we can say with confidence that 207, km/h > 1.3 yr or t,/h > 0.07 yr/km. If we accept the agreement at t=5¢, km/h between pre- dicted and observed strike-slip displacements along the Loma Prieta profile while overlooking the disagreement with fault-trace-normal motion there as well as possible disagreement with strike-slip motion along the Black Mountain profile, then we can conclude that 5t, km/h = 1.3 yr or t,/h = 0.3 yr/km. If we assume that postseismic slip is occurring within a fault zone with A < 1 km and take u=30 GPa, we then obtain an estimate of the effective viscosity of the deep aseismic portion of the fault zone; Mey < 3x10'7 Pa-s. This value is lower than any obvious interpretation of AFTERSHOCKS AND POSTSEISMIC EFFECTS laboratory measurements (Carter and Tsenn, 1987; Kirby and Kronenberg, 1987a, 1987b) of solid-state creep of crustal rocks at mid-crustal conditions (Lachenbruch and Sass, 1973). Li and Rice (1987) and Fares and Rice (1988) studied two-dimensional earthquake cycle models by comparing model calculations to the observed decay in strain rate that followed the 1906 San Francisco earthquake (fig. 1). Their physical model is equivalent to our Model 3 except that they assumed that the aseismic down-dip portion of the fault zone is freely slipping. As discussed previously, this assumption is equivalent to assuming that 1,// in the aseismic portion of the fault zone is much much less than in the lower crust, that is (t/h)fz << (t/h);,. It is likely that this assumption only affects their solutions at short- time following the earthquake and does not strongly af- fect their estimate of (1,/h);,. They obtained values in the range 0.2 yr/km < (t,/h);,. < 1.7 yr/km. The contemporary velocity field in the northern San Francisco Bay region (Prescott and Yu, 1986) requires values at the low end of this range (Fares and Rice, 1988). We, therefore, conclude that may be approximately uniform in the deep aseismic portion of the fault zone and in the lower crust and appears to take on a value of about 0.3 yr/km thickness. The apparent spatial uniformity of t,/ h indicates that the proposed deep postseismic slip that occurred in the 1.3 yr following the Loma Prieta earth- quake and the basal shear that followed the 1906 earth- quake might be controlled by the same physical process. We interpret the apparently consistent values of ( t/h)fz and (1,/h);,, and the corresponding low value of the effec- tive viscosity of the material in the deep aseismic portion of the fault zone as indications that postseismic relaxation may result from frictional sliding assisted by elevated pore pressure. This process may be taking place both in the deep portion of the fault zone and in horizontal shear zones in the lower crust. In the next section of this paper we evaluate this proposition by examining models of postseismic deformation that incorporate a constitutive law for frictional sliding. NON-LINEAR VISCOELASTIC MODELS: HOT-FRICTION We now represent the aseismically shearing regions with a nonlinear creep law derived from a steady-state version of the laboratory-derived rate- and state-dependent fric- tion law (Stesky, 1975, 1978; Dieterich, 1981; Ruina, 1983; Blanpied and others, 1991). At steady-state the resistance to sliding is described approximately by T= 10+c01n V/Vo, where V is the slip rate, c is a constant, 0 is the effective normal stress, and T, is the shear stress when V=V,. We MODELS OF POSTSEISMIC DEFORMATION AND STRESS TRANSFER ASSOCIATED WITH THE EARTHQUAKE interpret T, as the pre-earthquake shear stress across the aseismic portion of the fault and, in the case of Model 3, across the basal shear zone in the lower crust, 0 as the effective normal stress equal to total normal stress minus pore pressure, V as the slip rate across the shear zone, and V,, as the pre-earthquake slip rate. In using a steady-state form of the friction law, we implicitly assume that postseismic slip greatly exceeds the slip weakening dis- tance and so the state variable remains always at its steady- state value, appropriate to the current slip rate. The factor c is more commonly denoted as a - b (see, for example Blanpied and others (1991)). While we use the friction law to describe hot frictional slip occurring deep within the crust, the same law has been used in studies of shal- low aseismic creep on near-surface portions of a fault zone (Wesson, 1988; Marone and others, 1991). We make our calculations relative to the pre-earthquake state and define new parameters in that context. The fric- tion law becomes a V, +V T= , V 0 where 4 and V are the shear stress and slip speed mea- sured relative to the pre-earthquake state, and V =V -V,. In a narrow shear zone, a state of pure shear will dominate the deviatoric stress tensor (Rice, 1992). Therefore, from the previous section, q= J}? and €., =1//3 =V /~/3h, where q and &,, are now the Mises equivalent tensile stress and strain rate measured relative to the pre-earthquake state, and / is, as before, the thick- ness of the shear zone. Inverting the friction law we thus obtain £..(q)= 7V§Z(eqlfiw _ 1), so that the nonlinear creep law employed in our finite element calculations is gg @R a fartte .) We refer to the above formulation as our "hot-friction" model. Dimensional analysis tells us that the general form of the post-seismic slip rate across the aseismically slipping regions, measured relative to the pre-earthquake rate, will have the form P= v HL a £) coHl co H D267 where t is the time since the earthquake, At is a represen- tative coseismic static-stress change in the creep zone, V,, is a representative pre-earthquake slip rate, and x is position. The time-scale for postseismic relaxation is now governed by the quantities coH/uV,, Atco, and x/H. In our finite element calculations we assume values of H, such as rupture depth in the case of Models 1 and 2, and thickness of the elastic portion of the crust in the case of Model 3, just as with the previous linear models. We additionally assume that =30 GPa, as in all previous analyses. To constrain V,, we have studied two- dimensional earthquake cycle models in the manner of Li and Rice (1987) and Fares and Rice (1988) but using the steady-state friction law. Preliminary analyses in which we compare the decay in strain rate to that observed after the 1906 earthquake (fig. 1) indicate that 0.1 < Va/Vplate < 1.0 in the deep aseismic portion of the fault zone within the crust, where Vplaze is the average plate velocity. This range is consistent with values obtained by Tse and Rice (1986) and by Rice (1993), in which they incorporate the full rate- and state-dependent constitutive law throughout the crustal fault zone and thereby model both stable and unstable fault slip throughout the earthquake cycle. We use V0=Vplm=20 mm/yr, which is representative of esti- mates of the average slip rate along this segment of the San Andreas fault (Working Group, 1990). In our calcula- tions, smaller values of V0 would yield smaller values of postseismic slip rate since there is a direct trade-off be- tween V,, and our measure of time. Finally, laboratory measurements indicate that at mid-crustal conditions of elevated temperature with pore fluids present, 0.010 < c < 0.020 (Blanpied and others, 1991). These constraints on H, u, V,,, and c, together with comparisons of model pre- dictions and geodetic data, provide us a means of con- straining the effective normal stress, G. We analyze Models 1, 2, and 3 (fig. 2) while incorpo- rating the hot-friction constitutive law in the deep aseismic portion of the fault zone and lower crust. In figure 7 we plot profiles of postseismic displacement analogous to those in figure 6. The solution plotted is for Model 1 with co=0.5 MPa. Larger values of co yield greatly reduced postseismic slip rate and so also reduced displacement at the free surface. This observation is consistent with in- spection of the creep law and with our dimensional analy- sis. The former indicates that the initial postseismic slip rate will scale exponentially with the ratio Atco. Models 2 and 3 yield displacement profiles nearly identical to those in figure 7 at times less than or equal to 2 years. Only at greater times do the surface displacements from the three models differ substantially. Perhaps the greatest difference between the response of the nonlinear hot-friction model and the linear viscoelas- tic model is their time dependence. In figure 8 we plot the strike-slip components of displacement, from nonlinear AFTERSHOCKS AND POSTSEISMIC EFFECTS D268 'EdW O€ = SIO°O/MIW SO Jo 19p10 ay} uo st auoz jng; ay; yo uonod arustase daap ay) ut ssans feuniou aanoaJyo ay} Jey) sojeatput tjep ay} pure suonotpaid fapour ay} usomraq juawaise aAngjor ouf, 'ayenbyprea oy) soupe sreok ¢'1 01; puodsanos squowaor[dsip paarosqo ay} 'gq ans; ut sy 's1ayjo pur 0700 01 010°O aSue1 ay} ut st 2 feisnuo-piu je apeur sjuaurainseaur uj 'arnssord arod ay} snutu ssans feuniou [2403 ay} 01 jenba 'ssans reuiou aanoay}a ay} SI 0 alo '2dJA J0J st panord uonnjos ay J, uonoauy-j0y seaurjuou ay} s4aqo auoz jng] ay} Jo uotSa1 doop ay) ut reuorew onseJaoosta ay) Yoljm UI (Q7 '31J) | jopop Jo uots1aA e Joj ing 9 angy ut asoy 01 snoSofeue Jo (SMALIWOTDN) HDVML LIAVA AO LSYVAIHILMON AHDNV.LSIG 09 Of _ OZ 0 OT- 09 Of _ OZ 0 OT- 1 OLL T-r-r-T T 1 _L OL bae d uuu he) mmr dll w = o n Ng O pa d uue O w --i oof 09 Of _ OZ 0 OT- 09 Of OT 0 OT- ror-r-t - O&O he) ba ba fug 09 Of I i | 1 | 1 | 1 [ OO~ TTorTpfororT- or- | 1 1 _ I i | 1 1 COM LSZMHILNON (SMALAWITIIW) os LSVAIHILYON feranane ATIMO¥Ud NIVLNAOWN MOV TIG I i | 1 | 1 | i I COM ATMHMOHUd VLIL4Id VWNOTU MODELS OF POSTSEISMIC DEFORMATION AND STRESS TRANSFER ASSOCIATED WITH THE EARTHQUAKE and linear versions of Model 1, of a point lying on the Loma Prieta profile 13 km southwest of the model fault trace (see figs. 2C, 4, 6, and 7). We again use co=0.5 MPa in the nonlinear model, as in figure 7, and then choose the time-scale for the linear model so that the displace- ment at 2 years equals that of the nonlinear model. The nonlinear hot-friction model yields very rapid postseismic motion immediately after the earthquake in comparison to the linear model. This very rapid short-term motion re- sults principally from rapid postseismic slip that occurs near the edge of the coseismic rupture, where AT is large compared to co. Given time, this highly stressed region of the model relaxes to the degree that 7 c0. As a result, estimates of coseismic deformation based on geo- detic data may be contaminated by postseismic deforma- tion whenever the first post-earthquake survey lags too far behind the time of the earthquake. In figures 7 and 8, for example, over half of the displacement that accumu- lates in 1.3 years occurs during the first month following the earthquake. Note, however, that the displacements at issue here are small compared to the coseismic displace- ment. Returning again to figure 7, we compare the model predictions to the geodetic data. Along the Loma Prieta | Hot-Friction Model |___..." _ Linear Model | -- Fo & 30 | 3 8 a t < 2 F 2 a [ a go *" [ Q2 L. E G F m 1 |- [] -= - Ss T L - . x [ Z O_J;4‘AAAAI 0.0 0.5 TIME SINCE EARTHQUAKE (YEARS) Figure 8. -Time histories of displacement of the free surface for nonlinear versus linear versions of Model 1 (fig. 2C). The nonlinear hot-friction model used ca=0.5 MPa, as in figure 7. The time-scale for the linear model was chosen so that the displacement at two years equals that of the nonlinear model. The point for which displacement is plotted lies along the Loma Prieta profile 13 km to the southwest of the fault trace. Note that with the nonlinear hot-friction model, the displacement at 1 month after the earthquake is nearly one-half of the total that accumulates by the end of 2 years. D270 profile, the hot-friction model does reasonably well in comparison to the observed strike-slip component of dis- placement but does not generate displacements compat- ible with the observed fault-trace-normal displacements. The addition of a uniform fault-trace-normal component of compression would bring the predictions toward agree- ment with the data along the Loma Prieta profile. Along the Black Mountain profile, the computed strike-slip dis- placements appear to exceed the observed values, but, as discussed previously, the error bars do not include the uncertainty in the pre-earthquake velocity field, nor do they address the unconstrained rigid body motion associ- ated with the pre-earthquake EDM data. All of these com- A AFTERSHOCKS AND POSTSEISMIC EFFECTS parisons are much like those made with the linear models (fig. 6A). We can conclude that the effective normal stress in the deep aseismic portion of the fault zone is extremely low, provided that we accept the agreement between the pre- dicted and observed strike-slip displacements along the Loma Prieta profile while overlooking the disagreements both with fault-trace-normal motion there as well as with strike-slip motion along the Black Mountain profile. If we accept that c0=0.5 MPa and take c=0.015 from the labo- ratory measurements (Blanpied and others, 1991), we then obtain an estimate of the effective normal stress in the deep aseismic portion of the fault zone; 0=33 MPa. This MISES STRESS (MEGAPASCALS) 0.2 0.4 0.6 0.8% 1.0 1.2 Figure 9.-Contours of Mises stress in the deep aseismic portion of the fault zone for the nonlinear hot-friction version of Model 1 (fig. 2C) using co=0.5 MPa. The contour interval is 0.1 MPa and the maximum contour level is 1.2 MPa. A, Static coseismic stress state. B, Stress state after 2 years of relaxation. Note the development of a relaxation front that propagates from the edge of the rupture into the creep zone as relaxation proceeds. MODELS OF POSTSEISMIC DEFORMATION AND STRESS TRANSFER ASSOCIATED WITH THE EARTHQUAKE value is a factor of 10 smaller than the effective overbur- den stress that would exist at 18 km depth if pore-pres- sure was hydrostatic. Therefore, if deep aseismic slip is the cause of the observed postseismic strike-slip deforma- tion near Loma Prieta and if hot-frictional shearing is the mechanism by which this process takes place, then pore pressure in the deep aseismic portion of the fault zone must be substantially elevated above the hydrostatic level. DISCUSSION AND CONCLUSIONS ALTERNATIVE GEOMETRIC MODELS We have demonstrated that models of postseismic de- formation that include deep sources of viscous relaxation can generate displacement fields that agree with some, but not all, of the geodetic data. Our models appear to be capable of producing strike- slip displacements that are compatible with the observa- tions along the Loma Prieta profile but may exceed those made along the Black Mountain profile. We have exam- ined one model of postseismic deformation in which coseismic slip is non-uniform. The goal was to establish the degree to which postseismic strike-slip displacement along the Black Mountain profile could be suppressed by moving the centroid of coseismic strike slip away from Black Mountain. The non-uniform slip model is motivated by analysis of leveling data (Marshall and others, 1991) and by seismological estimates of the distribution of coseismic slip (Beroza, 1991; Hartzell and others, 1991; Steidl and others, 1991; Wald and others, 1991). We have examined an extreme model in which all of the strike slip is in the southeast half of the rupture and all of the re- verse slip is in the northwest half of the rupture. The moment and average rake are as in the uniform slip mod- els. The non-uniform slip model employs the linear vis- coelastic rheology in the deep aseismic portion of the fault zone and so falls into the category of Model 1 (fig. 2C). As expected, the non-uniform model does suppress postseismic strike-slip displacement along the Black Moun- tain profile in comparison to the corresponding uniform- slip model. At t=5t, km/h the strike-slip displacement along the Black Mountain profile is about 70 percent of that obtained with the uniform slip model and plotted in figure 6A. However, this reduction is more than compen- sated for by an accompanying reduction in strike-slip dis- placement along the Loma Prieta profile. To match the data there we must now choose 1.3 yr = t = 201, km/h, which corresponds to larger strike-slip displacements along the Black Mountain profile than those obtained with the uniform slip model at 1=5t, km/h. Thus, the non-uniform coseismic slip model does not help to suppress postseismic strike-slip displacements along the Black Mountain pro- file relative to along the Loma Prieta profile. D271 Our models also appear to be incompatible with the fault-trace-normal contraction observed along the Loma Prieta profile. This discrepancy could be suppressed by adding a uniform fault-trace-normal compressive strain to our models of magnitude about 0.3 ppm, or 0.2 ppm/yr. This strain would be interpreted as a regional postseismic change in fault-trace-normal strain rate, though we are not aware of any independent data that reveals such a feature. The addition of this regional fault-trace-normal compression would, furthermore, make our models less consistent with the data from the Black Mountain profile, if the change in velocity along the Black Mountain profile is indeed not measurably different from zero (Biirgmann and others, 1992). We conclude that alternative models of postseismic de- formation should be examined in future work. Two over- lapping viewpoints should be considered. First, our deep-slip models do a remarkably good job of matching the strike-slip component of postseismic displacement along the Loma Prieta profile, but they do not appear to be capable of generating the observed fault-trace-normal contraction there. Therefore, future models should include similar deep-slip sources, while additionally incorporat- ing sources that contribute a greater degree of fault-trace- normal contraction. Second, our deep-slip models may generate strike-slip displacements along the Black Moun- tain profile that exceed the observed displacements. There- fore, future work should also examine models that incorporate sources of deformation that are shallower than our deep-slip sources. By moving the deformation sources toward the free surface the width of the postseismic dis- placement field will be narrowed relative to those gener- ated with our deep sources. This narrowing of the displacement field will make it easier to satisfy the data from the Black Mountain profile, if indeed the current lack of postseismic signal holds up to further data analy- sis. A planned re-leveling across the Loma Prieta profile (W.H. Prescott, written commun., 1993) may also help to resolve the depth of the postseismic deformation source. Biirgmann and others (1992) have performed trial-and- error searches to determine the location