8 "
∞∞∞
* * * * * * * * * * * * : * > .
a * * * * * *
= e tº a s a s
; : * * * * * * * *
* * * * * * * * * * * * * * * * * * * , ; ; ,
* * * * * * * *
e º a e º ºs
esp a car a
º
is a .
| .
| a
; ; , , , , ; ;
...]
#:
to a
a * > . * * *
.* “ . . . . . .
*： < ：*： <-- ！=-
««^ ;：.***--sae:
g … ， ） ---- « = » • • • •
Hººaegass.sváſzawy nezrºzºrºaerosºru
?!！！！!
—~~~~ <æº-~ae º ºsºbºz，
£ € （~ ~ ！， -， *， -**， *） ≤ （ ººº ！！ ！！ ！！
، ، ، ، ،
~ ~ -…,- s -, -, → … ~ ~ ~ ~ ~ . ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~ ~ ~ ~ ~ ~ ~ ~
* s; }
saº. - ）
∞ √≠ ≤ ≥ ≡ ? » ：
xº º














A Landscape-Level Analysis of
Urbanization Influence and Spatial Structure in
Chaparral Breeding Birds of the
Santa Monica Mountains, CA f
Js’ * / ſ.
• .. *. w"
, , , ; 9. tº *
Łº # & Cº . .# wº
/ ; * cº 6 º'
tº
*...
- ...~
f & A" *
..f. ****, *.**
- sº >. ... "
Diana Stralberg // ... "
i M .” -- …
^- { }
A thesis submitted
in partial fulfillment of the requirements
for the degree of Master of Science
(Natural Resources and Environment)
in the University of Michigan
April 1999
Thesis Committee: -
Associate Professor Terry L. Root, Chair
Dr. Shuming Bao
Abstract
This study used point count censuses and GIS land cover data to assess the
relationships between chaparral breeding bird abundance and urban encroachment at the
landscape level. A set of chaparral-vegetated points across an urbanizing landscape in
the Santa Monica Mountains of southern California was surveyed for birds in 1996-97
and mapped in a GIS. I first analyzed the effects of surrounding urbanization on eighteen
individual species’ abundances using Poisson regression models with four measures of
Surrounding urbanization. Then, due to the spatial distribution of point count locations
over a 17,000 ha area, I also examined and modeled spatial dependence in four selected
bird species using geostatistical methods.
Thirteen out of eighteen species examined exhibited significant (p<0.05)
responses to at least one measure of surrounding urbanization, with bootstrap resampling
trials indicating relatively robust models. Eight “chaparral specialist” species and two
woodland-associated species exhibited significant negative relationships with
urbanization, while two urban-associated species and one “chaparral specialist” were
positively associated. Most of these species were associated with surrounding
urbanization proportion at large (> 250m) radius distances. Landscape pattern explained
little additional variation in abundance, although responses to urban edge proximity in
Some species appeared to be greater in highly fragmented areas.
Exploratory geostatistical analyses, using semivariograms and Moran’s I
correlograms, revealed large- and small-scale spatial dependence in all four species
examined. Large-scale trends were largely attributed to the urbanization gradient over
which points were sampled. Remaining small-scale autocorrelation in urbanization
model residuals was modeled with a spatial covariance structure, and results were
compared with de-trended and non-spatial models. Conclusions about urbanization
effects did not change for three of the four species examined. Semivariogram and
correlogram analyses also provided further insight about the spatial structure and ranges
of spatial autocorrelation in the species examined.
Acknowledgements
First and foremost, I wish to thank my advisor, Dr. Terry Root, for her unending
Support, advice and assistance with this research project for demanding scientific rigor,
and for encouraging my career interests and aspirations. She has served as a great source
of Scientific and personal inspiration for me and numerous others. I am also extremely
thankful to Dr. Shuming Bao, for introducing me to spatial statistics and providing
exciting new research opportunities, as well as countless hours of assistance and advice.
I am eternally grateful to Dr. Michael Burger, Kim Hall, Thomas Dietsch and
Tonya Haff for providing insightful and crucial comments on this manuscript. Other past
and present Root lab members, including Dea Armstrong, Bridget Fahey, Dr. Lisa
Sorensen, Amie Ottinger and Dr. Russ Butler, have provided invaluable support and
assistance at various stages in my research.
Many thanks for statistical consultation are given to Kathy Welch, who tackled
numerous statistical challenges with grace and determination, going above and beyond
the call of duty, and Dr. Al Jensen, who provided many insightful comments and
suggestions. I also extend my gratitude to Dr. Andrew Brenner and the SNRE GIS lab
for use of computers, software and helpful technical support. .
I am very grateful to Michelle Petit for providing excellent field assistance for
two seasons, as well as inspiration, enthusiasm and support. Daniel Cooper also
contributed valuable field assistance and birding expertise.
This research would not have been possible without the assistance of Dr.
Raymond Sauvajot and Denise Kamradt of the National Park Service, Santa Monica
Mountains National Recreation Area, who provided GPS equipment, GIS landcover data,
and valuable consultation. Funding was provided by the Southwest Parks and
Monuments Association, as well as the Population-Environment Dynamics Project,
Rackham Graduate School and School of Natural Resources and Environment at the
University of Michigan.
Finally, a special thanks to my parents, who never questioned my choices, and
continue to accept, respect and support my goals and ambitions.
ii
Abstract................................................................................................. i
Acknowledgements....... • * * * * * * * * * * * * * * * * * ~ e º e º e º e º e º e e º e o e º e º e dº e º e º e s e e e o e s e e s e e e s e e e º e c e º e o e o e s e e s e ii
Table of Contents.................................................................................... iii
Chapter 1. Effects of Urban Encroachment on Chaparral Breeding Bird Species in the
Santa Monica Mountains, CA
1.1 Introduction..…..…........…............................................. 1
12 Methods.’…...….......................................... 8
1.3 Results.…....…......................... 13
14 Discussion...........................…........................................................... 16
1.5 Conclusion............................................................................................... 30
1.6 Literature Cited....................................................................................... 32
Chapter 2. Examination and Modeling of Spatial Structure in Selected Chaparral Species
2.1 Introduction.............................................................................................. 61
2.2 Field Methods and Previous Results......................................................... 71
2.3 Spatial Statistics Methods and Results..................................................... 71
24 Discussion................................................................................................ 83
2.5 Conclusion................................................................................................ 89
2.6 Literature Cited.....................…............................................. 91
Table of Contents
iii
Chapter 1:
Effects of Urban Encroachment on Chaparral Breeding Bird Species
in the Santa Monica Mountains, CA
1.1. Introduction
In an ecological context, urbanization represents an extensive modification of the
landscape in order to support increased human habitation (McDonnell and Pickett 1990).
Ecological forcing functions created by the growth of cities and associated human
activities, which can collectively be referred to as “urbanization,” have become
increasingly common topics of ecological study, as “natural" systems become more
difficult to find. Hydrology, nutrient cycles, microclimate and species composition are
but a few of the ecological features that may be substantially altered by urbanization
influences.
1.1.1. Urbanization in Southern California
The traditional model of urban expansion consists of an urban core surrounded by
progressively less intrusive land-uses, such that agricultural lands often provide a
transition zone from urban to natural areas (Hall 1966). In contrast, development in
southern California has been uniquely different from this traditional model, with its
decentralized settlement pattern rooted in an anti-urban bias and a “faux-rural” suburban
ideal that was prevalent among early 20th century settlers and developers (Fulton 1997).
Unlike other major U.S. cities, metropolitan Los Angeles was formed out of a series of
centers that grew and merged together to form a “metrosea” of fragmented communities
(Davis 1992). Los Angeles County alone has a current population of 9.6 million people
(County of Los Angeles 1998).
Despite Los Angeles’ natural unsuitability as a major city—including the scarcity of
natural water supplies, unstable geology, and the lack of a natural harbor—the “growth
machine” set into motion around the turn of the century resulted in unprecedented land
- 1 -
- 2 -
development and population growth in the Los Angeles basin (Fulton 1997), replacing
orange groves and asparagus fields with tract homes and freeways. As the flat, coastal
plain approached saturation, developers turned to inland deserts and mountain areas,
building on increasingly marginal lands. Thus much of post-World War II development
has occurred on lands that are too steep, infertile or dry for agriculture (Scott 1995). In
the region’s mountains, this development pattern has resulted in an extensive urban-
wildland interface, and landscapes dominated by two major land-use classes: native
vegetation (mostly chaparral scrub and woodland) and residential development. Indeed,
metropolitan Los Angeles has the longest “wild” edge (roughly 1100 kilometers long, by
the crudest measure) of any major city in the U.S. (Davis 1998).
This type of landscape mosaic is well-represented in the Santa Monica Mountains, a
coastal mountain range containing a mosaic of interspersed suburban development and
native chaparral vegetation, generally increasing in urbanization from west to east.
Urbanization has occurred mostly along canyon bottoms, and often on ridgetops,
resulting in “fingers” of development that penetrate the native chaparral habitat. In
response to threats of large-scale developments and road construction, the United States
Congress established the Santa Monica Mountains National Recreation Area in 1978 to
conserve natural and cultural resources, as well as the “airshed” of the Los Angeles Basin
(National Park Service 1999). Notwithstanding several major federal and state land
acquisitions, most land in the Santa Monica Mountains is privately-owned. With some of
the highest real estate values in the nation, these private lands remain subject to heavy
development pressures.
Urban encroachment may have a diverse array of impacts on native wildlife and
vegetation. From mountain lions (Edelman 1992, Smallwood 1994) to native ants
(Human and Gordon 1997, Suarez et al. 1998), organisms of the chaparral and other
Mediterranean-type ecosystems in southern California have been adversely affected by
the direct and indirect effects of urban development, including habitat fragmentation,
exotic species invasion, and other human disturbances.
- 3 -
Chaparral vegetation, although diverse in plant species, is quite uniform structurally, and
contains relatively few bird species (Miller 1951, Small 1974). The birds that do occur
and breed in chaparral tend to be somewhat specialized in their habits. Many species are
ground-foragers with good running abilities, allowing them to navigate narrow alleyways
under the dense scrub canopy (Small 1974). According to Miller (1951, p. 547),
approximately 46 percent of bird species found in chaparral exhibit “primary or exclusive
adherence to chapamal ..a high proportion for a terrestrial formation.” Most of these
Species are year-round residents, characterized by small home ranges, low vagility (Soulé
et al. 1988) and short natal dispersal distances (Baker et al. 1995). As a result, these birds
may be particularly vulnerable to changes in the surrounding landscape, as they are
continuously exposed to urbanization influences. Furthermore, the abundance of
chaparral specialists may be closely tied to the ecological integrity of the chaparral
vegetative community. Easily observed and monitored, some avian species may serve as
chaparral indicators.
1.1.2. Avian Responses to Urbanization and Habitat Fragmentation
Studies along urbanization gradients have shown that total bird numbers generally
increase with urbanization (largely attributable to a few abundant species), while species
diversity decreases (Emlen 1974, Hohtola 1978, Beissinger and Osborne 1982, Jokimaki
and Suhonen 1993). Factors associated with species’ declines in urban areas include
changes in vegetation volume (Mills et al. 1989), foliage height diversity (Lancaster and
Rees 1979), pedestrian and vehicular traffic (Emlen 1974), and increased domestic dog
and cat predation (Soulé et al. 1988). In arid and semi-arid landscapes, Some species may
be drawn to water and cover provided by suburban landscaped gardens (Emlen 1974,
Guthrie 1974, Vale and Vale 1976), but negative relationships have been demonstrated
between species richness and housing density, as well as other measures of urbanization-
related habitat change (Emlen 1974, Germaine et al. 1998).
In addition to direct habitat modification, urban development has resulted in the
fragmentation of remaining natural areas. Referred to as “the single greatest threat to
- 4 -
biodiversity” (Noss 1991, p. 27), habitat fragmentation may result in the exclusion of
Species with large home ranges (Wilcove et al. 1986), as well as reductions in gene flow
between isolated populations (Gilpin and Soulé 1986). The equilibrium theory of island
biogeography (MacArthur and Wilson 1967), when extended to habitat fragments,
predicts that smaller, more isolated patches will support fewer species, due in part to a
higher probability of extinction when populations are small (Soulé et al. 1988, Kadmon
and Pulliam 1993). Although the application of island biogeography theory to habitat
fragmentation has been criticized as simplistic (Wiens 1994, Simberloff 1994), numerous
forest fragmentation studies have demonstrated the so-called “area effect,” or tendency
for larger habitat “islands” to support more species than similar smaller fragments (Blake
and Karr 1984, Forman et al. 1976, Lynch and Whigham 1984, Whitcomb et al. 1981).
Furthermore, habitat fragmentation in urbanized landscapes may be different than
fragmentation in (more widely studied) agricultural or timber production landscapes, in
that transitions are more abrupt, and urban development is likely to constitute a less
hospitable habitat for native species than forest clearings or agricultural fields.
Habitat fragmentation also expands the interface between natural habitats and urban
development, thereby providing increased opportunities for urban elements to affect bird
populations. Several studies of Neotropical migrant birds in Eastern U.S. forest woodlots
have demonstrated significant negative relationships between avian diversity and
surrounding urbanization density (Tilghman 1987, Dickman 1987, Friesen et al. 1995).
Widely-recognized mechanisms include edge-related changes in vegetation composition
(Saunders et al. 1991) and increased avian nest parasitism and predation rates (Robbins
1980, Brittingham and Temple 1983, Wilcove 1985, Langen et al. 1991, Engels and
Sexton 1994). Non-native plant invasions (Mills et al. 1989), human disturbance
(Dawson 1990) and various other features of urbanization (e.g., Johnsen and VanDruff
1986) have also been implicated in songbird declines.
Several recent studies have examined the effects of urbanization and habitat
fragmentation on birds of southern California’s chaparral and coastal sage scrub habitats.
- 5 -
Soulé et al. (1988) applied the species-area relationship of island biogeography theory to
chaparral canyon fragments in San Diego County, demonstrating that the number of
“chaparral-requiring” resident bird species increased with fragment size and decreased
with time since fragmentation. Bolger et al. (1997) revisited this area with a landscape
approach to habitat fragmentation, showing that landscape composition and urbanization
pattern were significant predictors of occurrence for many species examined, some of
which were identified as edge/fragmentation-sensitive. In the less-fragmented eastern
Santa Monica Mountains (examined in this study), Sauvajot (1997) found anthropogenic
vegetation disturbances to be more important than urban development proximity in
determining resident bird diversity and abundance.
1.1.3. Scale of Urbanization Influence on Birds
The effects of urbanization on breeding bird communities may be manifest at a variety of
Scales, with the importance of different ecological processes varying across scales
(Addicott et al. 1987, Turner 1989) and potentially interacting between scales (Levin
1992). Just as different bird-habitat associations have been found at different spatial
scales (Wiens and Rotenberry 1981, Riitters et al. 1997), the detectable effects of
urbanization on bird populations are likely to be sensitive to the scale of sampling and
analysis.
