DOC,"
Y 3,N88;
25/5250/v.l,
NUREG/CR-5250
UCID-21517
Vol. 1
Seismic Hazard Characterization
of 69 Nuclear Plant Sites
East of the Rocky Mountains
Methodology, Input Data and Comparisons to Previous Results
for Ten Test Sites
Prepared by D. L. Bernreuter, J. B. Savy, R. W. Mensing, J. C. Chen
Lawrence Livermore National Laboratory
Prepared for
U.S. Nuclear Regulatory
Commission
NOTICE
This report was prepared as an account of work sponsored by an agency of the United States
Government. Neither the United States Government nor any agency thereof, or any of their
employees, makes any warranty, expressed or implied, or assumes any legal liability of re-
sponsibility for any third party's use, or the results of such use, of any information, apparatus,
product or process disclosed in this report, or represents that its use by such third party would
not infringe privately owned rights.
UNIVERSITY OF -
ILLINOIS LIBRARY
AT URBANA-CHAMPAIGN
SOOKSTACKG
NOTICE
Availability of Reference Materials Cited in NRC Publications
Most documents cited in NRC publications will be available from one of the following sources:
1. The NRC Public Document Room, 1717 H Street, N.W.
Washington, DC 20555
2. The Superintendent of Documents, U.S. Government Printing Office, Post Office Box 37082
Washington, DC 20013-7082
3. The National Technical Information Service, Springfield, VA 22161
>lthough the listing that follows represents the majority of documents cited in NRC publications,
it is not intended to be exhaustive.
Referenced documents available foi inspection and copying for a fee from the NRC Public Docu
ment Room include NRC correspondence and internal NRC memoranda; NRC Office of Inspection
and Enforcement bulletins, circulars, information notices, inspection and investigation notices;
Licensee Event Reports; vendor reports and correspondence; Commission papers; and applicant and
licensee documents and correspondence.
The following documents in the NUREG series are available for purchase from the GPO Sales
Program: formal NRC staff and contractor reports, NRC-sponsored conference proceedings, and
NRC booklets and brochures. Also available are Regulatory Guides, NRC regulations in the Code of
Federal Regulations, and Nuclear Regulatory Commission Issuances.
Documents available from the National Technical Information Service include NUREG series
reports and technical reports prepared by other federal agencies and reports prepared by the Atomic
Energy Commission, forerunner agency to the Nuclear Regulatory Commission.
Documents available from public and special technical libraries include all open literature items,
such as books, journal and periodical articles, and transactions. Federal Register notices, federal and
state legislation, and congressional reports can usually be obtained from these libraries.
Documents such as theses, dissertations, foreign reports and translations, and non-NRC conference
proceedings are available for purchase from the organization sponsoring the publication cited.
Single copies of NRC draft reports are available free, to the extent of supply, upon written
request to the Division of Information Support Services, Distribution Section, U.S. Nuclear
Regulatory Commission, Washington, DC 20555.
Copies of industry codes and standards used in a substantive manner in the NRC regulatory process
are maintained at the NRC Library, 7920 Norfolk Avenue, Bethesda, Maryland, and are available
there for reference use by the public. Codes and standards are usually copyrighted and may be
purchased from the originating organization or, if they are American National Standards, from the
American National Standards Institute, 1430 Broadway, New York, NY 10018.
NUREG/CR-5250
UCID-21517
Vol. 1
Seismic Hazard Characterization
of 69 Nuclear Plant Sites
East of the Rocky Mountains
Methodology, Input Data and Comparisons to Previous Results
for Ten Test Sites
Manuscript Completed: November 1988
Date Published: January 1989
Prepared by
D. L. Bernreuter, J. B. Savy, R. W. Mensing, J. C. Chen
Lawrence Livermore National Laboratory
7000 East Avenue
Livermore, CA 94550
Prepared for
Division of Engineering and System Technology
Office of Nuclear Reactor Regulation
U.S. Nuclear Regulatory Commission
Washington, DC 20555
NRC FIN A0448
Abstract
The EUS Seismic Hazard Characterization Project (SHC) is the outgrowth of an
^.mlx^c ^^"^^ performed as part of the U.S. Nuclear Regulatory Commission's
NRC) Systematic Evaluation Program (SEP). The objectives of the SHC were-
(1) to develop a seismic hazard characterization methodology for the region
east of the Rocky Mountains (EUS), and (2) the application of the methodology
^^u ^^] locations, some of them with several local soil conditions The
method developed uses expert opinions to obtain the input to the analyses An
important aspect of the elicitation of the expert opinion process was the'
holding of two feedback meetings with all the experts in order to finalize the
methodology and the input data bases. The hazard estimates are reported in
terms of peak ground acceleration (PGA) and 5% damping velocity response
the
A total of eight volumes make up this report which contains a thorough
description of the methodology, the expert opinion's elicitation process
input data base as well as a discussion, comparison and summary volume
(Volume VI). -^
Consistent with previous analyses, this study finds that there are large
uncertainties associated with the estimates of seismic hazard in the EUS and
It Identifies the ground motion modeling as the prime contributor to those
uncertainties.
The data bases and software are made available to the NRC and to public uses
through the National Energy Software Center (Argonne, Illinois).
-Ill-
:'".-r^;o.:;-
Table of Contents
Volume I
PAGE
Abstract
Table of Contents
List of Tables and Figures
Foreword
List of Abbreviations and Symbols
Overall Executive Summary
Executive Summary: Volume I
SECTION 1 INTRODUCTION
SECTION 2 OVERVIEW OF THE METHODOLOGY USED
2.1 Introduction
2.2 Description of the Hazard Calculation Methodology
2.3 Uncertainty in the Hazard
2.4 Use of Experts' Opinion
2.5 Aggregation of Experts' Opinions
2.6 Hazard Analysis Outputs
SECTION 3 DEVELOPMENT OF HAZARD ANALYSIS INPUTS
3.1 Processes Used for the Elicitation of Expert Opinion
3.2 Development of the Seismic Zonation Maps
3.3 Seismicity and Upper Magnitude Cutoff
3.4 Documentation
3.5 Ground Motion Models
3.6 Random Uncertainty and Truncation of GM Estimates
3.7 Correction for Local Soil Conditions
SECTION 4 COMPARISON TO PREVIOUS RESULTS AND OTHER STUDIES
4.1 Comparison to Previous Results
4.2 Comparison to Other Studies
SECTION 5 REFERENCES
m
V
vii
xiii
xvii
XX i
XXV
1
9
9
9
13
15
15
16
23
23
32
36
40
41
54
58
66
66
68
80
APPENDIX A
APPENDIX B
APPENDIX C
Documentation Responses
Final Inputs
Details of the Methodology Used
A-1
B-1
C-1
-V-
List of Tables and Figures
Table 1.1
Table 3.1
Table 3.2
Table 3.3
Table 3.4
Table 3.5
Table 3.6
Table 3.7
Table 3.8a
Table 3.8b
Table 3.9
Table 3.10
List of Sites in Each Volume of this Report
EUS Zonation and Seismicity Panel Members
EUS Ground Motion Model Panel Members
Schematic Representation of the Flow of Operations
in the Elicitation of the Experts' Opinions
Summary of Model Choices of S-Experts
PGA Models and Weights Selected by the G-Experts
5 Percent Damped Relative Velocity Spectral Models
and Weights Selected by the G-Experts
G-Experts Self-Weights
G-Expertjs Choice for the Random Uncertainty,
Sigma ^^> for the GM models
G-Experts' Choice for the Truncation of the
Ground Motion Variation
Definition of the Eight Site Categories
G-Experts' Weights for Site Correction Approach
PAGE
3
26
27
28
39
44
46
48
55
56
61
62
Figure 1.1a
Figure 1.1b
Figure 1.1c
Figure l.ld
Map showing the location of the Batch 1
sites contained in Vol. II of this
report. Map symbols are given in
Table 1.1.
Map showing the location of Batch 2 sites
contained in Vol. Ill of this report.
Map symbols are given in Table 1.1.
Map showing the location of Batch 3 sites
contained in Vol. IV of this report.
Map symbols are given in Table 1.1.
Map showing the location of Batch 4 sites
contained in Vol. V of this report.
Map symbols are given in Table 1.1.
Vll
Figure 2.1
Figure 2.2
Figure 2.3
Figure 2.4
Figure 2.5
Figure 2.6
Figure 2.7
Figure 3.1
Figure 3.2
Figure 3.3
Figure 3.4
Figure 3.5
Figure 3.6
Four steps involved in a probabilistic
seismic hazard analysis.
Typical results of our seismic hazard
analysis. Shown are the 15th, 50th and
85th percentile PGA hazard curves for
the Braidwood site.
Map indicating the boundary of the four
regions of the EUS and by our S-Experts to
determine their self weights and by the
G-Experts to select ground motion models.
BEHC for PGA per G-Expert for S-Expert I's
zonation and seismicity parameters.
15th, 50th and 85th CPHC for PGA for
S-Expert 1 aggregated over all G-Experts
for the Braidwood site.
BEHC per S-Expert aggregated over all
6-Experts for PGA for the Braidwood site.
Comparison of the BEHC and AMHC for PGA
aggregated over all S and G-Experts for
the Braidwood site.
Typical example of a zonation map from
our S-Experts. Shown is the BE map from
S-Expert 1.
Alternative zones provided by S-Expert 1.
Map showing the location of the ten test
sites used in our previous analysis,
Bernreuter et al. (1984,1985).
Best estimate PGA models listed in Table 3.5
plotted for magnitudes of 5 and 7.
Rock base case.
Remaining PGA models listed in Table 3.5
plotted for magnitudes of 5 and 7.
Rock base case.
Best estimate PGA models corrected to generic
deep soil for magnitudes of 5 and 7.
PAGE
10
14
17
18
20
21
22
33
34
35
49
50
51
VI 11
Vyyy.
PAGE
Figure 3.7
Figure 3.8
Figure 3.9
Figure 3.10
Figure 3.11a
Figure 3.11b
Figure 3.12
Figure 3.13
Figure 4.1
Best estimate 5 percent damped relative 52
velocity spectra models listed in Table 3.6
plotted for magnitudes of 5 and 7 at a distance
of 25 km. Rock base case.
Remaining 5 percent damped relative velocity 53
spectra models listed in Table 3.6 plotted
for magnitudes of 5 and 7 at a distance of 25 km.
Illustration of the three types of models 57
considered to model the physical saturation
of ground motion. The random variation of
the logarithm of the ground motion parameter
(GMP) is modeled by a normal distribution with
mean GMP (m.R) and a standard deviation o.
Simple correction factors relative to rock. 63
Schematic representation of our computational 64
procedure to model site correction factors.
The physical parameters used in the 1-D 64
analysis are drawn from probability
distributions.
Smoothed median correction factors for the 65
till-like categories listed in Table 3.9
relative to rock. The PGA correction factors
are plotted at 0.01s.
Smoothed median correction factors for 65
the sand-like categories listed in Table 3.9
relative to rock. Also shown are the
correction factors for deep soil relative
to rock. The PGA correction factors are
plotted at 0.01s.
Comparison of the 15th, 50th and 85th 69
percentile CPHCs for PGA between the new
results based on the updated input from
the S-Experts described in Section 3 of
our previous results given in Bernreuter
et al. (1985) for the Braidwood site.
Only a single PGA model (modified Nuttli)
was used.
IX
PAGE
Figure 4.2
Figure 4.3a
Figure 4.3b
Figure 4.4a
Figure 4.4b
Figure 4.5
Figure 4.6
Figure 4.7
Comparison of the 15th, 50th and 85th 70
pecentile CPHCs for PGA between the
new results based on the updated
input from the S-Experts described
in Section 3 of our previous results
given in Bernreuter et al. (1985)
for the Millstone site. Only a
single PGA model (modified Nuttli) was used.
BEHCs for PGA for the Braidwood site per 71
S-Expert based on the updated input provided
by the S-Experts. Only a single PGA model
(modified Nuttli) was used.
BEHCs for PGA for the Braidwood site per 72
S-Expert based on the previous input given
in Bernreuter et al. (1985). Only a single
PGA model (modified Nuttli) was used.
BEHCs for PGA for the Millstone site per 73
S-Expert based on the updated input provided
by the S-Experts. Only a single PGA model
(modified Nuttli) was used.
BEHCs for PGA for the Millstone site per 74
S-Expert based onthe updated input provided
by the S-Experts. Only a single PGA model
(modified Nuttli) was used.
Comparison of the 15th, 50th and 85th 75
percentile CPHCs between the CPHCs obtained
aggregating over all the S and G-Experts and
the CPHCs obtained using a single PGA model and
aggregated over all S-Experts for the
Braidwood site.
Comparison of the 15th, 50th and 85th 76
percentile CPHCs between the CPHCs
obtained aggregating over all the S and
G-Experts and the CPHCs obtained using a
single PGA model and aggregated over all
S-Experts for the Millstone site.
Comparison of the 15th, 50th and 85th 77
percentile CPHCs aggregated over all S
and G-Experts between the new input and
the previous input from the S and G-Experts
for the Braidwood site.
W.
PAGE
Figure 4.8
Comparison of the 15th, 50th and 85th
percentile CPHCs aggregated over all S
and G-Experts between the new input and
the previous input from the S and G-Experts
for the Millstone site.
78
Figure 4.9
Comparison of the 1000 year return period
15th, 50th and 85th percentile CPUHS for
5 percent damping aggregated over all S
and G-Experts between the new input and
our previous input from the S and G-Experts
for the Millstone site.
79
XI
:^S
Foreword
The impetus for this study came from two unrelated needs of the Nuclear
Regulatory Commission (NRC). One stimulus arose from the NRC funded "Seismic
Safety Margins Research Programs" (SSMRP). The SSMRP's task of simplified
methods needed to have available data and analysis software necessary to
compute the seismic hazard at any site located east of the Rocky Mountains
which we refer to as the Eastern United States (EUS) in a form suitable for
use in probabilistic risk assessment (PRA). The second stimulus was the
result of the NRC's discussions with the U.S. Geological Survey (US6S)
regarding the USGS's proposed clarification of their past position with
respect to the 1886 Charleston earthquake. The USGS clarification was finally
issued on November 18, 1982, in a letter to the NRC, which states that:
"Because the geologic and
similar to those in other
that although there is no
have experienced strong ea
itself, sufficient ground
regions of strong seismic
Charleston in 1886. Altho
to an earthquake in any gi
seaboard may be very low,
the seismic hazard should
seaboard to establish the
facilities."
tectonic features of the Charleston region are
regions of the eastern seaboard, we conclude
recent or historical evidence that other regions
rthquakes, the historical record is not, of
for ruling out the occurrence in these other
ground motions similar to those experienced near
ugh the probability of strong ground motion due
ven year at a particular location in the eastern
deterministic and probabilistic evaluations of
be made for individual sites in the eastern
seismic engineering parameters for critical
Anticipation of this letter led the Office of Nuclear Reactor Regulation to
jointly fund a project with the Office of Nuclear Regulatory Research. The
results were presented in Bernreuter et. al. (1985), and the objectives were:
1. to develop a seismic hazard characterization methodology for the
entire region of the United States east of the Rocky Mountains
(Referred to as EUS in this report).
2. to apply the methodology to selected sites to assist the NRC staff in
their assessment of the implications in the clarification of the USGS
position on the Charleston earthquake, and the implications of the
occurrence of the recent earthquakes such as that which occurred in
New Brunswick, Canada, in 1982.
The methodology used in that 1985 study evolved from two earlier studies LLNL
performed for the NRC. One study, Bernreuter and Minichino (1983), was part
of the NRC's Systematic Evaluation Program (SEP) and is simply referred
hereafter to as the SEP study. The other study was part of the SSMRP.
At the time (1980-1985), an improved hazard analysis methodology and EUS
seismicity and ground motion data set were required for several reasons:
Although the entire EUS was considered at the time of the SEP study,
attention was focused on the areas around the SEP sites—mainly in the
xm
Central United States (CUS) and New England. The zonation of other
areas was not performed with the same level of detail.
The peer review process, both by our Peer Review Panel and other
reviewers, identified some areas of possible improvements in the SEP
methodology.
Since the SEP zonations were provided by our EUS Seismicity Panel in
early 1979, a number of important studies have been completed and
several significant EUS earthquakes have occurred which could impact
the Panel members' understanding of the seismotectonics of the EUS.
Our understanding of the EUS ground motion had improved since the time
the SEP study was performed.
By the time our methodology was firmed up, the expert opinions collected
the calculations performed (i.e. by 1985), the Electric Power Research
Institute (EPRI) had embarked on a parallel study.
and
We performed a comparative study, Bernreuter
understanding the reasons for differences in
EPRI studies. The three main differences wer
magnitude value of the earthquakes contributi
the ground motion attenuation models, and (3)
local site characteristics and EPRI did not.
1985 study and the application of the methodo
EUS. In recognition of the fact that during
of research in seismotectonics and in the fie
prediction, in particular with the developmen
vibration or stochastic approach, NRC decided
and have a final round of feedback with all o
the input to the analysis.
et. al. (1987), to help in
results between the LLNL and the
e found to be: (1) the minimum
ng to the hazard in the EUS, (2)
the fact that LLNL accounted for
Several years passed between the
logy to all the sites in the
that time a considerable amount
Id of strong ground motion
t of the so called random
to follow our recommendations
ur experts prior to finalizing
In addition, we critically reviewed our methodology which lead to minor
improvements and we also provided an extensive account of documentation on the
ways the experts interpreted our. questionnaires and how they developed their
answers. Some of the improvements were necessitated by the recognition of the
fact that the results of our study will be used, together with results from
other studies such as the EPRI study or the USGS study, to evaluate the
relative hazard between the different plant sites in the EUS.
This report is comprised of eight volumes:
Volume I provides an overview of the methodology we developed as part of
this project. It also documents the final makeup of both our Seismicity
and Ground Motion Panels, and documents the final input from the members
of both panels used in the analysis. Comparisons are made between the new
results and previous results.
Volumes II to V provide the results for all the active nuclear power plant
sites of the EUS divided into four batches of approximately equal size and
of sites roughly located in the four main regions of the EUS. A regional
discussion is given in each of Vols. II to V.
xiv
Volume VI gives some important sensitivity studies, in particular the
sensitivity of the results to correction for local site conditions and
G-Expert 5's ground motion model. Volume VI also contains a surrmary of
the results and provides comparisons between the sites within a common
region and for sites between regions.
Volume VII contains unaltered copies of the ten questionnaires used from
the beginning of the 1985 study to develop the complete input for this
analysis.
After the bulk of the work was completed and draft reports for Vols. I-VII
were written, additional funding became available.
Volume VIII contains the hazard result for the 12 sites which had some
structures founded on shallow soil. These results supplement the results
given in Vols. II to V where only the primary soil condition at the site
was used.
XV
m
List of Abbreviations and Symbols
A Symbol for Seismicity Expert 10 in the figures displaying the results
for the S-Experts
ALEAS Computer code to compute the BE Hazard and the CP Hazard for each
seismicity expert
AM Arithmetic mean
AMHC Arithmetic mean hazard curve
B Symbol for Seismicity Expert 11 in the figures displaying the results
for the S-Experts
BE Best estimate
BEHC Best estimate hazard curve
BEUHS Best estimate uniform hazard spectrum
BEM Best estimate map
C Symbol for Seismicity Expert 12 in the figures displaying the results
for the S-Experts
COMAP Computer code to generate the set of all alternative maps and the
discrete probability density of maps
COMB Computer code to combine BE hazard and CP hazard over all seismicity
experts
CP Constant percentile
CPHC Constant percentile hazard curve
CPUHS Constant percentile uniform hazard spectrum
CUS Central United States, roughly the area bounded in the west by the
Rocky Mountains and on the east by the Appalachian Mountains,
excluding both mountain systems themselves
CZ Complementary zone
D Symbol for Seismicity Expert 13 in the figures displaying the results
for the S-Experts
EPRI Electric Power Research Institute
EUS Used to denote the general geographical region east of the Rocky
Mountains, including the specific region of the Central United States
(CUS)
I.
xvn
g Measure of acceleration: Ig = 9.81m/s/s = acceleration of gravity
G-Expert One of the five experts elicited to select the ground motion models
used in the analysis
GM
HC
h
LB
LLNL
M
Ml
Mb
M5
MMI
Mo
NC
NE
NRC
PGA
PGV
PRO
PSRV
Ground motion
Hazard curve
Epicentral intensity of an earthquake relative to the MMI scale
Site intensity of an earthquake relative to the MMI scale
Lower bound
Lawrence Livermore National Laboratory
Used generically for any of the many magnitude scales but generally
%y m^C-g), or M|_.
Local magnitude (Richter magnitude scale)
True body wave magnitude scale, assumed to be equivalent to mu(Lg)
(see Chung and Bernreuter, 1981)
Nuttli's magnitude scale for the Central United States based on the
Lg surface waves
Surface wave magnitude
Modified Mercalli Intensity
Lower magnitude of integration. Earthquakes with magnitude lower
than Mq are not considered to be contributing to the seismic hazard
North Central; Region 3
North East; Region 1
Nuclear Regulatory Commission
Peak ground acceleration
Peak ground velocity
Computer code to compute the probability distribution of epicentral
distances to the site
Pseudo relative velocity spectrum. Also see definition of spectra
below
xvni
^?<.
Q Seismic quality factor, which is inversely proportional to the
inelastic damping factor.
Ql Questionnaire 1 - Zonation (I)
Q2 Questionnaire 2 - Seismicity (I)
Q3 Questionnaire 3 - Regional Self Weights (I)
Q4 Questionnaire 4 - Ground Motion Models (I)
Q5 Questionnaire 5 - Feedback on seismicity and zonation (II)
Q6 Questionnaire 6 - Feedback on ground motion models (II)
Q7 Questionnaire 7 - Feedback on zonation (III)
Q8 Questionnaire 8 - Seismicity input documentation
Q9 Questionnaire 9 - Feedback on seismicity (III)
QIO Questionnaire 10 - Feedback on ground motion models (III)
R Distance metric, generally either the epicentral distance from a
recording site to the earthquake or the closest distance between the
recording site and the ruptured fault for a particular earthquake.
Region 1 (NE): North East of the United States, includes New England and
Eastern Canada
Region 2 (SE): South East United States
Region 3 (NC): North Central United States, includes the Northern Central
portions of the United States and Central Canada
(SC): Central United States, the Southern Central portions of the
United States including Texas and Louisiana
Region 4
RP
RV
SC
SE
S-Expert
Return period in years.
Random vibration. Abbreviation used for a class of ground motion
models also called stochastic models.
Site factor used in the regression analysis for G-Expert 5's GM
model: S = for deep soil, S = 1 for rock sites
South Central; Region 4
South East; Region 2
One of the eleven experts who provide the zonations and seismicity
models used in the analysis
xix
SEP Systematic Evaluation Program
SHC Seismic Hazard Characterization
SHCUS Seismic Hazard Characterization of the United States
SN Site Number
Spectra Specifically in this report: attenuation models for spectral
ordinates were for 5% damping for the pseudo-relative velocity
spectra in PSRV at five frequencies (25, 10, 5, 2.5, 1 Hz).
SSE Safe Shutdown Earthquake
SSI Soil-structure-interaction
SSMRP Seismic Safety Margins Research Program
UB Upper bound
UHS Uniform hazard spectrum {or spectra)
uses United States Geological Survey
WUS The regions in the Western United States where we have strong ground
motion data recorded and analyzed
XX
Overall Executive Summary
This study, for the Nuclear Regulatory Commission (NRC), constitutes the state
of the art in terms of assessment of the seismic hazard for the locations of
all active nuclear power plant sites in the northern American region east of
the Rocky Mountains, which we refer to as Eastern United States (EUS).
Another similar study commissioned by the utility sponsored Electric Power
Research Institute (EPRI) is in progress whose results will be available
shortly.
Because of the importance of these two studies for the NRC, numerous reviews
and comparisons have been performed by both groups of investigators, and
generally it was found that the methodologies did not fundamentally differ in
their principles, but if differences existed in the results, they had to be
traced down to the various inputs used in the analyses. The results of this
study, performed by Lawrence Livermore National Laboratory (LLNL), provide the
NRC with the tools for characterizing the seismicity and ground motion in the
EUS. These tools are:
a. A data base of zonation and seismicity models of the EUS for
predicting the frequency of occurrence of earthquakes of any
magnitude (or intensity) at any location in the EUS.
b. A data base of ground motion attenuation models for predicting the
ground motion (in terms of peak ground acceleration (PGA) or the
response spectral amplitudes with 5% of critical damping - (PSRV) at
a site, when the size and location of the earthquake are known.
c. A seismic hazard model which, given the data bases in (a) and (b)
above, provides an estimate of the probability of exceedance, at the
sites, of any value of the ground motion (PGA and/or PSRV).
d. A data base of estimates of the seismic hazard at the 69 sites with
either presently operating nuclear power plants or plants seeking a
license.
Numerous studies, including several by NRC, have demonstrated the inherent
uncertainty in estimating the seismic hazard in the EUS. Thus one important
aspect of the methodology in this study was to attempt to capture this
uncertainty, including both the random (physical) uncertainty and the modeling
(knowledge) uncertainty. To this effect, expert opinion was used to develop
the data bases described in (a) and (b) above. Two panels of experts were
formed whose composition was carefully chosen to include experts with
knowledge spanning the entire EUS, academics, utility consultants, and the
United States Geological Survey (USGS), in the fields of geotectonics and
seismicity features of the EUS and ground motion modeling.
The methodology and description of the data bases are given in a separate
volume (Vol. I), and the results of the analysis are given in five other
volumes for the 69 plant site locations. Of these 69 sites, 38 were with rock
conditions, 14 were deep soil sites, and 17 were shallow soil sites. In
xxi
addition, 12 of the rock sites also had areas with additional power plant
units or other important buildings or constructions which were located on
shallow soil .
Our methodology handles the various soil conditions by characterizing each one
of them in either one of two ways:
(1) The soil site conditions are one of three: rock, deep soil or
intermediate
(2) The soil site conditions are one of eight: rock, deep soil, shallow soil
till-like (with three different depths) or shallow soil sand-like (with
three different depth).
The uncertainty in the hazard is estimated by a Monte Carlo simulation. All
the uncertain parameters (zonation, seismicity, ground motion models and site
correction models) are simulated, and several percentiles in the hazard are
calculated.
The results presented in this study consist of hazard curves at each site for
PGA for the 15th, 50th and 85th percentiles, constant percentile hazard curves
(CPHC), arithmetic mean (AMHC) and best estimate hazard curves (BEHC).
Uniform hazard spectra for various return periods are also presented.
We compared the
(Bernreuter et
studies were mi
data bases, inc
in the study,
zonation and se
minimal . On th
which were not
developed yet,
changes by the
was primarily t
higher spectral
smaller values
results of this analysis with one of our previous studies
al., 1985) for 10 test sites. The differences between the two
nor "tuning" on the methodology and a complete review of the
luding a round of feedback with all the experts participating
In spite of some changes in the opinions of several of the
ismicity experts, the effects on the results were found to be
e other hand, the random vibration (RV) ground motion models
used in our previous studies because they had not been
took a 50% weight in the present study, as a result of opinion
Ground Motion (GM) Expert Panel. The effect of these changes
change the average response spectral shape to allow for much
values at the higher frequencies (10 to 25 Hz) and relatively
at lower frequencies (1 to 2 Hz).
The following is a set of general conclusions drawn from the entire study, for
the most part described and summarized in Vol. VI of the report.
(1) There is substantial uncertainty in the estimate of the hazard. The
typical range in the value of the probability of exceedance between the
15th and 85th percentile curves for the PGA is on the order of 40 times,
for low PGA; it is more than 100 at high PGA values. This translates into
an approximation factor of 4 in ground motion for the 15th-85th range of
values in the PGA given a fixed return period, and similarly an
approximate factor of 4 in the ground motion for the range of values in
the PSRV for a given return period.
The range between the 15th and the 85th percentile hazard curves
represents the total uncertainty in estimating the seismic hazard at a
site due to two sources of uncertainty:
xxn
the uncertainty of each expert in the zonation, models and values of
the parameters of the analyses
the variation in the hazard estimates due to the diversity of
opinions between experts.
The latter, or inter-expert variation, is an important contributor to the
total uncertainty in the estimated hazard. Specifically, the magnitude of
uncertainty introduced by the diversity of opinions between experts is of
the same order, on the average, as the uncertainty in the hazard due to
the uncertainty of an individual expert in the value of the parameters.
However, at times the uncertainty between experts can be very large.
For a given acceleration value, the range of the median hazard values at
all the sites analyzed falls within the 15th-85th percentile range of any
one of those sites.
(2) The 50th percentile CPHC appears to be a stable estimator of the seismic
hazard at the site. That is, it is the least sensitive to changes in the
parameters, when compared to the other estimators considered in this study
(arithmetic mean, best estimate, 15th or 85th percentiles).
(3) The process of estimating the seismic hazard in the EUS is reasonably
stable. Comparison with our previous results indicated that there has not
been a major shift in results over the past few years, although there have
been some significant perturbations in the form of recent occurrences of
EUS earthquakes and the completion of several major studies of the
seismotectonics of the EUS. In the feedback performed in this study,
there were some changes introduced by members of both the Seismicity and
GM Panels. The computed hazard when aggregated over all experts did not
significantly change. However, the introduction of the "new" random
vibration models introduced a significant change in the spectral shape by
raising the spectral values in the high frequency range and lowering it in
the low frequency range.
(4) It is difficult to rank the uncertainties becaus
parameters of the recurrence models are hard to
our results indicate that the uncertainty in zon
models are more significant than the uncertainty
seismicity parameters. The largest contribution
comes from the uncertainty of the ground motion,
site effects is a significant contribution to th
introduced by the ground motion models. However
uncertainty introduced by zonation and recurrence is also significant and
of the same order of magnitude.
e zonation and the
separate. Nevertheless,
ation and ground motion
associated with the
to modeling uncertainty
The correction for local
e overall uncertainty
as already noted, the
(5) Based on comparisons between the results of our broad generic study and
site specific studies, we concluded in Bernreuter et al. (1985) that the
scale of our study is adequate. No major differences in zonation or
results occurred between our study and site specific studies.
xxm
(6) We found, consistent with the conclusions in Bernreuter et al. (1985),
that generally earthquakes in the magnitude range 3.75 to 5 would
significantly increase the estimated seismic hazard if they were included
in the analysis. Thus, it may be important to keep in mind that the CPHCs
and CPUHS presented in this study only include the contribution from
earthquakes with magnitudes of 5 and greater when assessing the seismic
safety of brittle components of nuclear power plant systems, e.g., such as
relays. In addition, it must be kept in mind that the PGA value is not a
good estimator of the loading that very stiff components will experience
in the EUS. The actual ground motion will be amplified.
(7) We found that the correction for the site's soil category had an important
effect on the estimated hazard. We concluded that the approximate
correction to be applied to the estimated hazard for rock site to estimate
the hazard for shallow soil conditions at the same site (the correction is
done by multiplying each PGA value on the rock curve, for a given
probability by a constant correction factor) would lead to an estimate of
the hazard of the soil site within less than 13% of its actual value had
we calculated the hazard for the soil site by our methodology. However,
we found that for some sites, multiplying the median hazard curve for rock
by the median correction factor would have given approximately the same
median hazard curve we obtained by performing the full analysis with our
probabilistic correction factors. Unfortunately, at the present time, we
have not been able to develop criteria to identify when performing such
operation is correct.
(8) Although the soil site correction is not region dependent, we found that
other complex interactions, with zonation seismicity and ground motion
models, made the site correction actually region dependent.
(9) We found that the input from some experts lead to either high or low
estimates of the hazard at most sites. In particular G-Expert 5's input
lead to results, in general, higher than when only the other 4 GM Experts'
input is used.
We found that the impact from any S-Expert did not show a consistent
deviation from the results of all the other S-Experts at all sites,
however, the results from some of the S-Experts were found to be either
high in some regions of the EUS (i.e., S-Expert 2) or low (i.e.. Expert
12, especially in the South West and Central U.S.).