of the rectangular dislocation with uniform slip that best fits the geodetic data. They conclude that the postseismic source probably lies at depths above the 18-km-deep Loma Prieta hypo- center. They further conclude that postseismic slip oc- curred not only in the Loma Prieta aftershock zone but also on a shallow thrust fault lying north of Loma Prieta. If their model is correct, then perhaps postseismic slip in the aftershock zone represents stress relaxation between patches that slipped seismically in the earthquake. This interpretation is similar to our deep-source models, but now creeping patches lie within the seismogenic zone and presumably represent aseismic "non-asperity" regions. This view is in accord with the interpretation that aftershocks concentrate in regions that have relatively small coseismic D272 slip (Oppenheimer and others, 1990) and are therefore most severely stressed by the earthquake. Analyses like ours could be performed by including shallow creeping patches within the seismogenic zone, whose location is constrained by seismologic estimates of the slip distribu- tion in the earthquake. The shallow thrust fault proposed by Biirgmann and co-workers may be confined to rather shallow depths and consequently may represent stable fault creep (Tse and Rice, 1986; Scholz, 1990) of the sort examined by Marone and others (1991). Such a model could be examined within the framework of our finite element analyses or through use of the boundary element method. COMPARISON TO STRAIN-METER OBSERVATIONS AT SAN JUAN BAUTISTA The strain rate within a few kilometers of the surface trace of the San Andreas fault near San Juan Bautista (SJB) has been monitored by a Gladwin borehole tensor strain meter since 1983. The observations surrounding the time of the Loma Prieta earthquake are summarized here (Gwyther and others, 1992). Approximately one year prior to the Loma Prieta earthquake, the right-lateral strain rate increased to about 1 uradian/yr. Immediately following the earthquake, a 2-month transient occurred in which the right-lateral strain rate actually became negative and about 0.5 uradian of left-lateral strain accumulated. Following that short transient, the strain rate increased to about 2 puradian/yr, right-lateral, and has remained at that level for about 2 years. These changes in strain rate presumably reflect variations in slip rate on the fault that may be associated with the earthquake. While our models do not attempt to include the effects of the shallow San Andreas creep zone that extends 150 km to the southeast from SJB, it is still interesting to compare our model predictions to the data. In the case of our nonlinear hot-friction models, the right-lateral strain rate near SJB is very high immediately following the earth- quake but rapidly decays to about 0.4 uradian/yr above the pre-earthquake rate when averaged over the interval 0.5 to 2.0 postseismic years. This temporal behavior is qualitatively consistent with the time history of displace- ment plotted in figure 8. None of our models predict an immediate postseismic transient in which left-lateral strain accumulates in the vicinity of SJB. Furthermore, the pre- dicted rate of postseismic strain measured relative to the pre-earthquake rate is only 40 percent of the observed value. The former is about 0.4 uradian/yr, while the latter is about 1 uradian/yr. We conclude that models that at- tempt to include the effects of the shallow San Andreas creep zone are required to obtain better agreement with the data from SJB. AFTERSHOCKS AND POSTSEISMIC EFFECTS ALTERNATIVE PHYSICAL PROCESS We have proposed that postseismic deformation may result from frictional sliding assisted by elevated pore pres- sure. Alternatively, postseismic transient deformation may result from transient creep (Carter and Kirby, 1978; B. Evans, oral commun.). The displacement across our deep aseismic fault zone, required to match the observed strike- slip displacement along the Loma Prieta profile at 1.3 years, is about 0.5 m at a distance of 2.5 km from the edge of the coseismic rupture. If this deformation occurs within a shear zone with thickness h < 1 km, then the corresponding transient strain is y,, > 0.5 x10 and its av- erage rate is y,, >2x10~" /s. SUMMARY We have demonstrated that postseismic relaxation can increase the stress change at intermediate distances from the edge of the coseismic rupture by an amount well in excess of the coseismic change. Postseismic relaxation can have this same effect in the far-field, but the magni- tude of the total stress change there is very small com- pared to typical earthquake stress drops. Along the Hayward and Calaveras fault trends, northeast to north of Loma Prieta, both coseismic slip and postseismic relax- ation reduce the right-lateral shear stress and the com- pressive stress. Our models of deep aseismic fault slip occurring be- neath the seismogenic zone are capable of producing strike- slip displacements consistent with the initial 1.3 years of observed postseismic deformation along the profile of GPS stations that crosses through the epicenter (Lisowski and others, 1991a). However, these models appear to be inca- pable of producing fault-trace normal contraction compat- ible with the observations there, and an additional source of deformation may be required. For example, Biirgmann and others (1992, and this chapter) investigate models that include postseismic fault slip on shallow thrust faults north of Loma Prieta, while Savage and others (1994) consider models that include postseismic collapse of the coseismic rupture zone in the direction perpendicular to the plane of the rupture. Linker (1993) has shown that the magnitude and time-scale of Savage and co-workers' pro- posed fault-trace-normal collapse may be consistent with a relaxation time-scale controlled by postseismic flow of pressurized fluid out of the fault zone and into the coun- try rock. Along the Black Mountain GPS profile, northwest of Loma Prieta, postseismic displacements predicted by these same deep-slip models may exceed the observations, though a careful treatment of all of the pre-earthquake and post-earthquake geodetic data is required to better MODELS OF POSTSEISMIC DEFORMATION AND STRESS TRANSFER ASSOCIATED WITH THE EARTHQUAKE establish a measure of the change in the velocity field associated with the earthquake (see Biirgmann and others, this chapter). The time-scale for postseismic relaxation appears to be sufficiently long that the average background tectonic stress rate is likely to exceed the stress rate resulting from relaxation, except perhaps close to the edge of the coseismic rupture. If the time-scale for relaxation of the lower crustal detachment zone turns out to be on the order of 16 years or less, then its contribution to the stress rate within about 20 km of the end of the Loma Prieta rupture will be at least comparable to the average background stress rate. In our linear viscoelastic models, the parameter con- trolling the relaxation time is the ratio of material relax- ation time to the thickness of the deforming region, t,/h. This parameter appears to have a value of approximately 0.3 yr/km thickness, at least in the deep aseismic portion of the fault zone. If relaxation takes place in a shear zone that is less than 1 km thick and if the shear modulus is 30 GPa, then the effective viscosity in the deforming region must be less than about 3x10" Pa-s. This value is at least one order of magnitude lower than any obvious interpre- tation of laboratory measurements of the steady-state creep of solid crustal rocks at mid-crustal conditions. The laboratory-derived hot-friction model can yield postseismic deformation of magnitude comparable to that observed, but only if dw/d(In V) is on the order of 0.5 MPa, where 7 is the resistance to sliding and V is the slip speed. Laboratory measurements indicate that dt/d(In V) can be written approximately as co (Stesky, 1975, 1978; Dieterich, 1981; Ruina, 1983), where 0 is the effective normal stress equal to total normal stress minus pore pres- sure and c is on the order of 0.015 for mid-crustal condi- tions (Blanpied and others, 1991). Therefore, to be consistent with our models, the effective normal stress in the deep aseismic portion of the fault zone must be ex- tremely low, perhaps on the order of 30 MPa. In the context of the laboratory-derived hot-friction model, the occurrence of postseismic deformation may be evidence that pore pressure in the aseismic portions of the fault zone and perhaps the lower crust is near lithostatic. ACKNOWLEDGMENTS We wish to thank R. Biirgmann, M. Lisowski, J.C. Sav- age, and P. Segall for early access to the GPS data and for helpful discussions. We additionally thank B. Evans and R.W. Simpson for helpful discussions. P. Segall and R.W. Simpson provided careful reviews that led to correction of a sign error. This work was supported by the United States Geological Survey, National Earthquake Hazards Reduction Program, under grants G1844 and G1788 to D273 Harvard University. Supercomputer services were provided by grants to Harvard University from the NSF National Center for Supercomputing Applications, Urbana Illinois. Some of the study was done through access to a network of workstations supported primarily by J. Bloxham, A.M. Dziewonski, G.A. Ekstrom, and R.J. O'Connell at Harvard. Hibbitt, Karlsson, and Sorensen, Inc. of Pawtucket, Rhode Island, provided ABAQUS through an academic license agreement. REFERENCES CITED Aki, K., and Richards, P.G., 1980, Quantitative seismology: New York, Freeman, 932 p. Allen, C.R., 1969, Active faulting in northern Turkey: California Insti- tute of Technology, Division of Geological Sciences Contribution 1577, 32 p. Anderson, R.S., 1990, Evolution of the northern Santa Cruz mountains by advection of crust past a San Andreas fault bend: Science, v. 249, p. 397-401. Bakun, W.H., Clark, M.M., Cockerham, R.S., Ellsworth, W.L., Lindh, A.G., Prescott, W.H., Shakal, A.F., and Spudich, P., 1984, The 1984 Morgan Hill, California, earthquake: Science, v. 225, p. 288- 291. Ben-Zion, Y., Rice, J.R., and Dmowska, R., 1993, Interaction of the San Andreas fault creeping segment with adjacent great rupture zones, and earthquake recurrence at Parkfield: Journal of Geo- physical Research, v. 98, p. 2135-2144. Beroza, G.C., 1991, Near source modeling of the Loma Prieta earth- quake: Evidence for heterogeneous slip and implications for earth- quake hazard: Bulletin of the Seismological Society of America, v. 81, p. 1603-1621. Blanpied, ML., Lockner, D.A., and Byerlee, J.D., 1991, Fault stability inferred from granite sliding experiments at hydrothermal condi- tions: Geophysical Research Letters, v. 18, p. 609-612. Brace, W.F., and Kohlsted, D.L., 1980, Limits on lithospheric stress imposed by laboratory experiments: Journal of Geophysical Re- search, v. 85, p. 6248-6252. Biirgmann, R., and Segall, P., 1991, Postseismic GPS monitoring NW of the Loma Prieta rupture zone: Eos, Transactions of the Ameri- can Geophysical Union, v. 72, p. 119. Biirgmann, R., Segall, P., Lisowski, M., and Svarc, J.P., 1992, Rapid aseismic slip on the Berrocal fault zone following the Loma Prieta earthquake: Eos, Transactions of the American Geophysical Union, v. 73, p. 119. Carter, N.L., and Kirby, S.H., 1978, Transient creep and semibrittle behavior of crystalline rocks: Pure and Applied Geophysics, v. 116, p. 807-839. Carter, N.L., and Tsenn, M.C., 1987, Flow properties of continental lithosphere: Tectonophysics, v. 136, p. 27-63. Chen, W., and Molnar, P., 1983, Focal depths of intracontinental and intraplate earthquakes and their implications for the thermal and mechanical properties of the lithosphere: Journal of Geophysical Research, v. 88, p. 4183-4214. Davis, J.L., Prescott, W.H., Svarc, J.L., and Wendt, K.J., 1989, As- sessment of global positioning system measurements for studies of crustal deformation: Journal of Geophysical Research, v. 94, p. 13,635-13,650. Dieterich, J.H., 1981, Constitutive properties of faults with simulated gouge, in Carter, N.L., Friedman, M., Logan, J.M., and Stearns, D.W., eds., Mechanical Behavior of Crustal Rocks: Washington, D274 D.C., American Geophysical Union, Geophysical Monograph Se- ries v. 24, p. 103-120. Dietz, LD., and Ellsworth, W.D., 1990, The October 17, 1989 Loma Prieta earthquake and its aftershocks: geometry of the sequence from high-resolution locations: Geophysical Research Letters, v. 17, p. 1417-1420. Doser, D.I., 1986, Earthquake processes in the Rainbow Mountain- Fairview Peak-Dixie Valley, Nevada, region 1954-1959; Journal of Geophysical Research, v. 91, p. 12,572-12,586. Du, Y., and Aydin, A., 1990, Stress transfer during three consecutive moderate earthquakes along the Central Calaveras Fault: Eos, Transactions of the American Geophysical Union, v. 71, p. 1652. England, P., and Molnar, P., 1991, Inferences of the deviatoric stress in actively deforming belts from simple physical models: Philo- sophical Transactions of the Royal Society of London, v. 337, p. 151-164. Fares, N., and Rice, J.R., 1988, Crustal deformation modeling for San Andreas fault earthquake cycles, abstract T32B-07: AGU Fall Meeting Program, p. 149. Fuis, G.S., and Mooney, W.D., 1990, Lithospheric structure and tec- tonics from seismic-refraction and other data: U.S. Geological Survey Professional Paper 1515, p. 207-236. Gwyther, RL., Gladwin, MT., and Hart, R.H.G., 1992, A shear-strain anomaly following the Loma Prieta earthquake: Nature, v. 356, p. 142-144. Hartzell, S.H., Stuart, G.S., and Mendoza, C., 1991, Comparison of L1 and L2 norms in a teleseismic waveform inversion for the slip history of the Loma Prieta, California, earthquake: Bulletin of the Seismological Society of America, v. 81, p. 1518-1539. Hibbitt, Karlsson, and Sorensen, Inc., 1991, ABAQUS, Vers. 4.8 and 4.9: Pawtucket, RL. Hudnut, K.W., Seeber, L., and Pacheco, J., 1989, Cross-fault triggering in the November 1987 Superstition Hills earthquake sequence, Southern California: Geophysical Research Letters, v. 16, p. 199- 202. Hyndman, R.D., and Shearer, P.M., 1989, Water in the lower continen- tal crust: modeling magnetotelluric and seismic reflection results: Geophysical Journal International, v. 98, p. 343-365. Kirby, S.H., and Kronenberg, A.K., 1987a, Correction to "Rheology of the lithosphere: selected topics" by S. H. Kirby and A. K. Kronenberg: Reviews of Geophysics, v. 25, p. 1680-1681. 1987b, Rheology of the lithosphere: selected topics: Reviews of Geophysics, v. 25, p. 1219-1244. Lachenbruch, A., and Sass, J., 1973, Thermo-mechanical aspects of the San Andreas fault system, in Kovach, R., and Nur, A., eds., Pro- ceedings of the conference on tectonic problems of the San An- dreas fault system: Stanford, Calif., Stanford University Publica- tions in the Geological Sciences, v. 13, p. 192-205. Lehner, FK., Li, V.C., and Rice, JR., 1981, Stress diffusion along rupturing plate boundaries: Journal of Geophysical Research, v. 86, p. 6155-6169. Li, V.C., and Rice, J.R., 1987, Crustal deformation in great earthquake cycles: Journal of Geophysical Research, v. 92, p. 11,533-11,551. Linker, MF., 1993, In-situ permeability constrained by observations of postseismic deformation associated with the Loma Prieta earth- quake: Eos, Transactions of the American Geophysical Union, v. 74, p. 564-565. Lisowski, M., Prescott, W.H., Johnston, M.J., and Savage, J.C., 1990, Geodetic estimate of coseismic slip during the 1989 Loma Prieta, California, earthquake: Geophysical Research Letters, v. 17, p. 1437-1440. Lisowski, M., Savage, J.C., and Prescott, W.H., 19912, Surface defor- mation after the Loma Prieta, California, earthquake: Eos, Trans- actions of the American Geophysical Union, v. 72, p. 119. 1991b, The velocity field along the San Andreas fault in central AFTERSHOCKS AND POSTSEISMIC EFFECTS and southern California: Journal of Geophysical Research, v. 96, p. 8369-8389. Lyzenga, G.A., Raefsky, A., and Mulligan, S$.G., 1991, Models of re- current strike-slip earthquake cycles and the state of crustal stress: Journal of Geophysical Research, v. 96, p. 21,623-21,640. Malvern, LE., 1969, Introduction to the mechanics of a continuous media: Englewood Cliffs, N. J., Prentice-Hall, 713 p. Marone, C.J., Scholz, C.H., and Bilham, R., 1991, On the mechanics of earthquake afterslip: Journal of Geophysical Research, v. 96, p. 8441-8452. Marshall, G.A., Stein, R.S., and Thatcher, W., 1991, Faulting geom- etry and slip from co-seismic elevation changes: The 18 October 1989, Loma Prieta, California, earthquake: Bulletin of the Seismo- logical Society of America, v. 81, p. 1660-1693. Mavko, G.M., Schultz, S., and Brown, B.D., 1985, Effects of the 1983 Coalinga, California earthquake on creep along the San Andreas fault: Bulletin of the Seismological Society of America, v. 75, p. 475-489. Meissner, R., and Strehlau, J., 1982, Limits of stress in continental crusts and their relation to the depth-frequency distribution of shallow earthquakes: Tectonics, v. 1, p. 73-89. Michael, A.J., 1991, Spatial variations in stress within the 1987 Whittier Narrows, California, aftershock sequence: New tech- niques and results: Journal of Geophysical Research, v. 96, p. 6303-6319. Michael, A.J., Ellsworth, W.L., and Oppenheimer, D.H., 1990, Coseismic stress changes induced by the 1989 Loma Prieta, Cali- fornia earthquake: Geophysical Research Letters, v. 17, p. 1441- 1444. Mogi, K., 1968, Migration of Seismic Activity: Bulletin of the Earth- quake Research Institute, University of Tokyo, v. 46, p. 53-74. Olson, J., 1990, Seismicity in the twenty years preceding the 1989 Loma Prieta, California earthquake: Geophysical Research Letters, v. 17, p. 1429-1432. Oppenheimer, D.H., Bakun, W.H., and Lindh, A.G., 1990, Slip parti- tioning of the Calaveras fault, California, and prospects for future earthquakes: Journal of Geophysical Research, v. 95, p. 8483- 8498. Oppenheimer, D.H., and Eaton, J.P., 1984, Moho orientation beneath central California from regional earthquake traveltimes: Journal of Geophysical Research, v. 89, p. 10,267-10,282. Oppenheimer, D.H., Reasenberg, P.A., and Simpson, R.W., 1988, Fault plane solutions for the 1984 Morgan Hill, California, earthquake sequence: Evidence for the state of stress on the Calaveras fault: Journal of Geophysical Research, v. 93, p. 9007- 9026. Poley, C.M., Lindh, A.G., Bakun, W.H., and Schulz, $.S., 1987, Tem- poral changes in microseismicity and creep near Parkfield, Cali- fornia: Nature, v. 327, p. 134-137. Prescott, W.H., and Yu, S.B., 1986, Geodetic measurements of hori- zontal deformation in the northern San Francisco Bay region, California: Journal of Geophysical Research, v. 91, p. 7475-7484. Reasenberg, P.A., and Simpson, R.W., 1992, Response of regional seismicity to the static stress change produced by the Loma Prieta earthquake: Science, v. 255, p. 1687-1690. Reches, Z., Schubert, G., and Anderson, C., 1994, Modelling of peri- odic great earthquakes on the San Andreas fault: Effects of non- linear rheology: Journal of Geophysical Research, v. 99, p. 21,983-22,000. Rice, J.R., 1980, The mechanics of earthquake rupture, in Dziewonski, A.M., and Boschi, E., eds., Physics of the Earth's Interior (Pro- ceedings of the International School of Physics "Enrico Fermi", Course 78, 1979): Italian Physical Society, North-Holland Publ. Co., p. 555-649. 