To date, a majority of studies examining the effects of urbanization on bird communities
have focused on direct local impacts, examining bird population responses within urban
areas, or immediately adjacent to urbanization activities (e.g., Scott 1993, Blair 1996,
Clergeau et al. 1998, Germaine et al. 1998). For landscape-scale analyses, however, the
extent of urbanization influence becomes an important consideration. The identification
of the appropriate scale of influence or “ecological neighborhood” (Addicott et al. 1987)
with respect to a particular process is an important part of predicting and understanding
the relationships between urbanization and bird communities.
- 6 -
Previous studies have varied with respect to scale of urbanization influence examined,
perhaps due to logistical constraints, and/or regional differences. Pearson (1993), in a
Study of Midwestern winter bird populations, did not detect landscape influences beyond
200m from a sampling location (maximum distance examined was 300m). Bolger et al.
(1997), however, found significant associations between bird abundance and landscape
composition at distances of up to 3000m. Given that a consistent pattern has yet to
emerge, the examination of multiple potential scales of influence should help illuminate
the ecological processes involved and how they may vary with scale.
1.1.4. Landscape Composition vs. Pattern
According to Opdam et al. (1995, p. S140), landscape pattern can be defined, for a
particular species, as “the spatial configuration of patches of habitat and those landscape
elements that influence dispersal.” In highly fragmented landscapes, these elements
include patch configuration, size and insularity. Equilibrium models based on island
biogeography and metapopulation theories predict species declines with increasing
habitat loss and isolation (MacArthur and Wilson 1967, Levins 1969, Brown 1978,
Opdam 1991, Hanski 1994). -
The effects of landscape pattern are less well-studied in landscapes with interconnected
suitable habitat still remaining. Andrén (1994) concluded, through a review of
fragmentation studies, that in landscapes with more than 30 percent of suitable habitat
remaining, the effects of habitat fragmentation are primarily due to habitat loss, rather
than fragmentation pattern (i.e., patch size and isolation). In other words, if the
proportion of remaining “habitat” in a landscape is above this critical proportion, then the
landscape is well-connected and species will be able to disperse; otherwise, dispersal
distances become important in determining species abundance and persistence (Gardner
et al. 1991, Metzger and Décamps 1997). Other studies have identified different critical
habitat thresholds (e.g. With and Crist 1995, Keitt et al. 1997). -
- 7 -
Thus, depending on the proportion of suitable habitat remaining in a landscape,
fragmentation pattern perse may be of varying importance to bird distribution and
abundance.
1.1.5. Urban Edge Proximity and Direct Human Influence
Habitat fragmentation may provide increased opportunities for wildlife to come into
contact with human activity and other urban elements (e.g., domestic animal predators,
invasive species, chemical pollution), via the lengthening of the urban-wildland interface.
Previous studies have indicated that in fragmented landscapes, certain chaparral bird
species exhibit edge-sensitivity—with respect to urban development, as well as other
habitat edges (Bolger et al. 1997, Sisk et al. 1997). In contrast, Sauvajot (1997), in a
similar, yet less fragmented area, found no effects of urban edge proximity on bird
abundance, but did find a significant reduction in both species richness and abundance of
resident species in areas of human disturbance (i.e., cleared and eroded surfaces, such aS
roadcuts and landslips).
Habitat edges of all kinds are known to attract certain bird species, while repelling others,
often due to high predation risk. Miller et al. (1998) found evidence that trails may serve
as habitat edges for some species, attracting more habitat generalists, and deterring
specialist species. Increased rates of nest predation were also found near trails. In
addition, recreational trails may serve as “conduits” for urbanſhuman influences into
natural areas.
1.1.6. Research Objectives
Wiens (1994) describes habitat fragmentation as a dynamic process in which habitat
continuity is disrupted, producing a variety of patterns, ranging from “small breaks in an
otherwise homogeneous habitat” to “widely scattered units of remnant habitat in a
transformed matrix.” The eastern portion of the Santa Monica Mountains represents just
such a fragmentation continuum (Figure 1.1). The variation in landscape composition—
from primarily chaparral to primarily urban development—allows for an examination of
- 8 -
landscape-level urbanization effects on chaparral ecosystems. Although the effects of
urban edge proximity on resident birds in the Eastern Santa Monica Mountains have been
examined previously (Sauvajot and Beuchner 1991, Sauvajot 1997), no previous
landscape-level analyses have been undertaken.
In this study, I investigate the effects of urban encroachment and habitat fragmentation on
breeding chaparral birds (year-round and summer residents) in the eastern Santa Monica
Mountains. While most studies of habitat fragmentation have focused on the last phase
of the fragmentation process (isolated habitat patches), my study also considers
urbanization in its beginning and intermediate stages, where much of the remaining
habitat is still interconnected.
My research questions were:
1. Is chaparral bird abundance affected by the proportion of urbanization in the
Surrounding landscape? If so, which species decline, and which species increase with
increasing urbanization? What characterizes urbanization-sensitive species?
2. At what spatial scale (between 250m and 4000m) does urbanization proportion affect
bird abundance? What is the maximum spatial extent of urbanization influence?
3. What is the importance of landscape pattern, when urbanization proportion is
constant?
4. What are the effects of urban edge proximity and hiking trail density on bird
abundance?
1.2. Methods
1.2.1. Study Area
This study was conducted in the eastern end of the Santa Monica Mountains, which
belong to the Transverse Ranges of coastal Southern California. Located in middle of
- 9 –
metropolitan Los Angeles (Figure 1.2), the study area is bounded by Topanga Canyon to
the west, the eastern boundary of Griffith Park to the east, Mulholland Drive to the north,
and western Los Angeles and Sunset Blvd. to the south. Two interstate freeways, the 405
and 101, intersect the study area.
“True” or “hard” chaparral is the dominant vegetation cover in the study area. This
Mediterranean-climate vegetation community is characterized by tough, waxy evergreen
shrubs that are adapted to drought and frequent fire (Schoenherr 1992). Other vegetation
types in the area include coastal sage scrub, oak woodland, walnut woodland, and
sycamore/oak riparian corridors (Figure 1.3).
The western section of the study area (approximately 10,000 ha) is largely unfragmented,
with urbanization occurring primarily at the edges and as enclaves within the dominant
chaparral and coastal sage scrubland. The middle section (approximately 5,000 ha) is
highly fragmented by residential development, although most remaining chaparral
patches are interconnected. In both areas, most urbanization took place more than 25
years ago, although new developments are still occurring, particularly in the westernmost
section. The eastern section, Griffith Park, is a large (approximately 2,000 ha) habitat
island surrounded on all sides by urbanization (maximum distance from urbanization
~1,200m), and located a mere ten kilometers from downtown Los Angeles. Adjacent
land use consists primarily of single-family residential development, but a range of
residential, commercial and industrial land uses also surround the study area (Figure 1.1).
1.2.2. Field Data Collection
All bird abundance data were collected along trails and fire roads in Northern Mixed
Chaparral vegetation (modified Holland 1986 classification) at least 100m from urban
development. Census locations (points) were selected from a geographic information
system (GIS) trail map of the study area created by the National Park Service, based on
Harrison (1994) and supplemented with GPS-recorded trails. A stratified random Sample
was used to obtain representation in each section of the study area. All points were
- 10 -
dominated by a combination of Laurel Sumac (Rhus laurina), Ceanothus spp. (C.
megacarpus and/or C. spinosus), and Chamise (Adenostoma fasciculatum). Each point
represents a 100m radius area of chaparral vegetation, not overlapping with any other
area sampled. I sampled 78 points in both 1996 and 1997, and an additional 29 locations
were added in 1997 to increase coverage of areas with high levels of urbanization. The
locations of all census points were recorded with a global positioning system (GPS) and
mapped in an Arcſinfo Geographic Information System (GIS) (Figure 1.3).
I used 100m fixed-radius point counts (Ralph et al. 1993) to estimate bird abundance at
* each census location. Each point count lasted for ten minutes, and was conducted
between 5:30 and 9:30 a.m. PST. I counted all terrestrial birds seen or heard, except
those that flew over without landing. For simplicity, I omitted swallows, swifts and
raptors, which were difficult to count accurately. In both 1996 and 1997, two counts
were conducted at each point between mid-May and late July—one early and one late in
the season. In 1996, counts were conducted from June 8 through July 28. In 1997, I
conducted counts between May 19 and July 11.
1.2.3. GIS Methods
I used GIS land-use data (1:24000, from 1990 aerial photos, courtesy of Southern
California Association of Governments and National Park Service) to obtain landscape
measures. Using Arcſinfo GIS version 7.1.1, I calculated, for each census location, four
different measures of urbanization influence in the surrounding landscape:
Urbanization Extent. For each census location, I first calculated the area within a 250m
radius circle consisting of urban land uses (including golf courses, urban parks and other
human-constructed green spaces). I then calculated the area of additional urbanized land
contributed by consecutively larger circles, each radius twice as large as the previous
(500m-1000m-2000m-4000m). For the 250m circle, and each additional buffer ring, I
divided the urbanized land area by total area to obtain an urbanization proportion (see
Figure 1.4).
- 11 -
Fragmentation Index. For these same buffer distances, I also calculated a fragmentation
index, using the total length of urban/wildland edge within that radius of the point,
standardized by the total area. This represents the degree to which the surrounding
landscape is penetrated by urbanization, or the “convolutedness” of the urban-wildland
interface.
Distance to Nearest Urban Edge. To examine species’ edge-sensitivity and response to
urbanization proximity, I calculated the distance (m) from the center of each census
location to the nearest urban development. Due to sampling constraints, each point was
located at least 100m from the urban/chaparral edge.
Trail Density. To examine the effect of human activity (i.e., hikers, cyclists, horses,
dogs) and/or micro-scale edge effects, I calculated the density of trails within a 250m
radius of each point (meters of trail per square kilometer). Only a 250m radius was used
because the patchy coverage of the existing trail maps did not allow me to examine larger
alſea.S.
1.2.4. Statistical Methods
I analyzed individual species abundance, rather than species richness or other diversity
measures, since, in general, the same set of species was found throughout the study area.
I restricted my analyses to species that were easy to count and for which twenty or more
individuals were counted during at least one field season (1996 or 1997). All abundance
measures represent number of singing males, except when the total number of singing
males counted in either year was less than twenty. In these cases, I analyzed total
individuals (visual and auditory detections), rather than the number of singing males, for
both years. To best represent the capacity of the habitat in each year, I used the
maximum of two counts as a measure of species abundance.
Although counts of animal abundance can often be approximated by a normal or
lognormal distribution, the Negative Binomial (White and Bennetts 1996) or Poisson
distribution is usually more appropriate (Dawson 1981). The Poisson is a discrete
- 12 -
function, often positively skewed, with variance equal to the mean. To investigate the
scale at which each species was affected by urbanization (if at all), I fit a set of nested
generalized linear models for each bird species. Using the generalized linear models
procedure in SAS version 7, I specified a Poisson distribution with a dispersion parameter
estimated by the deviance, and a log link function (SAS 1997). Generalized Estimating
Equations (GEE) (Liang and Zeger 1986) were used to model the repeated measures
(1996 and 1997) correlation structure in the data.
Using measures of urbanization proportion within each buffer radius (i.e., ‘250m, 250-
500m, 500-1000m, 1000-2000m, 2000–4000m), I assessed the added contribution of each
urbanization variable to the model, and thus the maximum scale at which surrounding
urbanization explained significant additional variation in bird abundance. Score tests
(Tarone 1988) were used to detect significant improvements in model fit with each
consecutive radius distance. I then added a measure of large-scale habitat fragmentation
(length of the urban/wildland edge, standardized by area) within the maximum response
radius, as determined above, to detect additional responses to landscape pattern (above
and beyond urbanization extent). The overall influence of urbanization and
fragmentation combined was evaluated using score tests (Tarone 1988) for the
significance of these parameters at the maximum response radius for each species.
Finally, using this maximum response radius, I constructed GEE Poisson regression
models for each species’ abundance (again using a repeated measures covariance
structure). I initially included the following independent variables and interaction terms:
YEAR Year (1996 or 1997)
URBX Urbanization proportion within maximum response radius X
FRAGX Level of fragmentation (total length of urban edge) within
maximum response radius X
EDGEDIST Natural log of the distance from a point to the nearest urban edge
TRAIL Trail density (m/km’) within a 250m radius
- 13 -
URBX*URBEDGE Interaction between urbanization proportion (within radius X) and
distance to nearest urban edge
The inclusion of the interaction term was to test the hypothesis that edge sensitivity varies
according to the level of urbanization in the surrounding landscape. The urbanization and
fragmentation variables were included only if they were found significant in the previous
Scale analysis.
I examined the significance of each parameter estimate in the model, and eliminated non-
significant variables to construct the “best” possible model for each species. I then
performed a pseudovalidation of each final model, using a “bootstrap” resampling
approach (Efron and Tibshirani 1986) to obtain confidence intervals for model parameters
and detect potential sampling biases (Crowley 1992). For each species, fifty random
samples of points, each equal in size to the original sample (n = 107), were taken with
replacement, and used to fit the final models. Mean parameter estimates and confidence
intervals were calculated according to Little and Schenker (1995), and parameter
estimates were judged robust when confidence intervals did not overlap zero. When
models with several urbanization terms were not robust, scaled-down models including
just the most significant urbanization variable were also fit with a bootstrap sample.
1.3. Results
1.3.1. Species Encountered
Of the 38 species detected in this study (Appendix 1.1), 18 species were easy to count and
relatively common in the study area, according to my criteria (Appendix 1.2). By far, the
most commonly encountered species was the Wrentit, followed by the Spotted Towhee,
California Towhee, Western Scrub-Jay and Bewick’s Wren. These species were virtually
omnipresent throughout the study area, with over ninety percent site occupancy in each
year of the study (Appendix 1.2).
- 14-
Of species encountered, only the Wrentit is a true chaparral specialist. In coastal areas,
however, several other species have a primary affinity for chaparral, including the Ash-
Throated Flycatcher, Bewick’s Wren, California Quail, California Thrasher, California
Towhee, Orange-crowned Warbler, Phainopepla, Rufous-crowned Sparrow, and Spotted
Towhee (Miller 1951, Garrett and Dunn 1981). The remaining species may be
considered habitat generalists with respect to chaparral, in that their primary habitat
affinities are for other vegetation communities. The Mourning Dove, Northern
Mockingbird and Western Scrub-Jay can be classified as suburban-associated species
(Marzluffet al. 1994, Blair 1996); the Black-headed Grosbeak and Northern Flicker are
primarily riparian woodland-associated; and the Blue-gray Gnatcatcher, Nuttall’s
Woodpecker, and Oak Titmouse may be considered oak woodland species (Miller 1951).
1.3.2. Responses to Surrounding Urbanization Proportion at Various Scales
The abundances of eleven (out of eighteen) species were significantly associated with
surrounding urbanization proportion (p < 0.05) at one or more scales (Table 1.1).
Negatively associated with urbanization were the Ash-throated Flycatcher, Bewick’s
Wren, Blue-gray Gnatcatcher, Black-headed Grosbeak, California Towhee, Phainopepla,
Rufous-crowned Sparrow, Spotted Towhee and Wrentit. The Northern Mockingbird and
Western Scrub-lay exhibited significant positive associations (p < 0.05) with
urbanization. The Bewick’s Wren and Western Scrub-Jay were only weakly associated
with urbanization proportion (p > 0.01).