Finally, it is difficult to assess if our results have either a conservative
or unconservative bias. We insisted that our panel members not introduce such
biases in their inputs and we spent considerable effort in developing a
methodology which would allow the experts to properly express their
uncertainty without having to introduce some conservative approximations.
This was particularly true in the area of regional ground motion modeling and
in the incorporation of multiple alternatives to account for any local site
amplification of the ground motion.
xxi V
V^^^^SV■,^VVf^^
Executive Summary: Volume I
This Volume is the first of eight volumes. It provides an overview of the
methodology that we developed as part of the project to incorporate the
judgement of experts and their quantification of the uncertainties about their
input into a seismic hazard analysis using a simulation approach. Our
methodology relies heavily on our previous work, Bernreuter et al. (1985) and
It highlights the improvements made in the final version of our methodoloav
and computer codes. -^
In Section 2 and Appendix C we provide the basis for the seismic hazard
analysis computer programs we developed to account for the uncertainty in the
estimation of the seismic hazard at all EUS nuclear power plant sites using
input from experts. This software incorporates a simulation approach to
provide the experts with a relatively simple but effective way to express
their uncertainty. The hazard methodology is based on a probabilistic model
of the occurrence and distribution of magnitudes of earthquakes and the
attenuation of the ground motion from a source to a site. It also includes
modeling of local site effects.
In Section 3 we describe our elicitation process which was developed to qive
each expert sufficient flexibility to define his best estimate for the various
parameters of interest, e.g., zonation, and to fully express his
uncertainty. As part of our elicitation process, we assembled two groups of
experts. One group, called our Seismicity Panel (S-Panel), is composed of
experts in the seismic zonation of the EUS. The other group is composed of
experts in ground motion estimation and we referred to them as our Ground
Motion Panel (G-Panel). Our elicitation process also included a number of
feedback loops to provide the experts with an understanding of the
implications that their assumptions and models have on the computed hazards
and to provide them with a simple formal way to update or change their
input. In Section 3 and Appendix B we provide a summary of the final makeup
of our panels and their input as it was used in our analysis.
'^?^^Moo ^° ^^^ ^""^"^ ^^^^ documented in our previous report, Bernreuter et
al. (1985), the Ground Motion Experts extensively revised their input. Of the
eleven S-Experts, three of them proposed totally new seismicity models, four
of them retained their original model and the rest made minor changes.
In Section 4 a special effort is made to highlight the differences in results
between our previous study, Bernreuter et al. (1985), and the present study.
The important conclusion is, again, the stability of the seismicity
modeling. We compared the new seismicity models with the previous ones by
calculating the seismic hazard at the same ten sites and using the same ground
motion models. The differences were quite small after combining over all the
experts in spite of the fact that the results for some S-Experts changed
significantly. ^
The final spectral ground motion models are significantly different from the
previous (1985) set of models. This is due primarily to the fact that a new
type of models have made its appearance, namely the random vibration,
stochastic models.
XXV
Another important difference between the results of this analysis and the
previous one, Bernreuter et al., (1985), is in the value of the minimum
magnitude of the earthquakes contributing to the hazard which is magnitude 5
in the present study as opposed to 3.75 previously. The effect of this change
is documented in the following Volumes II to V, by providing the best estimate
seismic hazard created by the earthquakes of magnitude between 3.75 and 5.
Volume I contains in Appendix A the responses of the experts to the
questionnaire on documentation. Appendix B provides all the final input data
for the analysis, and Appendix C gives a detailed account of the methodology
used in the analysis.
xxvi
SECTION 1: INTRODUCTION
The impetus for this study was in a large part the
discussions with the U.S. Geological Survey (USGS)
proposed clarification of their past position with
Charleston earthquake. The USGS clarification was
1982, in a letter to the NRC, which states that:
result of the NRC's
regarding the USGS's
respect to the 1886
issued on November 18,
Because the geologic and tectonic features of the Charleston region are
^u'"^! i^u° ^^°^^ ^" °^^^^ regions of the eastern seaboard, we conclude
that although there is no recent or historical evidence that other regions
have experienced strong earthquakes, the historical record is not, of
Itself, sufficient ground for ruling out the occurrence in these other
regions of strong seismic ground motions similar to those experienced near
Charleston in 1886. Although the probability of strong ground motion due
to an earthquake in any given year at a particular location in the eastern
seaboard may be very low, deterministic and probabilistic evaluations of
the seismic hazard should be made for individual sites in the eastern
seaboard to establish the seismic engineering parameters for critical
facil ities."
In response to this letter the NRC started several projects, one of which is
this study, to assist them in assessing the implications of the USGS position.
The objectives of this study are:
1. To develop a seismic hazard characterization methodology for the entire
[!^^nM.°t *^^ ^^"^ted States east of the Rocky Mountains (referred to as
the EUS in this report).
2. To apply the methodology to all nuclear power plant sites East of the
Rocky Mountains to assist the NRC staff in their assessment of the
implications in the clarification of the USGS position on the
Charleston earthquake, and the implications of the occurrence of other
recent eastern U.S. earthquakes.
In this study we developed a methodology which allows experts to express their
uncertainty about the seismotectonics of the EUS and to incorporate the
experts uncertainty into our analysis. We also provide a representative
sample of expert judgment about the seismotectonic parameters that influence
the estimates of the seismic hazard, in the form of strong shaking induced by
future earthquakes, at any particular sites. y y j'
The methodology that we developed as part of this study has been discussed in
detail in Bernreuter et. al . (1984,1985). However, for completeness and ease
of reference in Section 2 of this report we provide an overview of our
approach. In Appendix C, we also provide details of our methodology needed
tor an in depth understanding our our approach and results. In Section 3 we
review how the inputs needed for the analysis were obtained. The inputs used
-1-
in the analysis are given in the Appendix B of this report. In Section 4 we
compare our updated results to our previous results given in Bernreuter et al .
(1985).
Because of the large number of sites for which results are presented we have
broken them up into four batches of roughly the same number of sites. The
results for each region are presented in separate volumes- Volumes II-V.
Table 1.1 lists the sites contained in each volume and Figs. la,b,c, and d
give the location of each site. In selecting the sites for each batch an
attempt was made to make a logical regional grouping of approximately the same
size. Some compromises had to be made, particularly in batch 4. It can be
seen from Figs, la and b that the sites in batch 1 correspond to the Northeast
and batch 2 to the Southeast. Batch 3, as can be seen from Fig. 1.1c, covers
the central part of the North Central region and Batch 4, as noted, is a
potpourri of the remaining sites as can be seen from Fig. 1.1. d. There is a
suimiary volume (Vol VI) which compares the results from the four regions and
contains our overall conclusions and recommendations. Finally, in Vol. VII we
provide all of our Questionnaires. To make it easier for the reader interested
in only the results for a few sites, we have made volumes II-V independent of
each other. Thus the reader only needs volume I which, as outlined above,
gives the details of our methodology and inputs used in the analysis; the
volume which has the results for the site of interest; and Volume VI which
draws general conclusions and provides added sensitivity discussion. Needless
to say in the lay out here is some considerable repetition of information in
volumes I-V.
We started work
development of
Bernreuter et a
our methodology
update their in
(1985). The El
program in 1983
funded us to pe
results at the
the comparison
Section 4 of th
developing our methodology in 1982.
the inputs in 1983 and published our
1. (1984). After extensive peer revi
. We held extensive feedback session
put. We published our updated result
ectric Power Research Institute (EPRI
and provided their preliminary resul
rform an in depth comparison between
nine sites and EPRI's methodology and
were published in Bernreuter et. al (
is report.
We completed our initial
preliminary results in
ew we made improvements to
s with our experts to
s in Bernreuter et al .
) started a similar
ts in EPRI 1985. NRC
our methodology and
results. The results of
1987) and are discussed in
Because of the somewhat long time delay between our last update, completed in
1984 and because many of our experts were also members of the six EPRI Teams
we undertook a final round of feedback to update the experts opinions.
The results presented in this report are based on the updated responses by our
panel members from both the Seismicity and Ground Motion Panels. In this
sense they are final results. However, as judgment plays a very significant
role in developing the input data, it is likely, considering the large
uncertainties expressed by each our our experts, that in the future various
experts will modify their views thus leading to results which may differ from
those presented here.
-2-
TABLE 1.1
List of Sites in Each Volume of this Report
Batch 1 - Sites in Vol. I I
Plotted in Fig. 1.1a
Plot
Symbol
1
2
3
4
5
6
7
8
9
A
B
C
D
E
F
G
H
I
J
Site
Fitzpatrick
Ginna-1
Haddam Neck
Hope Creek
Indian Point
Limerick
Maine Yankee
Millstone
Nine Mile Pt.
Oyster Creek
Peach Bottom
Pilgrim
Salem
Seabrook
Shoreham
Susquehanna
Three Mile Island
Vermont Yankee
Yankee at Rowe
Plot
Symbol
1
2
3
4
5
6
7
8
9
A
B
C
D
E
F
G
H
Batch 2 -Sites in Vol. Ill
Plotted In Fig. 1.1. b
Site
Bellefonte
Browns Ferry
Brunswick
Calvert Cliffs
Catawba
Farley
Hatch
McGuire
North Anna
Oconee
Robinson
Sequoyah
Shearon Harris
Summer
Surry
Vogtle
Watts Bar
-3-
Plot
Symbol
1
2
3
4
5
6
7
8
9
A
B
C
D
E
F
G
Batch 3 - Sites in Vol. lY
Plotted in Fig. 1.1. c
Site
Beaver Valley
Big Rock Point
Braidwood
Byron
Clinton
Cook
Davis Besse
Dresden
Fermi
Kewaunee
LaSalle
Palisades
Perry
Point Beach
Quad Cities
Zion
Plot
Symbol
Batch 4 - Sites in Yol.Y
Plotted in Fig. l.ld
Site
1
2
3
4
5
6
7
8
9
A
B
C
D
E
F
6
H
Arkansas
Callaway
Comanche Peak
Cooper
Crystal River
Duane Arnold
Fort Calhoun
Grand Gulf
LaCrosse
Monticello
Prairie Island
River Bend
South Texas
St. Lucie
Turkey Point
Waterford
Wolf Creek
-4-
vS:Cs«vsv
Figure 1.1a Map showing the location of the Batch 1 sites contained in Vo"
il of this report. Map symbols are given in Table 1.1.
-5-
Figure 1.1b Map showing the location of the Batch 2 sites contained in Vol,
III of this report. Map symbols are given in Table 1.1.
Figure 1.1c Map showing the location of the Batch 3 sites contained in Vol
IV of this report. Map symbols are given in Table 1.1.
-7-
Figure l.ld Map showing the location of the Batch 4 sites contained in Vol. V
of this report. Map symbols are given in Table 1.1.
■8-
SECTION 2: OVERVIEW OF THE METHODOLOGY USED
2.1 Introduction
spars ity of
the Eastern
study was to
could be
Seismic hazard analysis has been limited by the poor quality and
the available seismicity and strong motion data, particularly in
United States (EUS). As noted in Section 1, the purpose of this
develop a methodology and a data base so that the seismic hazard
estimated at any site in the EUS. Our methodology is the outgrowth of
previous studies, in particular, the Systematic Evaluation Program (SEP)
study, Bernreuter and Minichino (1983) and the Seismic Safety Margins Research
Program, Bernreuter et al. (1983) In the SEP study, we proposed the use of
experts' opinions to supplement the sparse data. The results of the SEP study
were point estimates of the seismic hazard. In this new study, referred to as
The Seismic Hazard Characterization (SHC) project (Bernreuter et al. (1984,
1985) in this report, we incorporated several significant modifications to the
SEP methodology. An important extension was recognition and inclusion of
uncertainties in the analysis and its inputs. Thus, the results of the SHC
include an estimate of the hazard with uncertainty bounds. Other
methodologies now exist to perform the same tasks. In particular the EPRI
study, EPRI (1985, 1986), sponsored by the utility companies, is similar in
many respects to the SHC study. It, like SHC, combines experts' opinions with
historical data and includes uncertainties in the analysis.
This section describes the SHC study in some detail. Emphasis is placed on
eliciting the experts' opinions and the treatment of uncertainty.
2.2 Description of the Hazard Calculation Methodology
In the SHC study, the seismic hazard at a site is quantified by a seismic
hazard curve which describes the relation between the value of a ground motion
parameter, e.g., peak ground acceleration (PGA) and the probability it is
exceeded in one year. The methodology is similar, in many ways, to the well-
established methods developed by Cornell (1968), McGuire (1976), Algermissen
et al.(1982), Mortgat and Shah, (1979) and Der Kiureghian and Ang (1977). All
these studies involved four basic elements as described in Fig. 2.1:
Identification of seismic source zones (Fig. 2.1a).
A model describing the expected frequency as a function of
magnitude (Fig. 2.1b).
A model describing the expected value of a ground motion parameter
(e.g.., peak acceleration) as a function of the magnitude and
distance to the source (Fig. 2.1c).
Integration into a seismic hazard curve (Fig. 2. Id).
In the SHC, the ground motion parameters considered are the peak ground
acceleration (PGA) and the pseudo-relative velocity (PSV) of a 5% damping
response spectrum . We assume that the region affecting the ground motion at
the site can be partitioned into distinct areas of constant seismic
-9-
a) Geometry of homogenous source
zones (seismicity/tectonics)
Log N I ,
a —
b. Slope
»^M
(b) Magnitude recurrence model
(frequency vs size)
By integration
0.1 g 0.1 g
d) Seismic hazard curve
LogY
Log Y n
Ground
motion
intensity
' '
a, Dispersion
•►R
Log distance
(c) Ground motion prediction
model (attenuation)
Figure 2.1 Four steps involved in a probabilistic seismic hazard analysis.
■10-
characteristics (referred to as source zones). This partition is partly based
on geophysical information accumulated by each expert (such as tectonic
stresses, plate motions, geology) and partly on observed seismicity developed
by the individual expert's analysis of earthquake catalogs.
The following assumptions about the occurrence of earthquakes throughout the
EUS form the basis for the probability calculations.
Earthquakes occur randomly over time and space within a source zone.
Earthquakes are point sources, thus the fact that they are created by
rupture of tectonic faults is neglected.
The occurrence of earthquakes is independent between source zones.
The occurrence rate of earthquakes within a source zone is constant;
its value describes the seismic and tectonic conditions that
presently exist within the zone.
The expected number of earthquakes of magnitude m or greater, A(m),
per unit of area occurring within a zone is described by the
magnitude recurrence relation
log A(m) = H(m) Mq< m j< My
(2.1)
where Mq is the minimum magnitude of interest and My (upper magnitude
cutoff) is the maximum magnitude possible in the zone under the
present tectonic conditions. My and the functional form H(m) are
elicited from each of the experts.
Given these assumptions, the number N^(m) of earthquakes with magnitude
greater than M, m>MQ, occurring within a zone in a time period of t years is a
Poisson random variable with intensity parameter A(m). Thus, the probability
of exactly n earthquakes with magnitude greater or equal than m in t years is:
P [N^(m) = n] = t ^^^"^)]" e-^^('"J ,n » 0,1,2,3,
n!
(2.2)
Using the assumption that earthquakes are point sources which occur uniformly
through a zone, if N^(r,m) is the number of earthquakes in t years of
magnitude greater than m occurring at points which are a distance r to r+dr
(kilometers) from the site, then N4^(r,m) is a Poisson random variable with
intensity parameter
m,t
A(m) f^{r) dr
(2.3)
where f^{r) is the density function for the distribution of the distance from
the site to points within a source zone.
Given an earthquake of magnitude M > m at a distance (r,r+dr) from the site,
the ground motion parameter, e.g. ,PGA, at the site depends on the attenuation
of the source energy between the source and the site. This is modeled as a
random process. The expected value of PGA is described by a ground motion
model depending on m and r. Since a multitude of such models exists, a panel
-11-
of ground motion experts was used in the SHC project to select appropriate
models. The conditional probability of PGA exceeding the value a, given m,r,
is denoted P(A>a|m,r), where A represents the peak ground acceleration.
Let N^(a) be the random variable, the number of earthquakes occurring in a
zone in t years such that the PGA at the site is greater than a. The
probability that one or more earthquakes occur in t years resulting in the PGA
at the site exceeding a, denoted P (A^ > a), is given by
P{^^ > a) = P(N^{a) > 0).
(2.4)
Given the range of magnitudes (Mq, My), where Mm is the upper magnitude cutoff
for the specific zone, and distances r>0, N^{a) is a Poisson random variable
with intensity parameter At where
M
U
M,
J P{A>a|m,r) fp(r) dr dA(m)
r>0 ^
(2.5)
such
unit
F
ea
that
time
dA(m) =
An dF^(m|MQ,My) and A.
, of earthquakes with ma(
is the expected frequency, per
and area, ^f e'alrthqtiak^s with mclgnitude exceeding Mq, and
l^(m|MQ,M|.) denotes the distribution function of magnitudes given an
rthquake, conditional on minimum magnitude Mq and upper magnitude cutoff My
The probability that the maximum PGA at the site exceeds a, in a time period
of length t, due to earthquakes occurring in zone q, is given by the
complement to the probability of no such events, i.e., using the Poisson
distribution.
Pq(V3) - Pq(Nt(a) > 0)
1 - exp (- A^qt)
(2.6)
where A
causing
, given by Eq. (2.5) is the expected number of earthquakes per year
^ a PGA greater than a at the site from earthquakes occurring in zone q,
The distance density fp(*) and magnitude distribution F('|Mf,,M..) are
dependent on the zone.
Finally, under the assumption that events between zones are independent, the
seismic hazard in t years at a site caused by earthquakes occurring in all
zones is given by:
P(A^>a) - 1 - 5[l-Pq(At>a)] - 1-g expf-A^^t)
(2.7)
In the SHC analysis, the ranges of magnitude and distance were discretized and
Eq. (2.5) was approximated numerically by a series of summations. Several
ground motion models and distributions were selected by the experts to model
-12-
the conditional probability P(A>a|m,r). Also the magnitude recurrence
relationship, Eq. (2.1), was modeled by either a linear or bilinear truncated
exponential relation, where the truncation was based on the model of Weichert,
(1980) or a relationship developed in the SHC project.
2.3 Uncertainty in the Hazard
The limited historical data, empirical models, which lead to the uses of
experts' opinions, caused the resulting hazard estimates to be uncertain.
This uncertainty needs to be identified and included in the description of the
seismic hazard. Thus, the hazard can be described not by a single curve, as
in Fig. 2. Id, but typically by envelopes of percentiles of the hazard as shown
in Fig. 2.2.
The approach developed for the SHC, is based on simulation to develop a
probability distribution of the hazard. Using a Monte Carlo approach, each of
the uncertain parameters is sampled a large number of times from its
respective probability distribution describing the uncertainty in the
parameter. With each hazard curve resulting from a given simulation is
associated a weight, or probability of being the true hazard curve, which is
calculated as the product of the probabilities or weights of each of the
random parameter values used in that simulation. For each pair of seismicity
and ground motion experts (respectively S- and G-Expert) described in
Section 2.4, a typical simulation is as follows:
levels of correlation with a, as specified by the
Draw a map from the distribution of maps for this S-Expert.
For each one of the seismic sources in a sample map, draw a set of
seismicity parameters from their respective distribution, i.e.,:
a value for the a parameter of the recurrence law
a value for the b parameter of the recurrence law (b is allowed
to have three
S-Expert)
the value of the upper magnitude (or intensity) cutoff
Draw a ground motion model from the distribution of models.
Draw a value for the random uncertainty parameter, which is
associated with the selected ground motion, for the appropriate EUS
region (NE, SE, NC or SC).
Draw a site correction method.
The hazard is calculated for each of the seismic sources and combined for all
sources. Each simulation gives a possible hazard curve. For each site
typically 2750 such curves (50 simulations per G-expert times 5 G-experts
times 11 S-Experts) were developed. Percentiles, usually the 15, 50 and 85th,
are then used to describe the uncertainty in the hazard. This method,
relative to discrete approaches such as, EPRI (1985, 1986) and YAEC (1973),
provides more flexibility by allowing for a wider range of distributions to
describe the uncertainties in the parameters. It also has the advantage of
better sampling the tails of the distributions.
-13-
. f^
E.U.S SEISMIC HAZARD CHARACTERIZATION
LOWER MAGNITUDE OF INTEGRATION IS 5.0
PERCENTILES = 15.. 50. AND 85.
>-
oe.
UJ
a.
u
2
<
CO
<
m
O
or
a.
-1
10
-2
10
-3
10
HAZARD CURVES USING ALL EXPERTS
UJ 4
u 10
O
>-
10
-6
10
-7
10
o
+
►O ■« ID (D
ACCELERATION CM/SEC»»2
BRA I DWOOD
Figure 2.2 Typical results of our seismic hazard analysis. Shown are the
15th, 50th and 85th percentile PGA hazard curves for the
Braidwood site.
■14-
2.4 Use of Experts' Opinion
The calculation of the hazard, described in Section 2.2, relies on the
availability of data to develop the seismicity and ground motion models used
in Eq. (2.5) i.e., the functions ^n(r), A(m) and P{A>a|m,r). Only limited
historical data are available for the EUS. Specifically, the earthquake
catalogs cover only 200 to 300 years at the most, and must be used to make
predictions in the range of 1000 to 10,000 years. Consequently, various
interpretations are possible and the scientific community offers a diversity
of opinion with respect to seismicity and ground motion prediction for the
EUS. An important aspect of the SHC was to recognize this diversity which
exists in the scientific community and to incorporate it in the uncertainty of
the hazard. Thus, in the SHC the inputs, i.e., parameter values and models,
for the hazard analysis were derived by eliciting experts' opinions in the
fields of seismicity modeling and ground motion prediction modeling. To this
end, two panels were formed. The S-panel was made up of experts on seismicity
and zonation, and experts on ground motion prediction formed the G-Panel. As
discussed in Section 3 the composition of the panels varied over the long
course of this study. The individuality of the experts was emphasized by
encouraging them to use their own information and data bases. The intent was
to avoid the screening of non classical interpretations. Thus, we avoided
favoring any kind of consensus among the experts including a consensus in the
raw data or in the modeling. The methodology developed here was not intended
to lead to some kind of artificial consensus, but rather the display of the
full range of opinions was to be retained. The opinions of the experts were
elicited through a series of written questionnaires, feedback meetings, and
feedback questionnaires.
2.5 Aggregation of Experts' Opinions
When the opinions of several individuals are to be elicited, it is frequently
necessary to consider ways of combining the information provided by the
individuals into a single statement which represents, in some way, the
"average" or consensus opinion of the group of individuals.
Basically, there are two classes of methods of aggregating experts'
opinions. One class of methods is based on pooling some normalized
quantification of the experts' opinions. In this case the experts are queried
individually, are not expected to interact, and no attempt is made to reach a
consensus through dialogue. Consensus is represented by the pooled
quantification of opinions. The emphasis is placed on independence and free
expression. The second class of methods attempts to reach a consensus. It is
based on group interaction in which the experts are allowed to interact, with
or without feedback, and through dialogue. In this case the free exchange of
information is expected to result in a reduction in the range of views (Genset
and Zidak (1984), Pill (1971) and Winkler(1968)) , thus, seemingly, to imply a
greater state of knowledge. However, unrestricted dialogue can be misleading
since agreement may have been a result of strategic manipulation,
intimidation, and other factors which could lead to biased results. To be
effective the interaction must be well-planned and carefully directed.
-15-
BKI
The method used in the SHC is based on the former method, pooling of
opinions. However, feedback, group interaction, extensive analysis of the
responses, checks for consistency and gross errors, and a peer review were
part of the overall elicitation process to alleviate some of the drawbacks
associated with complete anonymity of the experts (e.g.., lack of
responsibility, arbitrary answers).
Retention of the diversity of opinions between experts was an important
consideration in the SHC project. Thus, individual hazard curves were
estimated for each expert and the diversity of opinion between experts was
included in the description of uncertainty.
In the case of the SHC the hazard is calculated for every pair of experts
(i.e., S-Expert and G-expert) and these are subsequently combined. The
combination rule is based on a normalized weighted average of the hazard
curves or individual hazards in the uncertainty analysis. The weights for the
G-experts were normalized values of self-weights the experts provided. The
weights for the S-Experts were themselves a weighted average of four regional
self-weights provided by the S-Experts, i.e..
W3 = Iw^^P(A=AJ.
w
(2.8)
where w is the single weight for the s-th expert, w^ is the self-rating of
the s-th expert for region w, and P(A-A^) is the probability that the maximum
PGA at the site results from an earthquake originating in the w-th region.
The four regions are indicated on Fig. 2.3. An appealing property of Eq.'
(2.8) is that it will provide a "high" value for w^ if the self-weight is
highest in the region with highest probability of producing the maximum PGA.
Conversely, it will be low if the weight is highest in the region with the
lowest probability of producing the maximum PGA.
2.6 Hazard Analysis Outputs
Generally, the hazard at a site has been described in terms of a hazard curve ,
i.e., a graph of the probability that within a period of one year the maximum'
value of a ground motion parameter, e.g., peak ground acceleration or
velocity, will exceed a given level, say A, as a function of a. A number of
different estimators of the seismic hazard exist and are described in detail
in Appendix C. One estimator produced by the SHC methodology is referred to
35 the best estimate hazard curve (BEHC). This is the hazard curve, for a
particular pair of seismicity and ground motion experts, based on using the
best estimate (BE) models and parameter values given by the experts. This
corresponds to the hazard curve that would be produced if only a single source
of seismicity and ground motion information were available and no uncertainty
information were elicited. The BEHC is not necessarily the "best estimator",
but is simply one possible estimator of the seismic hazard at a site. Figure
2.4 gives a typical set of BEHC for PGA at the Braidwood site. For the case
shown, S-Expert 1 was used and the five BEHCs (one for each G-Expert) are
plotted. It should be noted that G-Experts 3 and 4 both had the same BE PGA
ground motion model hence their BEHCs dre the same. See Appendix C for
further discussion.
-16-
Limit of this
analysis
Figure 2.3 Map indicating the boundary of the four regions of the EUS used
by our S-Experts to determine their self weights and by the G-
Experts to select ground motion models.
-17-
EUS SEISMIC HAZARD CHARACTERIZATION, SEPT 1987
LOWER MAGNITUDE OF INTEGRATION = 5.
q:
<
UJ
cc
Ul
a.
u
z
<
10
-2
10
-3
10
BEST ESTIMATES FOR SEISMIC EXPERT 1
HAZARD CURVES BY ATTENUATION EXPERT
o 10
CD
<
m
o
a.
10
10
-7
10
ACCELERATION CM/SEC* 'Z
BRA I DWOOD
Figure 2.4 BEHC for PGA per G-Expert for S-Expert I's zonation and
seismicity parameters.
A second type of estimator produced by the SHC methodology, referred to as
constant percentile hazard curve (CPHC), is based on using the uncertainty
information provided by the experts. As explained in Appendix C, the
percentiles derived here are the actual percentiles in the data sample
obtained in the Monte Carlo simulation. It is not the percentiles one would
obtain by fitting a probability distribution on these data. A probability
(uncertainty) distribution for the hazard at each value, a, is developed by
treating all the input models and parameters as uncertain variables and using
simulation combining the percentiles of the hazard over all levels (over the
range of a) gives a CPHC. The 15th, 50th and 85th CPHC's were most often used
in the LLNL studies. Just as the BEHC, the CPHCs can be produced for each S-
Expert and G-expert pair. Such curves describe the uncertainty expressed by a
particular pair of experts. CPHCs for each S-G-Expert pair were not
produced. CPHCs were produced for each S-Expert for all of the G-Experts. A
typical set is plotted in Fig. 2.5 for S-Expert 1. Of most interest are the
CPHCs obtained by aggregating over all S and G-Experts producing an
uncertainty distribution for the hazard which describes both experts'
uncertainties as well as diversity of opinions between experts.
In addition to generating a BEHC
methodology includes aggregation
curves were based on using the s
discussed in Section 2.5. One 1
BEHC over ground motion experts
a typical set of BEHCs, one for
These aggregated curves can also
second level of aggregation. An
in the LLNL methodology. It is
(AMHC). In Fig. 2.7 we compare
and AMHC for the Braidwood site.
for each pair of S and G-experts, the
s of curves. Such combinations of hazard
elf-weights provided by the experts, as
evel of aggregation consists in combining the
for each seismicity expert. Figure 2.6 gives
each of the S-Experts at the Braidwood site,
be combined over seismicity experts to form a
additional aggregated estimator is considered
the arithmetic weighted average of the hazards
the aggregated (over all S and G-Expert) BEHC
As was found in the analysis, the probability density function of the
probability of exceedance of a given ground motion value is, in general, close
to a lognormal probability distribution. Thus the arithmetic mean is expected
to be in general closer to the 84th percentile, and much higher than the best
estimate which is closer to the median (for a lognormal distribution, the
median or 50th percentile is also the mean of the logarithms of the
probability of exceedance).
-19-
-1
10
-2
10
LOWER MAGNITUDE OF INTEGRATION = 5.
15.0. 50.0. AND 85.0 PERCENTILES
HAZARD CURVE FOR SEISMIC EXPERT 1
<
Q- 10
2
l!J -4
o 10
o
>-
CD
<
m
o
or
a.
10
-6
10
-7
10
ro •<» IT)
ACCELERATION CM/SEC* '2
BRA I DWOOD
Figure 2.5
15th 50th and 85th CPHC for PGA for S-Expert 1 aggregated over
all G-Experts for the Braidwood site. yy^teydtea over
•20-
E.U.S SEISMIC HAZARD CHARACTERIZATION
LOWER MAGNITUDE OF INTEGRATION IS 5.0
-1
10
-2
10
^^
oe.
<
>-
a. 10
<
o
uj -4
u 10
O
-J 10
CD
<
CD
O
q:
a.
BEST ESTIMATE
FOR THE SEISMICITY EXPERTS
-6
10
-7
10
O
+
■* ID ID r~
ACCELERATION CM/SEC* *2
BRA I DWOOD
'''"" '•' BraldCd'^it^"' '''''''''' °''' ^1' «-^^P-'^ fo^ PGA for the
-21-
E.U.S SEISMIC HAZARD CHARACTERIZATION
LOWER MAGNITUDE OF INTEGRATION IS 5
-1
10
HAZARD CURVES USING ALL EXPERTS
ACCELERATION CM/SEC"2
BRA I DWOOD
Figure 2.7 Comparison of the BEHC and
rZri rT / . ^^^^ '"^ ^^"^ ^0^ PGA aggregated over all
S and G-Experts for the Braidwood site.
-22-
SECTION 3: DEVELOPMENT OF HAZARD ANALYSIS INPUTS
3.1 Processes Used for the Elicitation of Expert Opinion
There are a variety of ways in which expert opinion may be elicited (Mensing,
(1981)). Our approach combines several different methods. It is
characterized by the following key features:
Two panels of experts were formed. The S-panel provided input for
the zonation and seismicity of the EUS and the G-panel provided input
for the ground motion attenuation from EUS earthquakes, (see Tables
3.1 and 3.2)
Detailed questionnaires, requiring several days of effort by the
panelist to complete, were distributed.
Panel members were generally paid.
Follow-up discussions and a feedback meetings were held for each
panel .
The responses of each panel member were used in a separate hazard
analysis and combined at the last step with other experts.
The elicitation process and hazard analysis methodology were subject
to peer review.
Additional formal feedback loops were performed to finalize the input
data.
In designing the elicitation process one of our guiding principles was to make
sure that all experts had complete flexibility to develop their resources and
opinions independent of the other panelists. We wanted everyone to function
independently in formulating their opinions. Thus, we did not attempt to
structure the experts line of thinking about the issues relevant to EUS
seismicity. This allowed them the flexibility to use analytical methods as
well as personal intuition and insight to the degree they felt appropriate.
Overall, we wanted to assure that everyone could express their opinions
without regard whether a consensus was being formulated among the
participants. That is, we wanted everyone to feel free to express their
opinions, even if they differed from the opinions of the other panelists.