1992, Fault stress states, pore pressure distributions, and the weakness of the San Andreas fault, in Evans, B., and Wong, T.., MODELS OF POSTSEISMIC DEFORMATION AND STRESS TRANSFER ASSOCIATED WITH THE EARTHQUAKE eds., Fault Mechanics and Transport Properties of Rocks: Orlando, Florida, Academic Press, p. 475-503. 1993, Spatio-temporal complexity of slip on a fault: Journal of Geophysical Research, v. 98, p. 9885-9907. Rice, J.R., and Gu, J., 1983, Earthquake aftereffects and triggered seis- mic phenomena: Pure and Applied Geophysics, v. 121, p. 187-219. Richter, C.F., 1958, Elementary seismology: San Francisco, W.H. Freeman, p. 768. Ruina, AL., 1983, Slip instability and state variable friction laws: Journal of Geophysical Research, v. 88, p. 10,359-10,370. Saleeby, J.B., 1986, C-2: Central California offshore to Colorado Pla- teau: Geological Society of America Centennial Continent/Ocean Transect 10, scale 1:500,000, 2 sheets. Savage, J.C., 1983, A dislocation model of strain accumulation and release at a subduction zone: Journal of Geophysical Research, v. 88, p. 4984-4996. 1990, Equivalent strike-slip earthquake cycles in half-space and lithosphere-asthenosphere earth models: Journal of Geophysical Research, v. 95, p. 4873-4879. Savage, J.C., and Prescott, W.H., 1978, Asthenosphere readjustment and the earthquake cycle: Journal of Geophysical Research, v. 83, p. 3369-3376. Savage, J.C., Lisowski, M., and Svarc, J.L., 1994, Postseismic defor- mation following the 1989 (M=7.1) Loma Prieta, California, earth- quake: Journal of Geophysical Research, v. 99, p. 13,757-13,765. Scholz, C.H., 1977, A physical interpretation of the Haicheng earth- quake prediction: Nature, v. 267, p. 121-124. 1990, The mechanics of earthquakes and faulting: Cambridge, U.K., Cambridge University Press, 439 p. Seeber, L., and Armbruster, J.G., 1990, Fault kinematics in the 1989 Loma Prieta rupture area during the 20 years before that event: Geophysical Research Letters, v. 17, p. 1425-1428. Segall, P., 1991, Fault mechanics: Reviews of Geophysics Supplement, U.S. National Report to International Union of Geodesy and Geo- physics 1987-1990, p. 864-876. Segall, P., and Lisowski, M., 1990, Comparison of surface displace- ments in the 1906 San Francisco and 1989 Loma Prieta earth- quakes: Science, v. 250, p. 1241-1244. Sibson, R.H., 1982, Fault zone models, heat flow, and the depth distri- bution of earthquakes in the continental crust of the United States: Bulletin of the Seismological Society of America, v. 72, p. 151- 163. Steid1, J.H., Archuleta, R.J., and Hartzell, S.H., 1991, Rupture history of the 1989 Loma Prieta, California, earthquake: Bulletin of the Seismological Society of America, v. 81, p. 1573-1602. Stein, R.S., and Ekstrom, G., 1992, Seismicity and geometry of a 110- km-long blind thrust fault 2. Synthesis of the 1982-1985 Califor- D275 nia Earthquake Sequence: Journal of Geophysical Research, v. 97, p. 4865-4883. Stein, R.S., and Lisowski, M., 1983, The 1979 Homestead Valley earthquake sequence, California: Control of aftershocks and postseismic deformation: Journal of Geophysical Research, v. 88, p. 6477-6490. Stesky, R.M., 1975, The mechanical behavior of faulted rock at high temperature and pressure: Cambridge, Massachusetts Institute of Technology, Ph.D. thesis. 1978, Mechanisms of high temperature frictional sliding in Westerly granite: Canadian Journal of Earth Science, v. 15, p. 361-375. Sykes, LR., Kisslinger, J., House, L., Davies, J., and Jacob, K.H., 1981, Rupture zones and repeat times of great earthquakes along the Alaska-Aleutian arc, in Simpson, D., and Richards, P., eds., Earthquake Prediction, an International Review: Washington, D.C., American Geophysical Union, Maurice Ewing v. 4, p. 217- 274. Thatcher, W., 1983, Nonlinear strain buildup and the earthquake cycle on the San Andreas fault: Journal of Geophysical Research, v. 88, p. 5893-5902. - Thatcher, W., and Rundle, J.R., 1984, A viscoelastic coupling model for the cyclic deformation due to periodically repeated earth- quakes at subduction zones: Journal of Geophysical Research, v. 89, p. 7631-7640. Toppazada, TR., Reel, C.R., and Park, D.L., 1981, Preparation of isoseismal maps and summaries of reported effects for pre-1990 California earthquakes: California Division of Mines and Geology Open-File Report 81-11, 182 p. Tse, S.T., and Rice, J.R., 1986, Crustal earthquake instability in rela- tionship to the depth variation of frictional slip properties: Journal of Geophysical Research, v. 91, p. 9452-9472. U.S. Geological Survey Staff, 1990, The Loma Prieta, California, earthquake: An anticipated event: Science, v. 247, p. 286-293. Wald, D.J., Helmberger, D.V., and Heaton, TH., 1991, Rupture mod- els of the 1989 Loma Prieta earthquake from the inversion of strong-motion and broadband teleseismic data: Bulletin of the Seismological Society of America, v. 81, p. 1540-1572. Wesson, RL., 1988, Dynamics of fault creep: Journal of Geophysical Research, v. 93, p. 8929-8951. Working Group On California Earthquake Probabilities, 1990, Prob- abilities of large earthquakes in the San Francisco Bay region, California: U.S. Geological Survey Circular 1053, 51 p. Yonekura, N., 1975, Quaternary tectonic movements in the outer arc of southwest Japan with special reference to seismic crustal defor- mation: Bulletin of the Department of Geography, University of Tokyo, v. 7, p. 19-71. THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989; EARTHQUAKE OCCURRENCE AFTERSHOCKS AND POSTSEISMIC EFFECTS A SHEAR STRAIN ANOMALY FOLLOWING THE LOMA PRIETA EARTHQUAKE By MT. Gladwin, R.L. Gwyther, and R.H.G. Hart, University of Queensland, Australia CONTENTS Page Abstract D277 Introduction 277 Results 278 Discussion 281 Acknowledgments 283 References cited 283 ABSTRACT Borehole tensor strain instruments deployed along the San Andreas fault for the past 10 years have provided sufficient resolution and stability to sample regional tec- tonic processes. Data obtained from an instrument at San Juan Bautista in the near field region of the Loma Prieta earthquake provide the first high-resolution continuous shear strain observations associated with a large earth- quake. A change in fault-parallel shear-strain rate of ap- proximately 1 pug per year occurred about a year prior to the earthquake and persisted to the time of the event. The strain rate decreased immediately after the earthquake, but following the Chittenden sequence of earthquakes in April 1989, a new and higher rate of fault-parallel shear accumulation (0.84 ug per year relative to the 1989 rate) was established. This strain rate has continued through 1993. Associated creep-rate changes are apparent at a num- ber of sites on the surface trace of the fault within 30 km, indicating that the measured change of strain rate at the time of the earthquake has regional significance. We pro- pose that the observed strain accumulation results from increased slip around a nearby locked section of the fault arising from loading by the failed Loma Prieta source region to the north. This model is consistent with sugges- tions of an increased probability of a moderate earthquake near San Juan Bautista and with evidence that interactions between fault regions are important in earthquake pro- cesses. INTRODUCTION A Gladwin borehole tensor strainmeter (Gladwin, 1984) installed near the San Andreas fault at San Juan Bautista in late 1983 has provided continuous areal and shear strain data with sub-nanostrain resolution and long-term stabil- ity better than 100 ng per year (Gladwin and others, 1987). Raw data from the instrument consist of diameter changes in three directions at 120° to each other in the horizontal plane. These are reduced to areal strain €,, and engineer- ing shear strains y, and , (approximately parallel to and at 45° to the fault respectively), which are defined in terms of strain tensor components (€,,, Eyy > and Ey by C0 Sax eyy y2=2£xy where, as subscripts to £, the x direction is east and y is north. The strainmeter is grouted into the surrounding rock, and this instrument inclusion is softer to shear than to compression. Observed strain components are thus scaled by hole-coupling parameters (Gladwin and Hart, 1985) determined by tidal calibration. Data from borehole inclusions are initially dominated by grout compression of the instrument, by thermally controlled decay as the instrument site re-establishes equilibrium with its surroundings, and by an exponential recovery of the virgin stress field relieved at the borehole during the drilling process (Berry and Fairhurst, 1966; Berry, 1967). The exponential signals are characteristic of viscoelastic rheology as typified by Burghers solids (Jae- ger and Cook, 1976). They have no relevance to the moni- toring of regional strain changes and were removed by an exponential least-squares fitting procedure over the inter- val January 1984 to February 1988 (Gladwin and others, 1991). D277 D278 Raw data from the three instrument gauges commenc- ing in 1984 (3 months after installation) are shown in figure 1A. In determination of the exponentials to be re- moved from the raw gauge data streams, all data known to be contaminated were eliminated from the fit (Gladwin and others, 1991; Gwyther and others, 1992). The same intervals of data are used for determination of exponentials for all three gauges, and no linear trend has been removed. The resulting exponentials, determined from data during early 1984, and from May 1986 to April 1988, are shown also in figure 14 (offset for clarity). The data following the Morgan Hill earthquake in April 1984 and anomalous data associated with field experiments at the site in 1986 were excluded. The instrument was off line for 6 months during 1987. All data after April 1988 were also excluded and provide no constraint on the least-squares fitting. Fig- ure 1B gives the residuals from the determined exponentials. The residuals indicate stable gauge behav- ior from mid-1986, an emerging anomaly on all compo- nents beginning in late 1988, the strain offsets for the Loma Prieta earthquake in October 1989 and significantly differing behavior on all gauges since that time. The postseismic data indicate immediate postseismic recovery for about 3 months, and following the Chittenden earth- quakes in April 1990 the establishment of new, relatively linear trends. These residuals were then reduced to the strains €,, },, and 7, shown in figure 24. The data differ slightly from those presented in Gladwin and others (1991), owing to a refinement of the selection of data windows since that time. The dominant signals present are the coseismic strain offsets of the Loma Prieta earthquake. These have been documented elsewhere (Gladwin and others, 1991); the present discussion is confined to the observed strain-rate changes. In figure 2B the coseismic offsets have been removed from the data to make long term trends more apparent. It is important to investigate how choices made in esti- mating and removing the exponential borehole response can influence interpretation of the strain data determined from the gauge residuals. These effects are examined in figure 3. The representative strain data shown are pro- duced from residuals obtained for three different data win- dows marked a, b and c used in the fitting procedure. We are documenting an apparent change of strain rate at the time of the Loma Prieta earthquake. The critical issue is whether the choice of data interval significantly affects determination of the strain-rate change. It is evident from figure 2 that any reasonable choice of fitted window dem- onstrates that a gradient change occurred in the raw datasets. For the extremes of fitting intervals shown, the effect on the observed change of strain rate before and after the Loma Prieta earthquake are shown in table 1. AFTERSHOCKS AND POSTSEISMIC EFFECTS Interval c is inappropriate because it extends into the data which is to be used to determine the pre-earthquake gradient. In the following discussion, interval b has been used because it gave the best variance over the available 1987 dataset. RESULTS An anomalous change in 7, is apparent by late 1988, showing a remarkably linear strain accumulation of 1 ue per year relative to the pre-1988 rate (Gladwin and others, 1991). The azimuth of maximum shear for the accumulat- ing shear strain is approximately parallel to the local San Andreas fault strike. The long-term stability of the mea- surements is particularly evident in the areal strain data. Areal strain is estimated from the sum of the three com- ponents and is seen to be constant (with the exception of the coseismic offset at the Loma Prieta event) at better than 1 microstrain over the whole 10-year period. Immediately after the earthquake the fault-parallel shear- strain rate decreased for about 2 months and then gradu- ally returned to the pre-anomaly value. These data are shown in figure 4. In May 1990, following the Chittenden aftershock sequence (four magnitude 4 to 5 earthquakes on April 18, 1990, centered on an area approximately 15 km north west of SJT), the present linear y, shear accu- mulation rate of approximately 2 we per year with the original sense had been established. The absolute strain rates may of course differ from those indicated on these figures (approximately 1 u€ per year from late 1988 to the Loma Prieta event, and approxi- mately 2 ue per year relative to the pre-1988 rate from May 1990 through December 1993) because the data have been effectively detrended by the exponential removal procedure. The point at issue is that there are significant changes of strain rate documented in the data, one preced- ing the Loma Prieta earthquake and another following it. Both appear to be linear with time, and their relative mag- nitudes are not an artifact of the exponential removal pro- cedure. An alternative and useful means for determining the physical significance of these data is to plot 7; against y, (see fig. 5). The shear state at a particular time is repre- sented by a point in this shear space, its history is repre- sented by the locus of these points, and the shear required to move from one shear state to another is the vector » Figure 1.-A, Raw gauge data for the SJT site beginning 3 months after installation. Fitted exponential curves are shown offset for clarity. The recording system was nonoperational for 5 months during 1987. B, Residual gauge data produced by removal of the fitted exponentials. Units are nominal microstrain only. Microstrain Microstrain A SHEAR STRAIN ANOMALY FOLLOWING THE LOMA PRIETA EARTHQUAKE 80 0 1984 1985 _ 1986 _ 1987 1988 1989 _ 1990 _ 1991 1992 _ 1993 -6 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 D279 D280 Microstrain Nanostrain AFTERSHOCKS AND POSTSEISMIC EFFECTS -6 i i 1 1 i i u" 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 10 T T T T T T T T 1 8 |- -_ 6} _ O O: 4} ds A//\’8/Ch A : : o | A / raz iz ~. ; WA WHHMf/aflww/ 0 [ e . “HM e | _ ’\ I ‘ LP ___. Gamma 2 i : rrr rm ry "* 2 -~. B _ - __ LE2 ; _1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 A SHEAR STRAIN ANOMALY FOLLOWING THE LOMA PRIETA EARTHQUAKE connecting these points. With the local trace of the San Andreas approximately at N48°W, the y, axis also repre- sents approximately fault-parallel shear, and the 7, axis approximate extension normal to the fault. Time is marked at 3-month intervals, and direction of the coseismic offset of the Loma Prieta event is indicated in the upper and lower parts of the figure by the arrow. The Chittenden aftershock sequence is marked as CH on the lower part of the figure. The shear-strain step C-C', coseismic with the Loma Prieta event, is seen to increase both right-lateral shear across the fault (arising mainly from the y, component, +1840 ng) and normal compression (arising mainly from the 4, component, -3790 ng). This fault-normal compres- sion from shear alone must be combined with the contri- bution due to the areal strain step (+2140 ng) to find the effective change in fault normal strain in the vicinity of the instrument of approximately 1600 ng. Apart from a 5- month period immediately prior to the Loma Prieta event (see B-C in the figure) during which the maximum shear accumulation was oriented at N56°W, the predominant trend of both anomalies (see sections A-B and D-E in the figure) is a shear vector at angle E6°®S corresponding to a maximum shear accumulation at N42°W, close to the lo- cal San Andreas fault strike of N48°W. To verify instrument and coupling stability, response to earth strain tides was examined using the dominant, ther- mally uncontaminated tidal components O; and M,. Sixty- day data windows of 90-minute data were used to provide normalised tidal component amplitude every 30 days. The strain step of the Loma Prieta event and other easily iden- tifiable strain steps associated with earthquakes or creep events were removed from the record before the tidal analysis. Results are shown in figure 6 for the €, and 7; data sets. Error bars indicate the precision of determina- tion for each point, assuming equal partition of noise over all tidal components. It is clear that there have been no significant or systematic changes of tidal admittance on this instrument over the whole period under discussion. < Figure 2.