Different species exhibited different scales of urbanization influence (Table 1.1). All of
the above-listed species, except the Phainopepla, Spotted Towhee and Western Scrub-Jay
were significantly associated (p < 0.05) with urbanization proportion at the smallest scale
tested, a 250m buffer radius. The Rufous-Crowned Sparrow was affected only at this
scale. The remaining species exhibited significant responses to urbanization at larger
scales, with maximum urbanization response radii of 500m for the Wrentit; 1000m for
the Ash-throated Flycatcher, Bewick’s Wren, Black-headed Grosbeak, California Towhee
and Phainopepla; 2000m for the Northern Mockingbird and Spotted Towhee; and 4000m
– 15 -
for the Blue-gray Gnatcatcher and Western Scrub-Jay. Species exhibiting no significant
relationship (p > 0.10) with urbanization proportion at any scale included the California
Quail, California Thrasher, Mourning Dove, Northern Flicker, Nuttall’s Woodpecker,
Orange-Crowned warbler and Oak Titmouse.
1.3.3. Responses to Large-Scale Landscape Pattern
Only the Spotted Towhee exhibited an additional significant response (p<0.01) to
fragmentation of the surrounding landscape, as measured by length of the urban-wildland
edge within 2000m (Table 1.1). After accounting for the negative effect of urbanization
proportion, fragmentation had a positive effect on Spotted Towhee abundance.
1.3.4. Responses to Urban Edge Proximity and Trail Density
Several species exhibited significant responses to urban edge proximity. The Northern
Mockingbird and Orange-Crowned Warbler were negatively associated with distance to
nearest urban edge (Table 1.2), while the California Thrasher was positively associated.
In addition, for several species (Bewick’s Wren, Blue-gray Gnatcatcher, Phainopepla,
Northern Mockingbird and Wrentit), there was a significant interaction (p<0.05)
between urban edge distance and surrounding urbanization proportion, indicating a
landscape-dependent response to urban edges. This effect was negative for the
Phainopepla, and positive for all other species. For the mockingbird and Phainopepla,
urban edge distance overwhelmed the effect of urbanization proportion.
Trail density was only marginally negatively significant for three species (the California
Thrasher, Rufous-crowned Sparrow and Wrentit) in the presence of the other urbanization
measures (Table 1.2).
1.3.5. Annual Variations in Abundance
Most of the species examined had higher recorded abundances in 1997 than in 1996,
although the Phainopepla was more abundant in 1996 (Table 1.2). Species whose
- 16-
abundance did not change significantly from one year to the next include the Blue-gray
Gnatcatcher, California Towhee, Northern Flicker and Oak Titmouse.
1.3.6. Model Robustness
Bootstrap resampling trials for the full final models generally supported results obtained
with the original empirical models (Table 1.3). The more complicated models were less
robust, however, and interaction terms, in particular, tended to have large variances.
Urbanization proportion, at least in the scaled-down models, was a fairly consistent
predictor of abundance, with 95% confidence intervals large, but not overlapping zero,
for 7 out of 9 species. The interaction terms between urbanization proportion and edge
distance were less robust under resampling trials, with parameter estimates significantly
different from zero for only 3 out of 5 species. The resumping estimates for urban edge
distance and trail density were each significant for 1 out of 2 species. Although the effect
of census year on per site abundance was highly significant (p < 0.01) for most species,
the bootstrap parameter estimates varied widely and were not significantly different from
zero for any species.
Overall, urbanization models for the Ash-throated Flycatcher, Bewick’s Wren, Blue-gray
Gnatcatcher, Black-headed Grosbeak, California Towhee, Northern Mockingbird,
Phainopepla, Spotted Towhee and Wrentit were validated with the bootstrap resampling
trials (Table 1.3). Species whose models didn’t withstand the resampling
pseudovalidation included the California Thrasher, Rufous-crowned Sparrow and
Western Scrub-Jay.
1.4. Discussion
1.4.1. Urbanization-Sensitive Species
The primary objectives of this study were to determine which, if any, chaparral bird
species were associated with levels of urbanization in the surrounding landscape and to
identify common characteristics of these species. Thirteen of 18 species exhibited
– 17 -
significant responses (p < 0.05) to at least one measure of urbanization. Most chaparral
specialists examined (8 of 10) were negatively affected by surrounding urbanization,
while most of the remaining species either increased or exhibited no significant response
(6 of 8), suggesting chaparral habitat degradation as a mechanism for decline with
increasing urbanization. However, differential responses among species with similar
habitat requirements, as well as significant responses by non-chaparral specialists,
indicated that reduced habitat quality was not the only mechanism for decline.
Chaparral Specialists
Eight out often “chaparral specialists” exhibited significant negative responses to at least
one measure of urbanization in the surrounding landscape. Exceptions to this trend were
not difficult to explain. The California Quail, which exhibited no significant response to
urbanization measures, tends to utilize grassy, open areas, and may be found in residential
neighborhoods and golf courses (Ehrlich et al. 1988, Blair 1996). The Orange-crowned
Warbler, which exhibited increased abundance near urban development, is known to
prefer mixed chaparral/woodland habitats (Sogge et al. 1994), and may actually prefer
oak and walnut woodlands to chaparral in the Santa Monica Mountains (personal
observation), where breeding density is low (Yeaton 1974). Thus, neither of these
species are true chaparral specialists, despite their primary affinities for chaparral.
For the remaining chaparral-associated species, urbanization-associated declines in
abundance may be attributable to factors like habitat degradation, predation and
stochastic demographic effects (Soulé et al. 1988, Bolger et al. 1991). The biggest
anomaly was the California Thrasher, which despite its high predation rate and strong
shrub dependence (Cody 1998), was only weakly negatively associated with urban edge
proximity and trail density.
In a San Diego County study similar to mine, but with more habitat fragmentation and a
greater diversity of habitat types, Bolger et al. (1997) found the Rufous-crowned Sparrow
to be edge/fragmentation-sensitive, but not the Wrentit, Spotted Towhee, Bewick’s Wren,
- 18 –
California Thrasher or California Towhee—all of which were found to be negatively
affected by urbanization in my study. Bolger et al. (1997), found that most of these
“insensitive” species are significantly associated with various landscape characteristics
(e.g., level of fragmentation), but not when habitat characteristics (e.g., shrub coverage)
were entered in the models first.
While associations with surrounding urbanization levels detected in my study may also
be due to local habitat differences, these habitat variations are likely to be confounded
with landscape configuration. Disturbed areas such as roadcuts and clearings, which
Sauvajot (1997) found to Support fewer resident bird species, are more likely to occur in
highly fragmented, urbanized landscapes. Increases in surrounding urbanization and
resulting human disturbances may also reduce shrub density, a measure of habitat quality
and determinant of population density for several chaparral species (Miller 1951, Small
1974).
Three species in particular—the Wrentit (Erickson 1938), California Thrasher (Cody
1998) and Spotted Towhee (Jones and Diamond 1976)—are known to depend on dense,
shrubby vegetation, which provides cover from predators. For the Wrentit, Erickson
(1938) found an inverse relationship between territory size and brush density, with
interior territories smaller than territories bordering on roads, clearings, or non-scrub
vegetation. Although Erickson attributes the larger territories near edges to the
corresponding reduction in territory boundary to be defended, it may also be an indication
of poorer habitat quality near roads and clearings I found the Wrentit and towhee to be
negatively associated with surrounding urbanization proportion, as expected, although the
Spotted Towhee was also positively associated with urban edge length, or degree of
fragmentation (Table 1.2). Surprisingly, however, the thrasher was not significantly
associated with surrounding urbanization proportion, although it did exhibit weak
negative associations with urban edge proximity and trail density (Table 1.2). Of these
three species, the Wrentit is the only one not found in residential areas (Blair 1996),
which may explain its comparatively strong urbanization response.
- 19 -
Other species are less shrub-dependent, utilizing a wider range of vegetation densities.
The Rufous-crowned Sparrow, which has been officially designated a conservation
priority (Federal “Special Concern Species” and State of California “Species of Special
Concern” [California Department of Fish and Game 1998]), seems to prefer areas with
sparser vegetation coverage—usually “dry, rocky slopes with scattered scrub and patches
of grass and forbs” (Ehrlich et al. 1988, p.584). The Bewick’s Wren may be found in
sparse Sagebrush and desert Scrub, as well as suburban gardens (Blair 1996, Kennedy and
White 1997), while the California Towhee, despite its primary affinity for chaparral, is
often considered an edge-exploiting species (Ehrlich et al. 1988, Sisk et al. 1997) and has
been demonstrated to increase with fragmentation (Bolger et al. 1997). The Ash-throated
Flycatcher and Phainopepla are primarily desert scrub species that also breed in the
chaparral (Miller 1951). Thus, factors other than shrub coverage may be more important
in regulating abundances of these species.
Ash-throated Flycatchers are long-distance migrants, wintering in Central America, while
the Phainopepla is a local migrant, breeding first in the inland desert and later that year
along the coast, in chaparral scrub. While their habitat requirements are less strict than
those of other “chaparral specialists,” these wide-ranging, relatively mobile species may
cue in on larger landscape patterns, preferring less urbanized areas. Indeed, the Ash-
throated Flycatcher has been characterized as an “urban avoider,” and is generally absent
in urban/suburban areas (Blair 1996).
All of the chaparral specialists examined, except the Ash-throated Flycatcher and
Bewick’s Wren, are open cup nesters and therefore more vulnerable to nest predation
(Martin and Li 1992), which may increase near urban areas if scrub-jays and other corvid
nest predators are more abundant (Wilcove 1985, Cody 1998). Ground-nesting species,
such as the Rufous-crowned Sparrow and Spotted Towhee are also highly susceptible to
mammalian predation (Martin and Li 1992), which may increase near urban
development, due to the presence of human-introduced domestic cats (Felis catus) and
dogs (Canis domesticus) (Soulé et al. 1988). In addition, these mammalian predators
- 20-
may be responsible for higher adult mortality rates among ground-foraging species,
including the Rufous-crowned Sparrow, California Thrasher and both towhees (Cody
1998). -
Oak Woodland Species
Two of the oak woodland species examined—the Oak Titmouse and Nuttall’s
Woodpecker—exhibited no significant urbanization response. This may be due to the
spatial habitat arrangement in this study area, and the tendency for oak woodlands to
OCCUIT INCaſ development, especially in the eastern portion of the study area (see Figure
1.3), or to the cavity-nesting habits of these species, and their consequent lower rates of
nest predation (Martin and Li 1992). Furthermore, these species are considered edge-
exploiting (woodpecker) or edge-neutral (titmouse) (Sisk et al. 1997) and may be found
in suburban areas (Blair 1996).
The Blue-Gray Gnatcatcher, the only urbanization-sensitive oak woodland species
identified, may actually be better characterized as chaparral-associated in the Santa
Monica Mountains (personal observation), and may also be detrimentally affected by
changes in shrub cover. Although this species does not appear especially sensitive to
human activity near the nest, it is highly susceptible to nest predation, which accounts for
the large majority of nest failures (Root 1969, Ellison 1992). The gnatcatcher may also
have specific microhabitat requirements, such as riparian corridors and canopy openings
(Ellison 1992), which may be limited in urbanized landscapes. It is generally not present
in urban or suburban areas (Blair 1996).
Riparian Woodland Species
Of the two riparian woodland species, only the Black-headed Grosbeak was negatively
affected by urbanization. Grosbeak reproductive opportunities are thought to be limited
by availability of suitable habitat, which is primarily a factor of jay abundance and
predation threat (Hill 1995). The negative association with urbanization in this species
may also be related to riparian habitat requirements, as riparian corridors in more
– 21 -
urbanized areas have often been channelized, highly modified and/or replaced entirely by
housing developments. Finally, this grosbeak is migratory, wintering in Central America,
and may respond to large-scale landscape structure more readily than some of the resident
species, as has previously been noted (Hansen and Urban 1992).
The Northern Flicker, which exhibits no significant relationship with urbanization
measures, is a widespread generalist in many respects, with three subspecies occurring
across North America. It prefers open woodlands, savannas and forest edges, but is well-
adapted to human habitats and commonly breeds in urban and suburban environments
(Moore 1995).
Suburban-Associated Species
The species that exhibited the strongest positive association with urbanization was the
Northern Mockingbird, a common suburban bird whose preferred breeding habitat is
mowed lawns interspersed with trees and ornamental bushes (Arnold 1980, David et al.
1990, Derrickson and Breitwisch 1992). This result was expected, given the virtual
absence of mockingbirds in chaparral and woodland interiors (Derrickson and Breitwisch
1992, Blair 1996, Cody 1998). The Mourning Dove, however, showed no detectable
association with surrounding urbanization, which is consistent with other studies that
found this species to be insensitive to land cover type (Germaine et al. 1998) and
landscape characteristics (Bolger et al. 1997).
The Western Scrub-Jay, which is more abundam in urban and suburban settings than in
native oak woodlands (Blair 1996), exhibited a weakly significant (p > 0.01) association
with large-scale surrounding urbanization (Table 1.2). Nests of several species examined
in this study, including the wenti (Geupel and DeSante 1990), California Thrasher
(Cody 1998), Black-headed Grosbeak (Hill 1995) and Blue-gray Gnatcatcher (Ellison
1992), are known to suffer significant losses from scrub-jay predation. Thus enhanced
scrub-jay abundance near urban edges could detrimentally affect songbird reproductive
success, as has been demonstrated with the related Blue Jay and Golden-cheeked
- 22 -
Warblers in Juniper woodlands near urban developments (Engels and Sexton 1994).
Although Langen et al. (1991) actually found higher overall rates of predation on
artificial nests in the chaparral interior than at highly disturbed sites, they suggest that
avian predators, which are visually cued, may have been underestimated, particularly at
disturbed sites.
Summary
These results suggest that, although chaparral specialists seem most susceptible to
landscape changes, individual species respond differently to urbanization for multifaceted
reasons. The abundance of certain common, yet “shy,” cover-dependent chaparral
specialists like the Wrentit and Spotted Towhee may directly track habitat quality,
perhaps measured by shrub density. Less common species with specialized habitat
requirements, such as the Rufous-crowned Sparrow, may suffer from a lack of suitable
nest sites.
Other species with broader habitat associations, such as the Blue-gray Gnatcatcher and
Black-headed Grosbeak, may be sensitive to urbanization and landscape changes due to
possible related increases in nest predation, or a reduction in preferred breeding habitat,
such as riparian corridors. Highly mobile migratory species, such as the Black-headed
Grosbeak and Ash-throated Flycatcher, may respond to large-scale landscape patterns,
selecting areas with less overall urbanization in which to breed.
1.4.2. Scale of Urbanization Influence
My secondary interest was in the spatial extent of urbanization influence for each of the
species exhibiting a significant urbanization response. This varied according to species,
and cannot be precisely determined, due to the coarseness of the intervals examined. For
all species examined, however, responses to urbanization were consistent across scales—
i.e., there was no sign reversal (positive/negative) as scale increased. This may be due in
part to the relatively narrow range of scales examined, such that this study is best suited
– 23 –
for identifying ranges of urbanization influence, or ecological neighborhoods, rather than
specific scale thresholds.