Thus, we wanted to be able to capture the range of opinions that might exist
among knowledgeable individuals. We believe that our elicitation process has
followed this principle. Our elicitation procedure was based on the experience
gained during the SEP study and incorporates suggestions made by both the SEP
Peer Review Panel, Bernreuter and Minichino (1983) and the SSMRP Panel on
Subjective Inputs, Bernreuter et al. (1983) as well as other reviewers'
comments.
Initially fourteen well known geoscientists knowledgeable about the seismicity
and tectonics of the Eastern and Central U.S. formed one panel called the
S-Panel (EUS Seismicity Panel), the list of whom is given in Table 3-1.
Drs. Stevens and Wentworth subsequently resigned from the panel after
providing us with their zonation maps. Dr. Basham resigned after providing
his seismicity parameters, limited to Canada thus making his data incomplete
for use in our analysis. However he participated in the zonation seismicity
-23-
•-%>;
feedback meeting, providing many useful inputs and generating discussions on
the seismicity of Canada and the North East of the United States with the
other panel members. The remaining eleven experts provided input to develop
the overall earthquake occurrence model.
The second panel, the G-Panel (the EUS Ground Motion Modeling Panel),
initially included five members. Professor Nuttli left the G-panel in the
summer of 1986, due to illness, he died in 1988. Drs. Dwyer and Anderson were
added in Fall of 1986. Professor Toksoz attended our final workshops and
contributed to the discussion, but because of other pressing commitments was
unable to complete the final ground motion questionnaire, and thus was dropped
from the final analysis as a G-Expert, but remained as one of the S-Experts.
The list of members in the G-Panel (G-Experts) is given in Table 3-2.
We investigated the impact of this loss of continuity in the G-Panel make-up
by comparing the results presented in our previous reports Bernreuter et al
(1984, 1985) and this report (see Section 4).
As can be seen in the flow chart of Table 3.3, considerable interaction, both
formal and informal, took place between LLNL and the expert panel members.
However, following our approach at no time during the elicitation were the
experts forced or even encouraged to reach a consensus. As previously
discussed, this study was designed as an expert opinion sampler. Thus
initially we limited the amount of common information that we provided our S-
Experts to a sorting of the earthquake per their zonation. Only at the second
feedback level, Q9, did we provide them estimates of the a and b-values. The
SHC IS, as discussed in Bernreuter et al (1987) and Section 4, conceptually
different from other current studies, such as the one sponsored by the EPRI,
whose goals are to try and reach a consensus of opinion at some levels in the
analysis .
Our goal in eliciting subjective judgment in the manner outlined in Table 3-3
was twofold. First, we believed it would give an accurate representation of
the experts' views about parameters that affect seismic hazard. Second, it
enabled us to retain the diversity of opinion which may exist in the
scientific community. Ten questionnaires were designed and sent to the
experts in order to collect all the necessary data for the analysis. They are
the fol lowing: ^
Questionnaire 1
Questionnaire 2
Questionnaire 3
Questionnaire 4
Questionnaire 5
Questionnaire 6
Questionnaire 7
Questionnaire 8 •
Questionnaire 9 •
Questionnaire 10
Zonation Questionnaire (Ql)
Seismicity Questionnaire (Q2)
Questionnaire on Regional Self Weights (Q3)
Ground Motion Models Questionnaire (Q4)
Feedback-1 Questionnaire on Zonation/Seismicity (Q5)
Feedback-1 Questionnaire on Ground Motion Models (Q6)
Feedback-2 Zonation Questionnaire (Q7)
Documentation Questionnaire (Q8)
Feedback -2 Seismicity Questionnaire (Q9)
- Feedback -2 Ground Motion Questionnaire (QIO)
■24-
Questionnaires Ql , Q2, Q3, Q5, Q7, Q8 and Q9 pertain to the S-Experts on
zonation and seismicity. Q4, Q6 and QIO pertain to the 6-Experts. A copy of
these questionnaires is given in the Vol. VII of this report, in the form as
they were sent to the experts.
The original plan was to compute the seismic hazard at all EUS sites using the
methodology and expert input data as it existed with the experts' responses to
Q1-Q6. However, as the initial EPRI results were becoming available, it was
deemed worthwhile to compare the results of the SHC and EPRI studies and to
assess the meaning of the differences in estimates of the seismic hazard
between the two studies at the test sites before continuing the analysis for
all EUS sites. The results of this comparison are given in Bernreuter et al.
(1987). Because of the somewhat long delay between the answering of Q1-Q6 by
our experts and the starting of our final computations, new data and studies
had become available particularly in the area of ground motion modeling. In
addition many of our experts also participated in the EPRI study and attended
the EPRI workshops on modeling the seismicity of the EUS and the ground motion
modeling in the EUS. Thus NRC funded a second feedback round allowing experts
to update their answers to the questionnaires.
In the following sections, we briefly describe the intent and highlights of
the Questionnaires. In each case we desired not only an expert's opinion
regarding the "most probable value" of a parameter but also, whenever
possible, a measure of his uncertainty in determining the value of the
parameter.
The experts of both the S and G Panels were instructed to avoid cognitive
biases insofar as possible. For example, three points were emphasized:
Answers were to be based on experience, geologic, tectonic and
geophysical considerations, and all other available data.
The level of confidence each expert placed in his answers would be
explicitly considered. Therefore, since his input would undergo
filtering and weighting when combined with the opinion of other
experts, the expert was asked not to feel reluctant to express
nonclassical viewpoints.
The experts were urged to attempt answering all questions.
-25-
Notes
TABLE 3-1
EUS ZONATION AND SEISMICITY PANEL MEMBERS
(S-Panel)
Dr. Peter W. Basham^^)
* Professor Gilbert A. Bollinger^l)
* Mr. Richard J. Holt^^)
* Professor Arch C. Johnston
* Dr. Alan L. Kafka
* Professor James E. Lawson
* Professor L. Tim Long^^)
* Professor Otto W. Nuttl i (1)&(4)
* Dr. Paul W. Pomeroy^^)
* Dr. J. Carl Stepp
Dr. Anne E. Stevens^-^)
* Professor Ronald L. Street^^)
* Professor M. Nafi Toksoz ^^^^^"^^
Dr. Carl M. Wentworth^^)
(1) Also participated in the SEP Panels
(2) Only provided zones and seismicity parameters for Canada
(3) Only provided zonation-no seismicity parameters
(4) Also member of the Ground Motion Panel (Table 3-2)
(*) Final member of the S-Panel
-26-
TABLE 3-2
EUS GROUND MOTION MODEL PANEL MEMBERS
(G-Panel)
* Dr. David M. Boore^^^
* Dr. Kenneth Campbell
Professor Otto W. Nuttli
Professor Nafi Toksoz ^^'
* Professor Mihailo Trifunac^^
* Dr. John Anderson {^'
* Dr. John Dwyer ^ '
(1) (2) (3)
Notes:
* Provided the final sets of ground motion models used
(1) Participated as a member of the SEP EUS Ground Motion Panel.
(2) Also member of the Seismicity Panel (See Table 3-1), did not
complete QIO.
(3) Left the Panel in June 1986
(4) Added to the Panel in the Fall of 1986
-27-
Operations performed by
the expert members of
the ground motion panel
Operations performed
by LLNL
Operations performed by
the expert members of the
zonation/seismic 1 ty panel
Define methodology
1982
Design Ql
quest lonnaire
on zonation
Finalize interim
zonations ,
Check data
Design 02 and Q3
questionnaires on
seismici ty and
self weights
Finalize interin
seismic 1 ty
Design Q4
questionnaire on
ground motion
models
Finalize interim
ground motion
models ,
Check data
Obtain interim
resul ts of
hazard
♦ Mostly by phone or mail, but also meeting in person for a feu cases
Table 3.3. Schematic representation of the flou of operations in
the elicitation of the Experts' opinions.
-28-
Operations performed by
the expert members of
the ground motion panel
Operations performed
by LLNL
Operations performed by
the expert members of the
zonation^seismic 1 ty panel
Obtain interim
results of
hazard
FEEDBACK MEETING
• PI I S-Experts
• LLNL
• NRC
(Dec , 83)
Update
methodology ,
Update codes and
data files
Design 05 feedback
questionnaire on
seismic ity and
self ueights
Update zonations,
seismic 1 ties, and
self ueights
'
'
FEEDBACK MEETING
• fill G-Experts
LLNL
• NRC
•
(/)
4J
&.
a>
CL
X
LU
CO
L.
3
O
E
o
&-
M-
Q.
(^
E
c
o
•r- '
•
■4->
.— 1
03
C
4->
o
t.
N
(U
CL
flB
X
LU
«4-
1
O CO
Ol
E
r—
O
i-
4-
(O
X
Q.
Ol
»o
E
n*"
«o
LU
o
CO
•r"
Q.
(U
."^■^
\-
+->
3
C3>
-33-
(U
a.
X
LU
I
00
>-.
X)
■a
Qi
X3
•^
>
o
L.
Q.
0)
c
o
N
>
•r-
C
s.
CM
ro
(U
S-
cn
34
Key to Site Index Numbers
1.
Limerick
2.
Shearon Harris
3.
Braidwood
4.
La Crosse
b.
River Bend
6,
Wolf Creek
7.
Watts Bar
8.
Vogtie
9.
Millstone
10.
Maine Yankee
A-
\
Figure 3.3
Map showing the location of the ten test sites used in our
previous analysis, Bernreuter et al. (1984, 1985).
-35-
some extent the thinking of various S-Experts. As a result of this meeting Q-
5 was sent to the S-Experts in which they were asked to make any changes to
their zonation and seismicity parameters they felt were needed in light of the
initial results and the discussions at the meeting and in Questionnaire 5.
These updated results from the members of our S-Panel were used
with the updated input from the 6-panel {Q6) to re-compute the
ten test sites. These results were published in Bernreuter et
sent to our experts. We also provided our expert's with both a
final copy of our comparison report, Bernreuter et al . (1987).
the S-Expert Q7 asking them to make any changes they wanted to
zonation. The final maps are given in Appendix B represent any
changes they introduced. In summary Experts 3, 6 and 12 introd
new sets of zonation. Expert 7 introduced major changes and Ex
13 made some modifications to existing zones and/or added few n
Experts 1,2,5 and 10 did not change their zonation relative to
feedback.
3.3 Seismicity and Upper Magnitude Cutoff
In most hazard analyses, the seismicity of a zone (Eq. 2.1) is described in
terms of the number N of earthquakes per unit of area greater than a given
magnitude (or intensity). The number N is customarily related to magnitude by
an empirical magnitude recurrence model such as
in conjunction
hazard at the
al. (1985) and
draft and
We then sent
their
updates or
uced completely
perts 4, 11 and
ew zones,
the first
Log^Q N = a - b m (or I) ,
(3.1)
where the seismicity parameters a and b are constant for a given seismic zone,
m denotes magnitude, and I is the epicentral intensity. Generally, Eq. (3.1)
is modified to account for the fact that every seismic zone is believed to be
only capable of producing earthquakes with magnitudes (or intensities) bounded
above by some maximum value (called the upper magnitude cutoff. My).
In Q2 of the SHC (see Volume VII for details), the experts were asked to model
the seismicity of the EUS by providing the a, b, and Mm values for each of the
zones they identified in their maps. They were asked to provide a best
estimate value (the value which they believe is the most likely to represent
the true state of nature) and a range of values which represented their
uncertainty in estimating the values of these parameters. This information
was used to develop probability distributions which were used in the Monte
Carlo simulations.
An expert's estimates of a and b depend on the catalog of events the expert
used. In Q2 of the SHC, the experts were expected to choose their own catalog
of earthquakes and estimate a and b using whatever technique they deemed most
appropriate. The experts were also asked to decide on the type of correction
to apply for incompleteness and aftershocks. We did provide the experts with
a listing of earthquakes sorted by zones using a catalog we developed based on
several catalogs (Bernreuter et al . (1984).
-36-
At the feedback meeting and in the follow-up questionnaire (Q5) we put special
emphasis on carefully reviewing.
How we developed the maps to be used in the analysis from the data
provided by each panel members.
The definition and importance of the upper magnitude cutoff.
Both the desirable and undesirable features of the earthquake
recurrence model Eq. (2.1) used in the analysis at the time of the
feedback meeting (referred to as the LLNL model) as contrasted to the
truncated exponential model.
Our concerns about the large ranges of values given for the a and b
parameters of the earthquake recurrence model.
The possible need for correlation between the a and b parameters
during simulation and how such correlation could be introduced.
The need to correct the historical, catalog for incompleteness and
removal of aftershocks.
The importance of the experts' estimates of the seismicity in the
complementary zone (CZ).
The definition of self weights and confidence bounds to reach a
common understanding of their meaning.
After the meeting a feedback questionnaire (Q5) was sent to the panel. In
this questionnaire the topics covered at the feedback meeting were reviewed
and the panel members were requested to update their responses. In this
questionnaire the experts were asked to choose between the LLNL recurrence
model and the truncated exponential model. The experts were also asked to
indicate any correlation that might exist in their estimate of the coefficient
a,b. They were asked to choose between no, partial or full negative
correlation between and a and b-values of the earthquake recurrence model.
The responses were evenly divided between the LLNL model and the truncated
exponential model and between no and partial correlation.
The difference
impact on the r
(3.1) to be lin
truncated expon
from the above
discussed in Q5
correlation bet
details). It s
computations in
improved) than
between the LLNL and truncated exponential have only a minor
esults (Bernreuter et al . (1985)). The LLNL model forces Eq
ear over a range specified by the experts. Whereas the
ential model with a finite upper magnitude cutoff. My departs
relation near My. The differences between the two models is
and repeated in Q9 and a discussion of our model for partial
ween the a and b-values is given in Q9 (see Volume VII for
hould be noted that our final correlation model used in the
this report is slightly different (and in our opinion,
the model we used to obtain the results reported in Bernreuter
-37-
et al . (1985 and 1987). The correlation model also has only minor impact on
the computed hazard at a site.
In Q9 we asked , as in Q2, for the a and b-values for each of their zones
identified in their final maps. We provided, as before, a sorting of
earthquakes in each zone using either the LLNL or EPRI catalog. At the
suggestion of our Peer Review Panel, Bernreuter et al . (1985) we departed from
our previous approach and provided our S-Experts with estimates for the a and
b-values using a uniform methodology that we developed. Our method for
estimating the a and b-values is described in Bernreuter et al. (1987) and in
Q9. The main features of our method for estimating the a and b-values are:
(1) a probability of detection function (with parameters supplied by our S-
Expert in Q7) is used to correct for incompleteness; and (2) we convert
intensity to a probability distribution rather than to a single magnitude
value. The S-Experts had a choice of using the LLNL catalog or the EPRI
catalog. They could also choose the approach used to identify aftershocks.
We carefully explained to our S-Experts (both in Q9 and in one-on-one
meetings) both the good features of our methodology and its limitations. We
emphasized that we were providing the a-b-values only as one more source of
information, and that we really wanted their judgement.
In Table 3.4 we summarize the model choices made by the S-Experts. In
Appendix B we provide their complete input as used in our analysis reported
here.
-38-
TABLE 3.4
SU»f1ARY OF MODEL CHOICES OF S-EXPERTS
(1)
(2)
EXPERT NO.
CATALOG'
METHOD FOR' '
ID OF
AFTERSHOCKS
RECURRENCE
MODEL
TYPE OF
CORRELATION
a-b-Value
1
LLNL
Own
LLNL
Partial
Own
2
LLNL
LLNL
LLNL
None
Own
3
EPRI
EPRI
LLNL
None
Own (^)
4
LLNL
LLNL
LLNL
Full
Own
5
LLNL
Own
LLNL
Partial
Own
6
EPRI
EPRI
LLNL
Full
Own
7
EPRI
Own
LLNL
None
Own
10
EPRI
EPRI
LLNL
None
Own
11
EPRI
Own
LLNL
None
Own
12
EPRI
EPRI
T. Exp.
None
Own
13
EPRI
EPRI
LLNL
None
Own
(1) The catalog the expert selected for us to use to sort the historic
seismicity data in each of the expert's zones and in our uniform a-b-value
analysis. Several experts indicated that they often relied on other
catalogs.
(2) See Appendix B for criteria each S-expert specified to identify
aftershocks.
(3) S-Expert 3's a and b- values were very close to the values obtained by the
LLNL uniform methodology.
-39-
3.4 Documentation
As indicated in Section 3.1 in designing the elicitation process one of our
guiding principles was to make sure that all experts had complete flexibility
to develop their resources and opinions independent of the other panelists.
We wanted everyone to function independently in formulating their opinions and
have the flexibility to use analytical methods as well as personal intuition
and insight to the degree they felt appropriate. Overall, we wanted to assure
that everyone could express their opinions and not be hampered by the need to
defend their intuition and insight.
In following our philosophy of eliciting opinions, we did not neglect the need
to assure the quality of the experts' responses. We introduced several
quality assurance measures into the elicitation process:
In the initial choice of experts for inclusion as panelists.
As part of the development of the questionnaires used to elicit the
expert's opinions, careful consideration was given to the structure
of the questions.
By interacting with individuals to clarify potential
misunderstanding.
By having group discussions after the initial elicitation and
following these with feedback questionnaires.
By introducing qualitative and quantitative comparisons of the
expert's inputs with available earthquake data, with subsequent
clarification of significant discrepancies.
We are confident that these measures can assure the quality of the final
seismicity inputs and the applicability of the resulting estimated seismic
hazard. However, a criticism by several members of our Peer Review Panel
(Bernreuter et al . (1985)) was the lack of documentation relative to the
opinions of our experts. The reviewers felt that better control documentation
of the seismicity would add credibility to and facilitate verification of the
results of the project. It was suggested, also, that documentation would help
to reduce biases and eliminate inconsistencies in the opinions expressed by
the experts. It has also been argued that lack of documentation will make it
difficult to judge when the experts' present opinions will be outdated, i.e.,
opinions have changed significantly, so that an updated study should be done.
To meet these criticisms and to ensure that we have a product which is
credible and readily applicable to making decisions regarding the relative
seismic risks to nuclear power plants located in the EUS, we included a task
of developing additional substantiation of the inputs used in the seismic
hazard analyses.
To achieve this goal, we sent each of the S-Experts a documentation
questionnaire (Q8). In Q8 we explained in some detail our rational for
including documentation. We also told the experts that in developing
substantiation of the seismicity inputs we are not interested in detailed
-40-
technical justification for each of their choices. Rather, we are Interested
in understanding "how" they arrived at their judgements, i.e., what is the
general basis of their opinions. We recognized, of course, that the likely
situation is that the experts used multiple sources and methods to develop
their final opinions. Thus we wanted the experts' to document what their
primary sources were and what methods, e.g., graphical, analytical, and
logical implications they might have used.
After sending Q8 to each of the S-Experts, we met with them, or discussed by
phone, to ensure that they understood the questions and, in general, discussed
their responses and in some cases asked for added clarification. If the
expert sent in handwritten response we sent the typed version back to him for
his review. The S-Experts' responses are given in Appendix A along with our
comments relative to the process. These were no modification of the experts
input as a result of this documentation process.
3.5 Ground Motion Models
The function of a ground motion model (GM model) is to provide an estimate of
the ground motion at a site caused by an earthquake of a known magnitude at a
given location. It is very difficult, if not impossible, to develop such a
model only on the basis of theoretical principles of physics, mechanics, and a
knowledge of the geology and tectonics because many aspects of earthquakes and
wave propagation through the earth crust still remain poorly understood.
Also, the earthquake energy path is determined by the nature and geometry of
the various media (whose properties are very erratic in general) between the
source and the site. In addition, the local site characteristics (topography
and nature of the soil layers immediately under the site) can have a
considerable effect on the level of ground motion observed at a site. Some of
the GM models rely entirely on the available strong motion data and are
empirical in nature, e.g. Joyner and Boore (1981), Campbell (1981). The more
recent models combine a geophysical formulation with the available strong
motion data, e.g., Atkinson (1983), Boore and Atkinson (1987).
Our approach to model this large uncertainty in the estimation of the ground
motion and the correction due to local characteristics at any EUS site given
an earthquake of magnitude m at a distance d from the site was to use multiple
experts input. The choice of which GM models and which correction for local
soil conditions should be used and their weights relative to other candidate
models was left to our Ground Motion Panel members (G-Experts). Each G-
Expert, like the S-Experts, provided his individual opinion, and as described
in Section 2.5, only in the final step were the inputs from all G-Experts
combined. Each G-Expert was asked to provide the GM model which was in his
view the "best" model. He was also asked to provide up to six additional
models which, in his view, modeled his modeling uncertainty. We indicated
that the selection of GM models and weights was to model their uncertainty
about the location of the median estimate and not the variability between
earthquakes of the same magnitude. This variability was modeled by assuming
that the uncertainty in the estimate of the ground motion between earthquakes
has some distribution (generally lognormal) and the G-Experts provided the
-41-
distribution to
distribution.
use and estimates for the parameters defining the
As can be seen from Table 3.3 a number of informational meetings/workshops
were held with the G-Experts. The main purposes of these meetings were to
discuss our methodology and how we were using the input that the S and G-
Experts provided as well as to discuss the areas of greatest uncertainty,
controversial issues and to describe various GM models (both strong and weak
points). In our GM questionnaires (Q4, Q6 and QIO), in addition to a listing
of the various models we also provided the G-Experts with comparisons between
the models and additional discussion of main issues. Refer to Vol VII which
contains the questionnaires and, in particular, to Section 6 of QIO which
lists all of the GM models. In this report for ease of reference we identify
each GM model by the ID number given in QIO which is given in Volume VII.
In addition, as the final analysis showed that the ground motion Expert 5's
input lead to hazard estimates substantially different from the estimates
obtained with the other four G-Experts (see Volume VI, Section 2.3), we
performed an additional feedback with Expert 5. The result of this feedback
confirmed that the input given by Expert 5 had been correctly interpreted and
that the results were consistent with the Expert's thinking.
Table 3.5 lists the peak gr
the 5 percent damped relati
Experts from the tabulation
the G-Experts gave for each
identified in Tables 3.5 an
and by a simple description
at five periods: 0.04s, 0.
which GM models the various
model. As discussed in QIO
models for the four regions
weights and BE models for d
3.6.
ound acceleration (PGA) models and Table 3.6 lists
ve velocity spectral models selected by the G-
of GM models given in QIO, along with the weight
model. To help the reader, the models are
d 3.6, both by an ID, as in the questionnaire QIO,
. It should be noted that the spectra are computed
Is, 0.2s, 0.4s and 1.0s. In addition we denote
G-Experts considered the "best estimate" (BE) GM
the experts were allowed to provide different
shown in Fig. 2.3. Only G-Expert 2 gave different
ifferent regions as indicated in Tables 3.5 and
Both G-Experts 2 and 3 elected to provide their own parameters for model RV-
5A, RV-5V and RV-5RS. G-Expert 2 selected the Boore-Atkinson RV model but
with
Q - lOOOf^-^
and the relation between the seismic moment M and m^ as
log Mq- 2 % + 13.2
log Mq - m^ + 17.7
'"b - ^'^
mu < 4.5
6-Expert 3 set the relation between seismic moment M^ and corner frequency f,
as
%^c
3.5
3. X 10
23
and between moment magnitude M and mb
-42-
M
2.72 - 0.28mjj + .13 m^'
G-Expert 2 set the depth (D) to be used in the 6M models to be a function of
magnitude:
D - 2.5 mu -2.5 if m^^is greater or equal to 5.0
if m^ is less than 5.0
2.5 mu -2.5
D « 5 mfj -15
G-Expert 3 set the depth term for the response spectrum model to be a function
of period T:
D - 10.3 km T< 0.15s
2 3
D - 5.022 - 1.073 (4-) + 0.708 (-^) - 0.064 (-4-)
for '
0.159 _< T_< 1.05
In Tables 3.5 and 3.6 we also give the overall aggregated weight of each 6M
model normalized by the G-Experts self weights given in table 3.7.
The best estimate PGA models are plotted in Fig. 3.4 for magnitudes of 5 and
7. The remaining PGA models are plotted in Fig. 3.5 also for magnitudes of 5
and 7. It should be noted that the base case used for both Figs. 3.4 and 3.5
is rock. Site correction factors are discussed later, however, it is
important to note that for the PGA models, the median correction factor to
convert the models selected from rock to generic deep soil is 1.0 except for
the model selected by G-Expert 5, G16-A3, where the correction is significant
as shown by comparing Fig. 3.6 to Fig. 3.4.
The BE spectral models are plotted for magnitudes of 5 and 7 at a distance of
25 km in Fig. 3.7 and the remaining models are also plotted for magnitudes of
5 and 7 at a distance of 25 km in Fig. 3.8. All models are for the rock base
case.
It should be noted that all of the RV models listed in Tables 3.5 and 3.6 have
a complex form which cannot be directly used in hazard analysis programs. It
was necessary to use the analytical form for the various RV models to generate
values of acceleration, velocity or spectral velocity as a function of m^ and
distance. Then a model more suitable for use in hazard analysis programs,
e.g..
log a
Cj + C2
% + C3
m^
+ C4 m^^
+ C5 R + Cg log R
was fit to the computed values of ground motion by a least squares fit.
Generally our fits were in two parts, from to 100 km and from 100 km to 1200
km. All fits had less than a 5 percent error, generally much better.
However, to get the error term small some times required relatively complex
functional forms. These functional forms were developed by a trial-and-error
use of stepwise fitting packages and examination of the residuals of proposed
fits.
-43-
TABLE 3.5
PGA MODELS AND WEIGHTS SELECTED BY THE G-EXPERTS
G-EXPERTS
ID
MODEL (^)
ID
DEPTH (2)
km
XI
All
Regions
X2
Region Regions
1 2-4
X3
All
Regions
X4
All
Regions
X5
All
Regions
Total
Aggregated ^^
Weight
Reg. 1 iRea. 2-
RV-IA
8.
0.25
-
—
0.4BE
0.4BE('^)
0.2
0.2
RV-2A
8.
0.25
-
0.3
-
0.12 0.12
RV-5A (5)
(6), 8.
-
0.4BEI 0.3
0.3
—
_
0.17 1 0.12
G16-A3
Epicentral
-
-
-
-
_
l.OBE
0.20 0.20
SE-IA
SE-IA
(6)
8.
0.25
0.3
0.4BE
-
0.25
-
0.16
0.18
SE-2A
8.
0.25BE 1
- 1 -
-
0.25 1
—
0.1 0.1
Comb-IA 1
1
1
0.3 0.3 1
1
0.1
-
0.07 0.07
Notes
(1) Model IDs are the IDs given in Section 6 of QIO in Vol. VII.
(2) Some models only differ by the average depth used in the GM model to
compute the distance.
(3) Aggregation includes the G-Experts' self-weights given in Table 3.7.
(4) BE- Best estimate GM model.
(5) G-Experts 2 and 3 provided their own parameters for the RV-model as
described in the text of Section 3.5.
(6) G-Expert 2 set the depth as a function of magnitude- refer to the text
of Section 3.5.
(7) The combined weight is given as the models only differ by the depth
term. ^
-44-
TABLE 3.5 (Continued)
PGA MODELS IDENTIFICATION
(SEE QIO IN VOLUME YII)
RV-IA:
RV-2A:
RV-5A:
G16-A3:
SE-IA:
SE-2A:
Comb-IA:
Boore-Atkinson model - based on physical assumptions
source spectrum shape and random vibration theory
(rock model) .
Toro-McGuire - based on physical assumptions, source
spectra shape and random vibration theory. Different
physical values as in RV-IA (rock model).
Same as RV-IA or RV-2A, different parameter values
(rock model ).
Trifunac correlation of PGA versus epicentral
intensity and Gupta-Nuttli attenuation of the
intensity. Entirely data based. Applies to rock,
deep soil or intermediate.
Semi-empirical Nuttli (1986) model. For f^ - Mq
(corner frequency - seismic moment) slope of 4 (soil
model ).
Semi-empirical Nuttli (1986) model for f^, - M slope
of 3 (soil model).
An empirical model based on all available intensity
and strong motion data, developed by Prof. Veneziano,
1986. Applied to rock or soil.
-45-
TABLE 3.6
5 PERCENT DAMPED RELATIVE VELOCITY SPECTRAL MODELS
AND WEIGHTS SELECTED BY THE G-EXPERTS
G-EXPERTS ID
MODEL (^)
ID
BT7 — TBT
DEPTH
km
XI
All
Regions
X2
Region Regions
1 2-4
X3
All
Regions
X4
All
Regions
X5
All
Regions
Total
Aggregated
Weight
Reg. 1 iRea. 2
RV-IRS
RV-IRS
(3)
8.0
0.25
0.46E
3.4BE
0.20
0.20
RV-2RS
RV-2RS
(3)
8.0
0.25
~
—
0.3
-
-
0.12
0.12
RV-5RS(5)
(3), (6) 1
-
0.3 0.3 1 0.3
-
_
0.12 0.12
TL-RS 1
E 1
-
- 1 -
l.OBE
0.2 0.2
NH-SE1A,V
NH-SEIA.V
(6)
8.0
0.25
0.3
0.4BE
-
0.3
-
0.17
0.19
NH-SE2A,V
8.0 1
0.25BE
1 - 1 - 1
0.3
1
0.1 n 1
NH-(7)
(6)
1
0.4BEI 0.3 1 - 1
1
1
0.09
0.06
Notes
1.
(1) Model IDs are the IDs given in Section 6 of QIO in Vol VII.
(2) Normalization includes the G-Experts self-weights given in Table 3.7.
(3) G-Expert 3 made the depth a function of frequency - see text in Section
3.5.
(4) The combined weight is given as the models only differ by the depth
t e rm .
(5) G-Experts 2 and 3 provided their own parameters for the RV-model as
described in the text in Section 3.5.
(6] G-Expert 2 set the depth as a function of magnitude- see text in Section
3.5.
(7) G-Expert 2's RV-5A and RV-5V models are used to anchor the acceleration
and velocity Tegs of the Newmark-Hal 1 spectral shape.
-46-
TABLE 3.6 (Continued)
IDEKTIFICATION OF SPECTRAL MODELS
(SEE QIO IN VOLUME YII FOR MORE DETAILS)
RV-IRS: Boore-Atkinson model. Based on physical assumptions,
source spectrum shape and random vibration theory.
Applies to rock.
RV-2RS: Toro-Mc6uire model. Same as RV-lRS, different
parameter values. Applies to rock.
RV-5RS: Same as RV-IRS or RY-2RS, customized by G-Experts 2
and 3. Applies to rock.
TL-RS: Trifunac and Lee (1985) empirical model.
NH-SE1A,V: Nevymark-Hal 1 constructed with semiempirical models SE-
lA and SE-IV (same as SE-IA but for velocity) for
acceleration and velocity.
NH-SE2A,V: Same as NH-SE1A,V, but uses SE-2A and SE-2V for
acceleration and velocity.
NH- RV-5A and RV-5V (same as RV-5A but for velocity) are
used to anchor the Newmark Hall spectral shape.
>^.;,-;:
-47-
, 'V.'i*,'.
'■■vj'^^y.
>'.
-/, .
■■
f""'"
;/, ■
Hi
i^'
^
■^
■I'y
','■'
m^:--
.<>.'
r.*- ■ -
TABLE 3.7
G-EXPERTS
SELF-WEIGHTS
EXPERT
SELF-WEIGHT
1
1 10
2
1 9.5
3
1 9
4
1 7
5
1 9
-48-
Plot
Symbol
GM
Model No.
1
2
3
4
5
RV-IA
RV-5A (X2)
G16-A3
SE-IA
SE-2A
10
o
CO
O
I
o
10
10
o 10
<
10
^"^^^^^r
^■^"^^■^■^"^^■m^
DISTANCE-KM
Figure 3.4 Best estimate PGA models listed in Table 3.5 plotted
for magnitudes of 5 and 7. Rock base Case.
-49-
Plot
Symbo 1
1
2
3
4
5
6
GM
Model No.
RV-2A
RV-5A (X3)
SE-IA (X4)
SE-2A (X4)
SE-IA (XI)
Comb- 1 A
DISTANCE-KM
Figure 3.5 Remaining PGA models listed in Table 3.5 plotted for
magnitudes of 5 and 7. Rock base case
■50-
Plot
Symbol
1
2
3
4
5
GM
Model No.