-A, Areal strain and shear strains from the SJT borehole tensor strainmeter at San Juan Bautista near the San Andreas fault in northern California. Exponential trends have been removed from this data. A dominant feature is the coseismic strain step from the Loma Prieta earthquake in October 1989. B, Removal of this step reveals the details of the strain records, in particular the striking anomaly in the 7, component, the relative constancy of the other two strain components, and trend reversal on 7, for 3 months following the Loma Prieta event. All steps in the data can be associated with seismic events or nearby creep events. For example, the times of the two Lake Elsman earth- quakes are indicated (LE1, LE2), as are the times of the Loma Prieta earthquake (LP), the Chittenden earthquake sequence (Ch), and creep events also monitored on a nearby creepmeter (0) and documented in table 2. D281 This result has been confirmed elsewhere (Linde and oth- ers, 1991). DISCUSSION Apart from coseismic steps, the strain data show sev- eral readily identifiable steps which correlate closely with the main creep events on the closest creepmeter XSJ, 2 km to the east of our instrument. Events from late 1988 to 1992 are tabulated in table 2, and the detailed correspon- dence of the events with the main creep events gives fur- ther confidence that the data represent regional tectonic activity. Further, the events show (Gladwin and others, 1993) remarkable similarity suggestive of a small source (less than 0.5 kilometers in depth and at most a few kilo- meters in extent) directly under San Juan Bautista. The shear strain resulting from such a source cannot account for the size of the post Loma Prieta strain anomaly. Other explanations for this anomaly need to be examined. A non-tectonic source from the instrument or its immediate vicinity is unlikely because of the stability of the tidal response, consistency of our internal instrument checks, the long-term stability of the areal strain record (better than 1 ug over 10 years), and the detailed corre- spondence in time of all observed strain steps with either earthquakes or creep events on nearby creep meters. Though the strain signals could arise from small-scale processes in a nearby section of the fault, observations of anomalous creep events at three sites up to 30 km away indicate a more extended source. Figure 7 shows long- term creepmeter data for 5 sites covering 40 km of the fault south of San Juan Bautista with long-term trends (Burford, 1988). A distinct increase in creep rate follow- ing the Loma Prieta earthquake is evident on sites XSJ, XHR and CWC, spanning 16 km of the fault. The XFL site (29 km from XSJ) shows only a marginal in- crease, and the more remote site XMR (40 km) shows no effect. The creep anomalies in figure 6 are unusual, especially for CWC and XHR, and begin at the time of the Loma Prieta earthquake. The creep anomaly at the nearby XSJ begins about the time of the establishment of the new shear strain anomaly at SJT, which, given the causal time correspondence, suggests that these signals are not just the consequence of normal interactions between fault sec- tions in this creeping section, but are linked to the earth- quake. We conclude that the failure of the Loma Prieta source region transferred load to the San Juan Bautista region just to its south, resulting in increased creep rate. The simplest explanation of an increased creep rate is frictional response to the increased fault-parallel shear loading indicated by the coseismic y, step at San Juan Bautista. D282 AFTERSHOCKS AND POSTSEISMIC EFFECTS However, slip via creep does not itself result in linearly _ Prieta source region particularly after the Chittenden af- increasing elastic strain. We suggest that our linear shear- _ tershock sequence (April 1990). strain anomaly is best explained by continued aseismic The response at the site immediately after the Loma slip around a nearby locked section of the fault, the slip _ Prieta earthquake and prior to the Chittenden events seems being associated with loading transferred from the Loma __ to indicate that this load transfer towards the SJT site was Microstrain 10 i ; i i i i ; ; 1984 1985 1986 1987 1988 1989 1990 191 1992 1993 Figure 3.-Representative y, shear-strain data illustrating the effect of the choice of data window used for the exponential fitting. The intervals used in three separate fit sets are marked as a, b, and c. Estimates of the change of strain rate from before to after the Loma Prieta event are only marginally affected for the three extreme choices shown. The choices have some effect on the timing of the pre-event anomaly only. Note that for choice c the change of gradient is well defined before the end of the fitting interval, indicating that the end point of the fit interval is not producing the effects discussed. The changes of gradient that determine the anomalies under discus- sion are clear on all residual sets. A SHEAR STRAIN ANOMALY FOLLOWING THE LOMA PRIETA EARTHQUAKE D283 Table 1.-Average gradient of shear strain y, determined over three periods [Values are residual gauge data after removal of exponential functions fitted over the three selection intervals 1985/02 to 1988/04 (a), 1986/06 to 1988/04 (b), and 1985/05 to 1989/07 (c). The lowest row of data indicates that the change in strain rate after the Loma Prieta earthquake is very similar for the three fitting intervals shown.] Selection Interval 1985/02 1986/06 1985/05 to to to 1988/04 1988/04 1989/07 (a) (b) (c) 1986-1988 0.11 £.11 -0.03 £ .08 0.01 £.14 1989 to L.P. 1.56 £.08 1.16 £.09 0.75 £ .09 Post-L.P. 24 £.10 2.01 £.11 1.54 £ .12 Difference 0.84 £ .13 0.85 £.14 0.79 £ .15 initially prevented by the strength of an asperity region associated with the Chittenden aftershock sequence. When this region ultimately failed in 1990, the SJT site immedi- ately began to respond to the fault parallel shear expected from the Loma Prieta event in the sense of a continued afterslip at the Loma Prieta region. A range of simple dislocation models of slip around a locked region were considered. The model which resulted in a fault-parallel shear-strain rate most comparable to that observed was a locked region extending along the fault strike at depth of 1.5 to 5 km, with the surrounding fault surface slipping at the regional average of 14 mm per year. While this model indicated a locked region shal- lower than the study mentioned above, uncertainty in the absolute value of the observed 7, strain rate indicates that our data are broadly consistent with the presence of such a locked region at moderate depth on the fault. The data thus suggest an increased probability of a moderate earth- quake in the San Juan Bautista region. We have previously suggested (Gladwin and others, 1991) that the pre-Loma Prieta strain changes were re- lated to a broad regional effect, based on the timing of the Lake Elsman foreshock and a marginal geodetic anomaly (Lisowski and others, 1990) near the Loma Prieta source. The effects we are reporting also appear to have some regional expression in increased creep activity since the Loma Prieta earthquake. The anomaly in y, is now so large that it would be expected to be detectable by geo- detic observations in the area. Unfortunately, the major geodetic network is centered well to the north of San Juan Bautista, making even inversion for the determination of the southern extent of the Loma Prieta rupture zone itself difficult (Johnston and others, 1990). Recent data from the southern end of this network is not yet available, but a change of fault-parallel shear-strain rate of lue per year coincident with the Loma Prieta event should easily be identified in the observational period to date if it extended over the total geodetic network involved. These data are critical in providing constraints on the scaling of the anomaly in the northerly direction. Some data has been taken in a small HP3808 network extending from San Juan Bautista north to Pajaro Gap, particularly following the earthquake (Johnston, oral comm., 1993), but the re- sults from this network are not yet published. Data from this network was not taken for many years prior to the Loma Prieta earthquake, so there is no likely contribution on the issue of whether or not a change of gradient has occurred following the event. There is no associated areal strain anomaly indicated in the data, and so no corrobora- tion of the anomaly by reference to the dilatometer at Searle Road, approximately 5 km away, is available. The dilatometer shows no significant areal strain change. Scaling of the anomaly cannot be determined from a single site, and the significance of these data will remain unknown until it is confirmed or denied from measure- ments at other nearby instrument sites. There is no evi- dence that the observed anomaly extends sufficiently far north to be measured in the Loma Prieta geodetic net- work. Our modelling suggests the presence of a source 5 km long. This source would produce minimal deforma- tion in the geodetic network. Hence, although we are con- fident that the observed anomaly is not an instrumental artifact or of very localized origin, the only currently avail- able supportive evidence that this anomaly is of regional significance is the increased creep activity in stations south of SJT. ACKNOWLEDGMENTS This work was performed under a grant from the U. S. Geological Survey, Department of the Interior. Contents do not represent policy of that agency, and no endorse- ment of the agency is to be assumed. The instruments were previously developed under awards of the Austra- lian Research Grants Scheme and fabricated by R. Willoby and staff in-house. We thank R. Liechti for maintenance support, K. Breckenridge for data retrieval and assistance with creepmeter data, and Drs. A. Linde and M. Johnston for general support in the program. REFERENCES CITED Berry, D.S., 1967, Deformation of a circular hole driven through a stressed viscoelastic material: International Journal of Rock Me- chanics and Mining Science, v. 4, p.181-187. Berry, D.S., and Fairhurst, C., 1966, Influence of rock anisotropy and time-dependent deformation on the stress-relief and high-modulus inclusion techniques of in-situ stress determination: American So- ciety for Testing Materials, STP 402, p.190-206. Burford, R.O., 1988, Retardations in fault creep rates before local moderate earthquakes along the San Andreas fault system, central D284 California: Pure and Applied Geophysics, v. 126, p. 499-529. Gladwin, M.T., 1984, High precision multi component borehole defor- mation monitoring: Reviews of Scientific Instruments, v. 55, p. 2011-2016. Gladwin MT., Breckenridge, K.S., Hart, R. and Gwyther, RL., 1994, Measurements of the strain field associated with episodic creep events on the San Andreas Fault at San Juan Bautista, California: Journal of Geophysical Research, v. 99, no. B3, p. 4559-4565. Gladwin, MT., Gwyther, RL., Hart, R. and Francis, M., 1987, Bore- hole tensor strain measurements in California: Journal of Geo- physical Research, v. 92 , no. B8, p. 7981-7988. AFTERSHOCKS AND POSTSEISMIC EFFECTS Gladwin, M.T., Gwyther, RL., Higbie, J.W. and Hart, R., 1991, A medium term precursor to the Loma Prieta earthquake?: Geophysi- cal Research Letters, v. 18, no. 8, p. 1377-1380. Gladwin, M.T., and Hart, R., 1985, Design parameters for borehole strain instrumentation. Pure and Applied Geophysics, v. 123, p. 59- 88. Gwyther, RL., Gladwin, M.T., and Hart, R.H.G., 1992, A shear strain anomaly following the Loma Prieta earthquake: Nature, v. 356, p, 142-144. Jaeger, J.C. and Cook, N.G.W., 1967, Fundamentals of rock mechan- ics: London, Chapman and Hall. Microstrain 10 ; a ; i i i a 1989 1989.2 1989.4 1989.6 1989.4 1990 1990.2 1990.4 1990.6 1990.4 1991 Figure 4.-Detail of the strains immediately following the Loma Prieta earthquake. The 7, record shows that the post-earthquake strain rate shown in figure 2 did not begin until after the Chittenden sequence in April 1990. A SHEAR STRAIN ANOMALY FOLLOWING THE LOMA PRIETA EARTHQUAKE Johnston, M.J.S., Linde, A.T., and Gladwin, M.T., 1990, Near-field high resolution strain measurements prior to the October 18, 1989, Loma Prieta Ms 7.1 earthquake: Geophysical Research Letters, v. 17, no. 10, p. 1777 - 1780. Linde, A.T., Gladwin, M.T., and Johnston, M.J.S., 1991, The Loma Prieta earthquake, 1989 and earth strain tidal amplitudes: an un- 0.5 D285 successful search for associated changes: Geophysical Research Letters , v. 19 no. 3, p 317-320. Lisowski, M., Prescott, W.H., Savage, J.C., and Svarc, J.L., 1990, A possible geodetic anomaly observed prior to the Loma Prieta, California, earthquake: Geophysical Research Letters, v. 17, no. 8, p. 1211-1214. Oct 88 1 i O tin 1 1.5 2. 2.3 y, Microstrain y;, Microstrain ...... g na e een kee e eae n nea ee post Loma prieta 1 1 1 41 i -0.5 0 0.5 1 1.5 2 2.5 y, Microstrain Figure 5.-Pre- and post-Loma Prieta plots of tensor strain components 7, against y, in ug for the period July 1986 to April 1991. The shear strain 7, is a measure of shear strain approximately parallel to the San Andreas fault at this locality, while 7, is a measure of shear strain in approximately a north-south or east-west direction, and thus extension normal to the fault. Three monthly intervals are indicated on the figure. The initiation and completion of the step associated with the Loma Prieta earthquake is shown by L.P. The size of this step was 7,=+1840 ne and 7,=-3790 ng. The Chittenden aftershock sequence strain step is indicated by Ch. For any shear-strain change (Ay,, Ay,), the maximum shear strain change is Ay =, Aylz +Ayz2 and 9:1/2tan‘1[A72/Ayl] is the angle N of E for the axis of maximum extension. AFTERSHOCKS AND POSTSEISMIC EFFECTS D286 qyusuoduroo _ 4fIsea pug juaAa ewo7 ay} jo dors urens oy}, 'sXep og L1aaa apmitfdure juouoduoa fepy Aue 10 soweunoprdd juawnnsut ut adueyo ou st 'astou ueissnen Sununsse 'uon _ pastfeunou apiao1d 01 pasn alom wep onurui-9g JO smoputm Aep-Ajxig 'T6e [-pi 01 gg61 pousad ay) -eurundjap Jo uotsroaid ay} ajeotput sreq Jourg ay} wou; paaowat arom sdays urens argeynuapt _ 10} 4 urens seays pue "3 urens feare Jo sjuauoduwos fepyu 'o pue {pq ay) yo S 1661 1661 $0661 0661 S 6861 6861 S 8861 wwoofi N B UIBIJSOUP A SHEAR STRAIN ANOMALY FOLLOWING THE LOMA PRIETA EARTHQUAKE Table 2.-Surface creep events recorded at XSJ creepmeter at San Juan Bautista in 1988- 1992 period, and corresponding strain offsets observed at SJT [Each strain event preceded the surface creep event by less than 1 hour] No. Date Strain event Creep event Y p £4 Size ne ne ne mm 1 21.11.88 -100 ~40 -80 1.7 2 22.12.89 -77 -38 -60 0.5 3 27.01.90 -69 -41 -53 0.6 4 07.04.90 -90 -55 -75 1.9 5 09.06.90 -100 -50 -90 3.9 6 04.07.90 None 0.3 7 30.08.90 -120 -52 -91 3.2 8 20.11.90 -50 45 -70 2.1 9 08.03.91 -95 -50 -80 3.6 10 06.08.91 -120 -50 -90 5.0 11 27.12.91 -90 -50 -80 0.3 12 07.06.92 -110 45 -90 5.3 D287 D288 AFTERSHOCKS AND POSTSEISMIC EFFECTS \ Monterey # Bay L_ j 0 10 20 km XMR 250 200 400 E E 150 300 100 200 50 100 Chs O i R j R R 1970 1975 1980 1985 1990 Figure 7.-Surface-creep data (provided by K. Breckenridge, USGS) from creepmeters along the fault south of San Juan Bautista, with positions of creepmeters XSJ, XHR, CWC, XFL and XMR, tensor strainmeter SJT, and Chittenden aftershock Ch indicated in the inset map. The trend lines indicated are from Burford (1988). Note that for the XMR data, the scale at the right axis should be used. THE LOMA PRIETA, CALIFORNIA, EARTHQUAKE OF OCTOBER 17, 1989: EARTHQUAKE OCCURRENCE AFTERSHOCKS AND POSTSEISMIC EFFECTS A MAGNETOTELLURIC SURVEY OF THE LOMA PRIETA EARTHQUAKE AREA: IMPLICATIONS FOR EARTHQUAKE PROCESSES AND LOWER CRUSTAL CONDUCTIVITY By Randall L. Mackie and Theodore R. Madden, Massachusetts Institute of Technology; and Edward A. Nichols, University of California, Berkeley CONTENTS Page Abstract D289 Introduction 289 The magnetotelluric TM-mode coast effect ------------------------ 290 Geologic framework 291 The magnetotelluric data 292 Two-dimensional inversion of the data ----------------------------- 293 Summary Of MOGEIiNG @SUItS ono 303 Geophysical implications for the San Andreas fault system ---- 305 Conclusions 307 Acknowledgments 309 References cited 309 ABSTRACT We conducted a magnetotelluric survey around the Loma Prieta earthquake zone one year after the earthquake to look for possible electrical anomalies in the lower crustal part of the earthquake zone. We collected data at stations on each side of the San Andreas fault from Hollister to near Palo Alto. We used a two-dimensional transverse magnetic (TM) mode inversion algorithm to interpret the observed TM mode data since it is the TM mode that is sensitive to lower crustal resistivity variations near ocean- continent boundaries, whereas the transverse electric (TE) mode is not. Inversions of the data suggested a vertical zone of increased lower crustal conductance in the vicin- ity of the San Andreas fault near the region that experi- enced the main shock and aftershocks. The inferred zone of increased lower crustal conductance may be a perma- nent feature of this part of the San Andreas fault zone, or it may be a temporal feature related to the Loma Prieta earthquake. Because healing and ductile deformation pro- cesses in the lower crust should decrease its conductance with time, we speculate that this anomalous zone may be related to the Loma Prieta earthquake. With only data collected after the earthquake, we cannot confirm this speculation. However, because of the short time period between the Loma Prieta earthquake and our measure- ments, we think that fluids were already present there before the earthquake, but that the rupturing process of the earthquake increased the lower-crustal fluid connec- tivity. Regardless, the conductance values determined by our inversions correspond to lower-crustal connected fluid porosities of about 0.06 percent for a 5-km-wide leakage zone. Higher fluid porosities would be needed if the fluids were concentrated in narrower zones, but we can- not determine the exact geometry of these zones. INTRODUCTION More than 5,000 aftershocks occurred over a distance of 60 km along the San Andreas fault zone and up to 20 km in depth following the Loma Prieta earthquake (Oppenheimer, 1990). The main shock and aftershocks generally fell along a fault plane that dips 70° to the west and strikes approximately 130° from north. The earth- quake was unusual because the epicenter was at a depth of almost 18 km (Dietz and Ellsworth, 1990), which is deeper than expected for this region of the San Andreas fault (Furlong and Langston, 1990), and because it is de- batable whether or not the earthquake actually occurred on the San Andreas fault (Segall and Lisowski, 1990; Dietz and Ellsworth, 1990). Large-scale tectonic processes, such as earthquakes, may have a dramatic influence on the electrical conductivity of the crust. This is because these processes can affect the fluid regime in the crust (Torgensen, 1990), where the primary control of electrical conductivity is connected sa- line fluids. The crust, in general, becomes more resistive with depth, although some regions of the world appear to have conductive zones in the middle or lower crust (for example, Jones, 1987). At shallow depths, the increased resistivity is due to the narrowing of cracks (Brace and others, 1965). Deeper in the crust, temperatures are high D289 D290 enough so that rocks deform ductilely rather than brittlely (Heard, 1976), and this should eliminate much of the con- nected porosity (Brace and Kohlstedt, 1980). It is more likely, however, that the middle-to-lower crust actually deforms semi-brittlely (Carter and Tsenn, 1987; Evans and others, 1990), which may be partly responsible for the fact that finite resistivities are