Landscape-scale responses (i.e., greater than 250m) were recorded for all urbanization-
sensitive species except the Rufous-crowned Sparrow and California Thrasher,
Suggesting that responses to surrounding urbanization are occurring at the level of the
population, rather than the individual, and on an intermediate, rather than short-term time
Scale (see Wiens 1981). Temporal lags in population responses to environmental change
may also result in spatial lags in the relationship between bird abundance and
characteristics of the surrounding landscape. Alternatively, these large-scale responses
may indicate the presence of a habitat gradient coinciding with the urbanization gradient,
or other large-scale variations in bird abundance. This seems particularly likely for
species such as the Blue-gray Gnatcatcher, which was virtually absent from counts
conducted in the eastern portion of the study area, and is negatively associated with
Surrounding urbanization proportion at distances up to 4000m.
Conversely, the Wrentit, which was highly urbanization-sensitive, in terms of both model
fit and magnitude of urbanization effects, was affected only at distances up to 500m.
This species is “extremely sedentary” (Ehrlich et al. 1988, p. 420) and may not be
affected by habitat configuration at such large scales. Furthermore, its small territory size
(Erickson 1938) and adherence to chaparral, the primary vegetation type within the study
area, may ensure continued dispersal and habitat occupation, even in narrow habitat
corridors.
Highly mobile migratory species, such as the Black-headed Grosbeak, Ash-throated
Flycatcher and Phainopepla, may have a greater range of potential urbanization influence,
as well as a greater awareness of large-scale landscape change. But these species were
urbanization-sensitive only up to 1000m, suggesting that the larger-scale responses
observed in the Northern Mockingbird and Blue-gray Gnatcatcher may have been
artifacts of large-scale distribution patterns.
– 24 -
Finally, for some species, such as the Black-headed Grosbeak, local chaparral habitat may
have been less important than the availability of suitable riparian or other woodland
habitat in the surrounding area. The parallel north-south running canyons in this study
alſea are separated by distances of approximately 1000m, such that, where canyon
bottoms are not urbanized, most sampling sites are within 1000m of seasonal streams and
associated riparian vegetation (Figure 1.3).
1.4.3. Landscape Pattern vs. Composition
Until recently, most studies of habitat fragmentation in urbanized landscapes have
focused on distinct habitat islands (e.g., Soulé et al. 1988, Tilghman 1987, Dickman
1987, Friesen et al. 1995), rather than largely contiguous habitat interspersed with urban
development (but see Sauvajot and Beuchner 1991, Bolger et al. 1997). The results of
this study suggest that the effects of habitat fragmentation may be manifest before the
traditional fragmentation scenario—habitat islands in a sea of inhospitable
environment—is reached. In other words, urban encroachment may cause population
declines in areas that aren’t yet completely isolated by urbanization or other
“inhospitable” land uses.
With respect to landscape pattern, however, my results indicate that fragmentation per se,
as measured by total urban edge length within the maximum response radius, does not
contribute to additional declines in the abundance of urbanization-sensitive species. For
all species except the Spotted Towhee, the amount of urban edge present within the
maximum urbanization response radius did not significantly improve the abundance
model. The towhee’s positive association with fragmentation, controlling for the
negative effect of urbanization proportion, may indicate an attraction to suburban food
and water resources, as this species is often found in suburban gardens with sufficient
cover (Blair 1996, Greenlaw 1996).
These results should be interpreted with caution, however, given the narrow range of
urbanization conditions available (high correlation between fragmentation and
- 25 -
urbanization proportion). In this study area, most highly urbanized areas were also highly
fragmented; thus little of the variation in abundance not explained by urbanization
proportion could be explained by the fragmentation index. The variation in bird
abundance across presumably-equivalent suitable habitat within the same landscape
indicates that landscape configuration is influential, but cannot be distinguished from the
effects of landscape composition at the local level.
Furthermore, most remaining chaparral habitat in this study area is still interconnected,
and very few of the true habitat islands were sampled due to logistical constraints. The
primary exception, Griffith Park (approximately 2000ha), is much larger than the home
ranges of the species under study (e.g. -9ha for the gnatcatcher [Ellison 1992]). As long
as habitat patches are larger than the minimum area requirements of a species, landscape
configuration is less likely to affect species abundance (Andrén 1992). Birds, in
particular, may be less sensitive to spatial landscape configuration due to their relative
mobility (Wiens 1981).
1.4.4. Urban Edge Proximity and Human Use
With respect to the effects of proximity to urban development, this study suggests that
certain species are urban edge-averse or edge-affiliated, but that the relationship may
depend upon the level of surrounding urbanization. Only the California Thrasher, a
chaparral specialist, exhibited a weakly (p < 0.05) positive association with distance to
urban development (natural log-transformed), independent of urbanization proportion
(Table 1.2). This species was not significantly associated with surrounding urbanization
proportion, but was expectedly edge-Sensitive, given its specialized habitat requirements
and susceptibility to nest predation. Its eggs and nestlings are susceptible to scrub-jay
and raven predation, and juveniles are often taken by domestic cats (Felis catus) near
urban fringes (Murphy and Fleischer 1986). Furthermore, thrashers are thought to
compete with mockingbirds, which are similar in diet and foraging behavior, near urban
areas (Cody 1998).
- 26-
As also expected, the Northern Mockingbird, an abundant urban-associated species that is
virtually absent in chaparral and forest interiors (Derrickson and Breitwisch 1992, Cody
1998), exhibited a negative relationship with distance to urban development. The
Orange-crowned Warbler was also found to be negatively associated with distance to
urban development, perhaps due to its preference for mixed chaparral/woodland habitat
(Sogge et al. 1994), and the tendency for walnut woodlands, in particular, to occur near
development in this study area (see Figure 1.3).
Species that exhibited a significant positive response to the interaction between urban e
edge distance and urbanization proportion included the Wrentit, Bewick's Wren, Blue-
gray Gnatcatcher and Northern Mockingbird. For the Wrentit, Blue-gray Gnatcatcher and
Bewick’s Wren, which exhibited negative responses to surrounding urbanization, this
Suggests a dampened positive influence of urban edge distance, which becomes less
important as urbanization proportion decreases. This may be due to the habitat saturation
that must occur at higher abundances, and the “carrying capacity” of this environment. In
more urbanized parts of the landscape, there may also be more opportunities for predation
and other urban-associated influences from several directions, rather than from just one
edge in the large chaparral tracts. Furthermore, since individuals counted were not
necessarily on breeding territories, or may have had territories extending farther than
100m from the point count locations, maximum density may be lower near the
chaparral/urban interface, and even less so where urbanization is encroaching from many
sides.
For the Northern Mockingbird, which was positively associated with surrounding
urbanization proportion, the positive interaction effect is reinforcing, in that
mockingbirds are even less abundant (nonexistent) in chaparral interior when surrounding
urbanization proportion is low. Mockingbirds apparently penetrate farther into the
chaparral in highly urbanized and fragmented landscapes, whereas they are rarely found
more than 100m from urban development that borders on large, contiguous chaparral
areas (personal observation).
- 27-
The limited influence of urban edge proximity in this study is generally consistent with
Sauvajot (1997), who, in a less urbanized portion of this same study area, found no
significant relationships between proximity to urban development and resident bird
Species richness or abundance. Edge-insensitivity in largely unfragmented areas may be
due to the hardiness of chaparral vegetation, and its resistance to invasion by non-native
plant species (Knops et al. 1995), as well its relative impenetrability to humans (Sauvajot
1997). Indeed, much of the chaparral-development interface appears to be characterized
by a distinct “edge,” rather than transitional vegetation, although chaparral vegetation in
more fragmented areas may become degraded and invaded by exotic plant species over
time (Alberts et al. 1993). Furthermore, recent studies (e.g. Hansen and Urban 1992,
Tewksbury et al. 1998) suggest that western bird species may be less sensitive than
eastern species to landscape fragmentation and edge effects, based on differences in life-
history traits and predation pressures. These studies examined forested, rather than urban
landscapes. however, and may therefore be influenced by quite different edge-associated
pressures.
Another possible explanation is that reproductive success is indeed lower near urban
development, but source-sink dynamics keep abundances high. Previous studies have
suggested that habitat fragmentation reduces reproductive success of migratory songbirds,
resulting in Source-sink dynamics that may not be detected in simple surveys of species
abundance (Donovan et al. 1995, Brawn and Robinson 1996).
Hiking Trail Density
In terms of surrounding trail density, only three species exhibit significant responses (all
negative) after accounting for other urbanization factors: the California Thrasher,
Rufous-crowned Sparrow and Wrentit. Other studies have also recorded a reduced
probability of nesting near trails in some species, as well as increased rates of nest
predation and occurrence of "edge" species (Zande et al. 1984, Hickman 1990, Miller et
al. 1998).
– 28 -
Of all species examined, the Wrentit and thrasher are probably the most sedentary and
cover-dependent, the Wrentit so much so that it often refuses to cross firebreaks and trails
(Small 1974). It follows, therefore, that these species respond negatively to trail density,
even after factoring in urban edge proximity and surrounding urbanization proportion,
which are correlated with this variable. The Rufous-crowned Sparrow, however, nests
primarily on dry, rocky open slopes (Ehrlich et al. 1988), which actually tend to be
located near hiking trails, firebreaks and roadcuts in this study area. Thus the relationship
found in this species is initially surprising, although given its relatively low density and
patchy distribution (Ehrlich et al. 1988), this species may be particularly sensitive to
human disturbance.
1.4.5. Caveats
The increase in urbanization from west to east within the study area (Figure 1.1)
complicates the interpretation of my results. Although I kept vegetation cover constant to
the extent possible, the effects of urbanization and fragmentation are difficult to
disentangle from the effects of geographic location, given the potential variations in
climate generated by proximity to the Pacific Ocean. The vegetation in the eastern end of
the study area tended to be patchier and drier in general. The extent to which this is a
function of increased urbanization, as opposed to natural microclimate conditions, is
uncertain. Furthermore, small-scale variations in vegetation, and interactions among sites
that are close in spatial proximity may result in spatial dependence among Sampling
locations.
For this study, I ignored potential clinal and local variations in vegetation, assuming that
vegetation is constant across the sampling locations. Given the lack of variation in
avifauna among scrub vegetation types in southern California (Small 1974), this
assumption may indeed be reasonable. If there is considerable spatial dependence among
sampling locations, however, and model residuals are autocorrelated, conclusions about
urbanization may be biased (see Chapter 2).
- 29-
1.4.6. Research Recommendations
This observational study examines relationships between a set of independent
urbanization-related variables and abundance of several common chaparral birds.
Without experimental manipulation, causal relationships can only be inferred, and the
mechanisms responsible for these patterns remain speculative. Given the difficulties of
conducting experimental studies at such a large scale, the development and testing of
predictive models can produce results valuable to land-use planners and researchers alike.
In order to better formulate specific policy recommendations, however, more research on
underlying mechanisms should be conducted.
Most important for future research is the study of breeding biology and factors
influencing mortality and reproductive success in urbanization-sensitive species. While
many species examined here are known to be susceptible to nest predation, studies
monitoring individual pairs and territories, their breeding behavior, and reproductive
Success, could illuminate population dynamics with respect to various landscape and
habitat features. In particular, data on predation pressures should be collected at various
sites throughout the landscape to confirm or refute the findings of Langen et al. (1991),
who did not find nest predation to increase near urban areas. Furthermore, long-term
studies would help determine if populations are stable and to what extent source-sink
dynamics are occurring.
Another approach to determining mechanisms for decline is to quantify more specifically
the factors associated with increased urbanization and fragmentation. For example,
measuring shrub density, vegetation structure and diversity, leaf litter characteristics, and
invertebrate abundance might indicate whether there is indeed a decrease in habitat
quality with increased urbanization. Quantifying urban-associated predators, such as
dogs and cats, as well as human activities, such as mountain biking and hiking, may also
be informative.
– 30 -
Finally, the examination of a broader range of habitats, including urban areas, may help
complete a spatial picture of species distributions. It would be useful to quantify how far,
if at all, birds of the chaparral penetrate into urbanized areas, what habitat characteristics
are necessary, and whether these results are constant across space, or differ according to
overall landscape matrix characteristics.
1.5. Conclusion
The eastern Santa Monica Mountains are unique in representing an intermediate stage of
habitat fragmentation. The associated declines in abundance of common chaparral and
woodland species may not be immediately apparent in the field, as the species are all still
relatively common. But Poisson regression models helped detect these not-so-obvious
patterns. The results of this study suggest that several chaparral bird species may decline
in abundance as surrounding urbanization increases, even when apparently suitable
habitat is present.
Some species, including the Wrentit, Spotted Towhee, Black-headed Grosbeak and Blue-
Gray Gnatcatcher, appear particularly vulnerable to urbanization, and may serve as
indicators for chaparral quality and/or landscape integrity. Furthermore, certain urban-
associated species, such as the Northern Mockingbird and Western Scrub-Jay, appear
increase in abundance as surrounding urbanization increases, and may contribute to
declines of other songbirds, through competition and nest predation, respectively.
Scale of urbanization influence can be quite large for some species, indicating the
importance of population responses to landscape-level urbanization processes. In the
case of the Blue-gray Gnatcatcher, additional declines in abundance were detected with
changes in urbanization up to 4000m from census locations.
Although urbanization extent was shown to be a significant predictor of abundance for
several species, urbanization pattern did not explain significant additional declines or
increases. This does not mean that the birds studied are not affected by habitat
fragmentation, only that the total proportion of available habitat is an adequate measure
- 31 –
offragmentation, and that, in this largely interconnected landscape, birds do not appear to
distinguish between different habitat configurations when urbanization extent is constant.
Urban edge aversion and/or affinity varied, depending on levels of urbanization in the
surrounding landscape. Within less urbanized parts of the chaparral, few species were
sensitive to urban development proximity, and weakly so. Distance from the
urban/chaparral edge does appeared to be important in highly urbanized landscapes,
however.
Urban expansion is largely inevitable in the Greater Los Angeles area, where rapid
population growth and high property values have placed tremendous pressure on
remaining undeveloped lands. While much of the remaining natural habitat in this area is
protected in Federal and State public parkland, development has continued at a rapid pace
on privately owned lands. The influence of surrounding landscape urbanization on
chaparral bird abundance suggests that conservation and land use planning efforts would
benefit by incorporating regional and landscape, as well as local perspectives, to help
ensure that sufficiently large, intact areas of chaparral are maintained. Such planning is
possible because chaparral habitat still remains largely unfragmented throughout large
sections of the Santa Monica Mountains, and elsewhere in southern California. The
relationships observed in this study highlight that planning and forethought will help
preserve chaparral habitat before crisis proportions are reached.
-32 -
1.6. Literature Cited
Addicott, J.F., J.M. Aho, M.F. Antolin, D.K. Padilla, J.S. Richardson and D.A. Soluk.
1987. Ecological neighborhoods: scaling environmental patterns. Oikos 49:340-
346.
Alberts, A.C., A.D. Richman, D. Tran, R. Sauvajot, C. McCalvin and D.T. Bolger. 1993.
Effects of habitat fiagmentation on native and exotic plants in southern California
coastal scrub. Interface Between Ecology and Land Development in California.
Pp. 103-110 in Keeley, J.E., ed. Interface Between Ecology and Land
Development in California, Southern California Academy of Sciences, Los
Angeles.