RV-IA
RV-5A (X2)
G16-A3
SE-IA
SE-2A
DISTANCE-KM
Figure 3.6 Best estimate PGA models corrected to generic deep
soil for magnitudes of 5 and 7.
-51-
Plot
Symbo 1
1
2
3
4
5
GM
Model No.
RV-IRS
TL-RS
NH-RV5A,V{X2}
NH-SE1A,V
NH-SE2A,V
I o
PERIOD (SEC) o
Figure 3.7
Best
model
of 5 a
est mate 5 percent damped relative velocity spectra
s listed in Table 3.6 plotted for magnitudes
and 7 at a distance of 25 km. Rock base case
-52-
o
to
3
o
u
o
Plot
GM
Symbol
Model No.
. 1
RV-2RS
2
3
4
RV-5RS (X3)
RV-5RS (X2)
NH-SE1A,V
5
6
RV-lRS (XI)
RV-2RS
/■■yy-.\>:y\
:■'■"' ■••„:''.v5
PERIOD (SEC) 2
Figure 3.8 Remaining 5 percent clamped relative velocity spectra
models listed in Table 3.6 plotted for magnitudes of
5 and 7 at a distance of 25 km.
-53-
3.6 Random Uncertainty and Truncation of GM Estimates
Each G-
used in
local s
correct
the G-E
Table 3
paramet
possibl
motion
Expert was a
the ground
ite was not
ion approach
Xpert values
.8a refer to
er. Each G-
e truncation
possible bas
sked to provide an estimate of the random uncertainty to be
motion attenuation relationships. The influence of the
to be included if the expert chose to use the category
discussed in the next section. In Table 3.8a we provide
. Except where noted for G-Expert 5, the values listed in
the standard deviation of the natural logarithm of the GM
Expert was also asked to select a method to model the
of the uncertainty distribution and/or maximum possible GM
ed on the following choices:
Model 4:
Model 1: No truncation.
Model 2: There is an absolute maximum acceleration, independent of
magnitude and distance, which will not be exceeded. This is
the Type 1 saturation.
Model 3: The maximum acceleration is a function of magnitude and
distance; this is modeled by assuming the maximum
acceleration is a fixed number of standard deviations from
the mean in the lognormal distribution of the GMP's. This
is the Type 2 saturation.
For any magnitude and distance the maximum acceleration is
the minimum of an absolute maximum and a fixed number of
standard deviations from the mean; this is an envelope of
Type 1 and 2 saturation.
The 3 types of limits, drawn as a function of distance R for a fixed magnitude
m, are depicted in Fig. 3.9. It is observed from Fig. 3.9 that:
Type 1, an absolute maximum acceleration, ai , results in the
horizontal curve Ci.
Type 2, the maximum acceleration is a fixed number, n, of standard
deviations from the mean curve, a(m,R), results in curve Co
Type 3, the envelope of Type 1 and 2, results in the curve C3.
The G-Experts' choices for truncation are given in Table 3.8b
^
-54-
TABLE 3.8a
G-EXPERT'S CHOICE FOR THE RANDOM UNCERTAINTY, SIGMA ^^J FOR THE GM MODELS
Expert
BE
PGA
Range
PGA
BE
Spectra
Range
Spectra
1
0.35
0.3 - 0.4
0.35
0.3 - 0.4
2
0.55
0.4 - 0.7
0.55
0.4 - 0.7
3
0.5
0.4 - 0.7
0.6
0.4 - 0.8
4
0.5
0.35 - 0.65
0.5
0.35 - 0.65
5
0.7
0.7 - 0.9
(2)
(2)
(1) Standard deviation of the natural logarithm of the GMP.
(2) G-Expert 5's spectral model TL-RS is not a lognormal model.
-55-
TABLE 3.8b
G-EXPERTS' CHOICE FOR TRUNCATION OF THE GROUND MOTION VARIATION
Expert
1
2
3(1)
4
5
Method to
Truncate
PGA
Max PGA
(cm/s/s)
and/or N
Sigmas
Method to
Truncate
Spectra
Max Sy
and/or
N Siqmas
None
1
None
2500/2.5
3
/2.5
None
1
None
None
1
None
4
1(2)
None
'" iz:zit\:ir^.z\'e%':''''''''' "^*'°' ' "^ '^"""'^•°"- ^"^ ^*^ -*
(2) G-Expert 5 indicated that the spectral
^ u.^j., . ^ iMu Ldieu tnat tne spectral model he selected TL-RS is nnt ;,
llTZT.Z'f. ':' the distribution used in model TL-RS adequa eN
accounts for the truncation of the distribution. M^ciLeiy
-56-
GMP
(Log scale)
Type 1
Distance
Figure 3.9 Illustration of the three types of models considered
to model the physical saturation of ground motion.
The random variation of the logarithm of the ground motion
parameter (GMP) is modeled by a normal distribution with
mean GMP (m,R) and a standard deviation o.
-57-
3.7 Correction for Local Soil Conditions
fnnn^^"^ ^tl^ °^ our G-Experts to provide relative weights for each of the
following three possible correction approaches for local soil conditions:
i
1.
2.
3.
No correction applied
Apply a simple correction , either the site is soil or rock
^PP'y our categorical correction approach.
Then, as outlined in Section 2, for each G-Expert we used a Monte Carlo
The no correction approach requires no discussion. The simole correrUnn
'r'oc°k' r\' ?nil' "r? 'T.'t:- '"' '''''''' GM model iL'md'o °nher
With this methodology, the site carrf^ri-inn ic nnf f-!«^ *.
motion in particular.' However n the case of G Exnprt ? ""^^P^^'f": S^o^nd
motion model was selected rseeTabl«%?^r,3\^f'^^ ^ ""^J* °''^ Sromi
correc on factors were developed in the following manner Figures 3 11a and
j.llD ulustrate our nrnrprinro Tn ^^A- , i. ^ y "luiiMcr . r lyures j.iia and
analysis, we seleaed a set o?*?n ?iS! h T^-^^ ^'""^ histories to use in the
s^^^;?ra^^;a^1^p a^: dS i\^ Fa3°-
™ " " ensTtT^^ s^S^^rr'^tlo'lr^V^^^;- r^ ""-" "r-the shear
TI. 12. T3 and deep so,l categories was computed using the sSakE Prog^a"* '
charac?er"lst'?cs""wrL'*"/'".'" ™'"^^^' P-'^Pe^ties and earthquake
:;5» " iff SFSijp^
-58-
earthquake time history for each simulation. Variability in the dynamic
modeling was introduced by sampling sets of input parameters (mainly shear
wave velocities of soil and rock, damping ratio of soil and the depth of soil
deposit) from assumed probability distributions for each simulation. A
lognormal distribution for each input parameter was assumed for this study.
Using the above data we developed two median correction factors as a function
of spectral frequency (to correct the 6M model from rock to the site's
category or from generic soil to the site's category) and the uncertainty in
our estimate of the median in the following manner. For the rock base case,
we simply computed the correction factor by taking the ratio of the computed
spectrum at the top of each soil column to the input rock spectrum. For each
category this resulted in a set of 20 correction factors. It was found that a
lognormal distribution could be used to model the uncertainty in the estimated
correction factor with o •= 0.5 . This value constitutes our opinion of the
best value for this parameter. This choice came from the analysis described
above and some judgement on the amount of reduction to apply to the values
found to remove non site effects, such as path and source effects, after a
study performed for NRC under a separate project, where these effects were
specifically studied and quantified (Bernreuter, Chen and Savy, 1986). Figure
3.12 shows the resultant smoothed median correction factors for the till-like
categories relative to rock, and Fig. 3.13 shows them for the sand-like sites
relative to rock. The median correction factors are plotted at each of the
five spectral periods (0.04, 0.1, 0.2, 0.4 and 1.0 sec) and connected by
straight lines to make it easier to follow the general trends. Also shown in
Fig. 3.13 are the correction factors for deep soil relative to rock. Note
that the correction factors for PGA are plotted at 0.01 sec in both Figs. 3.12
and 3.13.
Our original plan (see QIO) was to carry along two sets of correction factors,
one set relative rock and one set relative to generic soil. We did not do
this because it would have required an extensive rework of the logic of our
computer program and only a single family of soil base case models (Nuttli's
latest model labeled SE-1 and 2 in QIO). The only difference between SE-1 and
2 is the scaling with magnitude which is selected using theoretical
considerations. The PGA models SE-IA and SE-2A were "converted" to a rock
base case using the median correction factor (1.0) found in our analysis for
PGA between generic soil and rock sites and the velocity model was converted
to a rock base case in a similar fashion, however, for the velocity a
correction factor of V^q^i/V^qj.]^ = 1.7, as found in our analysis, was used.
In the Monte Carlo uncertainty analysis when the categorical correction
approach is selected the correction factor is simulated based on the assumed
lognormal distribution for the particular category, with the median and the
standard deviation derived for that distribution.
Table 3.10 lists the site correction methods selected by the G-Experts and
their weights, given by the G-Experts for use in the Monte Carlo analysis. It
can be seen from Table 3.10 that the categorical approach is very heavily
weighted. This is a change from our previous study, Bernreuter et al. (1985)
where the categorical approach carried an aggregated weight of about 0.53 for
spectra and 0.46 for PGA. Thus the site effect will be more significant in
-59-
the results presented in this report than in our previous report, Bernreuter
et al (1985) .
-60-
TABLE 3.9
DEFINITION OF THE EIGHT SITE CATEGORIES
Generic Rock
(1)
Sand Like
(2) Sand 1
(3) Sand 2
(4) Sand 3
Till-Like
(5) Till 1
(6) Till 2
(7) Till 3
Deep Soil
(8)
CATEGORY
Rock
SI
S2
S3
Tl
T2
T3
DEPTH
N/A
Deep Soil
25 to 80 ft.
80 to 180 ft.
180 to 300 ft,
25 to 80 ft.
80 to 180 ft.
180 to 300 ft,
N/A
-61-
^B^A.
■■fK
•'-y-i ';■■•
'.• -y
'Iv-i'J 1-,
'■'■•■■-
TABLE 3.10
G-EXPERTS' WEIGHTS FOR SITE CORRECTION APPROACH
Expert
I
2
3
4
5
Aggregated
Weight
No
Correction
— (j:
Simple
Correction
Categorical
Correction
0.
0.
0.
0.
"07
0.
0.
0.
1.0
T7D"
1.0
1.0
1.0
0.
-62-
1000
3.0-
100
2.5--
2.0--
u
o
c
o
m
c
o
I 1.5+
c
o
•iH
10
o
o.
E
1.0-
0.5-
0.0-
I I 1 I 1 I I I
TPifunac-Andepson
Acceleration
Factors. 1977
♦flock
llnterHdlate
•Deep Soil
-I I ■_
■ III
0.001
0.010
Fnequency. hertz
10
0.10
-•-n-
I I I I I I I
EPRI. 1986
ShalloM
^
Trifunac-Lee. 1985
J^Oaep Soil
-•IntePHdlate
-♦Bock
■■'y
-1 I ' ' ' ^
.J I « « ■
0.100
Period, sec
1.000
10.000
Figure 3.10. Simple correction factors relative to rock.
-63-
K ■ :■.•>; vv. ..-x ■.
CALCULATED MOTIONS AT SOIL SURFACE
INPUT
r.OTIOK
•GIVEN AT
ROCK OUTCROP
RGCK FORMATION
GIVEN Vg AND Qs
liJA)
S, OR T:;
(180 TO 300 FT)
Figure 3.11a Schematic representation of our computational procedure to model
site correction factors.
Variation in input moti
ON
INPUT
ROCK -'-rr,
OUTCROP
Ground Surface OUTPUT
Variation in
Dynamic Rock Properties
Shear wave vel^V^ Q Factor
Variation in
Dynamic Soil Properties
Shear wave vel. q factor
mean
MEAN
IIB)
Variation in
Soil Thickness
Figure 3.11b The physical parameters used in the 1-D analysis are drawn from
probability distributions.
-64-
3.0-
2.8-
2.8-
t4--
U--
2.0- ■
«.:
S 16-
I
8 u-
a„.
10-
0.8-
0.8-
0.4--
0£-
0.0
0.001
.SiiooL-SMiaasiiu-
• T-1
■ T-a
♦ T-a
0.010
0.100
Ptrlod, »*c
tooo
10.000
Figure 3.12 Smoothed median correction factors for the till-like categories
listed in Table 3.9 relative to rock. The PGA correction factors
are plotted at 0.01s.
3.0
2.8-
2.8-
2.4-
2.2-
2.0-
S '•8-
£ 1.8 -
g 14
g
" 1.2
1.0-
0.8-
0.8-
0.4
0.2
0.0
0.001
• Deep Soil
■ S-1
♦ 8-2
« S-3
jn
^
y^^ ■- .^
\
0.010
0.100
P«rlod, sac
tooo
10.000
Figure 3.13 Smoothed median correction factors for the sand-like categories
listed in Table 3.9 relative to rock. Also shown are the
correction factors for deep soil relative to rock. The PGA
correction factors are plotted at 0.01s.
-65-
SECTION 4: COMPARISON TO PREVIOUS RESULTS AND OTHER STUDIES
4.1 Comparison to Previous Results
As Indicated In Section 3 only S-Experts 1,2 and 5 made no changes in their
zonation or seismicity parameters, whereas S-Experts 3,6 and 12 Introduced
completely new maps and naturally all new seismicity parameters. Experts 4,11
and 13 Introduced some small changes in zonation and a number of changes in
the seismicity parameters whereas S-Expert 7 introduced major changes, without
completely redoing his maps. Expert 10 introduced changes in his seismicity
parameters. Even more significantly there has been major changes in the input
from our G-Experts.
In addition to the changes introduced by our S and G-Experts it should be
noted that for this report the integration over magnitude starts at a lower
bound of 5.0 whereas in our previous study we started at a lower bound of
3.75. This change generally has a significant effect on the results as
discussed in Bernreuter et al . (1987). These three factors must be considered
in order to compare our current results to our previous results. In order to
isolate the source of any differences in the estimates of the seismic hazard
at our ten test sites between the updated results and our previous results we
first compare the estimate of the seismic hazard at each of the ten test sites
using the new input from our S-Experts to the results obtained using the input
from S-Experts given in Bernreuter et al (1985). The ground motion model used
was the modified Nuttli PGA model used in the comparison to EPRI results
discussed in detail in Bernreuter et al (1987). This model is very similar to
the model SE-IA selected by several G-Experts. Figures 4.1 and 4.2 show
typical comparisons of the constant percentile hazard curves (CPHCs) for PGA
between the new and previous results at the Braidwood and the Millstone
sites. Our new seismicity input from the S-Experts results in an increase in
the estimate of the seismic hazard at five sites (Shearon Harris, Braidwood,
Lacrosse, River Bend, Wolf Creek) and a decrease at the other five sites. The
largest differences are at LaCrosse and River-Bend and the smallest difference
is at Watts Bar. Given the uncertainty in the estimate, as measured by say
the 15th and 85th percentile CPHC there is relatively little change in the
aggregated results between our previous input and the new final updated input
from our S-Experts.
At the individual S-Expert level some very major difference have been
introduced by some of the changes introduced by the S-Experts. For example,
in Fig. 4.3a the best estimate hazard curves (BEHC) for PGA for each of the*S-
Experts for the Braidwood site based on the updated input given in Appendix B
are plotted and in Fig. 4.3b the BEHC based on the input given in Appendix A
of Bernreuter et al. (1985) are plotted. If these figures are compared
several notable differences are observed. First, it is noted that the BEHC
for S-Expert 11 (plot symbol B) is significantly lower in Fig. 4.3a (new) than
in Fig. 4.3b (old). This large change occurs because S-Expert 11 changed his
zone 10. Previously, zone 10 extended northward and included the site. In
the updated map (see Appendix B) zone 10 was reshaped and no longer includes
the Braidwood site. Thus the BEHC for S-Expert 11 is now significantly lower
than that resulting from the older input from S-Expert 11. It is also seen
from comparion of Figs. 4.3a and 4.3b that new BEHC from S-Expert 12 (plot
-66-
symbol c) is also significantly lower than before. Previously, S-Expert 12
had a zone 10 which included the Braidwood site. The contribution too the
seismic hazard from zone 10 was significant, thus here is a significant
decrease in the BEHC for S-Expert 12 with his current zonation which does not
have a zone near Braidwood as compared to his previous zonation near the
Braidwood site.
In Fig. 4.4a we plot the BEHC for the S-Experts based on their updated input
given in Appendix B for the Millstone site and in Fig. 4.4.b based on their
previous input given in Bernreuter et al (1985). The most significant change
at the Millstone site is for S-Expert 7. Previously, S-Expert 7's input
resulted in a PGA BEHC at a Millstone site that was one of the highest (Fig.
4.4.b) whereas based on his current updated input S-Expert 7's BEHC is the
lowest at the Millstone site. In both the old and the new zonation for S-
Expert 7 the Millstone is in a zone 24. Currently S-Expert 7 has set the
upper magnitude cutoff in zone 24 at 5.75 whereas previously he set it at
6.5. In addition, currently the absolute value of the b-value is higher than
before, this coupled with a new a-value results in the rate of magnitude five
events being approximately a factor of five lower than previously and at
magnitude 5.75 the difference in rate of events greater than 5.75 is even
larger.
Figures 4.1 and 4.2 show that the 15th, the 50th (median) and the 85th CPHC
also remain relatively stable. The 85th CPHC shows less change than the 15th
CPHC. The range of variation at the ten test sites is reasonably covered by
Figs. 4.1 and 4.2.
The above comparisons suggest that if common ground motion models are used
then the CPHCs are a reasonably stable estimate of the seismic hazard at a
site for the same set of experts over a period of time; however, individual
experts results might significantly change. This is in agreement with
conclusions reached in Bernreuter et al. (1985).
It is of some interest to examine the effe
in the GM models has on the results. In F
CPHCs at the Braidwood and Millstone sites
computed using all of the GM models and we
updated seismicity input given in Appendix
model used to generate Figs. 4.1 and 4.2.
differences in the 50th CPHCs between the
used and only one GM model is used is sole
single GM model used, hence no meaning can
between the 50th CPHCs. However, what is
between the bounds as represented by the 1
that uncertainty introduced by the GM mode
ct that introducing the uncertainty
ig. 4.5 and Fig. 4.6 we compare the
for the case where the CPHCs are
ights given in Table 3.5 and the
B to the case of the single GM
It should be noted that the
case when all of the GM models are
ly dependent on the choice of the
be attached to the difference
significant is the difference
5th and 85th CPHCs. It can be seen
Is is significant.
In Figs. 4.7 and 4.8 we compare the CPHCs obtained using the latest updated
input to CPHCs obtained using our previous input as reported in Bernreuter et
al. (1985), for the Braidwood and Millstone sites. The difference between the
old and new results on Figs. 4.7 and 4.8 are typical of the difference between
the old and new results at the other test sites.
-67-
mm:
Given all of the changes in both the seismicity and PGA models there is
remarkably little difference between the current and previous results.
However, when the spectral models are considered we see a much greater
difference between the current results as compared to our previous results as
is illustrated in Fig. 4.9. In Fig. 4.9 we compare the 1000 year return
period constant percentile uniform hazard spectra (CPUHS) at the Braidwood
site for the case when the GM models, seismicity input given in Appendix B,
and weights in Table 3.6 are used to the results given in Bernreuter et al .
(1985). There is a reasonable agreement in the average level of the spectra
between the old and new results. However, Fig. 4.9 shows clearly that the new
set of ground motion spectral models has introduced higher spectral values at
high frequencies (low periods) and lower spectral values at low frequencies
(high periods in the range of one second). The reason for this is the very
major shift in the shape of spectral models selected by the G-Experts. In
Bernreuter et al . (1985) all of the available models had a shape similar to
the Newmark-Hall models (models denoted by NH-in table 3.6) and the Trifunac-
Lee model (the model denoted by TL - Table 3.6); however, in the updated input
from the G-Experts, the random vibration models (the models denoted by RV- in
Table 3.6) carry a weight of 0.44 which means they are used almost half the
time. The RV models are generally much lower at the longer periods than the
Newmark-Hall type models as can be seen from Fig. 3.8. Thus it is not
surprising to observe a significant difference between the updated CPUHS and
our previous results in the longer period range.
4.2 Comparison to Other Studies
In Bernreuter
the results o
other studies
same GM model
previous resu
studies, EPRI
Company (1983
that our prev
obtained usin
al. (1985).
the updated r
Bernreuter et
et al . (1985 and 1987) a number of comparisons were made between
btained using the input from our S and G-Panels to the results of
. In Bernreuter et al . (1985, 1987) we concluded that if the
s and lower bound of integration are used then generally our
Its were in good agreement with the results obtained from other
(1985b), Algermissen et al . (1982), Yankee Atomic Electric
) ERTEC (1983) and Dames and Moore (1983). In addition we showed
ious results were in good agreement with the hazard estimates
g a historical method, Veneziano et al . (1983) and Bernreuter et
Because of the good agreement between our previous results and
esults there is no need to repeat the comparisons made in
al . (1985, 1987) and the same Conclusions are valid.
-68-
PERCENTILES = 15.. 50. AND 85.
-1
10
-2
10
Old Seismic ity
Input
Updated Seismic ity
Input
CM ro •* If) lO
ACCELERATION CM/SEC»*2
BRA I DWOOD
Figure 4.1 Comparison of the 15th, 50th and 85th percentile CPHCs for PGA
between the new results based on the updated input from the S-
Experts described in Section 3 and our previous results given in
Bernreuter et al. (1985) for the Braidwood site. Only a single
PGA model (modifed Nuttli) was used.
-69-
PERCENTILES = 15., 50. AND 85.
DC
<
CD
<
m
o
q:
Q.
-1
10
-2
10
-3
I I
Old Seismicity
• • Input
Updated Seismic ity
Input
Q- 10
u
z
<
o
t> 10
X
o
-5
10
10
-7
10
o
+
tn -^ \r> -
m
<
m
o
a:
a.
-4
10
10
-6.
10
-7
10
•-
u
t
o
X
o
>-
<
m
o
Q.
O
+
■* in «) r^
ACCELERATION CM/SEC^Z
MILLSTONE
Figure 4.4a BEHCs for PGA for the Millstone site per S-Expert based on the
updated input provided by the S-Experts. Only a single PGA model
(modified Nuttli) was used.
-73-
-1
10
-2
10
-3
10
-4
10
-5
10
-6
10
-7
10
GMM=NUTTLI 1984. M0=5.0. NO SITE CORRECTION
BEST ESTIMATE
FOR THE SEISMICITY EXPERTS
o
+
■* lO ID r^
ACCELERATION CM/SEC"2
MILLSTONE
Figure 4.4b BEHCs for PGA for the Millstone site per S-Expert based on the
previous input given in Bernreuter et al. (1985). Only a single
PGA model (modified Nuttli) was used.
-74-
E.U.S SEISMIC HAZARD CHARACTERIZATION
LOWER MAGNITUDE OF INTEGRATION IS 5.0
PERCENTILES = 15.. 50. AND 85.
q:
<
Ui
>■
Ol
<
-1
10
-2
10
-3
10
UJ *
o 10
X
>-
*Z -5
-J 10
m
<
m
o
-6
10
-7
10
-
UJ
<
O
CD
<
CD
O
DC
-1
10
-2
10
-3
10
-4
10
-5
10
-6
10
-7
10
Single GM
Model
All Updated
GM Models
— fN CM
ui ACCELERATION CM/SEC* •2
CO o>
MILLSTONE
Figure 4.6 Comparison of the 15th, 50th and 85th percentile CPHCs between
the CPHCs obtained aggregating over all of the S and G-Experts
and the CPHCs obtained using a single PGA model (modified Nuttli)
and aggregated over all S-Experts for the Millstone site.
-76-
E.U.S SEISMIC HAZARD CHARACTERIZATION
LOWER MAGNITUDE OF INTEGRATION IS 5.0
PERCENTILES = 15., 50. AND 85.
-1
10
-2
10
<
>-
1 .1 o
a. 10
u
z
< ■
o
o 10
X
!l -5
-J 10
CO
<
m
o
q:
OL
HAZARD CURVES USING ALL EXPERTS
-6
10
-7
10
O o Previous Results
Updated Results
-
<
X
o
>-
<
m
o
or
Q.
n I I
O O Previous Results
Updated Results
— (N
o
»o n IT) to
ACCELERATION CM/SEC'»2
MILLSTONE
Figure 4.8 Comparison of the 15th, 50th and 85th percentile CPHCs aggregated
over all S and G-Experts between the new input and the
input from the S and G-Experts for the Millstone site.
previous
-78-
E.U.S SEISMIC HAZARD CHARACTERIZATION
LOWER MAGNITUDE OF INTEGRATION IS 5.0
1000. -YEAR RETURN PERIOD CONSTANT PERCENTILE SPECTRA FOR
PERCENTILES = 15.. 50. AND 85.
10
u
o
UJ
>
10
o
UJ
^ 1
" 10
>-
10
-1
10
CM
I o
o oo Previous Results
Updated Results
o o
o o
o o
To
PERIOD (SEC)°o
BRA I DWOOD
"o
Figure 4.9 Comparison of the 1000 year return period 15th, 50th and 85th
percentile CPUHS for 5 percent damping aggregated over all S and
G-Experts between the new input and our previous input from the S
and G-Experts for the Millstone site.
-79-
SECTION 5: REFERENCES
1. S.T. Algermissen, P.C. Perkins, S.L. Thenhaus, Hanson, and B.L. Bender
Probabilistic Estimates of Maximum Acceleration and V elocity in Rock in
the Contiguous U nited States . u.i>. Geological <^nrvpy iipon-t-.M^ b^p^pt
82-1033, 99 pages (1982).
2. G.M Atkinson, "Attenuation of Strong Motion in Canada from a Random
Vibration Approach", Bull. Seism. Soc. Am. Vol. 74. No. 6 dd 2629-
2653 (1984). * ^'^*
3. D.L. Bernreuter, Seismic Hazard Analysis: Application of Me thodolnnv
Results and Sensitivity studies . NUt^E:G/C:ft-lRA? iwtdi -c.-if\7n ^
Vol. 4). (1981) ^
D.L. Bernreuter, D.H. Chung, and C.P. Mortgat, Seismic S afety Margins
Research Program P h ase I: Final Report-Deyelopment of S e smic I nput "
(Projectll , NUkbG/Cfe-^015. Vol 2 t\_ 4^5. My choice
of this value is based on the following reasoning: Most seismic source zones
have a b value of approximately one, and a history of at least 100 years.
Therefore the existence of m^j = 4.5 earthquakes in 100 years would suggest the
occurrence of m^ = 5.5 earthquakes in 1000 years. On the other hand, the lack
of mjj >^ 4.5 earthquakes in 100 years would suggest the lack of m^j >^ 5.5
earthquakes in the last 1000 years.
Second, I used basement structure to outline the boundaries of the areas
containing earthquakes of m^ >_ 4.5. High priority was given to rift zones,
next to suture zones (e.g, Appalachian, and Ouachita Wichita Mountains), third
to basement uplifts (e.g., Cincinnati Arch, Nemaha uplift) and finally to
basement basins or subsidence zones (e.g, Illinois Basin).
Third, I looked at the recurrence rates for these seismic zones. If they did
not indicate a seismicity rate different from the background areas, I put them
in the background zone.
QUESTION 3 Identify some of the factors, such as scaling, lower bound
magnitude, which influenced your development of the EUS zonation
maps. Rank the relative importance of each of these factors,
and, if possible, explain how these factors influenced your
choice.
In order of importance.
1. Existence of earthquakes of m|^ >_ 4.5
2. Existence of basement geologic structures
3. a and b values
My answer to Question 2 attempted to explain how these factors influenced my
choice.
A-3
m^
.'/•'/.
M
^^^t'.
'.".■/
m
i
I'M
It-
R-i
M:
pi;
^B^i'
.* .- .
QUESTION 4 Describe, briefly, the influence of, and your use of, these
features in the development of your zonation maps.
I believe that I answered this in my response to Question 2.
QUESTION 5 Did you feel that these two ways of describing uncertainty
provided you with adequate means of expressing your state of
knowledge regarding zonation of the EUS? Discuss the types of
uncertainties you had identified and attempted to model by
specifying a probability of "existence" of a zone and/or
alternative zone boundaries.
My answer to the first of these questions is "yes". In some cases the
distribution of earthquakes of m^^ >_ 4.5 could be associated with different
basement structures. That is, alternate boundaries could be drawn, each of
which could be related to particular geologic structures. The probability
assessment was an expression of my confidence in a particular zone (and its
corresponding geologic structure) as being the correct explanation of the
earthquake activity.
QUESTION 6 Discuss briefly what type of uncertainties you considered when
specifying the bounds for the seismicity parameters.
All recurrence curves must depart from linearity and bend down at some
magnitude. In attempting to estimate maximum magnitude for a region, the
A-4
question becomes: Do all recurrence curves bend down at the same magnitude
(say mjj = 7.5, an upper limit for global earthquake data) or does the bending
down occur at different m^ values in different source zones? If the answer is
that all bend down at a particular mj^j value, then that value should be the
maximum for all source zones. If, on the other hand, they vary by source
zones, as I believe they do, then the maximum m^ must be estimated for each
source zone. To do so I extended the recurrence curve linearly for a 1000
year return period and used that m^ value as the estimated maximum magnitude.
QUESTION 7 Outline, briefly, the sources of information which formed the
basis for your predictions of the largest magnitude.
Lack of knowledge of active faults, and their rupture lengths, prevents us
from estimating maximum magnitude on this basis. Also, on the basis of the
known lengths of the segments of the New Madrid fault and of the estimated
magnitudes of the 1811-1812 earthquakes, I believe the scaling relations, or
the relations between surface length and m|^ (or M^) , are different for eastern
North America than for plate-margin regions.
In my answer to question 6, I attempted to explain my source of information
used in predicting the largest magnitude. I forgot to mention that, in
constructing the recurrence curve, the area of the source zone must be
equalized to some value, the choice of which is rather arbitrary.
'.'.•'Vv; •-'■
[■ •:•
\'. ■'.'.' ■
.% •
k^ ;,
•:v>>-. ''
i ■':'■■
I'A'r-i—,
*. * ,
-<'j .*fi3B
*<*«9H
*•*- 'l^H
'4 •^^H
''^^1
;>,;;
''•'•'•^^1
* *'^i^l
•.'tv'
C*/»r35H
!>.<'
■'■ ' -l . ''
\\\
.'.*• \' ■'
QUESTION 8 Did you follow a consistent procedure for predicting My? If so,
briefly outline the procedure. Were you influenced by any
constraints, e.g., limits of measurements scale, when considering
your estimate?
■v- ■<■.';-
?^\
A-5
M^.^-'
■-'.•:■
I did follow a consistent procedure, which is described in my reply to
Question 6. My only constraint was an upper limit on the m^ value, which I
believe I took to be m = 7.5.
QUESTION 9 What are the primary sources of uncertainty that you were trying
to describe in determining the upper and lower limits for My?
Uncertainty in the recurrence curve, i.e., in the b and a values.
QUESTION 10 Discuss, briefly, the various issues, e.g.. clusters, catalog
incompleteness, related to using recorded earthquake data as a
basis for estimating the seismicity parameters (a.b).
For seismic source zones that have an adequate amount of data, clustering is
no big problem. When I encountered an obvious cluster, I tried to estimate
the energy released in the cluster, and then replaced the cluster with one
earthquake of m^ corresponding to the total energy released. (I also used
this method for the 1811-1812 New Madrid earthquakes).
Catalog incompleteness must be compensated for. I used a relatively simple
technique, as contrasted with some of those developed in the EPRI study.
I believe that aftershocks should be deleted when the recurrence relation is
to be determined.
QUESTION 11 Outline, and rank, the principal sources, e.g.. analysis based on
your chosen catalog, the results of the LLNL "uniform approach",
you used to develop your estimates of (a.b).
A-6
J
I used the LLNL catalog except for the central United States, where I used my
catalog contained in NUREG/CR-1577.