measured for the lower crust. Because the main shock of the Loma Prieta earthquake was at a depth of 18 km, brittle deformation occurred down to at least that depth in the Loma Prieta region, and possibly deeper. Furthermore, it is possible that semi- brittle deformation, where the main deformation mecha- nisms are crystal plasticity and microcracking (Evans and others, 1990), would exist to even greater depths. Such deformation might have caused an increase in the fracture and microcrack connected porosity in the lower crustal part of the fault zone and hence an increase in conductivity in that zone. (An alternate scenario is that fluids were brought into this zone after the earthquake, but we believe this is less likely because of the time scales required for fluids to percolate through a fairly imperme- able crust.) Fluids at high pore pressures in the lower crustal part of a fault zone have been proposed to play an important role in the earthquake cycle (Rice, 1992). The existence of elevated pore pressures in a very narrow permeable slip zone having the usual laboratory frictional values may explain why the San Andreas fault does not show a pronounced heat flow peak and why the principal stress directions are at large angles relative to the trace of the San Andreas (Rice, 1992). Fluids at high pore pressures in the fault slip zone would give the lower crust a special role in the earthquake cycle as the pore pressure there is probably already close to the overburden pressure. Magnetotelluric (MT) measurements, under certain geo- graphic circumstances, can provide information about ver- tical zones in the lower crust that have anomalous electrical properties. This occurs when currents are forced through these zones, which is the case near ocean-continental boundaries because of the mismatch of the current sys- tems as a function of depth. Because the San Andreas fault system is parallel and close to the coast, we can take advantage of the currrent system set up by the ocean- continent adjustments to study the electrical properties of the lower crustal part of the fault zone. Therefore, we collected MT data in the Loma Prieta earthquake region to investigate the lower crustal properties of that area. Unfortunately, since all of our MT data were collected after the earthquake, we will not be able to make any statements concerning time changes in conductivity across the fault zone that might have been due to the earthquake itself. AFTERSHOCKS AND POSTSEISMIC EFFECTS THE MAGNETOTELLURIC TM-MODE COAST EFFECT Electromagnetic fields are affected by an ocean-conti- nent boundary in two very distinct ways. One type of coast effect is due to the concentration of near-surface electrical currents running parallel to the coast in both the ocean and the oceanic upper mantle, which causes anoma- lies in the vertical magnetic field recorded near an ocean coast. In the work herein, we are concerned with another coast effect-one that perturbs the electric fields running perpendicular to the coast. Because the ocean has such a high conductance, most of the induced electrical current in oceanic regimes is concentrated in the ocean itself up to periods of 500 s, whereas in a continental regime the current is mostly in the mantle. Because electrical currents are divergence- free, there is continuity in the normal component of cur- rent at the ocean-continent boundary. This results in extra current perpendicular to the coast (the transverse mag- netic or TM mode) in the continental upper crust com- pared to what would be induced were the ocean absent. Consequently, the currents must readjust in some broad zone around the ocean-continent boundary. This readjust- ment takes place by leaking currents out of the upper crust, across the resistive lower crust, and into the mantle (fig. 1). In this readjustment zone, the current system is still dominated by the oceanic current system. The width of this readjustment zone depends on the conductance of the upper crust and the resistance of the lower crust (Ranganayaki and Madden, 1980). The dis- tance at which the current has been reduced by a factor of e by leaking into the mantle is termed the adjustment distance, and is equal to the square root of the product of the integrated upper crustal conductance and the integrated lower crustal resistance (Ranganayaki, 1978). The adjust- ment distance can be 100 km or more for continental regimes, and even at these distances the upper crustal current system is still contaminated by excess ocean currents. It actually takes several adjustment distances for the current system to return to normal continental values. Vertical zones in the lower crust that have an anoma- lously high conductance can dramatically alter the TM response across that zone (Park and others, 1991; Mackie, 1991). This is because more current is attracted out of the ocean by this leakage zone, but less current remains in the upper crust inland of this zone. This results in a dramatic shift in the TM-mode response across the zone. The trans- verse electric (TE) mode response would not be sensitive to this zone. This is because the component of the electric field parallel to conductivity boundaries is continuous across those boundaries, and therefore, the TE-mode re- sponses are not anomalously perturbed from their normal A MAGNETOTELLURIC SURVEY: IMPLICATIONS FOR EARTHQUAKE PROCESSES AND CONDUCTIVITY values. By carefully following the readjustment of the TM current system perpendicular to the ocean-continent boundary, one can map out anomalous vertical features in the lower crustal resistivity structure. (However, we would not be sensitive to anomalously conductive horizontal lay- ers in the lower crust since we can only determine the integrated vertical resistivity properties.) Because most of the geologic features in California are oriented parallel to the ocean-continent boundary, we are in a good position to look for anomalous zones in the lower crust in the region associated with the Loma Prieta earthquake. By mapping the TM-mode response across the fault at several different locations, we can look for anomalous vertical conductivity channels in the lower crustal part of the fault zone. GEOLOGIC FRAMEWORK California has been subjected to many tectonic events throughout its history, including arc-related compressional orogenies, arc volcanism and plutonism, and Cenozoic transform motion (Burchfiel and Davis, 1972, 1975; Dickinson, 1981). The geology in the area of the Coast Ranges of central California, as shown in figure 2, is char- acterized by three main lithologic sequences (Bailey and others, 1970; Page, 1981): (1) the Salinian formation (plu- tonic and metamorphic rocks from a magmatic arc envi- D291 ronment), (2) the Franciscan formation (a subduction zone complex), and (3) the Great Valley sequence (forearc ba- sin sediments). Interspersed among the outcrops of these formations are surface deposits that are predominately Cenozoic and Mesozoic marine and non-marine sediments and alluvium. The Salinian block, bounded on the northeast by the San Andreas fault, is composed of metasedimentary rocks that have been invaded by granitic plutons. It is believed to have been translated some 600 km northward from its original position as part of the Klamath-Sierra Nevada terrane. The resistivities of the granites of the Salinian formation are much more resistive than the sedimentary and metamorphic rocks to the east, and are typically around 500 ©2-m (Mazzella and Morrison, 1974). The Franciscan formation is an assemblage of both oce- anic and terriginous materials. It consists of graywacke, shale, mafic volcanic rock, chert, limestone, and. meta- morphic rocks of zeolite and blueschist facies. These com- ponents commonly occur in a pervasively sheared argillaceous matrix. Much of the Franciscan formation has undergone blueschist metamorphism at temperatures of around 200°C and pressures of around 6-8 kbar (Page, 1981). It is believed that the Franciscan formation origi- nated in a subduction zone along the western margin of the North American plate in the late Mesozoic era and that these rocks underwent metamorphism as the oceanic plate was subducted under the continental plate. Surface Electric current upper crust --» \\\ \\\ sais \\\\\ resistive lower crust \\ summ conductive mantle Figure 1.-Simple sketch showing the magnetotelluric TM-mode coast effect. The ocean induces extra current that gets trapped in the continen- tal upper crust. This extra current in the continental upper crust gradu- ally leaks off into the mantle as one moves away from the coast. This leakage effect is what makes one sensitive to the electrical properties of the lower crust. D292 electrical measurements show typical resistivity values of 5-50 ©2-m for the sedimentary and metamorphic rocks of the Franciscan formation (Mazzella and Morrison, 1974). Finally, the Great Valley sequence is composed of strati- fied terriginous clastic sediments, primarily gray-wackes and shales, derived from the Klamath-Sierra Nevada ter- rane. These sediments are believed to rest on an ophiolite basement, beneath which the Franciscan formation has been thrust, and are even more conductive than the Fran- ciscan rocks, typically being about 1-10 -m. THE MAGNETOTELLURIC DATA A magnetotelluric survey was conducted in conjunction with Electromagnetic Instruments, Inc. (EMI) from Octo- AFTERSHOCKS AND POSTSEISMIC EFFECTS ber 20-30, 1990, one year after the Loma Prieta earth- quake. EMI collected a total of ten sites in pairs of two simultaneous recordings, one on each side of the fault (except for site 10, which was on the fault because of logistical problems). Fig. 2 shows the locations of these stations. EMI collected E,, Ey, Hy, Hy, and H, data at each pair of stations using synchronized clock control. They recorded data in three bands, with sample frequen- cies of 500 Hz, 20 Hz, and 2 Hz, thus yielding coverage over a period range from approximately 0.005-1000 s. The data were processed to yield a 10 by 10 cross-power matrix per frequency (approximately 6 frequencies per decade) for each pair of stations. From this cross-power matrix, MT parameters could be computed using the horizontal magnetic fields at the second site as a remote reference. FRANCISCAN SALINIA N © ut station mem - mee os PROFILE o) 96 7a Figure 2.-Simplified geologic map (after Bailey, 1966) showing the locations of the Franciscan Formation, the Salinian Block, and major faults. In between the Franciscan and Salinian outcrops are surface deposits of marine and non-marine sedimentary rocks and alluvial deposits. Also shown are the location of the MT stations and the profiles used for the inversion analysis. A MAGNETOTELLURIC SURVEY: IMPLICATIONS FOR EARTHQUAKE PROCESSES AND CONDUCTIVITY Remote referencing was fairly successful for canceling out the effects of random noise disturbances in the data at most sites, but there were additional noise problems that the remote referencing was not able to deal with. Robust processing methods also were not able to deal with those noise problems (Egbert, oral commun., 1992). In general, the data were noisier the closer they were to the industri- alized San Francisco Bay area to the north. Sites 5 and 6, the furthest north, were the noisiest data of all and were basically unusable. Site 8 also was noisy, which we later discovered was because that site was located close to an electrical power sub-station. At many of the sites (1,2,3,4,5,6) the data in the band from 1 to 100 s seemed to be contaminated by some man- made noise source. There are two lines of evidence for this. First, the magnetotelluric responses at these sites in this frequency band are characterized by slopes close to 1 in log of apparent resistivity versus log of period, and a sudden decrease of the phase to near 0°, which is charac- teristic of local noise sources (Qian and Pedersen, 1991). This is clearly evident in the MT data from site 1, Calero, shown in figure 3. Second, and probably more important, there is a large anomaly in the tipper magnitude from 1 to 100 s at these same sites. (Figure 4 shows an example of the anomaly from site 1, Calero.) The tipper is a measure of the ratio of the vertical magnetic field to the horizontal magnetic field. Local industrial noise sources, which act like vertical magnetic dipoles and horizontal electric di- poles, can create large tipper magnitudes which are clearly non-MT. These sources generate coherent electromagnetic energy, but their impedances are different than those of natural MT signals at periods longer than 1 s. At periods greater than 100 s, however, the natural MT signals were stonger than the man-made noise. It is possible that this noise problem is due to the BART transportation system in the San Francisco Bay area, which acts like a large vertical dipole (Fraser-Smith and Coates, 1978; Ho and others, 1979), although any local industrial source could cause similar features in the data. At the same time that EMI was collecting data, we (Madden and Mackie) collected single-site, longer-period (10-1000 s) MT data at several additional stations, mostly along the San Andreas fault zone. We attempted to gather data at the same time EMI was collecting long-period data so we could use the EMI data for remote referencing (we synchronized our clocks with EMI's clocks every couple of days). However, because of logistical problems in setting up the stations, we were only successful in co- ordinating this at one site (site 11, shown on the map in fig. 2). At the other stations, the daytime man-made noise problems prevented us from obtaining valid impedance estimates. The data from the EMI group were generally less contaminated with the man-made noise because they typically collected their long-period data (greater than 1 s) during the night, when such noise is at a minimum, D293 whereas we collected all our data during the day when these noise problems were highest. Unfortunately, the natu- ral MT signals are also usually stronger during the day than at night, which reduces the advantage of night-time recording. TWO-DIMENSIONAL INVERSION OF THE DATA An eigen-state analysis (LaTorraca and others, 1986) was used to decompose the impedance tensors for each site. At all stations except site 3, the major electric eigen- vector direction, which corresponds to the maximum am- plitude eigen value of the impedance tensor, was approximately perpendicular to the coastline, and this is the mode we identify as the transverse magnetic (TM) mode. The TM mode near ocean-continent boundaries has lower phases at longer periods as compared with the TE mode. This is due to ocean electrical current perpen- dicular to the coastline being trapped in the continental upper crust (Mackie, 1991). At site 3, the maximum am- plitude eigen value corresponded to the TE mode, and the TM mode corresponded to the minimum eigen value. We suspect this is due to local inhomogeneities that caused a static shift in the electric field perpendicular to the ocean and probably gave us a mixed MT impedance. Data from sites 5, 6, and 8 were unusable because they were too contaminated by man-made noise. Data from the remaining sites were mostly usable, although we edited out the data at some of the sites (1,2,3,4) from 1 to 100 s that appeared contaminated by man-made noise sources. Finally, selective editing of outliers and obviously incon- sistent data was done for all stations prior to running two- dimensional inversions of the data. The remaining data were input to the two-dimensional TM-mode inversion algorithm described in Mackie and others (1988). We used only TM-mode inversions of the data for two reasons: (1) the lower crustal resistivity anomalies discussed earlier affect only the TM mode and (2) two-dimensional TM- mode interpretation is more accurate for profiles centrally located across elongate three-dimensional bodies (Wannamaker and others, 1984). We do, however, show the TE-mode data and model predictions even though the TE data were not used in the inversions. The inversion procedure we used to interpret the data is the one we term the maximum likelihood inverse (Mackie and others, 1988; Madden, 1990). The maximum likeli- hood inverse gives the solution that maximizes the joint probability of fitting the observed data, subject to data covariance constraints, and adhering to an a priori model, subject to model covariance constraints (Tarantola, 1987). Our formulation of the inverse problem holds the block geometry of the system fixed-only resistivities are al- D294 AFTERSHOCKS AND POSTSEISMIC EFFECTS EMI site 1, Calero 108 g 105 E 104 E ~ - ']E F (s) - i é“ 103 E > - f F op® o © |- ® 98 al: 5p I C F 000 0 ® I II I o H 0 00° I I & _| 800000 L I 424" mas 1 101; A A AAAAZ‘AAAAAQA 2 a ps ) 10° E 10—1 1 | L L OLL L 1 LLL L 1 DULL 1 | J LLLLH 1 1 1 CLLEL L 1 (OLL 0 -10} £18? G -20} 6 0 © g 91 I¥8%§8 %0°°@°8 $ _ -30f ‘I af1 th. © - -40} 3 ® A * ao ~ 0 A 0 4 z» 4A e 4 a , aAd® & -50 |- a £*° % § e00 # am 3 { k $ -60 } a - -70f 8 I _.80»- _90 1 OU L L JUL 1 1 1 OLL 1 1 0. 1_ J J LLL 1 1 1 LUuL 1 1 LLL f o - 150} 0 o & 100k L LL LLL J a 50 |- C 6 6 2222222222 2k R2 RER _E c_ ~ 6 | A H Homo HoH mH HoH HoH my Rgn HHnm m wy Hap babe 8 8 -50}- H thtlH H HHNH a? 100 TTTTTTITITITITITITITITITITITITITTCTTT I; """""""""" WH & E & -1s0 & EEEEEEEEEEEEEEEEEEEEFEEEEEEEEEEEE ‘a 1 1 Llfi El | L L 1 D L 1 DULL 1 1 DOLE Eilllllll | 1 10-3 10-2 10-1 10% 101 102 103 104 period (seconds) A MAGNETOTELLURIC SURVEY: IMPLICATIONS FOR EARTHQUAKE PROCESSES AND CONDUCTIVITY < Figure 3.-Eigen-analysis results of the MT data at site 1, Calero. Shown are the maximum and minimum apparent resistivities, associated E/H phases, and the electric field and magnetic field eigen-directions. Note the steep slope in the apparent resistivities and phases near 0° in the period range from 1 to 100 s. We believe these are not natural MT responses but rather are due to Bay Area Rapid Transit noise contami- nating the signals. This behavior was at most sites in our survey area. EMI site 1, Calero, Tipper D295 lowed to vary during the inversion. Additionally, in all our inversions, we gave equal weight to fitting the phases and In (amplitudes). The profiles along which we did inversions are shown in figure 2. Each profile used data from only two or three stations. For each profile, we ran many inversions of the data using different model constraints and a priori mod- els. The a priori models were all generally constrained using available geologic and geophysical data. First, we results 1.2f 1.0} I 0.8 |- tipper magnitude 8 8 Al 8 8 0.4 o 00°03 808881 0.2} 1 } (O I_} 1 LLU I_} L |_ ( LLU 14 3 1 L LU 0.0 L_ ul 180 160 |- o 140 - o 120} o 100 |- 00 80 |- o o 60 |- o 20 |- o o 1 0 1G uuu 104 G 1 Lllllg L G LLU 10-3 10-2 10-! 