Andrén, H. 1992. Corvid density and nest predation in relation to forest fragmentation: al
landscape perspective. Ecology 73:794-804.
Andrén, H. 1994. Effects of habitat fragmentation on birds and mammals in landscapes
with different proportions of suitable habitat: a review. Oikos 71:355-366.
Arnold, J.R. 1980. Distribution of the mockingbird in California. Western Birds 11:97-
102. -
Baker, M., N. Nur and G.R. Geupel. 1995. Correcting biased estimates of dispersal and
survival due to limited study area: theory and an application using Wrentits.
Condor 97:663-674.
Beissinger, S.R. and D.R. Osborne. 1982. Effects of urbanization on avian community
organization. Condor 84:75-83.
Blair, R.B. 1996. Land use and avian species diversity along an urban gradient.
Ecological Applications 6:506-519. -
Blake, J.G. and J.R. Karr. 1984. Species composition of bird communities and the
conservation benefit of large versus small forests. Biological Conservation
30:173-187.
Bolger, D.T., A.C. Alberts, and M.E. Soulé. 1991. Occurrence patterns of bird species in
habitat fragments: sampling, extinction, and nested species subsets. American
Naturalist 137:155–166.
– 33 -
Bolger, D.T., T.A. Scott and J. T. Rotenberry. 1997. Breeding bird abundance in an
urbanizing landscape in coastal southern California. Conservation Biology
11:406-421. -
Brawn, J.D. and S.K. Robinson. 1996. Source-sink population dynamics may complicate
the interpretation of long-term census data. Ecology 77:3-12.
Brittingham, M.C. and S.A. Temple. 1983. Have cowbirds caused forest songbirds to
decline? BioScience 33:31-35.
- Brown, J.H. 1978. The theory of insular biogeography and the distribution of boreal birds
and mammals. Great Basin Naturalist Memoirs 2:209-227.
California Department of Fish and Game. 1998. Special Animals List, March 1998.
Natural Heritage Division, Sacramento. http://www.dfg.ca.gov/whdablenddb.htm.
Clergeau, P., J-P. L. Savard, G. Mennechez and G. Falardeau. 1998. Bird abundance and
diversity along an urban-rural gradient: a comparative study between two cities on
different continents. Condor 100:413-425. .
Cody, M.L. 1998. California Thrasher (Toxostoma redivivum). Birds of North America
323. A. Poole and F. Gill, Eds. Philadelphia, PA. -
County of Los Angeles. 1998. County of Los Angeles Statistical Data.
http://www.co.la.ca.us/statistics.htm, last updated 5/12/98.
Crowley, P.H. 1992. Resampling methods for computation-intensive data analysis in
ecology and evolution. Annual Review of Ecology and Systematics 23:405-447.
David, N., M. Gosselin and G. Seutin. 1990. Pattern of colonization by the Northern
Mockingbird in Quebec. Journal of Field Ornithology 61:1-8.
Davis, M. 1992. City of Quartz. Vintage Books, New York.
Davis, M. 1998. Ecology of Fear. Metropolitan Books, New York.
Dawson, D.G. 1981. Experimental design when counting birds. Studies in Avian
Biology 6:392–398.
Dawson, K.J. 1990. Human predation on isolated nature reserves. Pp. 83-88 in Wildlife
Conservation in Metropolitan Environments, L.W. Adams and L.E. Dove (eds.).
National Institute for Urban Wildlife, Columbia, MD.
- 34 -
Derrickson, K.C. and R. Breitwisch. 1992. Northern Mockingbird (Mimus polyglottos).
Birds of North America 7. A. Poole, P. Stettenheim and F. Gill, Eds. Philadelphia,
PA. g
Dickman, C.R. 1987. Habitat fragmentation and vertebrate species richness in an urban
environment. Journal of Applied Ecology 24:337-351.
Donovan, T.M., F.R. Thompson, J. Faaborg and J.R. Probst. 1995. Reproductive success
of migratory birds in habitat sources and sinks. Conservation Biology 9:1380–
1395.
Edelman, P. 1992. Maximizing habitat potential through connectivity. Abstracts of the
Annual Meeting of the Southern California Academy of Sciences. Los Angeles,
CA. -
Efron, B. and R. Tibshirani. 1986. Bootstrap methods for standard errors, confidence
intervals, and other measures of statistical accuracy. Statistical Science 1:1:54–77.
Ehrlich, P.R., D.F. Dobkin & D. Wheye. 1988. The Birder’s Handbook: A Field Guide to
the Natural History of Birds. Simon & Schuster. New York, NY.
Ellison, W.G. 1992. Blue-gray Gnatcatcher (Polioptila caerulea). Birds of North America
23. A. Poole, P. Stettenheim and F. Gill, Eds. Philadelphia, PA.
Emlen, J.T. 1974. An urban bird community in Tucson, Arizona: derivation, structure,
regulation. Condor 76:184-197.
Erickson, M.M. 1938. Territory, annual cycle and numbers in a population of wren-tits
(Chamaea fasciata). University of California Publications in Zoology 42:247-
334.
Engels, T.M. and C.W. Sexton. 1994. Negative correlation of Blue Jays and Golden-
cheeked Warblers near an urbanizing area. Conservation Biology 8:286-290.
F orman, R.T.T., A.E. Galli and C.F. Leck. 1976. Forest size and avian diversity in New
Jersey woodlots with some land use implications. Oecologia 26:1-8.
Friesen, L.E., P.F. Eagles & R.J. Mackay. 1995. Effects of residential development on
forest-dwelling Neotropical migrant songbirds. Conservation Biology 9:1408-
1414. *
Fulton, W. 1997. The Reluctant Metropolis. Solano Press, Point Arena, CA.
– 35-
Gardner, R.H., M.G. Turner, R.V. O'Neill and S. Lavorel. 1991. Simulation of the scale-
dependent effects of landscape boundaries on species persistence and dispersal.
Pp. 76-89 in Holland, M.M., P.G. Risser and R.J. Naiman, Eds. Ecotones: The
role of landscape boundaries in the management and restoration of changing
environments. Chapman and Hall, New York. -
Garrett, K. and J. Dunn. 1981. Birds of Southern California. Los Angeles Audubon
Society, Los Angeles, CA.
Germaine, S.S., S.S. Rosenstock, R.E. Schweinsburg, and W.S. Richardson. 1998.
Relationships among breeding birds, habitat, and residential development in
greater Tucson, Arizona. Ecological Applications 8:680-691.
Geupel, Geoffrey R. and D.F. DeSante. 1990. Incidence and determinants of double
brooding in Wrentits. Condor 92:67-75. -
Gilpin, M.E. and M.E. Soulé. 1986. Minimum viable populations: processes of species
extinctions. Pp. 19-34 in M.E. Soulé, ed. Conservation Biology: The Science of
Scarcity and Diversity. Sinauer Associates, Sunderland, MA.
Greenlaw, J.S. 1996. Spotted Towhee (Pipilo maculatus). Birds of North America 263.
A. Poole and F. Gill, Eds. Philadelphia, PA.
Guthrie, D.A. 1974. Suburban bird populations in southern California. American Midland
Naturalist 92:461–466.
Hall, P. 1966. World Cities. McGraw Hill, London.
Hansen, A.J. and D.L. Urban. 1992. Avian response to landscape pattern: the role of
species' life histories. Landscape Ecology 7:163–180.
Harrison, T. 1994. Trail map of the Santa Monica Mountains East. Tom Harrison
Cartography, San Rafael, CA.
Hickman, S. 1990. Evidence of edge species' attraction to nature trails within deciduous
forest. Natural Areas Journal 10:3-5.
Hill, G.E. 1995. Black-headed Grosbeak (Pheuticus melanocephalus). Birds of North
America 143. A. Poole and F. Gill, Eds. Philadelphia, PA.
Hohtola, E. 1978. Differential changes in bird community structure with urbanisation: a
study in Central Finland. Ornis Scandinavica 9:94-100.
- 36-
Holland, R.F. 1986. Preliminary descriptions of the terrestrial natural communities of
California. State of California, The Resources Agency, Department of Fish and .
Game, Natural Heritage Division, Sacramento, CA. Unpublished report.
Human, K.G. and D.M. Gordon. 1997. Effects of Argentine ants on invertebrate
biodiversity in northern California. Conservation Biology 11:1242-1248.
Johnsen, A.M. and L.W. VanDruff. 1986. Summer and winter distribution of introduced
bird species and native bird species richness within a complex urban environment.
Pp. 123-127 in L.W. Adams and D.L. Leedy, Eds. Integrating Man and Nature in
the Metropolitan Environment, Columbia, MD.
Jokimaki, J. and J. Suhonen. 1993. Effects of urbanization on the breeding bird species
richness in Finland: a biogeographical comparison. Ornis Fennica 70:71-77.
Kadmon, R. and H.R. Pulliam. 1993. Island biogeography: effect of geographical
isolation on species composition. Ecology 74:977-981.
Keitt, T.H., D.L. Urban, and B.T. Milne. 1997. Detecting critical scales in fragmented
landscapes. Conservation Ecology [online] 1:4.
http://www.consecol.org/vol1/issl/art4.
Kennedy, E.D. and D.W. White. 1997. Bewick’s Wren (Thryomanes bewickii). Birds of
North America 315. A. Poole and F. Gill, Eds. Philadelphia, PA.
Knops, J.M.H., J.R. Griffin and A.C. Royalty. 1995. Introduced and native plants of the
Hastings Reservation, Central Coastal California: a comparison. Biological
Conservation 71:115-123.
Langen, T.A., D.T. Bolger, and T.J. Case. 1991. Predation on artificial bird nests in
chaparral fragments. Oecologia 86:395–401.
Levin, S.A. 1992. The problem of pattern and scale in ecology. The Robert H. MacArthur
Award Lecture. Ecology 73:1943-1967.
Levins, R. 1969. Some demographic and genetic consequences of environmental
heterogeneity for biological control. Bulletin of the Entomological Society of
America 15:237-240.
Liang, K.Y. and S.L. Zeger. 1986. Longitudinal data analysis using generalized linear
models. Biometrika 73:13–22.
- 37-
Little, R.J.A. and N. Schenker. 1995. Missing data. Pp. 64–67 in G. Armiger, C.C. Klogg
and M.E. Sobel, Eds. Handbook of Statistical Modeling for the Social and
Behavioral Sciences. Plenum Press, NY.
Lynch, J.F. and D.F. Whigham. 1984. Effects of forest fragmentation on breeding bird
communities in Maryland, USA. Biological Conservation 28: 287-324. -
MacArthur, R.H. and E.O. Wilson. 1967. The theory of island biogeography. Princeton
University Press, Princeton, NJ.
Martin, T.E. and P. Li. 1992. Life history traits of open- vs. cavity-nesting birds. Ecology
73:579–592.
Marzluff, J.M., R.B. Boone and G.W. Cox. 1994. Historical changes in populations and
perceptions of native pest bird species in the West. Studies in Avian Biology
15:202-220.
McDonnell, M.J. and S.T.A. Pickett. 1990. Ecosystem structure and function along
urban-rural gradients: an unexploited opportunity for ecology. Ecology 71:1232–
1237.
Metzger, J-P. and H. Décamps. 1997. The structural connectivity threshold: an hypothesis
in conservation biology at the landscape scale. Acta Oecologia 18:1-12.
Miller, A.H. 1951. An Analysis of the Distribution of the Birds of California. University
of California Press, Berkeley, CA.
Miller, S.G., R.L. Knight and C.K. Miller. 1998. Influence of recreational trails on
breeding bird communities. Ecological Applications 8:162-169.
Mills, G.S., J.B. Dunning, and J.M. Bates. 1989. Effects of urbanization on breeding bird
community structure in southwestern desert habits. Condor 91:416–428.
Moore, W.S. 1995. Northern Flicker (Colaptes auratus). Birds of North America 166. A.
Poole and F. Gill, Eds. Philadelphia, PA.
Murphy, M.T. and R.C. Fleischer. 1986. Body size, nest predation, and reproductive
patterns in Brown Thrashers and other mimids. Condor 88:446-455.
National Park Service. 1999. Fascinating facts about Santa Monica Mountains National
Recreation Area. http://www.nps.gov/samo/facts.htm.
- 38 -
Noss, RF . 1991. Landscape connectivity: different functions at different scales. Pp. 27-
39 in W.E. Hudson, ed. Landscape Linkages and Biodiversity. Island Press,
Washington DC.
Opdam, P. 1991. Metapopulation theory and habitat fragmentation: a review of holarctic
breeding bird studies. Landscape Ecology 5:93-106.
Opdam, P., R. Foppen, R. Reijnen and A. Schotman. 1995. The landscape ecological
approach in bird conservation: integrating the metapopulation concept into spatial
planning. Ibis 137:S139-S146.
Pearson, S.M. 1993. The spatial extent and relative influence of landscape-level factors
on wintering bird populations. Landscape Ecology 8:3-18.
Ralph, C.J., G.R. Geupel, P. Pyle, T.E. Martin, and D.F. DeSante. 1993. Handbook of
field methods for monitoring landbirds. U.S. Department of Agriculture, Forest
Service, Pacific Southwest Research Station, Albany, CA. Gen. Tech. Rep. PSW-
GTR-144.
Riitters, K.H., R.V. O’Neill and K.B. Jones. 1997. Assessing habitat suitability at
multiple scales: a landscape-level approach. Biological Conservation 81:191-202.
Robbins, C.S. 1980. Effect of forest fragmentation on breeding bird populations in the
piedmont of the Mid-Atlantic region. Atlantic Naturalist 33:31-36.
Root, R.B. 1969. The behavior and reproductive success of the Blue-gray Gnatcatcher.
Condor 71:16-31.
SAS. 1997. SAS/STAT” Software: Changes and Enhancements through Release 6.12.
SAS Institute Inc., Cary, NC.
Saunders, D.A., R.J. Hobbs and C.R. Margules. 1991. Consequences of ecosystem
fragmentation: a review. Conservation Biology 5:18-32.
Sauvajot, R.M. 1997. Edge effects, urban encroachment, and anthropogenic disturbance
in southern California chaparral mammal and bird communities. Ph.D.
dissertation, University of California, Davis. 154 pp.
- 39 -
Sauvajot, R.M. and M. Buechner. 1991. Effects of urban encroachment on wildlife in the
Santa Monica Mountains. Pp. 171-180 in Keeley, J.E., ed. Interface Between
Ecology and Land Development in California, Southern California Academy of
Sciences, Los Angeles.
Schoenherr, A.A. 1992. A Natural History of California. University of California Press,
Berkeley, CA.
Scott, T.A. 1993. Initial effects of housing construction on woodland birds along the
wildland urban interface. Pp. 181-187 in Keeley, J.E., ed. Interface Between
Ecology and Land Development in California, Southern California Academy of
Sciences, Los Angeles. -
Scott, T.A. 1995. Prefire management along California's wildland/urban interface:
introduction and session overview. Pp. 3-10 in Keeley, J.E. and T.A. Scott, eds.
Brushfires in California Wildlands: Ecology and Resource Management.
International Association of Wildland Fire, Fairfield, WA. .