I used a relatively simple procedure to convert for incompleteness. For a
given source zone, I divided the catalog into equal time intervals (e.g, 10 or
25 years and counted the number of earthquakes in a certain magnitude range
(e.g., 3.25 to 3.75). When these data are placed in tabular form the time
earlier than which the catalog is incomplete for a specific magnitude interval
is fairly well defined.
I also used the fact that in populated areas an earthquake exceeding a certain
magnitude was bound to be observed and reported upon, considering the large
area of perceptibility of eastern North America earthquakes due to low crustal
attenuation.
QUESTION 12 If you used a catalog of recorded events as a basis for
estimating (a,b), to what extent did you use analytical
procedures to adjust for the aftershocks, etc and for
incompleteness? Give a brief description of these procedures.
My method of estimating the duration of aftershock activity is subjective.
Earthquakes of m^j = 7.5 (e.g, 1811-1812 New Madrid) have an aftershock
interval of about 10 years, those of mi^ = 6 an interval of about one year and
those of m|3 = 4 an interval of about one month.
Unless one claims that all earthquake activity between large earthquakes are
aftershocks of the earlier large quakes, I don't believe that aftershocks are
that much of a problem in estimating a and b values.
QUESTION 13 Identify and discuss, briefly, some of the significant issues
impacting characterization of the seismicity of the EUS.
A-7
One important question is whether certain areas which have not displayed much
seismic activity in historic times might in the near future produce a large
earthquake. The Meers fault is a good example. Another is the great China
earthquake of 1556 that killed over 800,000 people, but in recent times
dosen't even show much microseismic activity. If this is a true or real
problem in the EUS, then the concept of a seismic budget, and the
meaningfulness of _a values, must be abandoned. It would make all our attempts
to estimate seismic hazard meaningless.
If we could accurately determine focal depth of micro-earthquakes and the
occasionally larger earthquakes that occur, I believe we could use this
information to estimate maximum magnitude events for a region. Only the
regions that have earthquakes of depth, greater than, say, 10 km would have
the potential for producing a very large earthquake.
A- 8
\
QUESTION 14 What do you consider to be the major sources of uncertainty in
predicting the seismicity parameters (a,b)?
1. Low rate of seismic activity
2. The historic record may be too short (see my reply in Question 13).
3. Poor determination of earthquake magnitudes as given in the
catalogs. I believe this may be a problem in northeastern U.S. and
possibly in southwestern Canada.
4. Incorrect identification of source zone boundaries.
QUESTION 15 Were the alternates presented in the seismicity questionnaires
for modeling potential correlation between (a,b) adequate for you
to express your views about your joint uncertainty about the
seismicity parameters? Discuss your views on correlation between
a and b.
I am not convinced that there is a correlation between a_ and b values. The b^
value should depend on the state of stress and strength of friction on the
fault surface. Although the friction may depend upon frequency of movement.
I don't believe the state and stress does.
A- 9
1
m
•?<:"■
H'
W:[-
m9,:
SEISMICITY INPUT DOCUMENTATION FOR EXPERT 3
QUESTIONNAIRE Q8
QUESTION 1 Discuss briefly the adequacy of the principal sources of
information, including physical characteristics and observational
data, as a basis for identifying source zones.
(1) Historical Seismicity (spatial and temporal) record
a. Too short to show any long term cycles or changes on the order
of millenia.
b. Long enough to show overall general stationarity between non-
instrumental data (centuries) and network data (decade). Those
variations seen are probably expectable dispersions about the
longer term mean rates.
c. Maximum events, e.g. those at New Madrid and Charleston, with
repeat time from seismicity and geology data that are several
times the historical record have fortuitously occurred in a few
places in the region. What surprises are waiting elsewhere is
unknown. This impacts maximum magnitude estimate.
(2) Pal eoseismi city data at New Madrid and at Charleston have given
estimates of return periods that were roughly comparable to those
derived by historical earthquake frequency plus documented the
repetitive nature of those source zones along with their stationarity
in terms of millenia. Mitigates somewhat the deficiency of the
historical seismicity record.
(3) Geological Data - Bedrock geological maps entirely adequate. Sub
surface control generally inadequate (Depth dimension missing,
especially at basement depths).
A- 10
(4) Potential Field Data {gravity + magnetic) generally adequate.
(5) Reflection Seismic data skimpy and of highly variable coverage, but
most useful for 3D information (impacts (3) above).
(6) Ditto for Stress data.
QUESTION 2 Outline your principal bases for identifying zones? To what
extent did you rely on tectonic and geophysical features as a
source of zonation? What, if any, historical seismicity and
other observational data did you consider in your development of
seismic source zones?
(1) Historical Seismicity used as primary baseline factor
Spatial: Lineations, clusters, diffuse distributions, virtually
Aseismic.
Energy Level: Historical maximum, general level over time with
respect to surroundings
Frequency: Persistence over time. Sporadic or continuous?
(2) Regional Geology Provincial structure approach, i.e., large scale
(lOO's km) architecture used to modify boundaries of historical
seismicity to develop seismotectonic provinces (zones); tectonic
history including accreted terranes postulates used in same manner.
(3) Regional Gravity and Magnitude Data - Used only slightly in an
"anomaly pattern" type of approach.
(4) Local Data (in the order of tens of kilometers) used explicitly
whenever available in an attempt to infer where else in one
surrounding region similar structures might be reasonably expected^
Geology, Reflexion Seismology, Pal eoseismi city.
A-11
(5) Vertical Distribution of Seismicity (focal depths)- important to
define thickness of brittle, seismogenic upper crustal layer.
QUESTION 3 Identify some of the factors, such as scaling, lower bound
magnitude, which influenced your development of the EUS zonation
maps. Rank the relative importance of each of these factors,
and, if possible, explain how these factors influenced your
choice.
(1) Size of Zones Except for the three locales in the EUS that have
probably had their maximum earthquake (New Madrid, Charleston, St.
Lawrence Valley). The zones should be relatively large (lOO's km) in
size because of lack of knowledge to be more detailed and because of
a given geologic province containing a family of "similar" structures
as candidates for strain accumulation. Encompass uniform historical
seismicity and similar regional geology.
(2) Shape of Zones - Should conform to the regional geologic and tectonic
"grain" or "fabric" because of last reason given in (1) above.
Encompass "uniform" historical seismicity.
The above comments are of equal importance [
QUESTION 4 Describe, briefly, the influence of, and your use of, these
features in the development of your zonation maps.
Because of "regional size" approach, felt no need for "complementary"
zones" .
"Complementary zone" was defined as that region without any identifiable
character or temporal persistence in the historical seismicity record and
association with a regional geologic province that had such
A-12
QUESTION 5 Did you feel that these two ways of describing uncertainty
provided you with adequate means of expressing your state of
knowledge regarding zonation of the EUS? Discuss the types of
uncertainties you had identified and attempted to model by
specifying a probability of "existence" of a zone and/or
alternative zone boundaries.
Yes
Types of uncertainty-
a. Spatial extent and configuration of zones
b. Location of boundaries between adjacent zones, i.e., the "change"
from one seismic habit to a different one.
QUESTION 6 Discuss briefly what type of uncertainties you considered when
specifying the bounds for the seismicity parameters.
Length and completeness of historical record
Size of geologic host province and activity in adjacent provinces and in
similar provinces elsewhere in world. Amount of detailed studies
(network/aftershock monitoring, reflexion seismology, deep bore holes,
etc)
Occurrence/non occurrence of maximum magnitude event.
QUESTION 7 What does the terminology "95 percent confidence or uncertainty
bounds" mean to you? How did you interpret it?
In repeated trials or calculations, 95% of the bounds will
contain the true value.
A-13
QUESTION 8 Discuss, briefly, the significant issues related to predicting
the largest magnitude earthquake that can be expected to occur
in a source zone.
The absolute size of historical maximum is one of the significant issues
with respect to predicting maximum magnitude size of historical maximum
with respect to size of geologic structure. Another issue is the tectonic
history of the area, probable repeated reactivations, horizontal and
vertical extent of past tectonic episodes, etc. Orientation of candidate
geologic structures with respect to the contemporary East North easternly
compressive stress field.
Also the thickness of the brittle failure, seismogenic layers; maximum
focal depths observed, and the size of probable fault "planes" from
network and/or aftershock monitoring.
QUESTION 9 Outline, briefly, the sources of information which formed the
basis for your predictions of the largest magnitude.
See Q8 above.
QUESTION 10 Did you follow a consistent procedure for predicting My? If
so, briefly outline the procedure. Were you influenced by any
constraints, e.g., limits of measurements scale, when
considering your estimate?
No
Outline of procedure used:
(1) Define background earthquake as m^ = 5 1/2
(2) Determine historical maximum; size and character of spatial
seismicity pattern; size and character of associated geologyj
geophysics reflexion seismlogy features.
A-14
(3) Judge compatibility of elements in (2):
(4) If compatible, use historical maximum as M
until compatibility is achieved.
y. If not, adjust upward
QUESTION 11
What are the primary sources of uncertainty that you were
trying to describe in determining the upper and lower limits
for Mu?
Very subjective, no real basis for procedure except for departure from
compatibility constraints (QIO) by allowing greater than average slip on
smaller faults and less slip than average on larger faults.
QUESTION 12 Discuss, briefly, the various issues, e.g., clusters, catalog
incompleteness, related to using recorded earthquake data as a
basis for estimating the seismicity parameters (a,b).
(1) Removal of Foreshocks and Aftershocks: Generally clear enough that
subjective removal is adequate. Statistical/analytical methods can
be employed but it is problematical if they can do a significantly
better job. In EUS, the number of significant sequences is small
enough that any errors related to their removal should be
insignificant without other errors present.
(2) Variations in detection, location and sizing thresholds with time due
to population levels and numbers of seismographs.
A serious problem with no real effective solutions available;
mitigated somewhat by combining historical data on larger earthquakes
with network data on smaller shocks properlyi
(3) Variations due to societal conditions. Civil War in South; lack of
settlement west of St. Louis in the 18th and early 19th centuries^
A-15
(4) Short length of record with respect to repeat times for larger
shocks. Implies larger events missed in some/many areas?
QUESTION 13 Outline, and rank, the principal sources, e.g., analysis based
on your chosen catalog, the results of the LLNL "uniform
approach", you used to develop your estimates of (a,b).
LLNL uniform approach. Rank as well as any other approach for a large
area/volume of data. That is, short of doing an exhaustive, special study for
each zone.
QUESTION 14 If you used a catalog recorded events as a basis for
estimating (a,b), to what extent did you use analytical
procedures to adjust for the aftershocks, etc and for
incompleteness? Give a brief description of there procedures.
Procedure for adjustment aftershocks and incompleteness. As discussed
earlier, none, except to subjectively delete lower magnitudes that depart from
the straight line formed by the larger magnitudes.
QUESTION 15
Identify and discuss, briefly, some of the significant issues
impacting characterization of the seismcity of the EUS.
Issues on log N = a-bM in EUS.
(1) Range of linear semi-log relationship
Worldwide and regional EUS data indicate OK from smallest magnitude
to somewhere in the upper magnitude range. Just where and how the
departure from linearity occurs is a major problem. Constraints
available are the characteristic earthquake model for individual
fault zones and then for a multiplicity of fault zones, and the
presence of a limiting size/stress level of individual seismogenic
structures based on their geometric and mechanical characteristics.
No clear cut choice among available techniques.
A-16
(2) Method of departure from lineary at the larger magnitudes.
Same as (1) above
(3) Short historical record, sparse data base, long return periods
( Pal eoseismi city)
(4) Subject to significant short term temporal variations in the long
term average rate implied by long N - a-bM. Results in spatial
variations also.
(5) Not enough Paleoseismicity data for independent estimate of return
periods for larger shocks.
(6) Should be 1.0 + 0.2
(7) Can vary from zone to zone because of different populations of
seismogenic structures
QUESTION 16 What do you consider to be the major sources of uncertainty in
predicting the seismicity parameters (a,b)?
(1) b easier to predict reliably
(2) a more difficult because tied to definition of study volume, length
of time considered and lower magnitude cutoff chosen, focal depth
distribution.
(3) a and b are independent parameters but are usually not treated that
way.
(4) Choosing M[_ and My levels for a given catalog.
(5) Definition of time period to be employed in the determination of log
N = a-bM.
i
A-17
QUESTION 17 Were the alternates presented in the seismicity questionnaires
for modeling potential correlation between (a,b) adequate for
you to express your views about your joint uncertainty about
the seismicity parameters? Discuss your views on correlation
between a and b.
Yes Discussed previously.
A-18
SEISMICITY INPUT DOCUMENTATION FOR EXPERT 5
QUESTIONNAIRE Q8
QUESTION 1 Discuss briefly the adequacy of the principal sources of
information, including physical characteristics and observational
data, as a basis for identifying source zones.
The principal sources of information which were used for zonation
were
1. maps of historical and recent instrumental epicenters
2. tectonic maps of North America
3. energy release maps for the Eastern United States
In addition, in specific areas, geologic maps and stress orientation maps as
well as geophysical information were utilized.
Treating the adequacy of the principal sources of information -
The historic maps of epicenters suffer the lack of accurate
information as to exact location and size. Of equal importance is
the lack of any information on depth of occurrence. The recent
instrumental results are useful because of their accuracy for
(mostly) small magnitude events. The general correspondence of the
recent and historical data in map view is encouraging but the
comparison of these two data sources strongly suggested spatial non-
stationarity in the data.
2.
Tectonic maps were adequate to define uniform geological
characteristics but, some known details were not included and of
course, data on more recently documented features such as the Meers
Fault had to be obtained from additional sources.
A-19
3. Energy release maps are simply derived from observed historic and
recent seismicity and subject to all the inadequacies noted above.
QUESTION 2 Outline your principal bases for identifying zones? To what
extent did you rely on tectonic and geophysical features as a
source of zonation? What, if any, historical seismicity and
other observational data did you consider in your development of
seismic source zones?
The principal bases for zonation were listed in Question 1.
In the mid-west, excluding the New Madrid area, equal emphasis was placed on
geophysical and mapped tectonic features and on the historical recent
instrumental seismicity. In the New Madrid zone, all possible sources (and
they are extensive) were used to define the zone. In the Eastern United
States, major emphasis was placed on the seismicity although geologic maps
were used as appropriate.
It should be noted that the alternative model that I proposed involving a
single zone along the Atlantic seaboard was based both on the seismicity
patterns and on the stress patterns. Future work may change my estimate of
the correctness of that alternative model.
QUESTION 3 Identify some of the factors, such as scaling, lower bound
magnitude, which influenced your development of the EUS zonation
maps. Rank the relative importance of each of these factors,
and, if possible, explain how these factors influenced your
choice.
A- 20
I have always felt that the size of the zone in any seismic zonation process
must be a direct reflection of the amount of information available to the
zoner and the degree of certainty of the zoner in the results. An extremely
large number of small zones in the Eastern United States are not justified by
the totality of currently available data. In fact, the alternative model that
I proposed suggests the nature of both the uncertainty and the information
available. Given the uncertainty in all the observational data (particularly
the locations and sizes of seismicity , and the data-cutoffs, the zonation has
to be broad. Certainly, none of the zonation maps by any of the experts can
be considered to be more than an educated guess until we understand the
causative mechanisms of intraplate seismicity. Given the lack of this
understanding, zonation, in my mind, must be general and subject to the caveat
that it is incorrect. The hope (and it should be clearly identified as such)
is that the totality of the expert zonation will bracket the uncertainty in
zonation. The increasing evidence for non-stationarity of occurrence may
invalidate that hope.
Different factors affected my advice of zones in different areas but the most
important factor was the amount of information available. Although lower
magnitude cutoff played a role, it was not a primary one except where a
"characteristic" earthquake on a particular geologic feature entered into the
zonation.
QUESTION 4 Describe, briefly, the influence of, and your use of, these
features in the development of your zonation maps.
See Question 3.
QUESTION 5 Did you feel that these two ways of describing uncertainty
provided you with adequate means of expressing your state of
knowledge regarding zonation of the EUS? Discuss the types of
A-21
uncertainties you had identified and attempted to model by
specifying a probability of "existence" of a zone and/or
alternative zone boundaries.
The two ways of describing uncertainty provided adequate means of expressing
my state of knowledge regarding zonation of the Eastern United States. All of
the uncertainties discussed earlier entered in my decisions regarding zonation
particularly the alternative zones. Certainly, the alternative zone may prove
to be the 'correct' one and, if so, its probability of existence would be
(and, of course may now be) 100%.
QUESTION 6 Discuss briefly what type of uncertainties you considered when
specifying the bounds for the seismicity parameters.
Uncertainties considered included:
1. Uncertainties in historic and recent (instrumental) data base
particularly intensities and/or magnitudes.
2. Knowledge unknowns - particularly lack of knowledge of the causative
mechanisms - e.g., if you don't know what causes these events, how
can you assign an upper magnitude limit?
3. Uncertainty in extrapolating seismicity parameters to larger than
observed and low magnitudes.
QUESTION 7 What does the terminology "95 percent confidence or uncertainty
bounds" mean to you? How did you interpret it?
Assuming a Gaussian or bell shaped curve centered on the 'best estimate'
value, the true value will lie within ±2s of the 'best estimate' value.
A-22
QUESTION 8 Discuss, briefly, the significant issues related to predicting
the largest magnitude earthquake that can be expected to occur
in a source zone.
There are several significant issues related to predicting the upper magnitude
cutoff.
1. Is there an upper magnitude cutoff?
2. The limitations of the historic record particularly its limited
duration relative to postulated return periods.
3. The applicability of the 'characteristic' earthquake model in the
Eastern United States.
4. Numerous others.
QUESTION 9 Outline, briefly, the sources of information which formed the
basis for your predictions of the largest magnitude.
Sources used in the prediction of the upper magnitude included the following;
1. seismicity parameter plots based on the historic catalog and recent
instrumental data - used independently
2. geologic and geophysical data (and tectonic interpretations)
3. physical constraints
A-23
QUESTION 10
Did you follow a consistent procedure for predicting My? If so,
briefly outline the procedure. Were you influenced by any
constraints, e.g., limits of measurements scale, when
considering your estimate?
The prediction of the upper magnitude cutoff involved the following procedure:
1. Select an initial M^j based on plots of historic and recent
instrumental seismicity (where available). A usual starting value
was one intensity unit greater than that observed to date in the
historic record.
2. Modify the initial value based on geologic and geophysical inputs
such as known fault characteristics (where available),
'characteristic' earthquake limits etc.
3. Modify for any effects of physical constraints. Among others, I am
convinced that most regions are capable of producing an intensity VI
event so that, in effect set a lower limit to I^^.
QUESTION 11 What are the primary sources of uncertainty that you were trying
to describe in determining the upper and lower limits for My?
The primary source of uncertainty that was described by the selection of the
upper and lower bounds of M^j (or l^) was whether such an entity existed.
QUESTION 12 Discuss, briefly, the various issues, e.g., clusters, catalog
incompleteness, related to using recorded earthquake data as a
basis for estimating the seismicity parameters (a,b).
A- 24
Ideally, to estimate seismicity parameters, one should have both a 'perfect'
catalog, i.e., one which has exact locations, sizes, etc., and a 'complete'
catalog, i.e., one with uniform size coverage as a function of time. The best
catalogs are, of course, poor approximations to these ideal catalogs.
Issues affecting the choice of seismicity parameters are:
1. present catalogs have large uncertainty and errors in location and
size
2. present catalogs are incomplete at different size levels at different
points in time.
3. present catalogs normally include aftershocks, ice-quakes, etc.,
which probably should not be included in choosing seismicity
parameters.
QUESTION 13 Outline, and rank, the principal sources, e.g., analysis based
on your chosen catalog, the results of the LLNL "uniform
approach", you used to develop your estimates of (a,b).
Analysis of modified catalog entries was used exclusively as the basis of
assigning a and b values.
QUESTION 14 If you used a catalog recorded events as a basis for estimating
(a,b), to what extent did you use analytical procedures to
adjust for the aftershocks, etc and for incompleteness? Give a
brief description of these procedures.
Starting with a modified catalog (based on our own catalog), the following
simple procedures were used:
A- 25
1. Aftershocks (greater than Iq IV) were removed if they occurred withii
a one year period following an I^ - VII or greater event.
2.
3.
Only events in the range I^ = IV and above were used in the analysis
Where known incompleteness was assumed at Iq IV, those data were
disregarded in the analysis. The same procedure was used at higher
intensities in earlier times.
QUESTION 15 Identify and discuss, briefly, some of the significant issues
impacting characterization of the seismicity of the EUS.
The most significant issues that impact characterization of the seismicity of
the Eastern United States are;
1. the lack of knowledge of the causative mechanisms of the seismicity
2. the lack of an adequate catalog
3. the lack of knowledge of 'correct' zonation, i.e., which events
should be included in the analysis
QUESTION 16 What do you consider to be the major sources of uncertainty in
predicting the seismicity parameters (a.b)?
The major sources of uncertainty in predicting seismicity parameters are:
1. the historic (and recent) data uncertainties
A-26
I
correct location
correct size
incompleteness of the data set
2. the possible non-stationarity of the seismicity , i.e., the Meers
Fault question, eastern Massachusetts, etc.
3. the uncertainty in zonation
QUESTION 17 Were the alternates presented in the seismicity questionnaires
for modeling potential correlation between (a,b) adequate for
you to express your views about your joint uncertainty about the
seismicity parameters? Discuss your views on correlation
between a and b.
The alternatives presented in the seismicity questionnaires for modeling
potential correlation between (a,b) were adequate to express my views of the
joint uncertainty about the seismicity parameters. My views on the
correlation are simply that a and b are independent parameters.
In the final analysis, empirical judgements or 'educated guesses' are used to
derive these parameters. Until the causative mechanisms of these intraplate
events are known, all derived seismicity zonation and parameters no matter how
involved with detailed mathematical analyses, will remain 'best estimates'.
A-27
SEISMICITY INPUT DOCUMEKTATION FOR EXPERT 6
QUESTIONNAIRE Q8
QUESTION 1 Discuss briefly the adequacy of the principal sources of
information, including physical characteristics and
observational data, as a basis for identifying source zones.
I think that the observed seismicity is the only really viable source of
information for identifying source zones in most (if not all) of the eastern
United States (EUS). In principle, physical characteristics should be the
most important source of information. However, so little is known about the
cause of earthquakes in the EUS that it is hard to say what physical
characteristics should be considered as significant for identifying source
zones.
Furthermore, the generally accepted hypothesis—that EUS earthquakes occur
when the stress locally exceeds the strength of preexisting zones of
weakness— doesn't help me very much in identifying source zones. I consider
that hypothesis to be true by definition, but I don't know how to apply it.
For example, what zones of weakness should I be looking for? How might these
zones of weakness be identified from geologic structures that outcrop on the
surface or from deeper earth structures inferred from geophysical data? I am
not even sure what scale of structures I should be looking for.
The part of the EUS that I am most familiar with is the northeastern region
(NEUS), and in that region the identification of specific active features has
proven to be quite difficult. Unlike the situation along plate boundaries, it
is not at all clear whether faults mapped at the earth's surface in the NEUS
are the same faults along which the earthquakes are occurring. I know that I
have allowed my familiarity with the NEUS to bias my source zonation of other
parts of the EUS, and it is hard for me to avoid that. Although I am not as
familiar with earthquake activity in New Madrid, MO area, my sense is that the
A-28
cause of the earthquakes in that area are better understood. Thus, my
zonation of the New Madrid area is more dependent on physical characteristics
than are my zonations of other parts of the EUS.
QUESTION 2 Outline your principal bases for identifying zones? To what
extent did you rely on tectonic and geophysical features as a
source of zonation? What, if any, historical seismicity and
other observational data did you consider in your development of
seismic source zones?
The principal basis for identifying source zones on my seismic zonation maps
was locations of historical and instrumental ly located epicenters. I compared
a number of different maps of seismicity in the EUS, including LLNL and EPRI
maps as well as a recent (preliminary) map of seismicity in the U.S. that will
eventually be published in one of the Decade of North American Geology
volumes. The seismicity maps that I examined were plotted on various
different scales and/or using different symbols (e.g., smaller symbols for
smaller events) .
I had more detailed maps of seismicity and geologic structures available to me
for the NEUS. Nonetheless, I tried to make the scale of source zones
homogeneous throughout the EUS, unless I had a particular physical reason for
characterizing an area on a smaller scale. Thus, my source zones for the NEUS
probably have less detail than they might have had if that was the only region
I was analyzing, and my zones for other parts of the EUS probably have more
detail than they would have had if I was unaware of certain details in the
NEUS.
In some cases tectonic and geophysical features were used to define
boundaries. In general, however, physical characteristics "took a back seat"
in my judgements (see response to Question 1).
A-29
QUESTION 3 Identify some of the factors, such as scaling, lower bound
magnitude, which influenced your development of the EUS zonation
maps. Rank the relative importance of each of these factors,
and, if possible, explain how these factors influenced your
choice.
(1) Scale of Details :
With the exception of a few specific areas (e.g., New Madrid, MO), I have
no compelling reason to believe that the scale of source zones should
vary across the EUS. Thus, I generally tried to make the scale of source
zones homogenous throughout the EUS unless I had specific physical
reasons to characterize an area with greater detail. To some extent, I
relied on my understanding of the scale of details in areas that I am
most familiar with (primarily the NEUS). At the same time, I tried not
to overemphasize details in areas that I am more knowledgeable about,
because the resulting map would appear to imply an unrealistic
distribution of scales of features. To that end, I tried to think of
what NEUS would look like to seismologists who study other parts of the
world, and from that perspective I tried to zone the rest of the EUS.
I suppose that this approach is inherently biased by what I know about
the NEUS, but I couldn't think of any way to avoid that bias.
(2) Lower bound magnitude :
I think that an mi^^g 3. 75 earthquake could occur just about anywhere in
the EUS, but I am not as convinced that an m^^ 5.0 event could occur
anywhere in the region. Most likely, that belief influenced my
development of zonation maps, but it is hard to say how.
A- 30
QUESTION 4 Describe, briefly, the influence of, and your use of, these
features in the development of your zonation maps.
(1) Background :
I was not really clear on the significance of the distinction between
background zone (EPRI) and complementary zone (LLNL) described on page 9
of Q7. I guess the idea is that in the case of the background you
consider "features", whereas in the case of complimentary zone it is just
a matter of "what is left over" after the rest of the zonation map is
finished. But, how does that translate into differences in assigned
seismicity parameters?
(2) Complementary Zone ;
In my zonation maps, the complementary zone is the part of the EUS that
was left over after all other zone boundaries were created. I was not
very confident in the zone boundaries that I was able to create from the
available source of information, so I tried to find a few different ways
to compensate for that uncertainty. For example, the maximum magnitude
for the complementary zone was large (My = 6.0 + 0.5), and I allowed for
two alternative maps of zonation. Thus, my zonation and seismicity
parameters were intended to allow at least some probability of a damaging
earthquake occurring anywhere in the EUS.
Given the currently available information about earthquake activity in
the EUS, I did not see any reason for dividing the complementary zone
into regional complementary zones.
QUESTION 5
Did you feel that these two ways of describing uncertainty
provided you with adequate means of expressing your state of
knowledge regarding zonation of the EUS? Discuss the types of
A-31
'•■•^':
■
Ut0^<
1
mm::.
'/•%'•',
&jp' \
uncertainties you had identified and attempted to model by
specifying a probability of "existence" of a zone and/or
alternative zone boundaries.
These two ways of describing the uncertainty generally seemed adequate, but I
think that the way I applied this aspect of the LLNL methodology in my
responses to Q7 was a little different from the way it was intended to be
applied. During the "one-on-one interview" regarding my responses to this
questionnaire (June 25, 1987), I discussed my approach regarding this aspect
of the LLNL methodology with J. Savy and D. Bernreuter. At the end of that
meeting, it seemed that they were able to incorporate my responses (and what I
had intended to model) into the LLNL methodology.
(1) Probability of Existence :
I tried to identify and model the probability that a particular zone was
seismically different from the area surrounding it.
(2) Alternative Zone Boundaries ;
Original Zonation Map (response to Ql): Given that a zone exists, I
tried to identify and model the probability that my best estimate map had
the correct boundaries for that zone. My lack of confidence regarding
the existence of the zones in that map was accounted for by proposing
alternative zone boundaries.
Revised Zonation Map (response to Q7): In my revised zonation map, I
used my original best estimate map as an alternative map rather than
specifying alternative zones for the revised map. Also, I added a third
map which consisted of one "mega-zone" including the entire EUS without
internal zonation boundaries. So in this case, I think that what I
really submitted was "alternative maps" rather than "alternative
zones". My level of confidence values for alternative boundary shapes
A-32
are, therefore, really level of confidence values for each of the three
maps.
QUESTION 6 Discuss briefly what type of uncertainties you considered when
specifying the bounds for the seismicity parameters.
The uncertainties that I considered when specifying the bounds for the
seismicity parameters were:
(1) Incomplete recording in catalogues.
(2) Temporarily and/or spatially varying seismicity parameters.
(3) The possibility that even if we could correct for incomplete recording
and if seismicity parameters are stationary within a given zone, we may
nonetheless be limited by observations over too short a period of time to
see the real, stationary seismicity parameters.
QUESTION 7 What does the terminology "95 percent confidence or uncertainty
bounds" mean to you? How did you interpret it?
95% Confidence Bounds: For a given parameter, p, there is a probability of
0.95 that the true value for p lies somewhere within the 95% confidence
bounds.
Uncertainty Bounds: I interpret this be a more general concept indicating the
limits of a range of values that contain the true value of p with some
(unspecified) level of confidence.
A- 33
QUESTION 8 Discuss, briefly, the significant issues related to predicting
the largest magnitude earthquake that can be expected to occur
in a source zone.
In principle, it makes the most sense to use physical characteristics (such as
fault length, rupture area, stress drop, etc.) to predict maximum magnitudes
for a given source zone. However, I don't think that such information is very
useful for most of the EUS, given our present lack of understanding of the
cause of EUS earthquakes. As in the case of other parameters, I relied
heavily on the observed record of seismicity in estimating maximum magnitudes.
Again, I consider the New Madrid, MO area to be an exception to this general
rule. In the case of New Madrid area earthquakes, the record of seismicity is
very well documented and the physical characteristics in the area have been
extensively studied. Thus, physical characteristics played a larger role in
my decision for maximum magnitude in this area than they did in other parts of
the EUS.
QUESTION 9 Outline, briefly, the sources of information which formed the
basis for your predictions of the largest magnitude.
As in the case of other parameters, the observed record of seismic activity
was the most important source of information that I used in estimating maximum
magnitudes. Generally, once I defined a source zone, the most significant
factor in choosing maximum magnitude was the largest earthquake in the
historical and instrumental record of that zone. Another significant factor
was the number of smaller earthquakes that occurred in that zone. This factor
was considered because an exceptionally large number of smaller earthquakes in
a given area may be indicative of the possibility of larger earthquakes.
A- 34
QUESTION 10 Did you follow a consistent procedure for predicting My? If so,
briefly outline the procedure. Were you influenced by any
constraints, e.g., limits of measurements scale, when
considering your estimate?
Yes, I followed a consistent procedure for predicting My. I began by defining
the following categories of zones:
(1) A zone that is capable of generating a great intraplate earthquake
(Mu=7.3).
(2) A zone that is capable of generating a large intraplate earthquake
{Mu=6.8).
(3) A zone that is capable of generating a moderate- to-large intraplate
earthquake (My=6.5).
(4) All other zones (My=6.0). This "background My" was chosen because I
didn't feel comfortable assuming that there is any place in the EUS where
a magnitude 6.0 earthquake could not possibly occur.
-■ ■ ' m' y,s"''
'\i>:^- ■■:■:■:
Once these categories were defined, I evaluated each zone to decide which
category it belonged to.
My definition of a great intraplate earthquake was limited by the size of the
1811-1812 New Madrid, MO earthquakes and the fact that I am not aware of any
intraplate earthquakes that are larger than the New Madrid events.
QUESTION 11 What are the primary sources of uncertainty that you were trying
to describe in determining the upper and lower limits for My?