10° 10" period (seconds) rotation angles (degrees) I I I 1 I 1 I I I I I I I I I U U 1 U 1 I 1 I I I I I I I b 1 I I I 1 } Figure 4.-Plot of the tipper amplitude at site 1, Calero. Shown are the amplitudes in the direction where H, is most corre- lated with Hy. Note the large anomaly in the period range from 1 to 100 s. We be- lieve this is evidence of Bay Area Rapid Transit interference in the natural signals. Similar anomalies were observed at most sites in our survey area. es D296 used long-period (greater than 1 hour) MT data from the Hollister area (Bennett, 1985; Mackie and others, 1988), shorter-period (10-10,000 s) MT data from central Cali- fornia (Park and others, 1991), and the results of a con- trolled-source EM measurement on the Pacific Ocean floor (Cox and others, 1986) to constrain the lower crust and mantle a priori resistivity values. Second, we used the seismic tomographic inversions and MT analysis of Eberhart-Phillips and others (1990) to help constrain the a priori upper crustal resistivity variations. Third, we used bathymetric contour maps for approximating the ocean floor topography. Finally, we used surface geologic infor- mation (Clark and Rietman, 1973) for setting the very near surface resistivity information. We show one a priori model for profile A in figure 5. Other a priori models were tested that included putting vertical conductive zones in the lower crust and making the entire lower crust much more conductive. The a priori models for the other pro- files are essentially the same except for minor upper crustal differences. The inversion procedure weights fitting the data much more strongly than adhering to the a priori AFTERSHOCKS AND POSTSEISMIC EFFECTS model (by a factor of 10° in the ratio of the data to model covariances), although it never completely looses sight of the a priori model. The inversion procedure itself is not biased, although there certainly are biases introduced by our choice of an a priori model and by the a priori model constraints. Profile A uses data from sites 1 and 2 and goes over the epicenter of the main earthquake shock. We feel that these data sites represent our best data, so we will focus on the results for this profile. In the following discussion, it should be understood that a lightly weighted smoothing operator and a small diagonal term were always included in the model covariance operator used in the inversion, and that an "unconstrained inversion" means no additional con- straints were used beyond those, whereas a "constrained inversion" means additional constraints were used. A gray- scale plot of the results of an unconstrained two-dimen- sional TM-mode inversion of the data from these sites is shown in figure 6. The horizontal scale of the plot is the square root of distance from the ocean-continent bound- ary. The vertical scale is the square root of depth, but we PROFILE A (aptos-calero), a priori model 0.3 1.0 20.0 50.0 200.0 500.0 resistivity in ohm-m Figure 5.-Grayscale plot of the a priori model used for inversions along profile A (fig. 2). Resistivities are given in -m. The horizontal scale goes as the square root of distance from the coastline. The vertical scale goes as the square root of distance and is broken up into two 2000.0 5000.0 depth (km) 20000.0 50000.0 - 200000.0 sections-one for 0-45 km and one for 45-600 km. On this and all other grayscale plots, the station locations are marked with asterisks and the San Andreas fault is marked with an arrow. See the text for a discussion of how the a priori values were set. A MAGNETOTELLURIC SURVEY: IMPLICATIONS FOR EARTHQUAKE PROCESSES AND CONDUCTIVITY have broken it up into two different sections, the first one for 0-45 km depth and the second for 45-600 km depth. The models actually extend much further out on either side, but we show only the central portions of each one as there were only minor variations beyond the section we show. The station locations are marked with an asterisk. The inversion resulted in a wide area of increased con- ductance in the lower crust around the fault zone, with the largest change (10 times more conductance) concentrated in the immediate vicinity of the fault zone. Additionally, the lower crust was made more resistive away from the fault zone both under the ocean and under the continent. These changes were made by the inversion routine in or- der to, first, attract enough current onto the continental upper crust, and second, to lose enough of that current between the two data sites to fit the data at those sites. We show the predictions from this model in figure 7 as solid lines. (We show the predictions for all periods be- tween 0.01 s and 2000 s even though the data from 1 to 100 s were not used in the inversion.) The TE-mode pre- dictions are also shown, although we did not use them in D297 the inversion analysis. (At some of the data sites there was a static shift between the observed TM- and TE- mode apparent resistivities. Therefore, in plotting out the predicted TE-mode data, we applied, where necessary, a static shift correction to the predicted TE data so that they agreed with the observed TE data at the higher frequen- cies.) As seen on this and following plots, the TE-mode predictions are actually quite good at most sites except site 7, Hollister, where they are the worst. This gives us increased confidence as to the validity of our two-dimen- sional interpretation. We ran a second inversion of the data from profile A starting with the same a priori model except that we con- strained the lower crust and upper mantle to remain at their a priori values. We show the resulting model in figure 8. For this inversion, we required all resistivity changes to be put into the upper crust. There are mostly minor changes in the upper crustal resistivity values as opposed to the results from the previous inversion, except just above the lower crust near the fault zone and under the Sierra Nevada. In both those places, the conductance PROFILE A (aptos - calero), unconstrained inversion 50.0 200.0 500.0 resistivity in ohm-m 2000.0 depth (km) 5000.0 _ 20000.0 . 50000.0 - 200000.0 Figure 6.-Grayscale plot of the result of a smoothed but otherwise unconstrained two-dimensional TM-mode inversion of the data along profile A (fig. 2). Note the broad area of leakage put into the lower crust in the vicinity of the San Andreas fault zone. The data predicted by this model are shown in figures 7 and 9. AFTERSHOCKS AND POSTSEISMIC EFFECTS D298 'sorouanbay; 1say31y ay} 12 mep apowu moys aM 'g angy ut jopour ay} woj} pajotpoid ay} are sour poysep ay} pur 'q amsy ut jppou -d, porjdsqo ay} yj1m paoiSe Aay: yey} os arom tep apour-g f, pojotpaid oy, 'sis{feue ay} wor; tep pajotpard ay} are piJOS aU, searpuy ueg ay} Jo isea) olafep '| ans ((Y3U) ay) ut pasn arom ep apour-Ji{ J ay} 4juo ySnoy} uaasa 'viep apowu-gqJ, pug -y J porotpaid ay} oq pue seaipuy weg ay} jo sam) sordy 'z ars (oj) 12 jep porotpaid pure parrasqo-'/ amily (spuodes) pousd (spuovdes) pousd 401 eOl z0l 101 001 1-01 z-OL g-O1 401 eOl 101 00} ;-OL g-0t EPP TOT OOT TCT cal WEEP TOT TOT WEHPPTOT OQI {08- 4{0g- 4oc- m. , 4104- € Y 1° m {os- fa or- « a {oc- ® L 60 "y J 401- H 06 0 o TOT T WEEP TOT wlOF TOT u “IOF E 00} m 1-0} apow 11 4 104 o 7 q0} - ~ { hel 1 a 1 1 - a 3 3 3 m 3 zo r+ 10} - - ® - - 9, - - & < - c - 3 <. 3 3 c0l T 4 201 - ~ - - 0.0 - 3 3 1 3 hand p 4 4,04 3 cOl 3 gOt E 401 - { 3 sor = gO} o sopdy 'Z oris (saaibop) asoyd (u-) Apmagsisai juaupddo A MAGNETOTELLURIC SURVEY: IMPLICATIONS FOR EARTHQUAKE PROCESSES AND CONDUCTIVITY was increased as compared to the previous inversion. This was most likely necessary to attract the additional current onto the continental upper crust and to pull it down from the surface to the bottom of the upper crust across the fault zone. However, the increased conductance under the Sierra Nevada is probably unrealistic (Park and others, 1991). We show the TM- and TE-mode predictions from this model in figure 7 as dashed lines. Looking at figure 7, it seems as if there are not many differences in the predictions between the two models. This is because it is difficult to see subtle changes in the slopes of the apparent resistivities, but the changes in the phase are more obvious since they show integrated effects of changes in the apparent resistivity curves. In figure 9, we plot the observed and predicted TM phases for the two models just discussed. In this figure, the circles corre- spond to the TM phase data from site 2, Aptos (nearest the ocean), and the triangles correspond to the TM phase data from site 1, Calero. The predicted data from each site are shown with solid and dashed lines respectively. The TM-mode data show an increase of about 5-10° in the magnitude of the phase from site 2, Aptos, to site 1, Calero, and with our choice of a priori model, this is fit only D299 when the inversion puts a conductive window in the lower crust. Unfortunately, the interpretation of these data is non- unique. With the resistive lower crust we put into the a priori models, the observed data cannot be fit without changes in the lower crustal resistivity between the two stations. However, there are other models that can fit the observed data without any lower crustal leakage paths and using only upper crustal resistivity variations (Eberhart-Phillips and others,1990; Stanley, oral commun, 1992). These models, however, must have lower crusts with much lower resistivities than we have put into our models (1,000 -m vs. 30,000 2-m). Although these mod- els will fit the observed data at the shorter periods, they fail to predict the longer period data that we have previ- ously analyzed at Hollister (Mackie and others, 1988). We have, however, used the results of that study to help constrain our interpretation in this study. This is because at the longer periods involved in the Mackie and others (1988) study, the electric fields perpendicular to the coastline at longer periods must be similar up and down along the coast since the curl of the electric field must be small at these periods and any significant changes in the PROFILE A (aptos - calero), 1D lower crust and upper mantle 300 km 1 100 30 1 1 | 20.0 50.0 200.0 500.0 resistivity in ohm-m 2000.0 10 30 100 300 km depth (km) 5000.0 20000.0 0000.0 200000.0 Figure 8.-Grayscale plot of the result of an inversion that kept the lower crust and upper mantle at its a priori values, thus forcing all changes to be put into the upper crust. The data predicted by this model are shown in figures 7 and 9. D300 0 -10F- -20 phase (degrees) -70 -80 - _90 L_ L( LLLUL L_ 1 J LLL L_ L J J LUULL L _L I LLLLL L __ 1 1 JU LLL L_ 1 J L LUUUL L__L 1 LULU o -10}- -20 phase (degrees) -70 - -80 _go L_ L J [ LLLL L_ 1 J JUL L_ 1 J J LLLL 1 __ 1 1 J LULL L_ L 1 LL ULL L_ 1 _J 1 LULLL L_ _L _J [LLL -30 -40 -50 |- d@+-A a aA -60 -30 -40 -50 -60 AFTERSHOCKS AND POSTSEISMIC EFFECTS Model 1, conductive window in lower crust aptos TM phase 3 e g I I + I T 107-3 10-2 10-1 109 10} 102 1053 104 period (seconds) Model 2, uniform lower crust and mantle aptos TM phase 4 e 2 I I 4 (P P| [ I ¢ P al B I 10-3 10-2 10-"! 10° 10" 102 103 10+ period (seconds) A MAGNETOTELLURIC SURVEY: IMPLICATIONS FOR EARTHQUAKE PROCESSES AND CONDUCTIVITY electric field would involve huge vertical magnetic fields. Therefore, any model we produce to fit the shorter period data of this study should also do a good job at predicting the longer period data of the Mackie and others (1988) study. The longer period data are especially sensitive to the lower crustal and mantle resistivity properties, and therefore, even though we did not explicitly include them in the inversion scheme, they played an important role in constraining the range of possible models that fit the data. The unconstrained inversion increased the conductance in the lower crust over a broad zone, but the data are not sensitive to the details of the geometry of the lower crustal leakage path between the two stations. Physically, it is probably more realistic that any anomalies in the lower crustal resistivity around the Loma Prieta earthquake area < Figure 9.-Plot of the observed and predicted TM-mode phases for the model of figure 6 and the model of figure 8. Note that the data are fit only when the inversion is allowed to put some leakage into the lower crust. If the lower crust and upper mantle are uniform, then the observed data cannot be fit. D301 would be concentrated in a narrower zone. The broad zone of lower crustal leakage shown in figure 6 has a horizon- tal dimension of approximately 50 km. If that lower crustal leakage zone was made 10 times narrower (5 km), then it should require 10 times more conductance to get the same amount of current leakage as the broader zone. For the model in figure 6, this corresponds to a lower crustal resistivity value somewhere between 100 and 1,000 -m for a zone 5 km wide versus 1,000-10,000 §2-m for a 50- km-wide zone. Therefore, we assigned a narrow conductive leakage path in the lower crust for the a priori model and ran an inversion. This inversion reduced the resistivity in the lower crustal fault zone from its a priori value of 1,000 to approximately 300 -m, and slightly increased the resistivity under the western edge of the Sierra Ne- vada (fig. 10). This was necessary to encourage leakage through the narrow zone. The predicted and observed TM-mode phases are shown in the top panel of figure 11. This model resulted in a little more separation in the phases than the previous unconstrained model in figure 6 be- cause the narrow zone loses more current across a short PROFILE A (aptos - calero), apriori narrow conductive fault zone 0.3 1.0 20.0 50.0 200.0 500.0 2000.0 resistivity in ohm-m depth (km) 5000.0 _ 20000.0 _ 50000.0 200000.0 Figure 10.-Grayscale plot of the result of an unconstrained inversion of the data along profile A where the a priori model had a narrow conductive (1,000 2-m) fault zone in the approximate position of the San Andreas. The inversion made the fault zone even more conductive, forcing it down to about 300 ©-m. The data predicted by this model are shown in the top panel of figure 11. D302 AFTERSHOCKS AND POSTSEISMIC EFFECTS apriori narrow conductive fault zone aptos TM phase phase (degrees) I_ LLL |_ J L tuu 1 L L_ L LLL L_ L 1 J L -90 nenaninnmiy 10-2 10-1 10° 101 102 105 10+ 10-3 period (seconds) apriori Great Valley leakage aptos TM phase o 2 phase (degrees) __9 |_ 3 4 CLE 1 1 4 LLM L_ LLL L P L LLL 1 4 LLU 10 . LLL 1 4 4 Luu 10-3 apriori 10-2 10-1 10° 101 102 105 10+ period (seconds) narrow conductive fault zone and GV leakage aptos TM phase phase (degrees) 1 4 d |_ 1 1 ( ( LUW 1 3 1 LLL 1 _} 1 LULA |_ 1 P LLULL L_ J LLL -90 10-3 10-2 10-1 10° 101 102 108 10+ period (seconds) Figure 11.-Plots of the observed and predicted TM-mode phases for three different inversion results. In the top panel are the predictions for the model in figure 10. The middle panel shows the results of an inversion with a priori Great Valley leakage, and the bottom panel shows the results of an inver- sion with a priori Great Valley leak- age and a narrow conductive fault zone. A MAGNETOTELLURIC SURVEY: IMPLICATIONS FOR EARTHQUAKE PROCESSES AND CONDUCTIVITY distance as opposed to the broad zone. (It should be noted that due to the smoothing constraints of our inversion scheme, it is difficult for the inversion to put in abrupt conductivity boundaries unless they were assigned a priori, which is why the unconstrained inversion put in a broad zone of increased conductance.) Although we did not have data further to the east in the Great Valley and Sierra Nevada, we felt it was important to see how the inferred leakage zones of Park and others (1991) on both sides of the southern Great Valley might affect the interpretation of our data. Park and others (1991) collected data within 100 km of our test area, but he ob- served longitudinal uniformity of the Great Valley resis- tivity structure over a distance of approximately 150 km, so it is likely that the inferred Great Valley leakage paths of Park and others (1991) would exist further to the north near our study area. Therefore, we ran an inversion where we a priori assigned the Great Valley leakage zones using the results of Park and others (1991) as a guide in setting the resistivities and widths of those zones. The Great Val- ley leakage has the effect of attracting more ocean current onto the continental upper crust. Consequently, the inver- sion did not need to make the oceanic lower crust as resistive as did the unconstrained inversion without Great Valley leakage (fig. 6). As before, however, it put in a broad area of increased conductance around the Loma Prieta earthquake zone, with electrical properties not much different than before. The observed and predicted TM- mode phases are shown in the middle panel of figure 11. This model does not fit the observed data as well as the previous model with an a priori narrow fault leakage zone. If, however, we run an inversion where we a priori assign the Great Valley leakage and a narrow conductive fault zone, we find that the resulting model predicts the ob- served data very well, and this is shown in the bottom panel of figure 11. The lower crustal resistivities of the narrow fault zone were again reduced to around 300 ©2-m from the a priori values of 1,000 2-m, but the inversion also put in increased conductivities around the narrow fault zone to account for the extra current from the Great Valley leakage. SUMMARY OF MODELING RESULTS Although the interpretation of our data is non-unique, we believe it is likely that there is a vertical lower crustal conductive zone in the vicinity of the San Andreas fault. In a smoothed but otherwise unconstrained inversion, a broad area of increased conductance in the lower crust was put in by the inversion. Another inversion of the data that fixed the a priori lower crust and upper mantle (all changes forced into the upper crust) was not able to fit the data as well and required some unrealistic resistivities under the Sierra Nevada. It is possible, however, to fit the data without any lower crustal leakage paths (Eberhart- D303 Phillips and others, 1990; Stanley, oral commun., 1992), but these models require much more conductive lower crusts than we put into our a priori models, and they do not give model responses that agree with the longer pe- riod data we have analyzed at Hollister (Mackie and oth- ers, 1988). The inferred Great Valley leakage of Park and others (1991) had only a minor effect on the results of the inversions, and that was mainly in the resistance of the lower crust away from the fault zone. Although we can- not determine the exact geometry of the lower crustal leakage, we suspect a narrow conductive zone is geologi- cally more realistic. For the remaining profiles, we tested the various hy- potheses for lower crustal leakage just described