Simberloff, D. 1994. Habitat fragmentation and population extinction of birds. Ibis 137:
S105-S1.11.
Sisk, T.D., N.M. Haddad and P.R. Ehrlich. 1997. Bird assemblages in patchy woodlands:
modeling the effects of edge and matrix habitats. Ecological Applications 7:1170-
1180.
Small, A. 1974. The Birds of California. Winchester Press, New York, NY.
Smallwood, K.S. 1994. Trends in California mountain lion populations. Southwestern
Naturalist 39:67–72.
Sogge, M.K., W.M. Gilbert and C. Van Riper III. 1994. Orange-crowned Warbler
(Vermivora celata). Birds of North America 101. A. Poole and F. Gill, Eds.
Philadelphia, PA. ... *
Soulé, M.E., D.T. Bolger, A.C. Alberts, R. Sauvajot, J. Wright, M. Sorice, and S. Hill.
1988. Reconstructed dynamics of rapid extinctions of chaparral-requiring birds in
urban habitat islands. Conservation Biology 2:75-92.
Suarez, A.V., D.T. Bolger and T.J. Case. 1998. Effects of fragmentation and invasion on
native ant communities in coastal southern California. Ecology 79:2041-2056.
- 40 –
Tarone, R.E. 1988. Score Statistics. Pp. 304-308 in Encyclopedia of Statistical Sciences,
volume 8. John Wiley & Sons, New York.
Tewksbury, J.T., S.J. Hejl and T.E. Martin. 1998. Breeding productivity does not decline
with increasing fragmentation in a western landscape. Ecology 2890–2903.
Tilghman, N.G. 1987. Characteristics of urban woodlands affecting breeding bird
diversity and abundance. Landscape and Urban Planning 14:481–495.
Turner, M.G. 1989. Landscape ecology: the effect of pattern on process. Annual Review
of Ecology and Systematics 20:171-197.
Vale, T.R. and G.R. Vale. 1976. Suburban bird populations in west-central California.
Journal of Biogeography 3:157-165.
Villard, M. and B.A. Maurer. 1996. Geostatistics as a tool for examining hypothesized
declines in migratory songbirds. Ecology 77:59-68. º
Whitcomb, R.F., C.S. Robbins, J.F. Lynch, B.L. Whitcomb, K. Klimkiewicz and D.
Bystrak. 1981. Effects of forest fragmentation on avifauna of the eastern
deciduous forest. Pp. 125-205 in R.L. Burgess and DM. Sharpe, eds. Forest
island dynamics in man-dominated landscapes. Springer-Verlag, New York, NY.
White, G.C. and R.E. Bennetts. 1996. Analysis of frequency count data using the
negative binomial distribution. Ecology 77:2549-2557.
Wiens, J.A. 1981. Scale problems in avian censusing. Studies in Avian Biology No.
6:513-521.
Wiens, J.A. 1994. Habitat fragmentation: island v landscape perspectives on bird
conservation. Ibis 137:S97-S104.
Wiens, J.A. and J.T. Rotenberry. 1981. Censusing and the evaluation of avian habitat
occupancy. Studies in Avian Biology No. 6:522–532.
Wilcove, D.S. 1985. Nest predation in forest tracts and the decline of migratory
songbirds. Ecology 66:1211-1214.
Wilcove, D.S., C.H. McLellan and A.P. Dobson. 1986. Habitat fragmentation in the
temperate zone. Pp. 237-256 in M.E. Soulé, ed. Conservation Biology: the
Science of Scarcity and Diversity. Sinauer Press, Sunderland, MA.
- 41 -
Yeaton, R.I. 1974. An ecological analysis of chaparral and pine forest bird communities
on Santa Cruz Island and mainland California. Ecology 55:959-973.
Zande, A.N. van der, J.C. Berkhuizen, H.C. van Latesteijn, W.J. terKeurs, and A.J.
Poppelaars. 1984. Impact of outdoor recreation on the density of a number of
breeding bird species in woods adjacent to urban residential areas. Biological
Conservation 30:1-39.
Table 1.1. Results from nested Poisson regression analyses. Sequential score tests were used to evaluate the contribution of
urbanization proportion and fragmentation at increasing radius distances (see Figure 1.4). Overall models include urbanization
proportion at the maximum significant radius and urban edge length at the same radius when applicable. Chi-square (p) values are
displayed in bold when significant (p < 0.05).
Species? Urbanization Proportion -- Radius Distance
§
Urban Edge -
(per site abundance) 0-250m 250-500m 500-1000m 1000–2000m 2000-4000m Length” Overall Sign
Chaparral Specialists
ATFL 4.23 1.56 4.28 0.16 0.90 13.52 tºº
(0.040) (0.21) (0.039) (0.69) (0.34) (0.0002)
BEWR 5.26 0.76 6.75 0.66 2.11 8.47 tº-e
(0.022) (0.38) (0.0094) (0.42) (0.15) (0.014)
CAQU 1.39 0.92 0.72 0.85 0.41:
(0.24) (0.34) (0.40) (0.36) (0.52).
CATH 1.22 0.08 0.51 0.00 0.00
. . (0.27) (0.77) (0.48) (0.95) (0.97)
CATO 2.39 0.06 6.56 0.03 8.48 tº
(0.12) (0.81) (0.010) (0.86) (0.0036)
OCWA 0.46 1.17 0.35 0.01 º
(0.50) (0.28) (0.55) (0.91) 3. *...*. .
PHAI 3.26 0.68 7.59 2.33 1.62 12.12 wº
(0.071) (0.41) (0.0059) (0.13) (0.20) (0.0005)
RCSP 7.20 0.42 0.29 0.60 0.30 7.20 tº-
(0.0073) (0.52) (0.60) (0.44) (0.58) (0.0073)
f See Appendix 1.1 for full species names and scientific names.
* Evaluated at maximum significant radius of urbanization proportion.




i
Table 1.1. (continued)
Speciest
(per site abundance)
Urbanization Proportion -- Radius Distance
0–250m 250-500m 500-1000m 1000–2000m 2000-4000m
Urban Edge
Length” Overall Sign
Chaparral Specialists
SPTO
WREN
Riparian Species
BHGR
NOFL
Oak Woodland Species
BGGN
NUWO
OATI
3.44
(0.064)
9.44
(0.0021)
9.11
(0.0025)
0.03
(0.87)
9.11
(0.0025)
0.04
(0.84)
0.16
(0.69)
0.10
(0.75)
4.02
(0.045)
0.02
(0.88)
0.43
(0.51)
0.05
(0.82)
3.56
(0.06)
0.10
(0.75)
2.71
(0.10)
0.38
(0.54)
4.27
(0.039)
0.47
(0.49)
8.98
(0.0027)
0.98
(0.32)
0.00
(0.94)
f See Appendix 1.1 for full species names and scientific names.
* Evaluated at maximum significant radius of urbanization proportion.
8.15
(0.0043)
3.63
(0.057)
2.08
(0.15)
0.76
(0.38)
2.57
(0.11)
0.06
(0.81)
0.80
(0.37)
2.86 9,55 13.93 -,+
(0.091) (0.0020) (0.0023)
0.83 2.04 11.12 -
(0.36) (0.15) (0.0009)
0.05 0.85 12.18 –
(0.83) (0.36) (0.0005)
5.73 3.13 19.67 -
(0.0167) (0.077) (<0.0001)
1.22
(0.27)
2 ... ."
& . . .


Table 1.1. (continued)
Speciest Urbanization Proportion -- Radius Distance Urban Edge .
(per site abundance) 0-250m 250-500m 500-1000m 1000-2000m 2000-4000m Length” Overall Sign
Suburban Species: -
MODO 0.00 0.63 2.23 0.11 3.12
(0.98) (0.43) (0.14) (0.74) (0.08) || ||
NOMO 7.60 5.96 4.66 3.86 0.98 0.33 12.07 F
(0.0058) (0.015) (0.031) (0.050) (0.32) (0.57) (0.0005)
WESJ 3.16 2.59 0.02 1.80 4.33 1.00 5.93 +
(0.076) (0.11) (0.88) (0.18) (0.037) (0.32) (0.015)
f See Appendix 1.1 for full species names and scientific names.
* Evaluated at maximum significant radius of urbanization proportion.



#
Table 1.2. Final GEE Poisson regression models: significant (p < 0.05) parameter estimates and score tests for significance of
combined urbanization variables and overall models.
Urbanization Urbanization Urbanization Total Model
Proportion tation Index Proportion x Score Test Score Test
Urban Edge Trail Density (buffer Edge Chi-Square Chi-Square
Speciest Year Distance (ln) (250m buffer) distance) Distance” (p-value) (p-value)
Chaparral Specialists
ATFL 0.42 -1.80 14.57 18.77
(Std. Error) (0.13) (0.43) (0.0001) (<0.0001)
1997 - 1996 (1000m)
BEWR 0.23 -6.93 1.30 7.59 17.91
(Std. Error) (0.07) (1.78) (0.35) (0.023) (0.005)
1997 × 1996 (1000m)
CALT –0.75 8.48 8.48
(Std. Error) (0.25) (0.0036) (0.0036)
(1000m)
CAQU
(Std. Error)
CATH 0.24 0.11 –0.025 7.11 12.71
(Std. Error) (0.08) (0.05) (0.01) (0.029) (0.0053)
1997 × 1996
OCWA 1.06 -0.48 9.90 9.90
(Std. Error) (0.43) (0.19) (0.0071) (0.0071)
1997 × 1996
f See Appendix 1.1 for full species names and scientific names.
* Urbanization proportion evaluated at maximum significant radius.
Table 1.2. (continued)
Urbanization Fragmen- Urbanization Urbanization Total Model
Proportion tation Index Proportion x Score Test Score Test
Urban Edge Trail Density (buffer (buffer Edge Chi-Square Chi-Square
Speciest Year Distance (ln) (250m buffer) distance) distance) Distance” (p-value) (p-value)
Chaparral Specialists
PHAI -1.19 –0.93 11.50 20.47
(Std. Error) (0.23) (0.31) (0.0007) (<0.0001)
1996 × 1997 :
RCSP 0.96 –0.10 -12.29 11.85 16.15
(Std. Error) (0.35) (0.040) (4.725) (0.0027) (0.0011)
1997 - 1996 250m
SPTO 0.14 –2.91 1.86 10.77 12.65
(Std. Error) (0.06) (0.58) (0.42) (0.0046) (0.0054)
1997 - 1996 (2000m) (2000m)
WREN 0.07 –0.02 -5.76 0.85 13.93 16.57
(Std. Error) (0.03) (0.007) (1.96) (0.38) (0.003) (0.0023)
1997 - 1996 (500m) g
Riparian Species
BHGR 0.43 -2.79 13.19 18.53
(Std. Error) 2. (0.14) (0.77) (0.0003) (<0.0001)
1997 × 1996 (1000m)
NOFL
(Std. Error)
f See Appendix 1.1 for full species names and scientific names.
* Urbanization proportion evaluated at maximum significant radius.
§
Table 1.2. (continued)
Urbanization Fragmen- Urbanization Urbanization Total Model
Proportion tation Index Proportion x Score Test Score Test
Urban Edge Trail Density (buffer (buffer Edge Chi-Square Chi-Square
Speciest Year Distance (ln) (250m buffer) distance) distance) Distance” (p-value) (p-value)
Oak Woodland Species
BGGN -12.47 1.50 21.50 21.50
(Std. Error) (4.29) (0.73) (<0.0001) (<0.0001)
(4000m)
NUWO
(Std. Error)
OATI
(Std. Error)
Suburban Species
MODO
(Std. Error)
NOMO 0.47 -1,02 0.44 22.25 24.79
(Std. Error) (0.19) (0.16) (0.14) (<0.0001) (<0.0001)
1997 - 1996
WESJ 0.28 0.38 4.02 19.42
(Std. Error) (0.08) (0.19) (0.045) (<0.0001)
1997 - 1996 (4000m)
f See Appendix 1.1 for full species names and scientific names.
* Urbanization proportion evaluated at maximum significant radius.
#
Table 1.3. Mean parameter estimates and 95% confidence intervals from bootstrap resampling (k=50) of repeated measures
Poisson models. Parameter estimates in bold are robust (significantly different from zero). Scaled-down models are also
shown when original models are not robust.
Urbanization Fragmentation Urbanization
Urban Edge Trail Density Proportion Index (buffer Proportion x
Speciest - Year Distance (ln) (250m buffer) (buffer distance) distance) Edge Distance”
Chaparral Specialists
ATFL 0.22 -1.89
(95% CI) (-0.30, 0.73) (-3.26, -0.51)
(1000m)
BEWR –7.92 1.50
(95% CI) (-13.08, -2.75) (0.50, 2.51)
(1000m)
CALT -0.78
(95% CI) (-1.534, -0.026)
(1000m)
CATH - 1 0.11 0.11 -0.03
(95% CI) (-0.17, 0.38) (-0.021, 0.246) (-0.06, 0.005)
CATH - 2 0.11 0.11
(95% CI) (-0.17, 0.38) (-0.03, 0.25)
PHAI -0.62 -1.03
(95% CI) (-1.96, 0.71) (-2.03, -0.02)
f See Appendix 1.1 for full species names and scientific names.
* Urbanization proportion evaluated at maximum significant radius.
#
Table 1.3. (continued)
Urbanization Fragmentation Urbanization
Urban Edge Trail Density Proportion Index (buffer Proportion x
Species? Year Distance (ln) (250m buffer) (buffer distance) distance) Edge Distance”
Chaparral Specialists -
RCSP - 1 0.46 –0.10 -14.04
(95% CI) (-0.70, 1.2) (-0.22, 0.02) (-31.96, 3.89)
(250m)
RCSP - 2 0.50 -15.22
(95% CI) (-0.74, 1.75) (-33.25, 2.80)
(250m)
SPTO 0.07 –3.23 2.05
(95% CI) (-0.11, 0.25) (-4.96, -1.50) (0.79, 3.31)
(2000m) (2000m)
WREN - 1 0.03 -0.02 -6.13 • 0.91
(95% CI) (-0.06, 0.13) (-0.04, 0.0002) (-14.09, 1.82) (-0.62, 2.43)
(500m)
WREN - 2 -0.03 -0.02 -1.54
(95% CI) (-0.12, 0.06) (-0.04, -0.0005) (-2.18, -0.89)
- (500m)
Riparian Species
BHGR 0.23 –2.87
(95% CI) (-0.30, 0.75) (-5.06, -0.67)
(1000m)
f See Appendix 1.1 for full species names and scientific names.
* Urbanization proportion evaluated at maximum significant radius.
§
Table 1.3. (continued)
Urbanization Fragmentation Urbanization
Urban Edge Trail Density Proportion Index (buffer Proportion x
Speciest Year Distance (In) (250m buffer) (buffer distance) distance) Edge Distance”
Oak Woodland Species
BGGN - 1 -12.09 1.40
(95% CI) (-24.17, -0.003) (-0.72, 3.53)
(4000m)
BGGN - 2 -4.03
(95% CI) (-6.04, -2.01)
(4000m)
Suburban Species
NOMO 0.25 -1.15 0.40
(95% CI) (-0.39, 0.89) (-1.64, -0.67) (0.003, 0.79)
(2000m) g
WESJ 0.14 0.34
(95% CI) (-0.19, 0.47) (-0.15, 0.84)
(4000m)
f See Appendix 1.1 for full species names and scientific names.