:^v^--'->>\
A- 35
■O # < /
■'.y/f'.
%
W-^'
">y .'
>'^^'
W^.
li
w/V.
/t^
'■'.■ .■■'■''!
. '.-,•'/-<
1 , ' ' -' ■ ■
"SSL
'• v.\ ,
Wk-:
>:•:■.);•:•;
^m:^.Hr.
The primary sources of uncertainty that I considered in determining Mnn and
M|j|_ were:
(1) I considered the magnitudes of the largest intraplate earthquakes of
which I am aware.
(2) I didn't feel comfortable with allowing maximum magnitude estimates to
get too low anywhere. Thus, my choice of lower limits for maximum
magnitude are probably "on the high side".
(3) In my choice of upper limits for the more active zones, I considered the
possibility that the interior of the North American plate could actually
generate an earthquake larger than the New Madrid events.
QUESTION 12
Discuss, briefly, the various issues, e.g., clusters, catalog
incompleteness, related to using recorded earthquake data as a
basis for estimating the seismicity parameters (a,b).
Adjusting for aftershocks, clusters and dependent events seemed like too
complex an issue for me to tackle given the amount of time that I had to
respond to the questionnaires. So, initially (as best as I can remember) I
didn't give much thought to the question of dependent events in the
catalogue. Later, after being given the opportunity of using the EPRI
catalogue, I decided to trust the identification of main events and dependent
events made by EPRI.
Incompleteness of catalogues is such an obvious problem that it was clear from
the start that I couldn't ignore it. In the first phase of this project
(responses to Q2), I dealt with incompleteness by developing a qualitative
method for correcting the activity rate in each zone for incompleteness.
A- 36
For the most recent phase of this project (responses to Q9), I used the LLNL
uniform approach to estimating (a.b) as an initial estimate, and then I
intuitively decided whether or not to believe these results or somehow
readjust them.
QUESTION 13
Outline, and rank, the principal sources, e.g., analysis based
on your chosen catalog, the results of the LLNL "uniform
approach", you used to develop your estimates of (a,b).
(1) An analysis based on my chosen catalogue and the results of the LLNL
uniform approach were the most important sources that I used to develop
my estimates of (a,b). These two sources were given approximately equal
weight.
(2) Another important factor was a sense that b-values of approximately 1.0
were reasonable.
(3) In some cases I concluded (qualitatively) that the data and/or the LLNL
uniform approach results just didn't "look right" to me, and so, I
adjusted the values of (a,b). This aspect of my procedure for estimating
(a,b) is very difficult to quantify, but it was used in a number of
cases.
QUESTION 14 If you used a catalog recorded events as a basis for estimating
(a,b), to what extent did you use analytical procedures to
adjust for the aftershocks, etc and for incompleteness? Give a
brief description of these procedures.
Adjusting for aftershocks, clusters and dependent events seemed like too
complex an issue for me to tackle given the amount of time that I had to
respond to the questionnaires. So, initially (as best as I can remember) I
A- 37
didn't pay too much attention to this issue. Later, after being given the
option of using the EPRI catalogue, I decided to trust the identification of
main events and dependent events made by EPRI.
In the first phase of this project (responses to Q2) , I dealt with
incompleteness by developing a method for estimating (qualitatively) an
"equivalent period of time for completeness", T. for each zone. Then I
assumed that the catalogue was approximately complete for the past T years. T
was estimated "visually" from the lists of earthquakes in each of my source
zones provided by LLNL. I also "visually" fit a line through the data in the
plots (provided by LLNL) of the cumulative number of earthquakes in different
magnitude ranges for each of my source zones. The "equivalent period of time
for completeness" was used to convert my a-value estimates into units of
number of events per year.
For the most recent phase of this project (responses to Q9) , I used the LLNL
uniform approach to estimating (a,b) as an initial estimate, and then I
intuitively decided whether or not to accept these results as they were or to
somehow readjust them. I suspect that in at least 50% of the source zones, my
results were quite similar to the LLNL uniform approach results.
QUESTION 15 Identify and discuss, briefly, some of the significant issues
impacting characterization of the seismicity of the EUS.
I think that the most significant issue impacting characterization of the
seismicity of the EUS is that for most (if not all) of this region, the cause
of the earthquakes is unknown. The historical and instrumental records of
seismicity are the only truly viable sources of information that we have.
Yet. I find it hard to accept that this record is anything more than just a
snapshot in time. It seems to me that when using the LLNL methodology, we are
to some extent assuming that the process that generates EUS earthquakes is
stationary. That is, a source zone is either active or not active; if it is
A-38
active, then it has a particular set of seismicity parameters. It is
possible, however, that the pattern of seismic activity in the EDS changes
with time. That would imply that an area of the EUS that has been very active
for the past several hundred years could cease to be active and a previously
inactive area could become active. I tried to address that possibility by
choosing the large mega-zone as an alternative map. That map represents the
hypothesis that "anything can happen" in the EUS, and the seismicity
parameters for the large mega-zone are intended to characterize the general
nature of seismic activity in the interior of the North American plate.
QUESTION 16 What do you consider to be the major sources of uncertainty in
predicting the seismicity parameters (a,b)?
(1) Do we have a good enough data base from which to estimate (a,b)? Is the
time period of this data base long enough? Is the catalogue completely
and accurately recorded?
(2) Is the process that generates earthquakes in the EUS stationary?
QUESTION 17 Were the alternates presented in the seismicity questionnaires
for modeling potential correlation between (a,b) adequate for
you to express your views about your joint uncertainty about the
seismicity parameters? Discuss your views on correlation
between a and b.
'%
I realized, after the first feedback meeting, that this was a difficult issue
to deal with. It seemed to me that unless I was very careful in estimating
the bounds on (a,b) for each case, there was a good chance that some of the
implications of my inputs would yield a much wider range of possible
recurrence relationships than I had intended. The alternatives presented for
modeling correlation between (a,b) helped me to understand the implications of
A- 39
the uncertainty bounds that I chose for (a,b). In the most recent phase of
this project (Q9). I chose fully correlated because I thought that option made
it easier for me to keep track of the range of possibilities that I intended.
A-40
SEISMICITY INPUT DOCUMENTATION FOR EXPERT 7
QUESTIONNAIRE Q8
QUESTION 1 Discuss briefly the adequacy of the principal sources of
information, including physical characteristics and observational
data, as a basis for identifying source zones.
In most parts of the Eastern U.S. neither the seismicity nor the understanding
of the geologic and tectonic features are adequate to outline well-defined
source zones.
QUESTION 2 Outline your principal bases for identifying zones? To what
extend did you rely on tectonic and geophysical features as a
source of zonation? What, if any, historical seismicity and
other observational data did you consider in your development of
seismic source zones?
a. Seismicity (historic and instrumental) is the single most Important
data set to identify a region as a source zone.
b. Geologic and geophysical data are used in addition to seismicity to
define the shape and boundaries of a source zone.
c. Only in one case Meers Fault (Oklahoma) that the zone could be
defined on the basis of geology.
QUESTION 3 Identify some of the factors, such as scaling, lower bound
magnitude, which influenced your development of the EUS zonation
maps. Rank the relative importance of each of these factors,
and, if possible, explain how these factors influenced your
choice.
A-41
The scale of the seismic zones is chosen for a regional hazard assessment.
The zones are too broad to be used for site-specific hazard calculation. This
latter effort requires micro-zonation. The lower bound magnitude made some
difference by broadening specific zones and reducing the areas of
"background". Had M|_b been 5.0, background would have been larger in area.
QUESTION 4 Describe, briefly, the influence of. and your use of. these
features in the development of your zonation maps.
Background helped account for diffused seismicity where fA^5 in general.
This took place of many poorly defined seismic zones. I used three scales
a.
A well defined relatively small source zone to define regions such as
New Madrid, La Moodis, etc. of known significant earthquakes and for
seismogenic features. These zones may be enclosed 1n zones described
1n (b).
b.
Larger zones which are characterized by diffused but distinctive
seismicity and tectonics. Many zones fall into this category. Some
of these zones are quite large.
c.
Background- all regions not included in (a) or (b). Three areas have
varying degrees of seismicity and could have significant potential.
Yet I cannot distinguish these as distinct on geological geophysical
data.
QUESTION 5
Did you feel that these two ways of describing uncertainty
provided you with adequate means of expressing your state of
knowledge regarding zonation of the EUS? Discuss the types of
uncertainties you had identified and attempted to model by
specifying a probability of "existence" of a zone and/or
alternative zone boundaries.
A-42
I assumed that each zone exits with equal likelihood. There is some
probability incorporated in an implicit way by the procedure outlined in 4.
Examples are the "well defined" zones such as New Madrid that are shown as a
specific zone within a larger zone. The "alternate zone" idea sounded great
at first when it was introduced in the EPRI study. However, this is practical
only for regional or site-specific studies. For the general eastern U.S.
study there is no practical way of doing this unless the level of effort is
raised by order(s) of magnitude.
QUESTION 6 Discuss briefly what type of uncertainties you considered when
specifying the bounds for seismicity parameters.
None other than described under questions 4 and 5.
QUESTION 7 what does the terminology "95 percent confidence or uncertainty
bounds" mean to you? How did you interpret it?
95 confidence bounds: Probability is 0.95 that the "true value" is within the
bounds of the parameter estimate (bounds).
QUESTION 8 Discuss, briefly, the significant issues related to predicting
the largest magnitude earthquake that can be expected to occur
in a source zone.
Most physical parameters (such as the lengths of fault or fault segments,
fault width, stress drop, in-situ stress, strain accumulation rate, etc.) are
difficult to estimate for the Eastern U.S.
QUESTION 9 Outline, briefly, the sources of information which formed the
basis for your predictions of the largest magnitude.
A-43
I relied primarily on two observed parameters:
a.
b.
Largest magnitude event in a given source zone.
Magnitudes of intraplate earthquakes over the globe, and I believe
there is a characteristic maximum earthquake for each zone.
QUESTION 10
Did you follow a consistent procedure for predicting My? If so,
briefly outline the procedure. Were you influenced by any
constraints, e.g., limits of measurements scale, when
considering your estimate?
For estimating upper bound magnitude (Myg) I followed the following proced
ure;
If the largest earthquake that occurred in a source zone is (M^^)*, then
for
(Mb) <5.0
5.0l(Mtj)* <6.0
6.0<(M5)*
MUB = (Mb) plus 1
MuB = (^b)* + 0-75
MuB = (Mb) + 0.5
The following are considered in the above formulation
a.
b.
c.
Uncertainty of (Mb)*, for small (Mb),* is large i.e., when {%*±b/0)
Probability of missing an event of (Mb)* <5.0) is significant.
The compression of mb scale as we go to large magnitudes.
Had we used an M^ or M^ (moment-magnitude), the added increments would have
been different.
QUESTION 11 What are the primary sources of uncertainty that you were tryi
to describe in determining the upper and lower limits for My?
ng
A-44
The uncertainty listed in 10 plus the uncertainty of defining My in terms of
the maximum magnitude observed event, given that the time period over which
data are available is limited.
QUESTION 12 Discuss, briefly, the various issues, e.g., clusters, catalog
incompleteness, related to using recorded earthquake data as a
basis for estimating the seismicity parameters (a,b).
We need to calculate earthquake hazard due to all events rather than for those
that obey the exponential Poisson distribution. Thus we should remove as few
events as possible from the catalog.
There are several approaches:
a. Take out as few events as possible to reduce severe effects of
clustering and still use the exponential-Poisson model.
b. Eliminate clustering, use exponential-Poisson model and then
calculate hazard independently for those events removed.
c. Use a model other than Poisson.
I find approach "a" is most practical if applied the following way: (1) Ignore
the foreshocks. They are too few. (2) Remove "aftershocks". Define
aftershocks as "events clustered" in time and space and with magnitudes at
least one magnitude less than the main shock. (3) After removing aftershocks,
add the normal "background" seismicity for that zone.
I
\tm^'
There are good examples of events migrating along a "fault zone" over several
months or years (e.g.. New Madrid, Gaza-USSR, Song Pan - China). A typical
de-clustering routine will take out such independent events.
QUESTION 13 Outline, and rank, the principal sources, e.g., analysis based
on your chosen catalog, the results of the LLNL ■uniform
approach", you used to develop your estimates of (a,b).
A-45
I used LLNL values except where LLNL fits to seismicity data appeared to be
poor (where data are few). In such cases I fitted the data generally with a
"bias" that b = 1.0.
QUESTION 14
If you used a catalog recorded events as a basis for estimating
(a.b), to what extent did you use analytical procedures to
adjust for the aftershocks, etc and for incompleteness? Give a
brief description of these procedures.
Used LLNL catalog.
QUESTION 15 Identify and discuss, briefly, some of the significant issues
impacting characterization of the seismcity of the EUS.
No information was given on this question.
QUESTION 16 What do you consider to be the major sources of uncertainty in
predicting (a,b)?
No information was given on this question.
QUESTION 17 Were the alternates presented in the seismicity questionnaires
for modeling potential correlation between (a.b) adequate for
you to express your views about your joint uncertainty about the
seismicity parameters? Discuss your views on correlation
between and and b.
No information was given on this question.
A-46
SEISMICITY INPUT DOCUMEHTATION FOR EXPERT 10
QUESTIONNAIRE Q8
QUESTION 1 Discuss briefly the adequacy of the principal sources of
Information, including physical characteristics and observational
data, as a basis for identifying source zones.
The principal elements for identification of source zones
consists of seismicity and geological/geophysical data. The
geological data at earthquake depths has to be inferred from
geophysical data or extrapolated regional geological data.
The seismicity data base has been substantially improved in the
last several years and for purposes of selecting source zones, is
adequate. With a few exceptions, the geophysical data base is
spotty on a regional basis and even more so on a local basis for
use in correlation to epicentral locations.
While many seismologists are willing to use the seismicity data
base exclusively since earthquakes are the best manifestation of
a source zone, the observational time period for earthquakes is
very short and extrapolation to longer time must be done on a
geological basis.
Identifiable tectonic structures associated with earthquakes are
the best criteria for identifying source zones, unfortunately,
the data base [geophysical] for doing this is variable.
Consequently, experts either ignore it because of its limitations
or use it; but without accounting for its variations. Areas
where data are absent often are those assigned a low probability
of occurrence, out of ignorance rather than knowledge.
A-47
QUESTION 2 Outline your principal bases for identifying zones? To what
extend did you rely on tectonic and geophysical features as a
source of zonation? What, if any, historical seismicity and
other observational data did you consider in your development of
seismic source zones?
Identified zones were selected on the basis of seismic,
geological, and geophysical characteristics. Geophysical data
and known tectonics, where available in sufficient detail, were
heavily relied on. Historical seismicity was also heavily
relied on. Of particular interest are those earthquake
epicentral areas which have comparative geological/geophysical
features.
QUESTION 3
Identify some of the factors, such as scaling, lower bound
magnitude, which influenced your development of the EUS zonation
maps. Rank the relative importance of each of these factors,
and, if possible, explain how these factors influenced your
choice.
Geophysical maps in the eastern United States [EUS] exist at
scales ranging from 1:5,000 to 1:1,000,000 and the resolution is
dependent on these scales.
Earthquake magnitudes below 3.5 were not considered unless the
earthquakes had a definable pattern and/or correlated with
identifiable tectonic features.
The factors considered were: 1] the presence of earthquakes and
the correlation of tectonic features to the larger earthquakes
[i 4.5]; 2] the presence or absence of geological/geophysical
data to identify tectonic features at an appropriate scale, i.e.
A-48
identification of tectonic features at 5-10 mile lengths or
greater; 3] orientations and history of the tectonic
feature[s], if known; 4] physical characteristics of the
feature versus the surrounding rock, for example, rigid versus
non-rigid crustal blocks. A numerical rank was not assigned to
these.
QUESTION 4 Describe, briefly, the influence of, and your use of, these
features in the development of your zonation maps.
The background or complimentary zones are generally a
manifestation of ignorance of "correlatable" tectonics and
seismicity. This was and is many times due to an inadequate
geological/geophysical data base. At the required scale, it
appears that similar tectonic features exist where earthquakes
have occurred and where they have not. This could be due to an
insufficient observational time for earthquakes or lack of
necessary detail on the tectonic features or both. The use of
background zones, while an attempt to express uncertainty, tends
to homogenize the risk of assigning an even distribution of
earthquake over a larger area than the individual structures
which causes them.
This either unduly penalizes a site which is not near or on an
earthquake structure, or optimistically considers the threat as
having a much lower probability than is real if it is on or near
an earthquake producing structure. Depending on the size of the
area chosen as compared to the "real" size of a causative
structure, the error in probability could be substantial.
QUESTION 5 Did you feel that these two ways of describing uncertainty
provided you with adequate means of expressing your state of
knowledge regarding zonation of the EUS? Discuss the types of
I'/
A- 49
uncertainties you had identified and attempted to model by
specifying a probability of "existence" of a zone and/or
alternative zone boundaries.
The seismicity parameters depend on the stress field, the size
of the structure, and the maximum dimensions of structure
involved in one event. Clearly, there is uncertainty with
respect to all of these elements. An estimate of an upper bound
earthquake is gleaned from the past seismic history and in
particular, the nature of the tectonic structures associated
with past earthquakes.
The uncertainties include: the length of the seismic history
compared to the length of extrapolation; the consistency of the
occurrence of given magnitudes in a region or on a structure;
the degree of correlation between the earthquakes and the
geological structures; the expected dimensions of fault failure
based on the estimated stress field amplitude, and the
orientation, nature, and rheology of suspect of identified
faults.
I do not believe that the uncertainty, particularly with respect
to an individual site, is adequately expressed by either
alternate zones, their existence, or change in boundaries.
QUESTION 6 Discuss briefly what type of uncertainties you considered when
specifying the bounds for the seismicity parameters.
QUESTION 7 What does the terminology "95 percent confidence or uncertainty
bounds" mean to you? How did you interpret it?
A-50
The 95 percent confidence bound meant little or nothing since
the uncertainties are so great. It was interpreted in terms of
a value which required a high degree of conservatism [not to
exceed numbers] in the estimated time for which they were
intended to apply, 50 years or thereabouts.
QUESTION 8 Discuss, briefly, the significant issues related to predicting
the largest magnitude earthquake that can be expected to occur
in a source zone.
The largest earthquake prediction depends on the size of the
structure, the distribution of stress [whether localized on a
few or over many structures] and the brittleness of the
structure [rheology] whether it would creep or break. Critical
to the upper magnitude is the time involved in stress built up
and dissipation time. If all dissipation occurs in say a few
hundred years, through minor events and creep, the maximum
earthquake would be small as compared to areas where there is
little dissipation and stress can accumulate for several hundred
years and result in rigid deformation.
QUESTION 9 Outline, briefly, the sources of information which formed the
basis for your predictions of the largest magnitude.
See answer to questions 3 and 5.
QUESTION 10 Did you follow a consistent procedure for predicting My? If so,
briefly outline the procedure. Were you influenced by any
constraints, e.g., limits of measurements scale, when
considering your estimate?
The only way a consistent procedure could be established for Mjj
would be to have consistent input data as given in 3, 5, and 8
A-51
above. Since these data are inconsistent, any procedure for
selecting M^j must be deficient in certain input elements and
therefore, the results would be inconsistent.
QUESTION 11 What are the primary sources of uncertainty that you were trying
to describe in determining the upper and lower limits for Mn?
The degree of knowledge [or ignorance] relative to the elements
given in 3, 5, and 8.
QUESTION 12
Discuss, briefly, the various issues, e.g., clusters, catalog
incompleteness, related to using recorded earthquake data as a
basis for estimating the seismicity parameters (a,b).
The traditional approach for estimation of seismicity parameters
is to deal with main shock events and with completely reported
segments of the earthquake catalog. We have attempted to accomplish
this by eliminating all obvious foreshocks/aftershocks
from our catalog. In addition, as a result of much prior work,
we have identified with a reasonable degree of confidence the
completely reported time intervals for various magnitude events
listed in the catalog.
QUESTION 13 Outline, and rank, the principal sources, e.g., analysis based
on your chosen catalog, the results of the LLNL "uniform
approach", you used to develop your estimates of (a,b).
Seismicity parameters were derived based on analyses of the
earthquake catalog maintained by Weston Geophysical
Corporation. This earthquake catalog is the result of more than
two decades of research on the more important seismic events and
a complete review and integration of other existing catalogs.
Entries in the other major catalogs, such as the LLNL and EPRI
A-52
catalogs, were routinely compared to the data base we maintain
to identify any major discrepancies. Major discrepancies were
handled by reverting back to the original accounts or sources
for the events in question and using this information to
determine the best set of earthquake parameters.
QUESTION 14 If you used a catalog recorded events as a basis for estimating
(a,b), to what extent did you use analytical procedures to
adjust for the aftershocks, etc and for incompleteness? Give a
brief description of these procedures.
Seismicity parameters were determined using a catalog wherein
obvious foreshock and aftershock sequences were flagged and
later removed during statistical analyses. Obvious fore- or
aftershocks included only those that were closely, spatially,
and temporarily linked to a main shock. Formal analytical
cluster-analyses were not employed. Completeness intervals for
magnitude groups were determined for various sub-regions. These
completeness intervals were derived given our understanding of
seismograph network deployments and population expansions. For
example, in the NEUS, significant advances in seismographic
instrumentation occurred in the mid- to late 1970 's.
Characteristics of the expanded networks suggest to us that all
magnitude 3 and larger events have been reported during this
time. Further, due to the distribution of early population
centers in the NEUS, it is concluded that major events 6 have
been completely reported for 250 to 300 years. Variations in
seismographic instrumentation and population expansions in other
regions of the EUS were considered in estimation of completeness
intervals.
QUESTION 15 Identify and discuss, briefly, some of the significant issues
impacting characterization of the seismicity of the EUS.
A-53
See answers to questions 3, 5, and 8.
QUESTION 16 What do you consider to be the major sources of uncertainty in
predicting the seismicity parameters (a,b)?
There are several principal sources of uncertainty in predicting
seismicity parameters. These are summarized below in perceived
order of importance. The most important are listed first.
a.
c.
Completeness interval for various magnitudes - The
completeness interval assumed for the purpose of estimating
earthquake frequencies can significantly affect estimates of
a- and b-values. Completeness intervals should therefore be
carefully examined in the context of timing of seismograph
deployments and population expansions.
Magnitude calculations or estimates - Because the earthquake
catalog for the EUS contains a mix of historical and
instrumental events, it is important to convert all events
to a uniform magnitude scale, such as mb. Empirical
relationships among the parameters mb vs. 10, mb vs. ML, mb
vs. mbLg, etc., can have an important effect on seismicity
parameter results; these relations should therefore be
verified for usage on a regional or local basis.
Magnitude grouping - The methodology used to group
earthquake magnitudes and the magnitude chosen to
characterize a particular group of magnitude, e.g. low
magnitude of cell vs. center magnitude or average magnitude,
can affect seismicity parameter estimates.
QUESTION 17 Were the alternates presented in the seismicity questionnaires
A- 54
'>^k.%.'MM«^1i>'
for modeling potential correlation between (a,b) adequate for
you to express your views about your joint uncertainty about the
seismicity parameters? Discuss your views on correlation
between a and b.
The largest concern that I have with the developed relationship
is that there is an implied functional relationship between the
earthquakes magnitudes. If one were dealing with a single
earthquake producing fault or fault systems then the
relationship between smaller earthquakes and larger earthquakes
may be controlled by the stress field and the rheologies,
asperities, etc. of the fault lithologies and a functional
relationship of small to large magnitudes would serve as a
predictor. However, in a region the smaller earthquakes may be
the result of smaller fault movement and the larger ones due to
larger faults independent of the smaller faults. Therefore, if
the curve is to be used in a regional sense, as applied to a
site, one should know whether or not the site is near a large
fault or smaller fault. The use of a regional functional
relationship to a larger active fault would badly underestimate
to probability of maximum earthquake occurrence.
The controlling factor in the establishment of the curve is the
size of area chosen; other elements, errors, incompleteness,
etc., are not significant contributors if a well documented
earthquake such as that produced by EPRI is used.
Also see answers to questions 3, 5, 8, 12, and 13.
A-55
SEISMICITY INPUT DOCUMENTATION FOR EXPERT 11
QUESTIONNAIRE Q8
QUESTION 1 Discuss briefly the adequacy of the principal sources of
information, including physical characteristics and observational
data, as a basis for identifying source zones.
The available seismicity catalogs are sufficient to present the historical
seismicity. The major question in the data is whether sufficient recording
time has been available to provide a complete statistical distribution of
historical seismicity. Other data (potential and seismic velocity) are
sufficient, but with considerable more seismic data, significant improvements
could be incorporated in the determination of zones. Potential data were used
extensively and no significant improvement in this data should be expected.
QUESTION 2 Outline your principal bases for identifying zones? To what
extend did you rely on tectonic and geophysical features as a
source of zonation? What, if any, historical seismicity and
other observational data did you consider in your development of
seismic source zones?
Seismic zones were defined on the basis of near-surface geology, crustal
composition, crustal thickness, and seismicity. Tectonic and Geophysical
features account for about 70 percent of the definition and seismicity the
remaining 30 percent. The contribution of a particular tectonic or
geophysical feature to the definition of individual zones was dependent on the
character of the seismicity. For example, in the Piedmont the earthquakes
occur only as shallow events with mechanisms similar to that of reservoir
induced events.
A-56
The general pattern of seismicity, available from most earthquake catalogs,
was used to suggest active zones. However, detailed aftershock and seismic
net studies were given considerable weight in determining the style of the
earthquakes occurrence. The style was then used to choose appropriate
tectonic and geophysical features to help define the border of the seismic
zones.
QUESTION 3 Identify some of the factors, such as scaling, lower bound
magnitude, which influenced your development of the EUS zonation
maps. Rank the relative importance of each of these factors,
and, if possible, explain how these factors influenced your
choice.
The scale of resolution was more than sufficient. A zone boundary is diffuse,
not narrow, since stress fields in the crust require more than 100 km decay to
the level of local components of the stress field. However, most seismic
zones are at least 100 km wide except where an axis of a relic rift was
identified. Seismicity was used primarily to decide if a zone was active, and
hence, the lower bound magnitude was not a significant factor.
For a specific site the microzonation would be strongly dependent on the
tectonic setting of the seismicity. In the Piedmont, a specific site analysis
would require an analysis of geologic parameters of near surface rocks as well
as seismic history. In contrast, in the southeastern Tennessee areas, an
analysis based on non-surface (5 to 20 km deep) crustal properties would be
integrated into the analysis.
The estimate of a and b values should include all available seismicity.
Elimination of less than m=5 events would probably bias (because of difficulty
in calculating estimates) the estimates in areas of sparse seismicity, but I
would not expect them to change.
A-57
\-^r. ...
^M'
';:^-'
QUESTION 4 Describe, briefly, the influence of, and your use of, these
features in the development of your zonation maps.
I avoided the background and complementary zones as much as possible.
Instead, I tried to relate the significant seismicity of a zone to a
characteristic dominant mechanism and use that as the basis for a zone.
QUESTION 5 Did you feel that these two ways of describing uncertainty
provided you with adequate means of expressing your state of
knowledge regarding zonation of the EUS? Discuss the types of
uncertainties you had identified and attempted to model by
specifying a probability of "existence" of a zone and/or
alternative zone boundaries.
Yes, the methods were sufficient. The probability of existence was largely
determined by my direct knowledge of the seismicity in each zone and the
number of events in catalog.
QUESTION 6 Discuss briefly what type of uncertainties you considered when
specifying the bounds for seismicity parameters.
The upper bound magnitude was determined by the tectonic setting of the events
and its uncertainty by the strength of the interpretation of the tectonic
setting.
QUESTION 7 What does the terminology "95 percent confidence or uncertainty
bounds" mean to you? How did you interpret it?
Estimate the standard deviation based on distribution of the central 60
percent of the data, and double it. Modify slightly depending on whether I
believe the data were normally distributed according to the amplitude or Log
(ampl itude) .
A- 58
QUESTION 8 Discuss, briefly, the significant issues related to predicting
the largest magnitude earthquake that can be expected to occur
in a source zone.
The maximum magnitude was:
5.75 for Piedmont type, shallow (type event is New Brunswick)
6.5 for closed rift-like continental structures that are open to oceanic or
extensional crust or show significant seismic activity.
QUESTION 9 Outline, briefly, the sources of information which formed the
basis for your predictions of the largest magnitude.
The largest observed earthquakes in their respective tectonic setting (see
Question 8) .
QUESTION 10 Did you follow a consistent procedure for predicting My? If so,
briefly outline the procedure. Were you influenced by any
constraints, e.g., limits of measurements scale, when
considering your estimate?
(see Question 8-9)
QUESTION 11 What are the primary sources of uncertainty that you were trying
to describe in determining the upper and lower limits for My?
Tectonic and magnitude determination.
QUESTION 12 Discuss, briefly, the various issues, e.g., clusters, catalog
incompleteness, related to using recorded earthquake data as a
basis for estimating the seismicity parameters (a,b).
A-59
The greatest uncertainty comes not from the recorded data (which can generally
be corrected for incompleteness) but from non-stationarity in the seismicity
(for example, Charleston before and after 18876). I think the strong removal
of aftershocks could remove some of this uncertainty.
QUESTION 13 Outline, and rank, the principal sources, e.g., analysis based
on your chosen catalog, the results of the LLNL "uniform
approach", you used to develop your estimates of (a,b).
Plots of numbers on events versus magnitude (provided) tempered by an analysis
of completeness and contamination by aftershocks.
QUESTION 14
If you used a catalog recorded events as a basis for estimating
(a,b), to what extent did you use analytical procedures to
adjust for the aftershocks, etc and for incompleteness? Give a
brief description of these procedures.
see Question 13.
QUESTION 15 Identify and discuss, briefly, some of the significant issues
impacting characterization of the seismicity of the EUS.
The most significant issues impacting characterization of eastern United
States Seismicity are the unanticipated events (surprises), lack of
understanding of aftershock processes, and a means of direct detection of
potentially active tectonic zones.
QUESTION 16 What do you consider to be the major sources of uncertainty in
predicting (a,b)?
A- 60
sparse data (less than 25 events) and non-stationarity of the tectonic
process.
QUESTION 17 Were the alternates presented in the seismicity questionnaires
for modeling potential correlation between (a,b) adequate for
you to express your views about your joint uncertainty about the
seismicity parameters? Discuss your views on correlation
between a and b.
Yes Discussed previously, a and b are correlated, but the effect of this can
be minimized if constrain values to a mid-point in the data.
A-61
SEISMICITY INPUT DOCUMENTATION FOR EXPERT 12
QUESTIONNAIRE Q8
QUESTION 1 Discuss briefly the adequacy of the principal sources of
information, including physical characteristics and observational
data, as a basis for identifying source zones.
The sources of information that I used to make the seismic source zone
interpretations are shown in Figs. 2 and 3 attached. The adequacy of these
data has to be evaluated in terms of their usefulness for defining sources and
in terms of their quality.
I will start with the stress data. I felt the stress data were useful and
adequate to make the limited determination that there are no unusual,
anomalous sources of lithospheric stress in the central and eastern parts of
the United States, and that overall the stress field is somewhat uniform. The
stress field beyond that is not very useful for making specific
interpretations, since it is not possible to resolve the three-dimensional
aspect of the stress at any specific location; the data samples are too sparce
and local contributions, stress amplification and other possible influences
could not be resolved.
With regard to the Bouguer gravity anomaly data, I made most use of these,
since I felt the principal basis for localizing stress would be
discontinuities reflected as lateral changes in density in the upper part of
the Earth's crust. In addition, these data in combination with the magnetic
anomaly data are very useful for defining broad geographic areas of
contrasting crustal rocks that could define areas having different
geomechanical responses to imposed stresses. I used the unfiltered gravity
data and the 125 and 250 high pass filter gravity data for my primary
interpretations.
With respect to the magnetic anomaly data, these were used to supplement the
A-62
gravity data for defining the boundaries of sources in certain areas.