for pro- file A. The results for profiles B and C were similar to those obtained for profile A in that unconstrained inver- sions resulted in broad zones of increased conductance in the lower crust around the San Andreas fault zone. As for profile A, these data could also be fit with models that resulted from inversions with a priori narrow lower crustal fault zones. For the sake of brevity, however, we will only show the model predictions from the results of two different inversions of the data for each of the remaining profiles. For the first inversion, we started with an a priori model that had a narrow conductive fault zone and Great Valley leakage. The upper mantle and lower crusts in these models were similar to those that resulted from inversions of the data along profile A, and for this inversion we forced all changes to be made only in the upper crust and in the lower crust in the vicinity of the San Andreas fault. This is a reasonable approach since all the profiles should see approximately the same lower crust and upper mantle away from the fault zone. For the second inver- sion, we took out the narrow fault zone and Great Valley leakage and forced all changes to be put only into the upper crust. The results of the inversions along profile B (figs. 12, 13) again suggest the existence of a conductive zone in the lower crust in the vicinity of the San Andreas fault zone similar to that shown in figure 10 for profile A. The statements made earlier about non-uniqueness and alter- native models apply along this profile as well. There is a change of almost 20° in the observed TM-mode phase between the two sites along this profile, although they are only separated by a distance of 17 km. (The amplitude of the phases at site 4, however, seem to be somewhat too low in comparison with the slope of the apparent resistivity curve, so the true change in phase across the zone may be a little less than 20°.) The only way the inversion can even come close to fitting the data is by making the narrow fault zone very conductive (approxi- mately 50 2-m over the 5 km width as compared to 300 Q- m for profile A). The results from the inversion that constrains the lower crust and upper mantle to remain fixed at their a priori values show that the current levels AFTERSHOCKS AND POSTSEISMIC EFFECTS D304 'sarouanbay} 1say31y ay: je gjep apour-qJ, ay} JIm pooisSe Aoy: Jey} os poyptys ones arom tep apou-qJ, parotpord ou, 'sis{feue ay) ut pasn arom wep apou-p[J, ay} 4juo ySnoy} uaia 'tiep opou-IJ, pue -JAJL porotpard ay) y1oq moys am soddn ay out paoro; arom sasuey> (spuodas) pousad 401 eOl z0L 101 001 1-OL z-Ot g-OL TTTT OT mm TTT mmTTT- T wm TCT TTT T T TmTTT T 06- o8- pmmTT T -T mmm TT mmm TTT TOT TTT TOT TTT T T spow 31 mmTTT T T 1 i 1 4 u 1 + UL 1 Bd 1 i UUU L_ ua _L 00 104 cOl gO yuu t t youpy spjoufay 'g oris 20} (saaibap) ospud (wu-y) Aiagsisai quaupddop re pue jopow woud » oy) ojur md sem auoz jng} ou aroym uoisroaut ue wor; eiep pojorpoid ay} are sourf paysep ay} pue 'ouoz jing] moureu Loz/d » ue yjim uotsroAut ue wou; grep pajotpaid oy) are soul piJOS oY, seaupuy weg oy) Jo jsea) youey spjouday 'g ans (14311) pue seaipuy weg ay) Jo jsam) puouwo7 yoo7 p alls (Jat) ie eep parotpard pue parjasqq-7J ainsiy (spuodas) pousd 401 eOl z0} ( pOL 1-OL z-OI g-OL THPETOT TOT oml lowl- "O IF o ® ® ~ o. ® w jnj ® & U ~ U TEPPTOTOT FIOW 3 3 a 3 0% + [+] - C C 7 [«] 7 "I 7 ® - 3 7 c+ - p - &, - ® 3 a. I <. i lead | to i I 3 3 f ~ 7 4014 3 eOl 3 $014 puowo] yoo} 'p oS A MAGNETOTELLURIC SURVEY: IMPLICATIONS FOR EARTHQUAKE PROCESSES AND CONDUCTIVITY and changes in those levels across the fault zone are in- correctly predicted. On the other hand, the data from the sites along profile C show very little systematic separation in phases across the fault zone (figs. 14, 15). The data at the stations along this profile did not have anomalous tippers in the band from 1 to 100 s, but the apparent resistivities and phases were not quite consistent with each other either (the data from 0.1 to 1 s were not used because of noise). For inversions where the data from 1 to 100 s were not used, as with profiles A and B, results were obtained that indi- cated that some lower crustal leakage was preferred by the inversions, but not nearly as much as indicated for profiles A and B. When we included the data from 1 to 100 s in the inversion, then lower crustal leakage similar to that in profile A was preferred. The model predictions for the inversion without the conductive fault zone are not too different than those with the narrow fault zone, except that there is more separation between the curves when the fault zone is included. The results along this profile are too ambiguous to make any definitive statements about lower crustal leakage here, but the inversions clearly fa- vored some lower crustal leakage when no constraints were placed on the inversions. Finally, we look even further south to profile D, lo- cated near Hollister. The town of Hollister is actually in the creeping section of the Calaveras fault. The transition from creeping to locked behavior along the San Andreas fault occurs near the town of San Juan Bautista just west of Hollister. Unfortunately, the data from EMI site 8 was unusable, so we only had data from site 7 to use for the inversion. The data for this site is a combination of the original EMI data and MIT data from our second trip out to the field. The data for periods greater than 10 seconds is the MIT data (this was more coherent and less noisy than the original EMI data), and the higher frequency data is the EMI data. With just one station, we cannot say anything definitive about the resistivity structure along this profile except that the general trend of the data is in keeping with the data further to the north. These data can be fit equally well with or without a conductive zone in the lower crust (fig. 16). Before we end this section, let us return briefly to the issue of the TE-mode data and model predictions. In all cases except site 7, Hollister, the predicted TE-mode data were actually fairly close to the observed data even though they were not used in the inversion analysis. It is well known that the TE-mode response can be dramatically altered because of finite strike lengths of bodies or other three-dimensional inhomogeneities, whereas the TM mode is much less affected (Wannamaker and others, 1984). Near the interior edge of a conductive three-dimensional body, the fields will be depressed because not enough current is induced in the body to bring the levels up to their two-dimensional values, nor is enough current gath- D305 ered into the body to do likewise. This depresses the MT response there as compared to what would be observed for a two-dimensional model with infinite strike. This flat- tens out the TE response, which is what we observe at the Hollister site. The fact that Hollister sits in a conductive valley and thus may be seeing the effects of finite strike lengths may explain the disparity at this site between its observed and predicted TE-mode data. GEOPHYSICAL IMPLICATIONS FOR THE SAN ANDREAS FAULT SYSTEM In general, the electrical conductivity and distribution of fluids in the lower crust has always been somewhat of an enigma. The fact that finite resistivities are measured for the lower crust most likely implies that it has a small, connected, and fluid-filled porosity (Brace and others, 1965; Lee and others, 1983; Shankland and Ander, 1983), although graphite has also been proposed as a possible conduction mechanism in the lower crust (Duba and Shankland, 1982; Frost and others, 1989). There are many lines of evidence suggesting that free water is present to at least moderate depths within the crust. These include electromagnetic field studies (Shankland and Ander, 1983), isotopic studies of batholithic rocks (for example, Taylor, 1977), geochemical studies (for example, Kerrich and oth- ers, 1984; Kerrich, 1986), analyses of metamorphic rocks (for example, Etheridge and others, 1984), and seismic studies (for example, Nur and Simmons, 1969; Jones and Nur, 1982). At greater depths within the crust, the issue is far more controversial since granulite facies rocks and anhydrous mafic silicates (such as garnets and pyroxenes) commonly associated with the lower crust indicate that it should be fluid-free (Newton, 1990). It is argued (Yardley, 1986) that any water introduced into a granulite facies lower crust would be consumed by rehydration reactions forming lower-grade minerals. However, it is also pos- sible that a small amount of water may be present in the lower crust in equilibrium with a hydrated mineralogy that is not present at the Earth's surface (Gough, 1986). Even if a very small amount of free water does exist in the lower crust, it may not necessarily exist in a con- nected form; rather, because lower crustal rocks are be- lieved to deform ductilely, it is conceivable that the water may exist only in isolated pockets or in absorbed hydrous phases unless the wetting angles are low enough (Watson and Brenan, 1987) or unless fluid pressures can be main- tained at close to the rock pressure for geologically sig- nificant times (Walder and Nur, 1984). Another problem with maintaining water in a permeable crust is that over a period of time, the water should migrate upwards and out of the lower crust, unless it is trapped by some imperme- able layer, because of buoyancy forces (Bailey, 1990). D306 AFTERSHOCKS AND POSTSEISMIC EFFECTS apriori narrow conductive fault zone 0 as -10}- 81:8 ® loch lomond TM phase -20 |- -30 T -40 |- -50 T b O -60 T P PP phase (degrees) T -70 -80 I _go L_ L I [LLL L_ 1 J L LLL L_ L J L LLL L__L L LUN L_ 1 J L LLL 1 _J J J LLU L __ 1 J LLL 10-3 10-2 10-! 109 101 102 105 10+ period (seconds) no fault zone allowed 0 6 -10 } $ " loch lomond TM phase 6 -20 -30 |- -40 -50 |- M ii/ A ~ -60} * Aya phase (degrees) -70 -80 1 _ 1 J JLL J __ 1 J L LLL L __ 1 J J LLL L_ _| I I LLL 1 _L J P LLL L __| _J J LLU L __ 1 J LLU -90 10-3 10-2 10-! 10° 10" 102 103 10+ period (seconds) A MAGNETOTELLURIC SURVEY: IMPLICATIONS FOR EARTHQUAKE PROCESSES AND CONDUCTIVITY This may not be an issue if the hydraulics of the Earth's crust are time-dependent (Nur and Walder, 1990), as is quite likely. Our inversions of the Loma Prieta magnetotelluric data suggest that there may be vertical zones of increased elec- trical conductance in the lower crust that broadly corre- late with the San Andreas fault zone in the area that was involved with the Loma Prieta earthquake. Our findings are contrary to those of Park and others (1991) and Eberhart-Phillips and others (1990) who found no evi- dence for an anomalously conductive San Andreas fault zone. However, in a study of the telluric field variations around the Palmdale section of the San Andreas fault (Madden and others, 1993), systematic changes in the tel- luric field relationships were attributed to small increases in the lower crustal conductance in a vertical zone associ- ated with the San Andreas fault (60 milli-Seimens over a S-year period, which is too small to be seen by MT mea- surements). The strain necessary to cause this increase in lower crustal conductivity would be less than could be measured by the usual surface-strain measurements. We believe that connected fluids in the lower crustal part of the fault zone are responsible for its high conduc- tance relative to the lower crust away from the fault zone. The question, then, is about what amount of fluid are we talking? Typically, porosity estimates for the lower crust are derived by using Archies Law with an exponent of 2 (Brace and others, 1965) and the laboratory values for saline solution resistivities at high pressures (Quist and Marshall, 1968), which probably results in overestimates of the fluid porosities. This is because, first of all, an exponent of 1 in Archies Law is probably more appropri- ate for lower crustal conditions. Looking at the data in Brace and others (1965), we see that the resistivity of crystalline rocks at pressures from 0 to 10 kbar is linearly dependent upon the crack volume of the rock, thus imply- ing that the resistivity is linearly related to the crack po- rosity. In lower crustal crystalline rocks, conduction is almost certainly through microcracks and along grain-edge tubules as opposed to pores in sedimentary rocks. The much more efficient connectivity of this pore geometry greatly reduces the amount of fluid needed to explain the resistivity of lower crustal rocks compared to Archies Law for sedimentary rocks. Secondly, the data from Quist and Marshall (1968) were for saline solutions only up to 0.1 molar, whereas deep crustal fluids are almost certainly much more saline (Orville, 1963; Roedder, 1972). Recent laboratory work on more concentrated saline solutions (Ucok and others, 1980) show fluid resistivities about an order of magnitude lower than the results of Quist and Marshall (1968) at lower crustal conditions. «( Figure 13.-Plot of the observed and predicted TM-mode phases for the inversion results with and without an a priori narrow conductive fault zone. Note that the data are fit only when there is lower crustal leakage around the San Andreas fault zone. D307 The results of our MT inversions suggest that resistivities in the lower crustal fault zone averaged over a 5-km-wide zone may be as low as 50 -m. Concentrated saline solu- tions (20 wt percent NaCl) at temperatures of around 300°C for the lower crust have resistivities of about 0.01 -m (Ucok and others, 1980). Using these values and an expo- nent of 1 in Archies Law yields a connected porosity of approximately 0.02 percent for a 50 2-m, 5-km-wide fault zone if the connections are efficient. We should really multiply this by three due to inefficiencies in tubule con- nection, thus yielding a porosity of 0.06 percent. (Com- pare this with a 0.0001 percent porosity estimate for the lower crust away from the fault zone with a resistivity of 30,000 -m.) If the same amount of conductance were put into a fault zone 1 km wide, it would need to have a resistivity of 10 -m to give the same leakage effect. This translates to a porosity of 0.3 percent. Similarly, a 100-m-wide fault zone zone would require a connected porosity of 3.0 percent. If the lower crustal part of the San Andreas fault zone is as conductive as our inversions suggest, then this will have important implications for the fluid regime of the lower crustal part of fault zones. The question we cannot answer is whether the inferred high conductance of the lower crustal part of the fault zone is an inherent feature of fault systems, or whether it results from the earthquake itself. We do not believe that it is a permanent feature of the fault zone because non-elastic healing processes over time would close up much of the porosity connectivity, thus decreasing the conductivity. We believe a more likely scenario is that the hydraulics and conductivity of fault zones are time-dependent. While brittle failure occurred down to at least 18 km depth, any pore-pressure buildup before the earthquake associated with the closing of per- meable pathways would increase the likelihood of semi- brittle failure down to depths below the main shock of the earthquake. This may have increased the efficiency of fluid connectivity in and around the lower crustal part of the San Andreas fault zone. This hypothesis is based on the difficulty of bringing pore fluids into the lower crustal zone in such a small time scale (one year). However, it also requires that these fluids were inefficiently connected before the earthquake, which would probably lead to high fluid pressures (lithostatic) and a weakening of the me- chanical strength of the lower crustal part of the fault zone. This is a very important issue for understanding the earthquake cycle (Rice, 1992), but without having ana- lyzed data from before the earthquake, we cannot be any more than speculative about these interpretations. CONCLUSIONS We conducted a wide-band MT survey around the Loma Prieta region one year after the Loma Prieta earthquake. The survey was designed to look for lower crustal electri- AFTERSHOCKS AND POSTSEISMIC EFFECTS D308 s9y31Y ay} je gjep apou-gJ, poAJosqo ay) yJIm paoide jey} os poyiys one}s arom tJep apour -4L porotpard ou, 'stsAreue ay} ut posn arom giep apour-jy J, ay) Ajuo ySnoy pug -L porotpaid ay) jog moys am soddn ay: ojut paolo} aram saSueyo [je pue japou iod » ay} (spuoses) poued 404 Fds 301 104 404 3-01 BIO—ill mmm -mmmm--mimem- -mm mmr 408- 40c- 4os- 408, Aor {0¢- doz dow 0} 1 epow 31 5 104 ¥y322 E *J a: 3 m Z 3 z01 6 3 guia a annua g a moras a s04 span 's ous _} (geeibep) esoyd (w-p) Ampsises jueaoddo 'sorouanbay} (spuoses) poued l £01 304 104 Cda 4-04 3-01 g-01 mmr mrt 06 os- ~! - epow nL H«««4q mmm t anouu a a a on aie cl ot jie a ot . a ang a Cda 10% 204 404 aO} puuoppp IN '0l 21s s04 (seeibep) esoyd (w-y) Amgsises queanddo 401 (spuooes) poued £0} 201 104 CB 1-OL 2-01 ror -mm mmm os- os- oc- os- os- or- oc- oz- ol~ agua co accu a a amelia a woul on unul a . iva an ina a a 1-04 CB 10% a g04 a|IAuos;0M '|| 91S 1IW ojur ind sem auoz ou agam ue woj; eep pojotpaid ay) are sour; paysep ay} pue 'auoz ing} moireu rioud » ue Im uotsJoAut ue wou} pajotpoid ay) are saurf pijos ay J, '(ijne} searpuy weg oy) Jo jsea) sean '§ ars pue '(ine; ay) reau) "4 'of aus (appt) '(ijng} searpuy ueg ay} Jo jsam) offlauosieM '[ | alls (oJ) ie wep porotpoid pue porrasqqo-p| ainsi (see1bap) espyd (-y) Apagsises A MAGNETOTELLURIC SURVEY: IMPLICATIONS FOR EARTHQUAKE PROCESSES AND CONDUCTIVITY cal resistivity anomalies both across the fault zone and along it (from north to south). Local industrial noise seemed to contaminate the data at some sites in the band from 1 to 100 s, and at those sites, we did not use the data in that band in the analyses. The results of our two-dimensional TM-mode inver- sions of the data suggested that vertical zones of enhanced lower crustal conductivity exist in the vicinity of the San Andreas fault. If we forced the changes to be put into a 5- km-wide zone, then we obtained lower crustal resistivity values as low as 50 -m in that zone. To explain such dramatic conductivities relative to the lower crust away from the fault zone, we believe that fluids were present in the lower crustal part of the fault zone before the earth- quake, but were inefficiently connected, and that the earth- quake and its associated aftershocks increased the efficiency of fluid connectivity in that zone. An alternate interpretation is that the conductance of the fault zone is an inherent feature of that zone, but laboratory work sug- gests that non-elastic healing processes would eliminate much of the porosity connectivity, thus reducing the con- ductivity of that zone. Although our data alone cannot distinguish between the two interpretations, we believe