* Urbanization proportion evaluated at maximum significant radius.
*śji====
iſ 㺠.
Mºś
%. A d #º. # º º
º
º º *Q #. ſº- º T ºº::= E. ------- -
Multi-Family Residential
U. Single Family Residential ;
º Rural Residential º
Commercial/Industrial º
Transportation/Communication ;
Open Space and Recreation
2000 0 2000 4000 Meters
Figure 1.1. Study Area Land Uses. Green areas contain chaparral and other native vegetation types.
Point count locations censused in 1996 and 1997 are shown in black; locations in white were added in 1997.























:
* \ Q2 Ž arretore value,
º *_ º - º
To
LA5 */ -
clancaster
- - º -
º º - - º - * AMGELES º
º, º a tº ºf: º º NAICNA is
tºoval roºtsi º fºº"
º * - º, - \ oCastnic º - -(?) --
2.º.º. º
zº, º
O sent starre a Los Angeles county
Ojai º … º Fillmore waitey Santa -
°Palmdale
CT
-
º o Gy- o, o Clariº
º - Valencia - º * …
º Santa Paula - º - - -
º ventura county Sirm 21: #.do j. º 5AN 8: Kºga RDºug NATIONAL FOREST º
Hºſº sº "Pº" vºie, , o FGR::T º: * a s
º is? –- - o ſº-san Fernando - Sºn Gotºriei º º - º - ºr 'º
Ventura @ º tº Northridge value ºf - *ountains º - % º jº º
º ºr- - SAN BERNARDINo county
Glendale - - 15.) º *
Gº º Fasa demº Mºnrº, º Q3) San Ç ſº son Bernardº º
6. Ci rsäſº -Q) 9 º' - oGlendora - Bernardino o Mountain; 2. º
Beverly. Sºº-ºº: º Alhambra @ Azusa - *º. ofontana a º º * *-
- -- -
$4ts: 7. ---
- ... .ºcaayo Thousandº C C C. © . (9.Burbank
Camarillo Tº 2,8. º Sanoga
o, - O Agourai ** - - -
-
-
**
º
` His
c
-->
- -arº - - -- º
| Topanga Hillsóº. S&N Gº. - Coltono - º -
Malibu ºt.” * . Los A -Qſ). -- ºvatiéy 6 Ontario.o - o --> ... ---------
alibu º - - ºf Downtown L.A.: - Fomona ---------- Redlands C - ſ -º-º:
- Cº- º:$9 o - Hur - -(º, Covina -jºº *-ºſ -º- - Yucaipa -
ity”. - Huntington 50 º … . * - ----------- º
Santaº, º ºf NTRººp." - *- - -
- - º -- - * Ri - - º -
NInglewood ©ºus ºries / Whittier / º (º) Riverside - * Bºnn
- * o - -
- C
C -- --- -
º a
º) Compton? Gº º "Fuerton º o º
º - - -
c - 9 - ſº-
Beaumont
- º - sºººººººwº
º º: - * Corona º ºrial roºt
Redondo Beacº ©. . N-
- -
Torrance Y → Y / Anaheim” . . . . Riverside county.ºs
pºwds. " II & Yorºrise egº NTY. CŞan Jacinto
º *- - - º
º --
- - - (z2) - ity: A **i. Perris º - *
- º sº - 27 range City º º º * * *
- Ç) Santº oTustin * *… . 15 - ^º º
Ana Gº - º º --- -* º, - -- sº acinto
º --- Ç) º -- Zºº, o'- º º i tºº. Nº Ni ... ." -
º º **Lake NATIONAL, Lake *A*-i- ºr
hº **, Forestº Elsinore º,
- 2, º º - - s - --- º / º * * -
Newport bººs X. º . º -
Mission Z gºes - - -
coronadettºs
* ºº Murret.” º
º, º ſº
- - -- - º - -
----- Temecula o º - - º º
30 Kłł OMETERS A/ º
Los ANGELES METROPOLITAN AREA
…
--- ---
-------------------
Ç) º
SArt DiºGo º
count Y vº. º














































#
8000 Meters
Vegetation Category
D Coastal Sage Scrub
Sage Scrub / Chaparral
L. Northern Mixed Chaparral
| Chamise Chaparral
Non-Native Grassland
Rock Outcrops
| Coast Live Oak
| Walnut
Sycamore-Oak
[T] Non-Native Conifer/Hardwood
| Coastal Strand
[T] Urban Development
Figure 1.3. Vegetation classification within study area. Point count locations indicated in red.




- 54 -
1000m-2000m-4000m buffer
on index.
on and fragmentati
d to calculate
rings use
tion proporti
aniza
Figure 1.4. 250m-500m-
urb

- 55-
Appendix 1.1. Avian species recorded during study (swifts, swallows and raptors not
included), in alphabetical order, according to AOU code.
AOU Common Name Migratory Years Years Groups
Code (Scientific Name) Status Observed Counted Counted”
ALHU Allen’s Hummingbird Long- 1996, 1997 T
(Selasphorus sasin) distance 1997
migrant
AMCR American Crow Resident 1996, 1996, T
(Corvus 1997 1997
brachyrhynchos)
ANHU Anna’s Hummingbird Resident 1996, 1997 T
(Calypte anna) 1997
ATFL Ash-Throated º Long- 1996, 1996, T, S
Flycatcher (Myiarchus distance 1997 1997
cinerascens) migrant
BCHU Black-Chinned Long- 1996, 1997 T
Hummingbird distance 1997
(Archilochus alexandri) migrant
BEWR Bewick’s Wren Resident 1996, 1996, T, S
(Thryomanes bewickii) 1997 1997
BGGN Blue-Gray Gnatcatcher Resident 1996, 1996, T
(Polioptila caerulea) 1997 1997
BHGR Black-Headed Long- 1996, 1996, T, S
Grosbeak (Pheucticus distance 1997 1997
melanocephalus) migrant
BLPH Black Phoebe (Sayornis Resident 1996, 1996, T
nigricans) 1997 1997
BUSH Bushtit (Psaltriparus Resident 1996, N/A
minimus) 1997
CAQU California Quail Resident 1996, 1996, S
- (Callipepla californica) 1997 1997
CATH California Thrasher Resident 1996, 1996, T, S
(Toxostoma redivivum) 1997 1997
CALT California Towhee Resident 1996, 1996, T
(Pipilo crissalis) 1997 1997
* T = total individuals, S = singing males.
- 56 -
Appendix 1.1. (continued)
AOU Common Name Migratory Years Years Groups
Code (Scientific Name) Status Observed Counted Counted
CANW Canyon Wren Resident 1996, 1996, T, S
(Catherpes mexicanus) 1997 1997
CORA Common Raven Resident 1996, 1996, T
(Corvus corax) 1997 1997
COYE Common Yellowthroat Long- 1997 1997 T, S
(Geothlypis trichas) distance
migrant
HOFI House Finch Resident 1996, 1996, S
(Carpodacus 1997 1997
mexicanus) -
HOOR Hooded Oriole (Icterus Long- 1996, 1996, T, S
cucullatus) distance 1997 1997
Migrant -
HOWR House Wren Resident 1996, 1996, T, S
(Troglodytes aedon) 1997 1997
HUVI Hutton’s Vireo (Vireo Long- 1996, 1996, T, S
huttoni) distance 1997 1997
migrant
LEGO Lesser Goldfinch Resident 1996, N/A
(Carduelis psaltria) 1997
LOSH Loggerhead Shrike Long- 1996 1996 T
(Lanius ludovicianus) distance
migrant
MODO Mourning Dove Resident 1996, 1996, T
(Zenaida macroura) 1997 1997
NOFL Northern Flicker Resident 1996, 1996, T
(Colaptes auratus) 1997 1997
NOMO Northern Mockingbird Resident 1996, 1996, T, S
(Mimus polyglottos) 1997 1997
NUWO Nuttall’s Woodpecker Resident 1996, 1996, T
(Picoides nuttallii) 1997 1997
OATI Oak Titmouse Resident 1996, 1996, T
(Baeolophus inornatus) 1997 1997
* T = total individuals, S = singing males.
- 57 -
Appendix 1.1. (continued)
AOU Common Name Migratory Years Years Groups
Code (Scientific Name) Status Observed Counted Counted
OCWA Orange-Crowned Resident 1996, 1996, T, S
Warbler (Vermivora 1997 1997 -
celata)
PHAI Phainopepla Resident / 1996, 1996, T, S
(Phainopepla nitens) Local 1997 1997
migrant
PSFL Pacific Slope Long- 1996, 1996, T, S
Flycatcher (Empidonax distance 1997 1997
difficilis) migrant
RCSP Rufous-Crowned Resident 1996, 1996, T, S
Sparrow (Aimophila 1997 1997
ruficeps)
SOSP Song Sparrow Resident 1996, 1996, T, S
(Melospiza melodia) 1997 1997
SPTO Spotted Towhee (Pipilo Resident 1996, 1996, T, S
maculatus) 1997 1997
WEKI Western Kingbird Long- 1997 1997 T
(Tyrannus verticalis) distance
- migrant
WESJ Western Scrub-Jay Resident 1996, 1996, T
(Aphelocoma 1997 1997
californica)
WETA Western Tanager Long- 1996 1996 T
(Piranga ludoviciana) distance
migrant -
WREN Wrentit (Chamaea Resident 1996, 1996, T, S
fasciata) 1997 1997
YEWA Yellow Warbler Long- 1997 1997 T
(Dendroica petechia) distance
migrant
* T = total individuals, S = singing males.
3.
Appendix 1.2. Commonly-Occurring Species, in Order of Total Abundance. Summary Statistics.
1996 (n = 78) 1997 (n = 107)
2. 3 Pl, 3 >
$2 O
# # ă ă # # . . . #
3. ă ă ă ă ă ă ă ă ă ă ă ă ă ă.
Tº # 3 ; ; ; ; ; ; ; ; ; ; ; ;
Species ºf ;3 tº j ż, , ; ; ; ;3 ; ; ; ; ;
WREN (S) 1821 Count 1 327 78 100% 4.19 0.19 4 8 490 107 100% 4.58 0.17 5 9
Count 2 409 77 99% 5.24 0.22 5 10 595 106 99% 5.56 0.20 6 10
Maximum 78 100% 5.68 0.19 6 10 107 100% 5.99 O.19 6 10
SPTO (S) 898 Count 1 254 74 96%. 3.26 0.19 3 9 385 106 99% 3.60 0.16 3 9
Count 2 71 61 79% 0.91 0.14 0 6 188 95 89% 1.76 0.14 2 6
Maximum 75 97% 3.31 0.20 3 9 106 99% 3.70 0.16 4 9
CALT (T) 740 Count 1 129 0.9 95% 1.65 0.15 1 5 176 92 86%. 1.64 (). 12 1 5
- Count 2 181 66 85% 2.32 0.19 2 6 254 95 89% 2.37 0.17 2 9
Maximum 74 95% 2.69 0.18 2 6 103 96% 2.79 0.15 3 9
CATH (T) 600 Count 1 52 35 45% 0.67 0.10 0 3 150 83 78%, 1.40 0.11 1 5
Count 2 150 63 81%, 1.92 O.17 2 6 248 96 90% 2.32 0.14 2 7
Maximum 71 91%. 2.13 0.15 2 6 104 97% 2.64 0.13 3 7
WESJ (T) 584 Count 1 87 53 68%. 1.12 0.12 1 3 178 88 82%. 1.66 0.11 2 5
Count 2 107 58 74%. 1.37 0.13 1 5 212 92 86%. 1.98 0.13 2 6
Maximum - 71 91%. 1.81 0.12 2 5 104 97% 2.45 0.12 2 6
* T = total individuals, S = singing males.
f See Appendix 1.1 for full species names and scientific names.
3.
Appendix 1.2. (continued)
1996 (n = 78) 1997 (n = 107)
2 GD (D -
5 § 3. $2 GD É 3. (D GD
O : 'ºz, ...t. red 3 : ...tº
le 5 Q0 G. 5 5 * 5 Uſ) G. * S *
3. # # 3 # 3 & # # 3 # # 3 &
… 3 # 3 ; ; ; ; ; ; ; ; ; ; ; ;
Species* H Bº is tº 3 & X > tº # 7, X & X >
BEWR (S) 571 Count 1 103 65 84%. 1.32 0.12 1 4 219 105 98% 2.05 0.12 2 5
Count 2 108 70 91%. 1.38 0.14 1 5 141 101 94%. 1.32 0.11 1 4
Maximum 77 100% 1.96 0.12 2 5 107 100% 2.49 0.10 2 5
ATFL (S) 167 Count 1 30 25 32% 0.38 0.07 0 2 38 53 50% 0.49 0.08 0 3
Count 2 38 36 47% 0.49 0.08 0 3 61 63 59% 0.57 0.07 0 3
Maximum 47 61% 0.72 0.09 1 3 82 77%. 1.07 0.08 1 3
BHGR (S) 160 Count 1 45 42 55% ().58 0.07 0.5 2 86 65 61% 0.80 0.08 1 3
; : Count 2 4 8 10% 0.05 0.03 0 2 25 28 26% 0.23 0.04 0 2
Maximum 45 58% 0.60 0.07 1 2 73 68% 0.90 0.08 1 3
MODO (T) 143 Count 1 33 25 32% 0.42 0.08 0 4 41 33 31% 0.38 0.07 0 4
Count 2 22 15 19% 0.28 0.09 0 5 47 32 30% 0.44 0.07 0 3
Maximum - 35 45% 0.64 (). 11 0 5 53 50% 0.70 0.08 0 4
NOMO (S) 127 Count 1 21 16 2.1% 0.27 0.07 0 4 51 32 30% 0.48 0.09 () 5
Count 2 16 13 17% 0.21 0.06 0 3 39 27 25% 0.36 0.07 0 4
Maximum 19 25% 0.35 0.09 0 4 39 36% 0.58 0.09 0 5
* T = total individuals, S = singing males.
f See Appendix 1.1 for full species names and scientific names.