Generally, I did not specifically use the magnetic data to identify
boundaries, but only to supplement the gravity data. One additional use of
the magnetic data was to define general areas of the upper part of the crust
that have contrasting rock material properties, that is, granitic rocks vs.
basic rocks. The magnetic data are particularly useful to define lateral
density discontinuities.
The next set of data I used was cumulative seismic moment. The seismic moment
data were used in rather limited cases to help define boundaries of sources
where other data were not adequately definitive. The seismic moment data was
also used in the upper Mississippi embayment to differentiate between source
boundaries in detail. In some situations, I used historic seismicity plots to
help position source boundaries in areas where there was not definitive
information in the geophysical data.
The tectonic map of the United States was used to refine source boundaries and
to delineate tectonic structure specific sources.
QUESTION 2 Outline your principal bases for identifying zones. To what
extend did you rely on tectonic and geophysical features as a
source of zonation? What, if any, historical seismicity and
other observational data did you consider in your development of
seismic source zones?
The principal basis for defining sources was the combination of Bouguer
gravity anomalies and magnetic anomalies. These data were supplemented by the
seismicity data and by geologic data in some situations where I felt those two
sets of data added to the source definition. I relied heavily on the
assumption that earthquake sources are associated with definable tectonic
features of the Earth's crust, and that these features would be identifiable
either in the tectonic map of the United States or by geophysical anomalies.
A-63
I considered historical seismicity only to the extent that it supplemented
definition of boundaries in some areas where the geophysical and tectonic data
were not definitive.
QUESTION 3 Identify some of the factors, such as scaling, lower bound
magnitude, which influenced your development of the EUS zonation
maps. Rank the relative importance of each of these factors,
and, if possible, explain how these factors influenced your
choice.
With regard to scale of resolution of the zonation maps, I was somewhat
influenced by my view that at the scale of approximately 1:5 million, with
which I was working, I did not need to look in detail at specific geological
features to determine the zonation boundaries. In other words, I think that
given a much larger scale map, say 1 to a million to work from, the zonation
boundaries would have required a more detailed look at local geology than I
felt was necessary for this small scale. So, I would think that the zonation
boundaries are relatively less precisely located at this scale than they would
be at a larger scale. But, it is not clear to me that this contributes to the
uncertainty in the hazard computation using maps of this scale.
The lower bound magnitude did not influence my zonation in any way. I
attempted, first of all, to define seismic zones on the basis of geophysical
and geological data without consideration of the maximum or lower bound
magnitude that these zones might be able to generate. With regard to the
lower bound magnitude, however, I did make the assumption that the defined
seismic zones would not define all of the seismicity; the background and
complementary sources in my interpretation are meant to pick up the level of
seismicity that would not be included in individual source interpretations.
A- 64
QUESTION 4 Describe, briefly, the influence of, and your use of, these
features in the development of your zonation maps.
For development of the zonation map, the background and complementary sources
were used to identify areas of seismicity that I could not associate with
identifiable tectonic features at the scale with which I was working. I made
the assumption that identifiable tectonic features would have some potential
of being distinct seismic sources. This assumption was made independently of
whether the rates of activity in those sources or the maximum or minimum
magnitudes in them differed from adjoining sources. So, in my interpretation,
many of the sources will have similar rates of activity, similar maximum
magnitudes and so on. I simply meant to distinguish them as being
identifiable features in the Earth's crust that could localize earthquakes.
The background and complementary sources simply represent activity in those
areas where I could not identify specific tectonic features.
QUESTION 5 Did you feel that these two ways of describing uncertainty
provided you with adequate means of expressing your state of
knowledge regarding zonation of the EUS? Discuss the types of
uncertainties you had identified and attempted to model by
specifying a probability of ^existence" of a zone and/or
alternative zone boundaries.
The structure of the interpretation approach is amenable to expressing
alternative interpretations. The background and complementary zones are
adequate in my judgement to permit expression of the alternative
interpretations that one could wish to propose. This, of course, may require
developing multiple maps.
The types of uncertainties that I attempted to express in modeling the
probability of existence of the sources were simply my judgements of the
likelihood that the source was, in fact, active in the contemporary stress
regime. The alternative, in my interpretation, is express as the background
or complementary zone.
A-65
QUESTION 6 Discuss briefly what type of uncertainties you considered when
specifying the bounds for seismicity parameters.
My response here considers the parameters a and b of the frequency-magnitude
equation, and the upper bound magnitude.
First of all, I have assumed that the basic properties of an earthquake source
are:
1. seismicity parameters are constant throughout the source; and
2. the upper bound magnitude applies to the entire source.
With respect to the parameters a and b, I have made two governing assumptions:
First, with respect to the seismicity parameter b, I have assumed this
parameter to be stable over large tectonic regions. I would permit, for
example, differences in this parameter between a stable continental
interior, a Paleozoic erogenic system such as the Wichita or the
Appalachian systems, and continental boundary sources. The data are
adequate to make some minor distinctions of the parameter among these
regions, and I used these distinctions to make subjective interpretations
of what the value of the parameter should be for seismic sources within
those general tectonic environments. With respect to the uncertainty on
the parameter b, I have assumed that the parameter is determined with
reasonable certainty. I have allowed an uncertainty of + .15 depending on
my confidence in the value within the general crustal environment of the
seismic source.
Second, with respect to the parameter a, I have assumed that this value
can be determined from the historic seismicity sample for each source, and
A-66
I have allowed it to vary from source to source depending on the historic
rate of occurrence of earthquake activity within that source. The value
is determined by anchoring the frequency magnitude curve to the lowest
magnitude level for which I considered the reporting to be complete, and
then the rate of activity is computed from the frequency magnitude
formula. I have assumed an uncertainty of jfl/2 magnitude in my assessment
of the magnitude for which the reporting is complete and have taken the
resulting range in a to be the uncertainty.
With respect to maximum magnitude, I have made strong use of the historic
seismicity data for each source of analogies with worldwide data for similar
tectonic environments. My principal governing data are tectonic similarities
with other regions and the size of tectonic features within the source. For
example, the uncertainty in the maximum magnitude estimate for the New Madrid
seismic source is relatively low, based on a world wide observation that
earthquakes of this size simply are exceedingly rare in intraplate regions.
Based on this observation, I assume that this is a maximum event for this
seismic source with little uncertainty. For other areas, for example, sources
within the stable platform tectonic environment are given higher levels of
uncertainty. The higher level of uncertainty generally is related to my
assessment of the order of importance of the tectonic feature with which the
source is associated.
QUESTION 7 What does the terminology "95 percent confidence or uncertainty
bounds" mean to you? How did you interpret it?
I interpret the terminology to mean a very high degree of confidence that the
actual value given for the parameter in general, falls within the range of the
upper and lower bound stated.
A- 67
.Sj^t^V
QUESTION 8 Discuss, briefly, the significant issues related to predicting
the largest magnitude earthquake that can be expected to occur
in a source zone.
The principal issues related to the largest earthquake that might be expected
within a source are the tectonic order and size of the tectonic features. I
have assumed that large scale, first order tectonic features equate to larger
maximum earthquakes in general. However, this is very much modulated by
tectonic analogy with experience worldwide. I generally put higher degrees of
uncertainty on my assessments of maximum magnitude where larger tectonic
features existed and worldwide experience indicated that only moderate size
earthquakes might be expected.
QUESTION 9 Outline, briefly, the sources of information which formed the
basis for your predictions of the largest magnitude.
The sources of information are basically those that I used to define the
seismic sources themselves, plus a compilation of worldwide earthquake
activity associated with intraplate regions developed by EPRI under project
2556-12. This document, although in limited release, has been distributed to
scientists working in this subject area.
QUESTION 10 Did you follow a consistent procedure for predicting My? If so,
briefly outline the procedure. Were you influenced by any
constraints, e.g., limits of measurements scale, when
considering your estimate?
With respect to the first part of this question, I refer back to Question No.
8. I was not limited by the magnitude scale used. All of my estimates are
made in terms of m^j, which I have assumed to saturate at about magnitude 7 to
7 1/2.
A- 68
QUESTION 11 What are the primary sources of uncertainty that you were trying
to describe in determining the upper and lower limits for My?
The uncertainty on M^, reflects my uncertainty on the processes of earthquake
occurrences in intraplate tectonic regions, the uncertainty on the orientation
of structures within a specific source, the relationship of the overall stress
regime to earthquake occurrences within a source, and limitations of data from
worldwide analogous regions.
QUESTION 12 Discuss, briefly, the various issues, e.g., clusters, catalog
incompleteness, related to using recorded earthquake data as a
basis for estimating the seismicity parameters (a,b).
I have assumed that clustering of activity results in some error to the rates
of activity determined from the historic catalog, but this was not taken into
specific account in my estimates of the a value. The incompleteness of the
catalog is, of course, the most serious difficulty in determining a and b
values from historic record. Second to this is the length of the data base
itself. I have accepted that the incompleteness in the catalog has been
properly accounted for by the Lawrence Livermore model. For most of the
sources, I have made the assumption that the data sample is not adequate to
determine, independently, the a and b values. For this reason, I have placed
very strong priors on b, and have allowed a to be determined from the
magnitude value above which the historic seismicity is assumed complete.
QUESTION 13 Outline, and rank, the principal sources, e.g., analysis based
on your chosen catalog, the results of the LLNL "uniform
approach", you used to develop your estimates of (a,b).
A-69
I referred to question 12 with this answer.
QUESTION 14
If you used a catalog of recorded events as a basis for
estimating (a,b), to what extent did you use analytical
procedures to adjust for the aftershocks, etc and for
incompleteness? Give a brief description of these procedures
I refer to my response to question 12.
QUESTION 15 Identify and discuss, briefly, some of the significant issues
impacting characterization of the seismcity of the EUS.
Again, I refer my response to question 12. I believe that the most
significant issue is the length of sample available to determine seismicity
parameters within a restricted source for regions of such low seismicity rates
such as the eastern North American Continent. The issue of whether b is
essentially a constant value that approximately equals to 1 or whether it is a
value that varies significantly from one source to another is of considerable
importance. I have assumed with a high. degree of certainty, in my mind, that
b is essentially a constant parameter, approximately equal to 1. The issue of
variation of seismicity rates from one source to another is also of
considerable importance. Typically, one would like to assume that seismicity
rates are related to tectonic strain rates. Within a model of a flawed, rigid
lithospheric plate driven by boundary forces, the validity of this assumption
becomes less clear. That is, there is no indication in the available data
that tectonic rates should vary from one location to another within the
region. Notwithstanding, the uncertainties about variations in rates of
activity in the intraplate region, I have made the assumption that the best
available information to determine rates of earthquake activity for a source
are the historic earthquakes. Where possible, the a value has been determined
from the historic sample within a source. Where the sample was not adequate
in my judgement for this purpose, I estimated rates from analogous sources.
A-70
I
QUESTION 16 What do you consider to be the major souces of uncertainty in
predicting the seismicity parameters (a,b)?
The major source of uncertainty, in my mind, is the fundamental lack of
knowledge about strain release mechanisms. Other important sources of
uncertainty relate to the length of the data sample. If we had either a
longer sample or a better understanding of the process, the uncertainty could
be reduced.
QUESTION 17 Were the alternates presented in the seismicity questionnaires
for modeling potential correlation between (a,b) adequate for
you to express your views about your joint uncertainty about the
seismicity parameters? Discuss your views on correlation
between and and b.
The answer to the first part of this question is yes. I had no difficulty
expressing my views about the uncertainty of seismicity parameters based on
the discussions and alternatives presented to me. Generally, I have made the
interpretation that a and b are not correlated.
A-71
APPENDIX B
This appendix provides a complete account for the input data as it was used in
the analysis.
The data is presented for each S-Expert independently in the following format:
Table Bi.l: Information on the zonation given by Expert i.
Table Bi.2:
Table Bi.3
Figure Bi .1
Figure Bi .2
Figure Bi.3:
List of alternative zones or clusters of boundary
zones.
Seismicity data for Expert i.
Seismic zonation base map for Expert i.
Map of alternative seismic zonations to
Expert i's base map.
Map of alternative seismic zonations to
Expert i's base map.
In addition some comments are given when the above information was not
sufficient to entirely define an Expert's input. This is the case for Expert
6 for example.
Table Bi.l:
Table Bi.2:
Gives the name of Expert i's zones as they appear
in his maps, their area in km^, and the
confidence he associates with their existence as
well as the name of the host zone (see Section 2
and Appendix C for details). The sequential
index at the left of Table Bi.l is a dummy index.
Provides an account of how the alternative
zonation maps are constructed.
It gives the list of the zones to be replaced,
the list of the replacing zones, and the name of
the host zone.
For each cluster, the top line corresponds to the
zones to be replaced and the bottom line
identifies the replacing zones (OLD/NEW zones in
cluster). BACK is the name of the host zone, and
the two numbers specify the probability
associated with each alternative.
B-1
Table Bi.3;
Figure Bi .1
Figures Bi.2 and Bi.3;
For example, in the case of Expert 1, in the
third cluster zone 4 and zone 5 have .6
probability of being in the zonation map, given
that a zone exists at that location, and zone 25
has .4 probability of being there.
Provides the information necessary to define the
earthquake frequency distribution within each
zone. This includes for each seismic zone the
best estimate, lower and upper bound for the a-
and b-values, the upper magnitude cutoff. It
also includes the type of variable used for the
zone (magnitude or intensity), and the range of
magnitude (or intensity values) for which the
occurrence model is linear when the LLNL
occurrence model is used. In addition, the very
first line of the table indicates the total
number of zones identified by the Expert, the
self weights that the Expert gave for the four
regions of the EUS (northeast, southeast,
northcentral , and southcentral) , and finally the
type of correlation the Expert chose for the
correlation between the a- and b-values.
Gives the best estimate zonation map (BEM) for
Expert i .
Give additional alternative zonation inputs for
Expert i. In some cases the Expert has given
entire alternative maps (e.g., Expert 1 or Expert
6) and in other cases the alternative maps
provide additional input on alternative zones to
replace zones of the BEM (e.g.. Experts 5, 10,
and 13). Experts 2, 3, 4, 7, 11, and 12 provided
only a BEM. In these latter cases, the
uncertainty in the zonation is modeled only by
the probability of existence of each zone as
shown in Table Bi.l.
Note that Table Bi.2 does not appear for the Experts who did not provide
alternative zone boundaries (i.e., for Experts 2, 3, 4, 7, 11, and 12).
B-2
U.
o
Hi
UID
<
pq
W
pq
<
Q
U
U.
u
Q
M
UJi-
z
Ol-
NQ£
UJ
HO.
=>X
DLU
n
<
>-
zn
D
M
I-
<
E
oe
o
u.
V)
M
X
lU
. Vi
ZLU
oo
QUI
U.
Ul
o •
z»-
UJ(0
Qm
MX
ILIII
z
DZ
Om
UJ
z
o
NliJ
> z
I-
oe
UJ
a.
X
UJ
X
UJ
Q
l^>.IOOI^OlNlO^IO^I^I^IOI^Ir-lf-I^OIOOIOIOI^OIO^I(SIII/^l
ieoi^ivAioiroiroivAi^i(Mi|v.ir^ir>.ir-ieoir^ii/>it-iovi^ihOi
l^OlOllAl(sll^Ol^l^oll-l^OlOlOl^OlOl^cl^OlOloOl^Jleol<^l
ir^lvOIOI^Ir-IOI(SilvOlr-lsOlr-l(SIISOir«-ir^l'Olr-irOirOION|
1^1 iovi«-ioir-ir^irvi\«it-iooii^isciNiir)iNivoioieoioi
i(sii iv«ioiONir^iovicvli i^iirti«-iK)ivoiioiroiONioiroivoi
lOI li-lVlCSII 1 iroi 1 1 1 1 1 lvOlf-IK)li-lCSII 1
1 r~ I 1 1 1 1 1 1 1 1 1 1 1 1 1 r- 1 1 1 1 1 1
1 1 1 1 1 itni iini 1 1 1 iini i iiAiinimiini i
• 1^1 1 1 Ir-I Ir-I 1 1 1 1^1 1 1,-ir-lr-lT— 1 1
1 lUJi 1 1 lUJi lUJi 1 1 1 lUJi 1 lUJlUJlUllUJl 1
• 'Zl 1 1 IZI IZ Zl 1 IZlZlZlZl 1
1 IDl 1 1 IDl lOl 1 1 1 lOl 1 IDIDIDIDI 1
I iNl 1 1 iNi iNi 1 1 1 iNl 1 iNlNiNlNi i
lOIOlOIOIOIOIOIOlOIOIOIOIOIOIOIOIOIOIOIOI
ioiooioioioir<.ioimioioioioiv«ioioirv.imirvit>.ioi
Ir-I Ir-lt-lr-l If-I |,-|^ir-l<— 1 It-tr-l 1 1 1 1^1
1 I 1 1 1 t 1 I 1 IOIl-lNl^Ol^llAlvOlr>■IOOIO^IOI
lr*)NltOl^linivOir^lOOIONlT— Ir-lr-lr-lr-lr-lr-li— l>— It— ICJI
lUJiUJiUJiUJiUJiLLllliJillJllUllUiUJiUJiLiJiUJiUJiUJiUJiUJiLlJllUl
IZIZIZIZIZIZIZIZIZIZIZIZIZIZIZIZIZIZIZIZI
iDtOlDlDIDlDlDlDlOlDlDlDlDlDlDlDlDlOlDlDl
iNlNiNlNiNiNlNiNiNiNiNiNiNiNlNlNlNiNiNiNi
l«-lNlK)l^lml^olr^leol(^lOlr-lNl^Ol^llnl^cl^^lool<^lOl
1 1 1 1 1 1 1 1 1 1 I — lr-*lv~l^l^l^"l^«l^l^lM-lf\tl
B-3
u.
D
HI
X
Ui
ZUI
tuz
COD
DU
QUI
M
U.
UJ
O •
ZK
Qm
MX
U-UJ
z
DZ
Om
UI
z
c
Nil)
(0-icsii^iviK)iooiaNi
IO^I^Ola^lr-l(Ml^OI^OlC^ilU^lvOll/^I^OlOIVOlT-l(^IO^l0^ltOloOI
IO^IOIsOlT-llrtlO^I^Ol^vl^Olr-ICT^Ilf>l<^l^-lr-I^Ol<^leOI»l<^l
ll«l^lC\IIOVlCMI^ieOICOIr-lr-ir~ieOIOIpvir^lOlr-IMlsOlini
l^Ola^ll^llOlIOIOIOIOIOIOIOIOIOIOIOIOIOIOIOIOIOIOIOI
I 1 1 1 I 1 < 1 n 1 1 1 1 1 1 1 1 1 1 1 !! !
lf-lNl^Ol^imi\OlvOI|N.ICOiaviOli-l(MirOl^llAlvoll^lOOIONI
l(Ml(MlNl(MI(SIICSIirgirsll(MlNlK)IK)IK)IK)irOIK)irOl>OirOllOl
lUllUllUllUllUllUllUllUllUllUllUllUllUllUllUllUJiUllUllUllUJi
•ZlZlZlZlZlZlZlZlZlZlZlZlZlZlZlZlZlZlZlZl
lOlOlDlDlDlOlOlDlDlOlDlDlDlDlDlOlDlDlDlDl
iNiNiNiNiNiNiNiNiNiNiNiNiNiNiNiNiNiNiNiNi
ir-i(siiroiTriu^ivoir^iooiavioi>-ifMiroi^iinivoirvicioiovioi
i(MiNi(Sli(SJi(Sli(Ml(SJi(SliNiiOirOiK0irOirOl»OirOiroiK0i>OiVI
I
B-4
(0
UJ
z
D
N
>-
OS
«
Q
Z
3
D
Of
111
H
CM
pq
H
UJ (O
Of
UJ
w
o
V)
Ml
D
N
UJ
Z
D
N
Ul
>
M
<
Z
(le
UJ
H
«
U.
V)
_l I UJ
C I Z
I
X
o
UJ
X
Ul
Q
Of
UJ
H
oo
Ul
z
D
N
Ul
D
N
00
UJUJ
DO
NN
V0
UIUI
DO
NN
i-N
UJUI
CO
NN
UJ
z
D
N
oo
\n\r\
vOfO
N
UJ
o
N
tON
UJUJ
oo
NN
UJ
o
N
oo
in ID
N
UJ
z
D
N
VN
UJUI
DO
NN
Ul
D
N
oo
oo
to
N
Ul
o
N
UJUJ
zz
oo
NN
UJUJ
OO
NN
Ul
o
N
oo
oo
^0>
.-N
UJUJ
ZZ
oo
NN
UJ
D
N
oo
oo
UJ
o
N
eotn
r-KO
UJUJ
oo
NN
rs.'T
r-tO
UJUJ
zz
oo
NN
soro
»-to
UJUJ
zz
oo
NN
iflN
T-tO
UJUJ
zz
oo
NN
UJ
Z
O
N
oo
in in
soro
orv
NM
UJUJ
oo
NN
N
N
UJ
Z
o
N
oo
oo
mm
N
Ul
z
o
N
o
o
B-5
I
-.''.A ■•
.'■.•if''. '
t- fi
y/-A-
c
o
m
u
PQ
<
o
o
1 ON
1 to
UJ
z
D
N
NO
(MOO
(Mro
UJ
z
D
N
UlUj
zz
oo
NN
OO
oo
B-6
en
pq
z*
H
:;
9!
] !
9E
3 :
H
1
t
*
1-
1
4:
4c
4c
t-
4c
4c
t-
o*
M
00:jc
M
ooi:
M
^o4c
M
vo*
M
oes
m4e
Elft^(^*
Emr^oo*
EOV0^4e
Eifieo^*
ENino
M
:'^.^.t
M
•O^*
M
..-in 4c
M
•r^o4c
M
•Csiin
-JO
^C
> • •*
-JOv • •4c
-JON • •4c
_JC
> • •
-i4c
N 1 i
r"
•N i *
K) 1 4e
Vf>^4c
II UJ
Er-
r>>.ON
^
O*
Q
M ■
so NO*
Q
M •
sorr*
Q
M •
Mf^4e
Q
M >
MOO 4c
Q
M •
ON£r-
Sf
UJO
-I0<
• •*
UID
-lOv
• •*
WD
-JOO
I . • ^
UJO
-IOC
1 • •4c
UID
-JON
H
^1
zz:
(SI 1 *
EE
.- 1 *
££:
N 1 4^
^:^
K) 1 4:
S.^
(M 1 1-
OC
<
GC
*
<
o:
*
<
a:
4c
<
Of
1
<
Of
(^
UJ
-J*
z
UJ
*
Z
Ul
*
z
Ul
4c
z
UJ
Z
UJ
UJ
OL
51
z
*
z
*
z
4c
z
4c
z
Q.
X
UJ>-s
D
1
Ui'^
D
1
Ul^
D
4c
UJ^
D
4:
UJ^
D
X
UJ
-1*
Zm
_J
ZM
_l
ZM
-1
ZM
^
4:
ZM
^
UJ
*
o*
*
o*
jff.
D4c
04c
4c
04c
0£
♦
NCQ
*
N«
*
Nn
:)(
NCQ
4c
N«
Of
O
u.
o|
(L-
•*
1
1
1
*
*
1
4c
4c
4c
4c
4c
4:
>.
M
*
1
1
*
*
*
4c
4c
i
4c
4c
4c
4:
o
o*
*u.
1
*u.
*
4:U.
*
4:U.
4c
4c U.
o
o;l
4eO
*o
1
4cO
1
♦ O
1
4=0
M
E
• -4:
<
(/)On
• •4c
^1
UJ
z
UJ
N 1 *
Ul
z
UJr-
*" ' i
Ui
z
UJ
N 1 *
UJ
z
UJ
Ml 1 4c
Ul
z
Ul
(M 1 4c
*
M
H
3K
M
>-
4c
M
1-
4c
M
h-
4c
M
1-
4c
yi
-J
(/)
:|e
_i
M
ac
-J
v>
X
mJ
in
4c
_J
(0
4c
H*
UJ
4(
UJ
jjc
UJ
T
UJ
4:
UJ
4c
X4c
n
i
m
*
m
4c
»
4c
n
:|e
O*
N
(VJ
4c
M
jk
N
4c
^r
1
M*
>-
*
V
4c
>-
¥
>-
4c
>.
^1
K
♦
J-
K
¥
y-
4c
H
4c
OM
*
OM
4c
OM
ijc
Dm
4c
Om
■ J
Z(0
*
ZV)
*
zw
4c
ZVi
4c
Z
*
o
0.3
4c
^4:
Sr^eor^*
£:iAr^(M4:
SMvDi-
^
M ■
■lf>M*
M '
•so CM 4:
M '
•r-N4c
M >
rvr<-4c
M •
•ON 00
9JC
-IIT
1 . .*
-J IT
1 . .4:
-JIT
1 ■ -4;
-I IV
, . .4:
•J|V
sic
Vr-*
Vi-4:
lA,-4:
N 1 4c
N 1
4c
oc
1 *
OC
1 4c
e£.
'1
0£
4c
(^
4^
o
UJ
*
Ul
*
Ul
UJ
4:
Ul
>k
*
Ol
*
TT
a.
4c
f
0.
4:
lA
Q.
4c
r"
Q.
*
•
0.
*
•
a.
4:
•
0.
4c
•
CL
4c
Ol
*
in
■=>
*
lA
3
4c
lA
3
4c
r>>
3
4c
r*.
3
*
*
*
4c
4c
*
*
4:
%
*
*
i
1
I
?
*
sfc
4>
4c
*
*
*
:)c
4:
4:
*
o
*
4c
4c
4c
*
lA
*
U\
4:
in
4c
lA
4c
kA
jjc
*
•
*
•
4c
•
4:
•
4:
4c
*
ro
4c
K)
4:
to
4c
ro
4c
ro
4c
*
4c
*
*
4c
4:
1
4c
4:
4c
1
4c
4c
*
4:
4c
4e
4c
4c
*
vo
4:
1^
f
eo
4;
ON
4c
4c
*
4:
4c
4:
4:
r-
*
4c
4c
4c
4:
*
UJ(n
4=
UJ(A
4c
UJ(/)
4c
UJtO
4c
UJ(0
*
ZM
4:
Zw
4c
ZM
4:
ZM
4c
ZM
*
o
1-
4:
D
H
4c
1-
4:
D
H
1
D
H
*
N-i
M
004:
N_J
M
004c
N_J
M
004:
N-I
M
0VO4:
N-i
M
00
*
II Ul
StO
T-1^4:
II UJ
Er/)^r>.4c
II Ul
SfO|v.r>.4:
11 Ul
HM
ii-r-4c
II UJ
Z:c^'r^>.
*
Q
M •
^in4:
a
M •
iniA4:
Q
M •
OIA4:
a
M •
^C0 4c
Q
M •
l^eOr-
*
UJO
-Jl/^
• -4:
UJO
-tin
1 • •4c
UJO
_IIA
i • •4c
UIO
-irv
• •4c
UJO
-J so
*
zs:
K7t-4:
zz.
t^r-*
SZ
T,-*
Sit:
Csl 1 4c
z::e
(M 1 t-
*
■
*
*ll.
*
4c U.
4:
4:U.
4c
4:U.
Hf^
4:U.
1-
*
*D
4c
4CO
4c
4cO
i
4:0
4c
♦ D
M
*
*
4c
4:
4c
4:
4c
4c
4:
*
*IU
4c
4c UJ
4c
4c UJ
4c
4:UJ
4c
4c UJ
M
*
O
4c
4c
4:
1
z:
♦
z
4c
z
4c
z
4:
Z
Z
(0
*
<
Ul
4c
o4:
cc
z:
00
*
«£
►HTfeooo*
•I
Mtr
.-eo4:
-
*
M
UJ
^r-4:
M
UJ
Vr-4:
M
Ul
^r-4:
M
Ul
(M 1 4:
M
UJ
N 1 4c
*
—I
1 4:
-1
'1
—1
1 4:
—1
-J
4:
jk
h-
4c
H
H-
4c
H
%
H
1
M
V)
4:
(O
4c
V)
4:
Vi
V>
M
Ul
4t
UJ
1
Ul
4c
UJ
*
UJ
X
4c
09
4c
n
ta
4c
ta
*
GQ
X
i
'J-
4:
TT
4:
^
4c
V
rr
X
UJ
4c
Ul
4c
Ul
4:
UJ
4c
UJ
5
*
Q
4:
Q
4:
Q
Q
4c
Q
*
C3
4c
D=>
4c
03
03
4c
D3
«
Z(-
i
Zl-
4c
ZH
1
Zt-
4c
ZH
*
l-l
4c
t-i
4:
M
M
4c
M
*
cz
4c
oz
4c
OZ
*
OZ
4c
OZ
*
UJO
4c
UIO
4c
UIO
UIO
4c
UIO
*
Q:
*
s:<^^^oo*
ENOOO*
Eooeo*
EoiftT
i!
M •
.- !>>.!»!
M •
OVO*
M •
•eor-*
M •
■(>o*
M •
fsiro
)t
-J SO
• '9^
-JIT
1 . .*
-JVC
, . .*
-JVC
1 • •*
— isfi
91
]E
0£
to 1 *
Q£
'"Tl
Q£
Iftr-*
1 9tE
a:
TTr-*
1 *
Of
Vfir-
1
: :
o
UJ
*
o
UJ
*
O
UJ
*
o
UI
*
O
UJ
i t
r*
a.
*
r^
0.
*
o
O.
9tC
IX
0.
*
CO
a.
::
•
Ol
*
•
0.
*
•
Ol
*
0.
*
D.
: E
9E
IE
:e
9E
9E
*«
3
*
u\
3
*
VO
Z>
*
*
*
9|t
*
*
*
Ift
3
*
Ift
3
9|C
9 :
o
9(C
o
Hi
o
t
o
*
O
9|C
9 :
in
♦
IT)
*
m
*
Ift
*
Ift
9|e
9E
•
4(
•
t
•
*
•
*
*
9tE
9E
•o
i
M
*
ro
*
M
*
lO
9|C
9E
9e
*
*
*
*
9|E
%
*
*
*
9|e
9E
*
*
*
*
9lC
*
*
*
*
*
9 E
r-
*
M
*
M
*
^
*
Ift
*
9E
9 E
^-
*
*
r-
*
*
>—
*
*
"-
*
*
•-
9 E
UlVi
*
UlVi
*
UlbO
*
UICO
*
UJ(0
: E
ZM
*
ZM
*
ZM
9|C
ZM
*
ZM
9E
D
H-
*
o
1-
*
D
1-
*
D
H
*
D
1-
9E
N-J
M
tot
N-i
M
00:k
N-i
M
OO*
N-I
M
TO*
N-I
M
TO
9 :
II Ui
S.ir\NVt
II UJ
Ein(MO:4c
iiUI
SvOrOON*
II UJ
SiftlOT}-*
II UI
•EfxIftT
9E
a
M •
r-Ov*
Q
M •
hOsO*
Q
M •
OOlft*
Q
M •
TT*
Q
M ■
Ow^r*
9 E
UJO
-ISO
• •*
UJO
-lift
• •*
UJO
-lift
• •*
UID
-lift
• •*
UIO
-J Ift
JE
£S
N 1 *
EE
to,-*
£:£:
lOf-*
Es:
lO,-*
s:z
T.-I-
9!
<
Of
-
9 E
i^lL.
*
9*EU.
*
*\1.
*
*U.
*
*u.
h-
9 E
$0
*
iko
*
*o
*
*o
*
*D
M
9 E
9iC
i
9|t
*
*
*
*
*
*
o
9E
9|CllJ
*
9|CUJ
*
*lli
*
*UJ
*
*UJ
M
9E
O
9k
o
*
o
*
o
*
o
s
9 !