the time-dependent interpretation is more consistent with the mechanics of the lower crust. Without having ana- lyzed data from before the earthquake, it is impossible to test this hypothesis. ACKNOWLEDGMENTS We would like to thank Frank Morrison, Electromag- netic Instruments Inc., and the students from U.C. Berke- ley for the tremendous job they did in collecting the MT field data and putting up with poison oak, thick brush, and long hours in the field. Useful discussions were held with Dal Stanley, Frank Morrison, Brian Evans, Gary Egbert, Steve Park, Jim Larsen, and Jim Rice. Reviews by Dal Stanley, Paul Reasenberg, and an anonymous reviewer greatly improved the manuscript. The CRAY X-MP EA/ 464 at the MIT supercomputer facility was used for run- ning all the inversions. This work was supported in full by NSF grant 9011919-EAR. REFERENCES CITED Bailey, E.H., ed., 1966, Geology of northern California: California Di- vision of Mines and Geology, v. 190, 508 p. Bailey, EH., Blake, Jr., M.C., and Jones, D.L., 1970, On-land Meso- zoic oceanic crust in California Coast Ranges: U.S. Geological Survey Professional Paper 700-C, p. C70-C81. Bailey, R.C., 1990, Trapping of aqueous fluids in the deep crust: Geo- physical Research Letters, v. 17, no. 8, p. 1129-1132. Bennett, BR., 1985, A long-period magnetotelluric study in Califor- nia: Massachusetts Institute of Technology, masters thesis, 115 p. D309 Brace, W.F., and Kohlstedt, D.L., 1980, Limits on lithospheric stress imposed by laboratory experiments: Journal of Geophysical Re- search, v. 85, no. B11, p. 6248-6252. Brace, W.F., Orange, A.S., and Madden, TR., 1965, The effect of pressure on the electrical resistivity of water-saturated crystalline rocks: Journal of Geophysical Research, v. 70, p. 5669-5678. Burchfiel, B.C., and Davis, G.A., 1972, Structural framework and evo- lution of the southern part of the cordilleran orogen, western United States: Americal Journal of Science, v. 272, no. 2, p. 97- 118. ---1975, Nature and controls of cordilleran orogenesis, western United States: extensions of an earlier synthesis: American Journal of Science, v. 275-A, p. 363-396. Carter, N.L., and Tsenn, M.C., 1987, Flow properties of continental lithosphere: Tectonophysics, v. 136, no. 1/2, p. 27-63. Clark, J.C., and Rietman, J.D., 1973, Oligocene stratigraphy, tectonics, and paleogeography southwest of the San Andreas fault, Santa Cruz Mountains and Gabilan Range, California Coast Ranges: U.S. Geological Survey Professional Paper 783, 18 p. Cox, C.S., Constable, S.C., Chave, A.D., and Webb, S.C., 1986, Con- trolled-source electromagnetic sounding of the oceanic lithos- phere: Nature, v. 320, no. 6057, p. 52-54. Dickinson, W.R., 1981, Plate tectonics and the continental margin of California, in Ernst, W.G., ed., The Geotectonic Development of California: Englewood Cliffs, N.J., Prentice-Hall, p. 1-28. Dietz, L.D., and Ellsworth, W.L., 1990, The October 17, 1989, Loma Prieta, California, earthquake and its aftershocks; geometry of the sequence from high-resolution locations: Geophysical Research Letters, v. 17, no. 9, p. 1417-1420. Duba, A.G., and Shankland, T.J., 1982, Free carbon & electrical con- ductivity in the Earth's mantle: Geophysical Research Letters, v. 9, no. 11, p. 1271-1274. Eberhart-Phillips, D., Labson, V.F., Stanley, W.D., Michael, A.J., and Rodriguez, B.D., 1990, Preliminary velocity and resistivity models of the Loma Prieta earthquake region: Geophysical Research Let- ters, v. 17, no. 8, p. 1235-1238. Etheridge, M.A., Wall, V.J., Cox, S.F., and Vernon, RH., 1984, High fluid pressures during regional metamorphism and deforma- tion: implications for mass transport and deformation mechanisms: Journal of Geophysical Research, v. 89, no. B6, p. 4344-4358. Evans, BR., Fredrich, J.T., and Wong, TW., 1990, The brittle-ductile transition in rocks: recent experimental and theoretical progress, in Geophysical Monograph Series 56, The Heard Volume: American Geophysical Union. Fraser-Smith, A.C., and Coates, D.B., 1978, Large-amplitude ULF electromagnetic fields from BART: Radio Science, v. 13, no. 4, p. 661-668. Frost, B.R., Fyfe, W.S., Tazaki, K., and Chan, T., 1989, Grain-bound- ary graphite in rocks and implications for high electrical conduc- tivity in the lower crust: Nature, v. 340, p. 134-136. Furlong, K.P., and Langston, C.A., 1990, Geodynamic aspects of the Loma Prieta earthquake: Geophysical Research Letters, v. 17, no. 9, p. 1457-1460. Gough, D.I., 1986, Seismic reflectors, conductivity, water and stress in the continental crust: Nature, v. 323, p. 143-144. Heard, H.C., 1976, Comparison of the flow properties of rocks at crustal conditions: Philosophical Transactions of the Royal Soci- ety of London, v. A283, p. 173-186. Ho, A.M. -H, Fraser-Smith, A.C., and Villard, 0.G., Jr., 1979, Large- amplitude ULF magnetic fields produced by a rapid transit sys- tem: close-range measurements: Radio Science, v. 14, no. 6, p- 1011-1015. Jones, A.G., 1987, MT and reflection: an essential combination: Geo- physical Journal of the Royal Astronomical Society, v. 89, no. 1, p. 7-18. D310 AFTERSHOCKS AND POSTSEISMIC EFFECTS Model 1, conductive window in lower crust 0 -10}- -20 -30 -40 |- -50 -60 F- uvas TM phase phase (degrees) -70}- -80 _90 L_ L J L J_ 1 J J LLU L_ L L L LLL L_ 1 J J LLL L_ 1 _J L LLL L __ 1 J L _LLU 1 _ 1 1 1 LLU 10-5 10-2 10-! 10° 101 102 105 10+ period (seconds) Model 2, uniform lower crust and mantle 0 -10 f- watsonville TM phase -20 -30 _4p| mt madonna TM phase -50 ~:_t~‘—F-+~+-—¥-"’,’ -60 - phase (degrees) uvas TM phase -70 |- -80 __90 L_ 1 I I LLULL L_ L_ I JL LLU L_ 1 J D LLLL L_ L_ J JLU L_ |_ I I LUO 1 _| J JUL |__ L_ I 1 LLU 10-3 10-2 10-1 109° 101 102 105 10+ period (seconds) A MAGNETOTELLURIC SURVEY: IMPLICATIONS FOR EARTHQUAKE PROCESSES AND CONDUCTIVITY Jones, TD., and Nur, A., 1982, Seismic velocity and anisotropy in mylonite and the reflectivity of deep crustal fault zones: Geology, v. 10, no. 5, p. 260-263. Kerrich, R., 1986, Fluid transport in lineaments: Philosophical Trans- actions of the Royal Society of London, v. A317, p. 219-251. Kerrich, R., LaTour, TE., and Willmore, L., 1984, Fluid participation in deep fault zones: evidence from geological, geochemical, and *Q/*O relations: Journal of Geophysical Research, v. 89, no. B6, p. 4331-4343. LaTorraca, G.A., Madden, TR., and Korringa, J., 1986, An analysis of the magnetotelluric impedance for three-dimensional conductivity structures: Geophysics, v. 51, no. 9, p. 1819-1829. Lee, C.D., Vine, F.J., and Ross, R.G., 1983, Electrical conductivity models for the continental crust based on laboratory measurements on high-grade metamorphic rocks: Geophysical Journal of the Royal Astronomical Society, v. 72, no. 2, p. 353-371. Mackie, RL., 1991, Three-dimensional magnetotelluric modeling and inversion with applications to the California Basin and Range Province: Massachusetts Institute of Technology, Ph.D. disserta- tion, 238 p. Mackie, RL., Bennett, BR., and Madden, TR., 1988, Long-period magnetotelluric measurements near the central California coast: a land-locked view of the conductivity structure under the Pacific Ocean: Geophysical Journal, v. 95, no. 1, p. 181-194. Madden, TR., 1990, Inversion of low-frequency electromagnetic data, in Oceanographic and Geophysical Tomography: Elsevier Science Publishers, p. 377-408. Madden, TR., LaTorraca, G.A., and Park, S.K., 1993, Electrical con- ductivity variations around the Palmdale section of the San An- dreas fault zone: Journal of Geophysical Research, v. 98, no. B1, p. 795-808. Mazzella, A., and Morrison, H. F., 1974, Electrical resistivity varia- tions associated with earthquakes on the San Andreas fault: Sci- ence, v. 185, no. 4154, p. 855-857. Newton, R.C., 1990, Fluids and shear zones in the deep crust: Tectonophysics, v. 182, no. 1/2, p. 21-37. Nur, A., and Simmons, G., 1969, The effect of saturation on velocity in low porosity rocks: Earth and Planetary Science Letters, v. 7, p. 183-193. Nur, A., and Walder, J., 1990, Time-dependent hydraulics of the Earth's crust, in The Role of Fluids in Crustal Processes: National Academy Press, p. 113-127. Oppenheimer, D.H., 1990, Aftershock slip behavior of the 1989 Loma Prieta California earthquake: Geophysical Research Letters, v. 17, no. 8, p. 1199-1202. Orville, P.M., 1963, Alkali ion exchange between vapor and feldspar phases: American Journal of Science, v. 261, no. 3, p. 201-237. Page, B.M., 1981, The southern Coast Ranges, in Ernst, W.G., ed., The Geotectonic Development of California: Englewood Cliffs, N.J., Prentice-Hall, p. 329-417. Park, S.K., Biasi, G.P., Mackie, RL., and Madden, TR., 1991, Magnetotelluric evidence for crustal suture zones bounding the Figure 15.-Plot of the observed and predicted TM-mode phases for the inversion results with and without an a priori narrow conductive fault zone. It is difficult to tell much difference between the two models within the spread of the data, but we believe that data are fit slightly better when there is a somewhat conductive fault zone. D311 southern Great Valley, California: Journal of Geophysical Re- search, v. 96, p. 353-376. Qian, W., and Pedersen, L.B., 1991, Industrial interference magnetotellurics: an example from the Tangshan area, China: Geophysics, v. 56, no. 2, p. 265-273. Quist, A.S., and Marshall, W.L., 1968, Electrical conductances of aqueous sodium chloride solutions from 0 to 800° and at pressures to 4000 bars: Journal of Physical Chemistry, v. 72, no. 2, p. 684- 702. Ranganayaki, RP., 1978, Generalized thin sheet approximation for magnetotelluric modeling: Massachusetts Institute of Technology, Ph.D. dissertation, 204 p. Ranganayaki, RP., and Madden, TR., 1980, Generalized thin sheet analysis in magnetotellurics, an extension of Price's analysis: Geophysical Journal of the Royal Astronomical Society, v. 60, no. 3, p. 445-457. Rice, J.R., 1992, Fault stress states, pore pressure distributions, and the weakness of the San Andreas fault, in Evans and Wong, Fault Me- chanics and Transport Properties in Rocks: Academic Press, p. 475-503. Roedder, E., 1972, Composition of fluid inclusions, in Fleischer, M., ed., Data of Geochemistry (sixth edition): U.S. Geological Society Professional Paper 440-JJ, 164 p. Segall, P., and Lisowski, M., 1990, Surface displacements in the 1906 San Francisco and 1989 Loma Prieta earthquakes: Science, v. 250, p. 1241-1243. Shankland, T.J., and Ander, ME., 1983, Electrical conductivity, tem- peratures, and fluids in the lower crust: Journal of Geophysical Research, v. 88, no. B11, p. 9475-9484.m Tarantola, A., 1987, Inverse problem theory: Methods for data fitting and model parameter estimation: New York, Elsevier. Taylor, H.P., Jr., 1977, Water/rock interaction and the origin of H,O in granitic batholiths: Journal of the Geological Society of London, v. 133, p. 509-558. Torgensen, T., 1990, Crustal-scale fluid transport: Eos, Transactions, American Geophysical Union, v. 71, p. 1-4. Ucok, H., Ershaghi, I., and Olhoeft, G.R., 1980, Electrical resistivity of geothermal brines: Journal of Petroleum Technology, v. 32, p. 717-727. Walder, J., and Nur, A., 1984, Porosity reduction and crustal pore pressure development: Journal of Geophysical Research, v. 89, no. B13, p. 11539-11548. Wannamaker, P.E., Hohmann, G.W., and Ward, S.H., 1984, Magnetotelluric responses of three-dimensional bodies in layered earths: Geophysics, v. 49, no. 9, p. 1517-1533. Watson, E.B., and Brenan, J.M., 1987, Fluids in the lithosphere, 1. Experimentally-determined wetting characteristics of CO,-H,0 fluids and their implications for fluid transport, host-rock ph-ysiéal properties, and fluid inclusion formation: Earth and Planetary Sci- ence Letters, v. 85, no. 4, p. 497-515. Yardley, B.W.D., 1986, Is there water in the deep continental crust?: Nature, v. 323, no. 6084, p. 111. Figure 16.-Observed and predicted data at site 7, Hollister (east of the San Andreas fault). The solid lines are the predicted data from an inver- sion with an a priori narrow fault zone, and the dashed lines are the predicted data from an inversion where no fault zone was put into the a priori model and all changes were forced into the upper crust. We show both the predicted TM- and TE-mode data, even though only the TM- mode data were used in the analysis. Both models fit the data equally well. D312 AFTERSHOCKS AND POSTSEISMIC EFFECTS Site 7, Hollister 106 T T IIHTT! 105 T TTT T 104 T T TTT 103 Io T 102 T apparent resistivity (Q-m) 10" T T TTT I 10° T T TTT 10—1 L_ L J L L_ [ L LLL L_ L 1 LL L_ 1 1 L LLL L_ 1 I L LLL L_ L _I L LLLL L_ LL I I I | B (G No- o 06 o o I I I I phase (degrees) I w O T _90 L_ J ( 1 __1 I LLL L_ _L I LLL L_ 1 LLL L_ L L DLL L_ 1 I [JUL L_ L J JUL, 10-5 10-2 10-1 109 101 102 1035 104 period (seconds) * U.S. GOVERNMENT PRINTING OFFICE: 1997 - 573-046 / 20077 REGION NO. 8 SELECTED SERIES OF U.S. GEOLOGICAL SURVEY PUBLICATIONS Periodicals Earthquakes & Volcanoes (issued bimonthly). Preliminary Determination of Epicenters (issued monthly). Technical Books and Reports Professional Papers are mainly comprehensive scientific reports of wide and lasting interest and importance to professional scientists and engineers. Included are reports on the results of resource studies and of topographic, hydrologic, and geologic investigations. They also include collections of related papers addressing different aspects of a single sci- entific topic. Bulletins contain significant data and interpretations that are of lasting scientific interest but are generally more limited in scope or geo- graphic coverage than Professional Papers. They include the results of re- source studies and of geologic and topographic investigations; as well as collections of short papers related to a specific topic. Water-Supply Papers are comprehensive reports that present sig- nificant interpretive results of hydrologic investigations of wide interest to professional geologists, hydrologists, and engineers. The series covers investigations in all phases of hydrology, including hydrology, availabil- ity of water, quality of water, and use of water. Circulars present administrative information or important scientif- ic information of wide popular interest in a format designed for distribu- tion at no cost to the public. Information is usually of short-term interest. Water-Resources Investigations Reports are papers of an inter- pretive nature made available to the public outside the formal USGS pub- lications series. Copies are reproduced on request unlike formal USGS publications, and they are also available for public inspection at deposi- tories indicated in USGS catalogs. Open-File Reports include unpublished manuscript reports, maps, and other material that are made available for public consultation at de- positories. They are a nonpermanent form of publication that may be cit- ed in other publications as sources of information. Maps Geologic Quadrangle Maps are multicolor geologic maps on to- pographic bases in 7 1/2- or 15-minute quadrangle formats (scales main- ly 1:24,000 or 1:62,500) showing bedrock, surficial, or engineering geology. Maps generally include brief texts; some maps include structure and columnar sections only. Geophysical Investigations Maps are on topographic or planimet- ric bases at various scales, they show results of surveys using geophysical techniques, such as gravity, magnetic, seismic, or radioactivity, which re- flect subsurface structures that are of economic or geologic significance. Many maps include correlations with the geology. Miscellaneous Investigations Series Maps are on planimetric or topographic bases of regular and irregular areas at various scales; they present a wide variety of format and subject matter. The series also in- cludes 7 1/2-minute quadrangle photogeologic maps on planimetric bases which show geology as interpreted from aerial photographs. The series also includes maps of Mars and the Moon. Coal Investigations Maps are geologic maps on topographic or planimetric bases at various scales showing bedrock or surficial geology, stratigraphy, and structural relations in certain coal-resource areas. Oil and Gas Investigations Charts show stratigraphic informa- tion for certain oil and gas fields and other areas having petroleum poten- tial. Miscellaneous Field Studies Maps are multicolor or black-and- white maps on topographic or planimetric bases on quadrangle or irreg- ular areas at various scales. Pre-1971 maps show bedrock geology in re- lation to specific mining or mineral-deposit problems; post-1971 maps are primarily black-and-white maps on various subjects such as environ- mental studies or wilderness mineral investigations. Hydrologic Investigations Atlases are multicolored or black-and- white maps on topographic or planimetric bases presenting a wide range of geohydrologic data of both regular and irregular areas; the principal scale is 1:24,000, and regional studies are at 1:250,000 scale or smaller. Catalogs Permanent catalogs, as well as some others, giving comprehensive listings of U.S. Geological Survey publications are available under the conditions indicated below from USGS Map Distribution, Box 25286, Building 810, Denver Federal Center, Denver, CO 80225. (See latest Price and Availability List.) "Publications of the Geological Survey, 1879-1961" may be pur- chased by mail and over the counter in paperback book form and as a set microfiche. "Publications of the Geological Survey, 1962-1970" may be pur- chased by mail and over the counter in paperback book form and as a set of microfiche. "Publications of the U.S. Geological Survey, 1971-1981" may be purchased by mail and over the counter in paperback book form (two vol- umes, publications listing and index) and as a set of microfiche. Supplements for 1982, 1983, 1984, 1985, 1986, and for subse- quent years since the last permanent catalog may be purchased by mail and over the counter in paperback book form. State catalogs, "List of U.S. Geological Survey Geologic and Wa- ter-Supply Reports and Maps For (State)," may be purchased by mail and over the counter in paperback booklet form only. "Price and Availability List of U.S. Geological Survey Publica- tions," issued annually, is available free of charge in paperback booklet form only. Selected copies of a monthly catalog "New Publications of the U.S. Geological Survey" is available free of charge by mail or may be ob- tained over the counter in paperback booklet form only. Those wishing a free subscription to the monthly catalog "New Publications of the U.S. Geological Survey" should write to the U.S. Geological Survey, 582 Na- tional Center, Reston, VA 22092. Note.-Prices of Government publications listed in older catalogs, announcements, and publications may be incorrect. Therefore, the prices charged may differ from the prices in catalogs, announcements, and pub- lications.