9E
z
ac
z
*
z
*
z
*
z
(O
<
JJ
9|E
.9t£
UJ
1- •
inro!|t
UJ
1- •
TTCSI*
UJ
>- •
MCSJ*
UJ
1- •
SOT*
9E
z
V>\0
• •*
z
l«
• •*
z
(Oso
• •*
z
(Olft
• •*
z
(Olft
• •*
9E
M
UJ
N 1 t
M
UI
'«r,-9tE
M
UJ
TTt-*
M
UJ
T.-*
M
UJ
Ift,-*
9E
-1
i
-1
I *
-1
1 *
^
1 *
—1
1 *
9E
H
h-
*
h-
*
H
*
1-
*
9E
V)
:k
Vi
*
9k
o=>
*
D=>
D3
*
03
9E
ZK-
4c
ZH
*
ZH
*
Zl-
*
zt-
9 E
M
*
M
*
M
*
M
*
M
9E
P.=
*
OZ
*
OZ
*
OZ
*
OZ
9E
UJO
9k
UIO
*
UJO
*
UIO
*
UJO
9 E
0£
*
(OO
0.3
*
(OO
0.3
*
(OO
0.3
B-9
c;
o
en
<
4c
:|e
1
♦
H- *
K *
1-
M 00 4:
M VO*
l-t ^o
E»OlAr-*
SOr-VO*
^rtt^^r^
M -M^*
M 'NV
i_im • '^c
-JO • •
ac ""Tl
»o 1 *
f-N 1
DC *
0^
UJ 4:
o
lU *
o
Ul
a. *
t/\
0- *
a.
^ 1
•
0- *
.
a.
so
3 *
o
3
O
►-
N-J
M
oo
II UJ
SOOOfO
Q
M
•O^U\t-
UJD
-JOS • •
s:z
r- 1 h-
^
It
<
1-
B-10
4J
a
o
o
en
■ ' - V- •> ■
•!'x^->^ii;'^^
B-11
u
4J
c
o
u
W
pa
<
*
*
*
*
*
*
*
*
*
*
*
*
*
*
9|C
*
*
*
*
*
Sf!
H
*
H
*
t-
*
H-
*
1-
#
M
r^o*
M
oo*
M oo*
M
00*
M 00
3^
SvO
1-NO*
EOOTTM*
STt^ir^vo*
£0000^*
Sr^TON
4^
M •
VO'T*
M •
(MsO*
M -ov*
M •
r-TT*
M •OO'O
*
_IO
• •♦
-lOO
•
•*
_l<^ •
•*
-J<^
• •*
_IO •
t
*
I—
r- 1 *
tn
*
M
*
N 1 *
,— r.
t
*
D£
*
a:
*
q:
*
q:
*
on
*
O
LU
*
o
LU
*
O
lU
*
UJ
*
LU
*
eg
D.
*
r^
a.
*
to
Q.
*
Ov
Q.
*
(^
O.
*
•
(L
*
•
a.
*
•
a.
*
a.
*
0.
*
o
3
*
CO
3
*
Ov
3
*
00
3
*
(^
3
*
r-
*
*
*
*
*
*
*
*
*
*
*
*
*
*
♦
*
*
*
*
*
*
*
•¥
*
*
*
*
:|e
*
*
O
*
o
*
o
*
*
*
*
O
*
o
*
o
*
*
*
*
•
*
•
*
•
*
•
*
•
4c
*
V
*
^
*
V
*
^
*
TT
*
*
*
*
*
*
*
*
*
*
*
*
Hf
*
*
*
*
*
a|(
*
.
*
oo
*
<^
*
4c
*
(M
*
N
*
CM
*
(SJ
*
M
*
*
*
*
*
*
HI to
*
iuo*
IlLU
Ettooi^*
II LU
£OOv^*
II LU
HMvOO*
II lU
2: in 00
r-
*
Q
H •
r-lft*
Q
M •
Vl^*
a
M •inin^:
Q
M •
(Or>.*
Q
M 'r-
Lfti-
*
UJO
-J0>
• •*
LUO
-iOO
•
•*
LUO
_J0> •
•*
LUO
-100
• •*
LUO
_ION •
.
*
££
,- 1 *
ZZ
N
*
SH
r-
*
sis:
r- I *
SH
»—
IH
*
-
*
*
*
*
*
*
*
*
*
*
H
*
*
*
*
*
*
*
*
*
*
M
*
*u-
*
♦ LL
*
♦ u.
*
*u.
*
*LJL
u
*
*D
*
*o
*
*o
1
*o
*
*D
M
*
*
*
*
Z
*
(U
*
UJ
*
LU
*
111
*
111
to
*
o
LU
*
o
tu
*
O
LU
*
111
*
LU
M
*
z
l-
*
z
1-
*
z
1-
*
z
H
*
Z
H
LU
*
-(Mvo*
MlOOr-*
MC^(^vo*
M0^^^^^
*
0£
1- •
MTT*
tm
1- •
r^vo*
Oi
H -oom*
u.
H •
(7Mf>*
on
H -vo
1-*
*
-
*
>-
*
>-
*
>-
*
>•
*
1-
*
1-
*
H-
*
H
*
H
3|C
*
OIH
*
DM
*
DM
I
DM
*
DM
*
zco
*
Z(/)
*
z«o
zoo
*
Z(0
*
z
«
Z
*
Z
*
z
*
Z
*
OUJ
*
OLU
*
OUJ
*
OLU
*
OLU
*
UJt-
*
LUI-
*
LUH
*
LU(-
*
LUH
*
Q^Z
*
QiZ
*
a:z
*
a:z
*
OCZ
*
n
*
M
*
M
*
M
*
M
♦
z
*
z
*
z
*
z
*
z
*
HZ
*
MZ
♦
MZ
*
MZ
*
MZ
*
M
*
M
t
M
#
M
*
M
*
o
*
O
*
O
*
*
t
D-J
♦
D_l
*
D_i
*
O-l
*
D-l
*
_ilU
♦
_lLU
*
_jLU
*
-ILU
t
-ILU
♦
Q
*
a
♦
Q
*
Q
a
*
D
*
o
t
D
*
*
D
£
*
s:
£
*
z:
I
£
i
♦
*
*
]|t
UJ
*
LU
*
LU
4c
lU
*
LU
*
SOU
ck:o
♦
r-O
Q^O
*
OOO
ck:o
*
OvO
0:0
*
00
ono
♦
(SJZ
LUU
*
MZ
LUU
*
(vjz
lUU
4:
*
(OO
a.3
*
(OO
a.i>
*
(OO
0.0
t
(OO
a.z>
B-12
4J
a
o
u
CO
pq
H
]^
*
«
*
*
1
*
*
1
*
*
*
sje
1-
*
1-
*
H
*
t-
*
\-
^
M
004:
M OO*
M
OO*
M
oo*
M 1
r-O
9|:
S^oov*
ENOr-*
ECMOr-*
Sttmv*
SOOM
f
M •
mr^*
i-H .rs-in*
M •
r^tn*
M •
so in*
M •(
JnN
*
-ir^
• •*
-Jr- •
•*
_lt-
• •*
-ir«-
• •*
-IPJ
• •
*
K) 1 ±
t-K)
1 *
r—
.- 1 *
r- 1 *
1
*
ee.
Of
*
ck:
*
q:
*
OS*"
*
o
Ul
*
o
Ul
*
o
UJ
*
o
UJ
*
O
UJ
*
CM
0.
*
M
a.
*
(M
Q.
*
*
^
3
r-
*
r-
*
*
,—
1
1
*
*
*
1
*
1
1
*
1
*
*
*
1
o
*
O
*
O
o
*
O
*
in
I
O
1
O
1
in
*
*
o
1
1
K)
*
^
*
^
*
ro
*
V
*
*
%
*
*
*
*
*
*
*•
*
i.
*
*
*
*
4e
*
1
*
*
*
3|e
*
t—
M
*
ro
*
V
*
in
*
*
»o
*
M
*
K)
*
M
*
to
*
*
*
*
*
*
UJ(^
*
UlVi
*
ujcn
*
UJ(0
*
UIW
*
2M
*
ZM
%
ZM
*
ZM
*
ZM
*
D
1-
*
O
1-
o
H
*
o
1-
*
D
H
*
N-i
M
00:|c
N-J
M OO*
N-l
M
oo*
N-J
M
oo*
N-l
M 1
i-O
*
II HI
SOCsTT*
II Ul
Zv£NO*
II Ul
STvOCJO*
II UJ
SOO>>0*
II UJ
Sr^eor*.
i
Q
»-l •
t^Ov*
a
M -I^IN.*
a
M •
r>«.r>-*
O
M •
OsO*
Q
M -1
^roi-
*
LJD
-ll>.
CM 1 i
UJO
_40v •
•*
UIO
-JON
• •*
UJO
-ir>-
• •*
UJD
-JON
• •
*
£:z:
s:s:
M
1 *
£:z
1 *
rs
O 1 *
£:z
1 H
%
-^
D
♦
Ul^
o
X
*
z;£
^
*
ZM
_J
*
ZM
—1
*
ZS.
_J
*
ZM
—J
UJ
*
o*
*
o*
*
o*
*
D*
*
o*
*
NCP
*
NM
*
NM
*
NCQ
*
NCQ
0£.
*
1
*
1
*
1
*
1
*
1
D
*
0--
*
♦u.
*
*
*
*
*
*u.
*
*
y-
*
*D
1
*
*
*
*
*o
*
*
M
*
*
♦ u.
*
*u.
*
*
*
*u.
O
t
1
*o
*
*
*o
*
*
*UJ
o
*
*
*D
M
*
z
*
Ul
*
Hi
*
z
*
UJ
V)
*
-eo
*
UJ
»- •
ON CO*
ce.
H -ONSO*
Q£
1- •
Ovvo*
UJ
H •
tOvO*
Oi
1- -TTCM*
*
z
v)r>.
• •*
«t
V)0 •
•*
«I
(no
• •*
z
V)t^
• •*
*
>-
*
Ul
*
>-
*
Q
*
1-
*
K
*
a
*
H
*
s
03
*
DM
DM
*
on
*
Dm
*
21-
*
zw
:^
ZV1
*
ZH
*
ZV)
$
M
*
Z
:(c
Z
*
M
*
Z
*
OZ
*
OUI
*
OUI
*
OZ
*
OUJ
*
tuo
*
UJH-
*
UJt-
*
UJO
*
UIH
^
o:<
*
Q£Z
*
cit:z
1
o:
*
*
(OD
0.13
*
tno
0.3
B-13
1
*
1
1
*
1
3|C
4^
*
4c
4c
*
cao *
4(
h-
*
1-
*
t-
4c
H
*
y.
♦
M
coo*
M
OO*
M
00*
M
00*
M ^
III
£0000*
SMMr-*
SMVO*
ZToo^o*
^N^voro *
♦
M •
•OCM*
M
•o^«
M
•l^ifl*
M
• ir\u\±
M '
■ sotf*
!l
-iCs
1 • •*
-J,- . .*
_io . .4:
-JON • •4c
-10 • •*
r-r- 1 4c
Oi *
*
Of
■-'1
1-N 1 *
..-^'1
02
N 1 *
*
o
UJ
*
o
Ul
*
O
UJ
4c
UJ
*
UJ
4c
4:
*
V
a.
*
oo
0.
*
o
0.
*
V
Ol
*
N
CL
*
•
a.
*
•
a.
*
•
Q.
*
Q.
*
Q.
*
*
*
*
T-
3
*
o
Zi
*
o
3
*
ON
3
*
3
^
4c
f—
*
r-
4c
*
4c
*
*
*
4:
*
*
*
*
*
*
*
4c
*
*
4:
4c
*
*
*
*
O
*
*
o
*
*
O
4c
*
*
*
*
o
i
o
*
O
4c
*
*
•
*
•
*
•
*
4c
1
*
*
*
*
*
%
4c
4:
*
*
1
*
*
*
*
*
*
*
*
*
*
*
*
OO
*
*
*
*
*
4c
ON
4:
*
4c
4:
4:
4:
NO
4:
4c
*
*
*
4:
*
*
to
UJbO
*
*
*
to
UJcn
*
*
*
ro
UJ(0
4c
4c
4c
to
UJ(0
4c
*
*
CM
UICO
(O
UJ
2
*
2M
*
ZM
*
ZM
4:
Zm
*
ZM
*
*
D
l-l
*
coo*
D
N-J
M
*
Oo4:
D
N_J
»-
M
*
004:
D
N-i
M
*
004:
D
N-J
1-
M ,-
*
N
%
II tu
SOI^SO*
II UJ
s: to OS ON*
II Ul
SincoT-*
II UJ
s.-
•0N«-4C
II UJ
Seo^o*
Q
tH • ' ■
^to*
a
M •
U^V*
Q
M ■
•r-^O*
Q
M •
' 00 NO 4c
Q
M «
CNJf^*
*
UJO
^r-
• •*
UIO
-lO
• •4c
UJO
_IO<
• •*
UJO
-JV
' • -4:
UIO
—ION
.:.-|
*
s:s:
t—
1 *
£:£:
.- 1 *
ZZ
N 1 *
sz:
r- 1 *
sz
*
<
Of
*
-
1-
M
M
t
z
H
*
z
1-
*
z
1-
*
z
y-
*
z
H-
*
s:
m
*
<
-
*
*
>■
*
*
'~>-
*
*
"">•
4:
4:
CnI
>-
1
*
z
Ul
4c
*
Dm
4c
*
OM
*
4c
1-
CM
1
Om
*
*
Om
1
*
*
UJ
so
*
*
Zl/>
z
1
Z(/)
z
4:
*
z
*
*
zco
z
1
z II
q:x
Ul<
I
rOZ
uju
*
MZ
UIO
*
toz
UIO
♦
»02
UIO
*
vz
UIO
*
Q.E
LU
t-
*
Ol
H
*
Ul
H
*
UJ
H
*
Ul
)—
*
x«-»
#
not:
LUO
*
4tQ£
UIO
*
«Qi
ujo
*
«0£.
UIO
*
^to:
UJO
*
UlM
*
a:
s:<
s^
>^>
V>>. .^,*> '
>;'•:
B-15
o
Ul
z
»-
0£
UJ
Q.
X
UJ
X
UJ
a
iNi<-iini(vjir«'i<9'iv«ioisoi'OirviNir^ir>.ir^ir>>ii-isoiiAivoi
iONirvioimirsiiooiu^ir^iONiu>io^iosi^ir^ivi(Miroimii/^ia\i
l(^lNlNlVlT-l^vlOl^Ol(Mlv£loOlooliAllnl<-l^Ol(Mlc^lr^l^x>l
• ^lNlvOlT-lr-llrtlOsiNl»ir«.lNtVllf>lMieOlvOir>-IOviMir/)l
Ir-lr-l'fl^ll/^l^lCgiNlOICOlNfiliniOlvOliniTrii-IIAli-lNl
1 1 1 irOlr-l 1 IVIr-IOOl IvOlVltOlNOIl^lr-ir^ieOIOI
1 1 1 fSlIN INI l«-l
lUJi 1 lUJi 1 1 lUJiUJiLLllUJiLJJl 1 iLUlUJlLJlUJlUJtUJl
IZI 1 IZI 1 1 IZIZIZIZIZI 1 1 Z 1 Z 1 Z 1 Z 1 Z 1 Z 1
lOl 1 IDI 1 1 IDIDlOlOlDl 1 IDIDIOIDIDIDI
iNi 1 1 N 1 vo 1 "9- 1 SO 1 N 1 N 1 N 1 N 1 N 1 i i N i N i N i N i N i N i
ItLI 1 IOLIUJIUJIUJIOLIQ.IQ.ICI.IQ.1 1 lQ.lCLlQ.ia.lQ.lQ.1
lEi 1 IZ IZ IZ IZ 1 E iZ: 1 E 1 Z I £ 1 1 1 Z 1 S IS 1 S 1 £ 1 S 1
IDl 1 IDIDIOIDIDIDIOIDIDI 1 IDIOIOIDICIDI
lOi 1 1 O 1 N 1 N 1 N 1 O 1 O 1 U 1 O 1 O 1 1 1 O 1 O 1 O 1 O 1 O 1 O 1
lOIOIOIOIOIOIOIOIOIOIOIOIOIOIOIOIOIOIOIOI
lOlOlOl(^l^«l^oll/^ltnl^«loOlt'^ltrtlOlOloolU^ll'^lr^l^OllAl
lr-l»— It— 1 1 1 1 1 1 1 1 1 l^-lr-l 1 1 1 1 1 1
i<-iNi>OiViu>irviooi(siivo(roi^iirtiooio>ioir-i(Mir^iooiONi
1 1 1 1 1 1 1 IT-I Ir-li-li-li-lr-lNlNlNlCSIlNlNl
lUJlUllUllUllUJlUtlUJlUJlUllUJlUJlUJlUJlLlJlUJlUJlUJlUJlUJllUl
IZIZIZIZIZIZIZIZIZIZIZIZIZIZIZIZIZIZIZIZI
IDlDlDlDlDlDlOlDlOlDlDlDlDlDIDlDlDIOlDlOl
iNiNlNiNiNiNlNiNiNiNiNiNiNiNiNiNiNiNiNiNi
lr>l(MirOl^ltAlsOir^lOOIOsiOlT-lNlhOl^llAlv«i|v.iOOIOviOI
1 1 1 1 1 1 1 1 i l^lv— Iv— l^li la l^l^l^«lw»ir^ll
B-17
u.
D
-i
i(Mivoiroiooir^iavioir>iNiooimii/>i(vjir^i
i(Mioiiniovir>-iNiooir-iNiosiooiK?i<-ieoi
IOlr-lr-lr-irOlt/)l(MlOtOIVIO\|(Ml(Mlini
1 .- 1 j; 1 r-v 1 so 1 1 »o 1 <^ 1 -o- 1 r>. 1 NO 1 ^ 1 r- 1 Ki 1 M 1
1 1 N 1 1 vo 1 1 1 1 1 1 1 1 1 1 ,
' ■ I 1 r^ 1 1 1 1 r 1 1 1 1 1 1
lUJlUJlUIi 1 1 lUJlUliUJlUllUJiUJlUllUJl
IBIS'S' ' ' 'Z'Z'Z'Z'ZIZIZIZI
iDlDlDl 1 1 IDlDlDIDlOlDlDlDl
NiNiNi 1 vo 1 vo 1 N 1 N 1 N 1 N 1 N 1 N 1 N 1 N 1
Q. 1 Dl 1 OL 1 1 UJ 1 Ul 1 Q. 1 a! 1 Ol 1 (L 1 OL 1 a.* 1 CL ! CL 1
SiSiSi 1 Z 1 Z 1 S IE 1 £ IS 1 E IE IS 1 S 1
DiDlOl lOlDlOlOlDlDlDlOlDlDl
OlOlOl 1 N 1 N 1 O 1 O 1 O 1 O 1 O 1 O 1 O 1 O 1
OIOIOIOIOIOIOIOIOIOIOIOIOIOI
r^ir^ir^ioi»oivi^ihOitoiK)i^iroiTrivi
1 1 1 UJ 1 1 1 1 1 1
1 1 1 Z 1 1 1 1
1 1 1 D 1 1 1 1 1 1
oif-iNiNiosioit-ivoir^irOiVimivoi»oi
Kl 1 ro 1 K) 1 1 1 r- 1 .- 1 r- 1 f- 1 CSI 1 M 1 N 1 N 1 1^ 1
UllliJillliQ,iLlJlUJlUJlUJlUJlUJlUJlUJlUllUJl
Z.ZlZlSlZiZlZlZlZiZlZlZlZiZl
DiDlOlDlDlDlDlDlDlDlClDlDlOl
NiNiNiOiNiNiNiNiNlNiNiNiNiNi
•-lCVJiKOi^llAivoi|^iOOiO«lOi,-l(SJl»Ol^l
NlNiNi(SllNlCSIirgi(MirsllK>irOiK)irOiK)l
B-18
CM
<
|k
::
*
t
*
X
9!
*
*
*
1
J!
1
t
*
*
z*
1-
3 :
H-
*
H
*
1-
*
1-
Oij'
M
'«■'»■*
M
VON*
M
CNir^*
M TnO*
M
(SIV
MX
EinoNoo*
£01000*
srocsiM*
Eoocoo*
E^0O(^
H*
M •
ONOOH
M •
iftr>.*
M •
NOlft*
M -inoN*
M •
IN.V
1
V
=>
*
*
*
*
*
m
3
i
I
1
*
*
*
O*
o
o
*
o
*
o
*
o
*
z*
o
o
*
o
*
r«-
*
r-
*
i
•
•
*
•
*
*
*
N
CO
*
(SI
*
fmm
*
(Sl
*
4>
*
*
*
*
*
*
*
*
*
*
*
*
1
-
N
1
*
K)
*
*
*
V
*
1
in
1
X
*
*
*
x
UJ(/)
UJto
1
Miin
*
UIOO
*
UJ(^
X
Zm
Zm
Zm
*
Zm
*
ZM
^
o
1-
D
»-
*
D
H
*
O
H
*
D
h-
-J*
N_i
l-t
VO*
N-i
M
vo*
N_i
M
NO*
N_l
M VO*
N-I
M
(SIO
UJ:|c
II UJ
Sor^TT*
II Ul
EtnoNrs.*
II UJ
Eifteor^*
II UJ
Zinvrv*
II UJ
Ein
i-V
(SI
Q*
Q
M •
ircsi*
Q
M •
r-(SI*
Q
M •
VIO*
Q
M -ror-*
Q
M •
MSOCSI
O*
UJD
-INO
• •*
UJD
-Itn
UJO
-J in
• •*
UJD
_iin •
•*
UJD
-iin
H
£3|C
z:^
(NJr-*
Hs:
NK)*
ZZ
r-.-*
Es:
Nr
-*
s:s:
(NJr-H
0£
-1
<
Df
1 *
-
ON*
*
1
*
*
*
*
*
*
*
>-
h-
*
*u.
*u.
*
*u.
*
*U.
*
*u.
1-
M
*
*o
*
*D
*
*D
*
*D
*
*o
M
u
°i
*
^
*
*
*
*
*
*
*
o
M
•*
*UJ
i
*UJ
*
*UJ
*
*UI
*
*UJ
M
E
NO*
o
o
*
O
*
o
*
o
E
.*
Of
Z
*
w
*
to
*
V)
*
(0
*
H*
UJ
*
UJ
*
UJ
*
UJ
*
Ul
*
I*
n
*
eq
*
m
*
CO
*
n
*
O*
*
*
'J-
*
V
*
V
*
V
*
M*
Ul
*
UJ
*
UJ
*
UJ
*
UJ
*
UJ*
o
*
Q
*
Q
*
a
*
Q
*
3*
CIS
1
D3
*
DZ>
*
OS
*
03
. *
ZH
Zl-
*
ZH
*
ZK
*
ZH
U.*
l-«
*
M
*
M
*
M
*
M
-*!!!
02
*
OZ
*
OZ
*
OZ
*
OZ
Ul*
UJO
*
UJO
*
UIO
*
UJO
*
UJO
v>*
o:
B-19
4e
#
4c
4c
^
i
*
4e
4:
4c
*
*
4c
4c
4c
4c
*
4c
»-
*
1-
4c
1-
4c
H
*
(-
9|e
M
rr,-*
M
MIA 4:
M
(MO*
M
(MO*
M
cao
J
ZTOr-TT*
SOOyo4:
Zleoino*
:E.\t\^f^it.
£0000
9|:
M
•MCO*
M
•On ON*
M
•oo>*
M
•,-NO*
M
•ODO
1
-ISO • •*
-ir-. • •4c
_jin • •*
-Jin • •*
_JV4) . .
CM 1 *
N 1 4c
to,-*
V0N4C
^ 1
4c
Oi
♦
0£.
4c
OC
1 4c
Q£
1 4c
0£
*
o
Hi
*
o
UJ
4:
UJ
4c
Ul
4c
UJ
*
o
a.
*
o
0.
4c
If)
a.
4:
IT)
0.
4:
a.
*
•
a.
4:
•
a.
4c
•
a.
4c
•
a.
*
CL
*
lA
z>
4=
*
3
4:
V
3
4c
T
3
*
«£
3
*
*
4c
4:
1
*
4c
4:
4c
*
1
*
4c
4c
4c
4c
*
4:
4c
4c
4c
4c
*
o
*
o
4c
\f\
4c
*
4c
4:
rv
4=
in
4:
t^
4c
*
1
*
"
4=
•
4:
•
4:
•
*
*
*
f—
4:
r-
4c
t^
4c
(M
4:
'a-
1
*
4:
4c
4c
4e
*
^
s
s
4e
4c
1
1
4c
4e
4c
4c
4c
4c
4c
1
*
v«
4c
r>»
4:
00
4c
0>
*
4c
*
4:
4:
4c
*
^
*
4:
4:
4c
4c
*
UJW
4:
UlV)
4:
UJ(0
4:
UJM
*
UJy)
*
Zi-l
4:
Zm
4c
ZM
4c
Zm
4c
ZM
*
o
h-
4c
D
1-
4c
D
H-
4c
H
*
D
h-
4:
N-l
M
VOy*
N-J
M
(MO*
N-l
M
CMO*
N-l
M
(MO*
N— 1
M
C3C3
*
II III
SlftTTO*
II UJ
Zir
.-0*
II Ul
Hocvlin*
II UJ
z:a
»-v0 4c
II UJ
Z:inoo
*
Q
M
i-eo*
a
M
S0(M*
Q
M
COON*
M
00 CO 4:
Q
M
0(M(M
*
UJO
_jin
• '4:
UJO
_jir
• •4c
UJO
_iir
• •4c
UJD
-iir
• •4:
UJD
_JU1
• •
*
ES
M 1 4:
ss:
Mr-*
z:z
Ni-*
z::^
^(M*
z:z:
tOf-l-
-^
*
«
e:
4c
-
<
*
*U-
4c
4:U.
4:
4c U.
4c
*U.
4c
4:U.
H
*
*o
4c
4:0
4c
4:0
4c
4:0
*
4:D
M
*
*
4:
4c
4c
4:
*
4c
*
4c
*
*UJ
4:
4c UJ
*
*UJ
4c
*UJ
4c
*UJ
M
4:
o
4c
4c
4c
1
z:
*
z
4c
Z
4c
z
4:
Z
Z
V)
*
eo*
•X
Min
T-ro*
4c
D3
*
D3
*
ZK
4c
Zl-
1
ZH
4c
Zt-
4c
ZK
*
l-l
4c
M
t-l
*
M
4c
M
*
OZ
4c
OZ
4!
OZ
1
OZ
4:
OZ
*
UJO
4c
UJO
4:
UJO
UJO
I
UJO
*
o:«i
4:
0£..4c
M r-i
04:
M r-
04c
M OvO
:(c
ZrOl>.M4:
Sin (MX. 4c
EMr-.04:
Eeooo4:
Emino
*
M -f—
so 4:
M •eoin4c
M -t-i
CO 4:
M -r-
D0 4C
M •rrtc
*
_ivO •
•4c
_|SO •
•4c
-ISO •
•4:
_jin .
•*
-ISO • •
*
K)
1 4:
*
■1
^
1 4:
ro
1 4c
ro 1
*
(k:
4c
q:
Q£
4c
o:
4:
OC
if
o
UJ
4c
o
UJ
4c
o
Ul
4c
Ul
4!
Ul
1
o
a.
4:
in
0.
4c
in
Q.
4c
in
0.
4c
in
0.
*
0.
4:
•
0.
4c
•
Q.
4c
•
0.
4c
•
0.
:(c
u\
3
4c
in
3
4c
in
3
4c
in
3
4c
in
3
1
4c
4c
4c
4c
4:
4c
4c
Hfi
*
*
4c
4c
%
4c
2
*
*
4;
4c
4c
*
4c
4c
*
*
$
*
o
4c
o
4c
o
4:
4c
4c
*
in
4c
in
4c
o
*
4=
*
*
4c
•
4c
•
4:
•
4:
•
4c
*
*
K>
4c
4c
ro
1
^
4c
4c
^
4c
4c
TT
4c
4:
*
4c
4c
4:
*
4:
4c
4:
*
4:
4c
4:
4c
4c
4:
*:
4c
4c
3|:
r-
*
CSI
4c
KJ
4:
rt
4c
in
4c
1
f^
4c
f
4c
^
4c
r*
4c
r—
*
4c
4c
4:
4:
\11V>
*
uicn
4:
UJ(/0
4=
\n-
PQ
<
4c
4c u.
4c
4:U.
4e
4c U.
4c
4:U.
4:
4:U.
H-
4c
4:D
4c
4cO
4c
4:0
*:
4=0
4c
4:D
M
4c
4c
4c
4c
4c
4c
4:
4c
4c
4:
u
4c
*lii
4c
4:UJ
4:
4:UJ
4:
4:Ui
4:
4c UJ
M
4c
O
4c
O
4c
4c
4:
E
4:
Z
4:
z
4c
Z
4c
Z
4c
Z
.04:
oe.
E r-
■04c
Q£
E ovo
1
tf
IHOr-r>-4c
<
MOv« •
•4:
z
4c
D3
4c
oz>
4c
D3
4c
ZH
jjc
ZH
ZH
4c
ZH
4:
Zt-
4c
M
3|C
M
i
M
4:
M
4c
M
4:
oz
4:
OZ
OZ
4c
OZ
*
OZ
4:
LUO
4c
UIO
4:
UJU
4:
UJO
$
UJO
1
(^
<:z.
4c
•Z>
r'.::
ii;v^:'
g^^^:-:-:--'.>
IkI^ • : ■
mm^^
.'.«
m-
m
V'.''.-'
•■'./.-:•;
■■•X'« • •
JE
N 1
1 *
fo
1 3tc
M
'1
*
1 *
ro 1
: :
Q£
1
0:
*
oe.
0£.
*
Q£
9E
UJ
UJ
*
Ul
*
UJ
St:
UJ
: E
in
0.
*
CL
*
a.
*
in
0.
3te
(L
:e
•
Ol
*
.
a.
*
•
a.
*
•
0.
3tC
•
0.
: E
9E
IE
IE
:e
in
Z>
*
10
z>
%
u\
3
*
*
t
in
3
1
*
in
3
: E
Hi
*
*
9E
9E
%
%
*
1
*
*
IE
9E
*
%
in
*
9E
4e
.
*
•
•
*
•
♦
•
9E
9E
V
*
*
%
M
1
^
:4c
!k
^
9E
9E
; :
i
*
1
*
9E
9 E
3 E
J
*
*
*
3|E
9E
9E
9E
^
*
CM
*
1^
*
*
3k
in
9E
9E
(SI
*
CM
N
*
N
9|C
N
3:
*
1
*
9E
UJco
*
UJ(0
%
UJ(/)
Ul<0
*
UiVt
9E
ZM
*
ZM
ZM
3tE
ZM
*
ZM
9!
H
*
y-
*
H
*
H
*
h-
3 !
N-J
M OOMjC
N_i
M t^OltE
N-i
M 004:
N-J
M OOltc
N-i
M r-O
9 E
II UJ
£inooi/>3ic
II Ul
:^inavO:t:
II UJ
SfOOO*
II Ul
Eoooitc
II UJ
£inoo
9 E
Q
l-l 'r-0\t
Q
M -OMst:
Q
M -NNJft
Q
M -NlOt
Q
M -ONN
9E
UJD
_iin •
•*
UIO
_llf> .
•*
UIO
_jin •
•*
UIO
-ISO •
•*
UJO
-im . .
3 E
E£
(M {
1 *
s:s:
(Mr- 3*:
zz
Nt-*
EE
Nr-*
:e:s.
Nr-H
9E
9E
<
D£
*
9E
Q.<
I E
IE
0.4
1 E
IE
0.4
3 :
IE
(L<
f^
9E
<«
9E
<»-
IE
9E
9ED
IE
IE
9EO
3 E
IE
IE
1:0
1 :
3E
3E
1*°
IE
IE
i^
M
u
3E
3EUJ
IE
seUJ
3 E
i:UI
: E
%\ll
IE
♦ UJ
M
3 E
9E
IE
3|E
3|C
E
3 E
z
IE
z
9E
Z
3k
3|C
z
±
34c
z
Vt
9E
NO •
•*
z
coin •
•*
z
(/>sO .
z
V)sO . •34c
3E
3C
M
UJ N 1 *
M
UJ CMi
F-*
M
UJ Nr-«
M
Ul roi
— :4e
M
Ul N<-3k
-J
♦
^
1 *
^
1 Ik
-1
1 *
^
3):
H
*
H
H
*
h-
*