§
Appendix 1.2. (continued)
1996 (n = 78) 1997 (n = 107)
2. 8 P. 3 Pl,
GD
# § § { # # # #
: B Ú/) B. 5 S $– 5 C/D G. 5 3 $–
Neº’ # # 3 iſ ; ; ; ; # 3 # # 5 #
3 # 3 ; ; ; ; ; ; ; ; ; ; ; ;
Species*f H tº its ºf > 07 - > ë is 7, 3 & X >
PHAI (T) 117 Count 1 57 30 38% 0.73 0.13 0 5 17 12 11% 0.16 0.05 0 3
Count 2 29 16 2.1% 0.37 0.10 0 5 14 12 11% (). 13 0.04 0 2
Maximum 42 39% 0.72 0.11 0 5 20 19% 0.24 0.05 0 3
NOFL (T) 95 Count 1 19 12 15% ().24 0.08 0 4 22 21 20% 0.21 0.04 0 2
- Count 2 25 22 28% 0.32 0.06 0 2 29 28 26% 0.27 0.05 0 2
Maximum 28 36% 0.49 0.09 () 4 42 39% 0.41 0.05 0 2
RCSP (T) 83 Count 1 7 6 8% ().09 0.04 0 2 38 28 26% 0.36 0.08 0 6
Count 2 10 6 8% (). 13 0.06 0 3 28 21 20% 0.26 0.06 0 2
- Maximum 12 15% 0.22 0.09 0 6 39 36% 0.51 0.08 0 6
NUWO (T) 78 Count 1 9 9 12% 0.12 0.04 0 1 21 20 19% 0.20 0.04 0 2
Count 2 21 15 19% ().27 0.07 0 3 27 26 24% 0.25 0.04 0 - 2
Maximum 22 28% 0.36 0.07 0 3 39 36% 0.38 0.05 0 2
BGGN (T) 67 Count 1 19 17 22% 0.24 0.06 0 2 17 14 13% 0.16 0.04 0 2
Count 2 11 9 12% 0.14 0.05 0 2 20 18 17% 0.19 ().04 0 2
Maximum 25 32% 0.37 0.03 0. 1 29 27% 0.32 0.05 0 2
* T = total individuals, S = singing males.
f See Appendix 1.1 for full species names and scientific names.
Chapter 2: -
Examination and Modeling of Spatial Structure
in Selected Chaparral Bird Species
2.1. Introduction
Increasingly, many large-scale ecological studies involve an explicit spatial component.
With the growing availability and sophistication of Geographic Information Systems
(GIS), as well as portable, inexpensive Geographic Positioning Systems (GPS), ecologists
can now georeference their sampling locations relatively easily and accurately. 6.
Oftentimes, however, this extra information is not used to its full potential. In ecology,
GIS has been used primarily as a map presentation tool, facilitating the visualization of
Spatial variables and relationships among them. Commonly, spatial data are analyzed
using simple statistical methods that ignore the underlying spatial structure and often
improperly assume independence among sampling locations. This tendency is likely due
to the statistical complexity and computation-intensiveness of many spatial statistical
techniques, as well as their lack of widespread acceptance. But the omission of spatial
dependence in statistical analyses can be problematic for the interpretation of results, and
may result in biased model parameter estimates (Robertson 1987, Anselin 1989) and false
detection of significant relationships (Legendre 1993).
This is not to say that spatial patterns in nature have not been widely studied. Indeed,
there is a long history of quantitative spatial analysis in ecology, with the description and
explanation of regional and continent-wide vegetation distributions dating at least back to
Von Humboldt (1807). Large-scale biogeographic vegetation trends have been explained
largely by variation in soils, topography, and especially climate (e.g., DeCandolle 1874,
Warming 1895, Merriam 1898, Gleason 1917, Whittaker 1956, Peet and Loucks 1977,
Delcourt and Delcourt 1988, Davis and Zabinski 1992, Holling 1992). Factors controlling
the distributions of animal species (Andrewartha and Birch 1954, Simpson 1964,
MacArthur 1972, Bock 1984, Root 1988), and the relationships between distributions and
– 61 -
-62-
abundance (Cook 1969, Brown 1984, Bock 1987, Maurer 1990, Brown et al. 1995), have
also been studied and modeled. Avian range distributions, in particular, have received
considerable attention (Terborgh 1971, Bock 1984, Root 1988, Maurer and Heywood
i 993, Lawton et al. 1994).
At the local scale, spatial dependence in the distribution of plants and other sessile
organisms is well recognized by theoretical ecologists. A variety of heterogeneity- and
aggregation-promoting mechanisms have been proposed, modeled and field-tested,
including disturbance-recovery dynamics (Paine and Levin 1981), dispersal limitation
(Levin et al. 1984), spatially subdivided resource competition (Roughgarden 1974, Tilman
1994), and predator-prey dynamics (Levin 1976, 1978). Mobile organisms are, almost by
definition, more difficult to describe spatially. With the increasing speed and power of
raster GIS technology, however, spatially-explicit population models (SEPMs),
incorporating (cell-based) movement as well as spatial location of individuals, are
becoming more sophisticated and useful (e.g., Pulliam et al., 1992, Turner et al. 1993,
Dunning et al. 1995, Holt et al. 1995).
Until recently, however, there has been little focus on intermediate scales—those much
Smaller than a species range, yet large enough that individual dynamics become obscured
by aggregate patterns (but see Watt 1947). For many species, this intermediate scale
occurs at the landscape level—an admittedly broad term, with different definitions in
different contexts. The landscape scale has significance, however, because it is inherently
human-defined, and often coincides with the scale of anthropogenic land-use change.
Within the last decade, often with the help of GIS technology, greater efforts have been
made to describe the spatial patterning of organisms across landscapes (e.g., Swanson et
al. 1988, O’Neill et al. 1991, Ward and Saltz 1994, Cherrill et al. 1995, Fortin et al. 1996)
and the processes driving those patterns (e.g., Kareiva and Wennergren 1995, Turner et al.
1997). When spatial pattern is not the primary focus, however, it is often ignored
altogether. This is especially true of studies examining the effects of human disturbance.
– 63 -
This study addresses some of the above-named shortcomings in the analysis of spatial
ecological data. My primary goal is to determine how selected passerine bird species
respond to increasing levels of urbanization (residential development) in a mountain
landscape facing encroachment on three sides by metropolitan Los Angeles. Previously, I
used Standard statistical analyses to answer this question (Chapter 1). Here I also examine
large-scale spatial trends and Small-scale heterogeneity in a landscape-level spatial
statistical analysis of avian responses to urbanization. Most of the bird species under
Study are particularly well-suited to this type of analysis, as they are year-round residents
with low vagility and high territoriality (Soulé et al. 1988), and can therefore be modeled
as fixed points in space.
2.1.1. Types and Sources of Spatial Dependence in Animal Populations
Spatial dependence can occur at different scales, which may be indicative of the
process(es) driving the observed spatial patterns. Generally speaking, however, spatial
autocorrelation in a data set indicates the presence of small-scale spatial dependence due
to the interdependence of neighboring sites (Cressie 1991, MathSoft 1996). Small-scale
spatial dependence is generally stochastic in origin and therefore difficult to predict.
Conversely, large-scale spatial dependence refers to a predictable pattern or spatial trend -
(Cressie 1991), also known as “structural” spatial dependence (Rossi et al. 1992), which is
often related to underlying physiographic features.
Large-scale spatial dependence in a species’ distribution and abundance may reflect
“continuous biogeographical dispersion routes” (Carroll and Pearson 1998), as well as
underlying trends in the abiotic environment (e.g., topography, substrate, climate) that are
generated slowly over geologic time (Brown 1984). Small-scale dependence tends instead
to reflect patchiness in the biotic environment, stemming from stochastic disturbance
events or complex species interactions that occur on an ecological time scale (O’Neill et
al. 1991, Wiens 1981). Adding to the complexity of spatial dependence are intraspecific
factors like non-selective dispersal, time lags in population dynamics, and density-
dependent behaviors (e.g., territoriality, flocking) (Brown et al. 1995). These small-scale
– 64-
biological interactions may introduce temporal and/or spatial lags, decoupling a system
from underlying geophysical factors (Wiens 1989) and making spatial variability difficult
to predict. At large scales, however, physical processes often overshadow the biological
(Wiens 1989).
As a result of these and other sources of spatial dependence, animal populations are likely -
to be non-uniformly distributed across a landscape, and may contain variations in
abundance that are not easily explained by environmental factors. The existence of spatial
heterogeneity suggests that sampling locations across a landscape may not be equally
independent of one another. Rather, sites that are closer in spatial proximity are likely to
be more similar in species abundance (Brown 1984, Rossi et al. 1992). Thus, a
fundamental statistical assumption is violated when populations exhibit significant spatial
dependence. This leads to an over-specification of degrees of freedom, and an increased
probability of rejecting a valid null hypothesis (type I or o error), meaning that results
may too often be declared significant (Legendre 1993).
One must therefore control for spatial location when analyzing relationships between
spatially distributed ecological variables. The detection of spatial autocorrelation in
model residuals indicates the presence of an additional explanatory variable, which may or
may not be measurable. When this variable is not measurable, the spatial autocorrelation
may instead be explicitly modeled. Spatial regression models eliminate correlation in the
residuals, so that assumptions are met and significance tests are valid. Large-scale trends,
which are easily misinterpreted as autocorrelation, should be investigated before spatial
autocorrelation is modeled (Legendre and Fortin 1989). Semivariograms, correlograms,
and other “structure functions” (Yaglom 1957) can help identify these trends.
The purpose of this study is to investigate the importance of both small- and large-scale
spatial dependence in a landscape-level spatial data set, thereby providing an example of
how simple spatial statistics may be used to construct improved models of species
abundance.
– 65 -
2.1.2. Tools for Investigating Spatial Structure and Autocorrelation
Several methods have been used to examine spatial structure and detect scales of influence
in ecological data. Scale detection has received particular attention among landscape
ecologists (e.g., Carlile et al. 1989, Turner 1989, Wiens 1989, Moloney et al. 1991,
O'Neill et al. 1991, Cullinan and Thomas 1992, Levin 1992), with respect to both
simulated and actual landscapes. Methods used to examine scales of spatial pattern
include spectral analysis (Ripley 1978), variance ratio analysis (Carlile et al. 1989), Hill
analysis (Hill 1973), and fractal dimension (Sugihara and May 1990), each of which has
its strengths and weaknesses. No one method seems to provide consistently good
estimates of scale, and consequently the use of more than one method is recommended
(O’Neill et al. 1991, Cullinan and Thomas 1992).
Another class of tools for analyzing spatial structure comes from the field of geostatistics
and the generalized regional variable model (Matheron 1963). Geostatistics theory is
most commonly used in “kriging,” a method for interpolating the values of a variable over
the geographic area from which the sample was taken by incorporating a semivariogram
model for spatial correlation. Related to the semivariogram are the covariogram and the
correlogram. Each of these “structure functions” may be used to describe the spatial
dependence between sampling points at various “lag distances” or distance classes, and to
detect spatial trends in the data.
The Semivariogram
The semivariance function M(h) (Matheron 1963) is estimated by one-half the average
squared difference between points separated by a distance h, calculated as:
|NOh)||N(h) 2
(z , -z y)
h) = ++ 2 2.1
y(h) 2| N(h)| (2.1)
- 66 -
where N(h) is the set of all pairwise Euclidean distances i-j = h, N(h) is the number of
distinct pairs in N(h), and zi and zj are data values at spatial locations i and j, respectively
(Cressie 1991, MathSoft 1996).
A Semivariogram is a plot of the semivariance function for increasing lag distances.
Standard semivariograms are omnidirectional, but directional semivariograms, where yis
a function of the direction as well as magnitude of h, may also be examined if the spatial
autocorrelation of a variable changes with direction—a property referred to as
“anisotropy” (lack of “isotropy”). In this case, the set of data pairs, N(h) is defined by
direction as well as distance.
The semivariance represents the portion of the population variance that is not explained by
Spatial autocorrelation. Thus in a system with no spatial dependence, the semivariance is
constant and equal to the population variance (o’). For most spatial data, however, the
semivariogram increases with distance until it converges at a “sill,” which corresponds to
the population variance. The distance at which the sill is reached and data are no longer
autocorrelated is called the “range” (Cressie 1991; see Figure 2.1). The range may be
interpreted as a maximum autocorrelation distance, while autocorrelation is generally
strongest at the distance for which the semivariogram slope is steepest (Robertson 1987).
A semivariogram that fails to converge to a sill indicates that the variable of interest is
non-stationary (non-constant mean and variance) and exhibits a large-scale trend over the
area of interest (Legendre and Fortin 1989). The “nugget” of a semivariogram represents
all unaccounted-for spatial variability at distances smaller than the smallest sampling
distance, which may be due to micro-scale variation and/or measurement error (Rossi et
al. 1992; see Figure 2.1). Several functions are typically used to model theoretical
semivariograms, depending on the spatial structure of the variable. Some of the more
commonly used functions are the spherical, exponential, linear and Gaussian models (see
Figure 2.2).
– 67 -
An empirical semivariogram provides a graphical description of the autocorrelation
Structure in a sample of a particular variable, and can be used to estimate the theoretical
semivariogram for the entire population, which is required for kriging. It does not lend
itself to statistical hypothesis testing, however; nor is it standardized, so comparison of
different models may be difficult. In addition, the semivariogram is sensitive to local
mean and variance differences and is strongly affected by outliers, so it may provide an
incomplete picture of spatial pattern (Rossi et al. 1992). Meisel and Turner (1998) found
that the semivariogram acts best as a “high pass filter,” detecting coarse but not
necessarily fine scale patterns, and that it is not particularly well-suited to the study of
multiple scales of pattern. They also found it to be highly sensitive to data gaps.
Despite a relatively long kriging tradition in the mining and geology fields (e.g., David
1977, Journel and Huijbregts 1978), semivariograms have only recently been applied to
ecological studies, where they have been useful for describing spatial structure and
interpolating the values of ecological variables over space (e.g., Grieg-Smith 1983, Taylor
1984, Villard and Maurer 1996, Robertson et al. 1997, Meisel and Turner 1998).
The Covariogram and Correlogram
The empirical covariance function, C(h), represents the portion of the population variance
that is explained by spatial autocorrelation at lag distance h. It is defined by:
|NOh)||N(h) - e-
(z, –Z)(z, –Z)
C(h) = cov(2,2,.) = −. | N(h)| (2.2)
When the population mean and variance are constant over the sampling space (i.e., the
second-order stationarity assumption is met), then for any lag distance h, the
semivariance, Mh), and the covariance, C(h), add up to the population variance, o” (Rossi
et al. 1992). Thus the semivariogram and the covariogram (a plot of the covariance
function against increasing lag distances) are inversely related. Under these same
- 68-
conditions, the correlogram, p(h), or standardized covariogram (C(h)/o”), is also directly
related to the semivariogram (Rossi et al. 1992).
The term “correlogram” may also be used to refer to any plot of autocorrelation versus
Spatial lag distance (or temporal legs. in the case of time-series analysis). Thus
correlograms for spatial data may also be constructed using other standardized measures
of autocorrelation. A commonly used autocorrelation index, related to p(h), is Moran's I
statistic (Moran 1950). Moran’s I, which generally ranges from -1 to 1, is similar in
interpretation to a Pearson’s Product Moment correlation statistic, with the numerator
consisting of a sum of cross-products (Legendre and Fortin 1989). For a given distance
class h, Moran’s statistic, I(h), is defined as:
º X. w; (z, -Z)(zy - *
I(h) = } H. º Pl 9 (2.3)
3×36-y
where wif takes on the value 1 when the pair (i,j) pertains to distance class h, and 0
otherwise; n is the number of sampling locations (Fortin 1989). Because they are
Standardized, correlograms may easily be compared. A Moran’s I correlogram can also be
tested for statistical significance, making it less subjective than the semivariogram for
describing spatial structure and detecting autocorrelation. Q
In terms of large-scale patterns, positive autocorrelation at small distances coupled with
negative autocorrelation at large distances indicates the presence of a gradient, or large-
scale trend. Negative values at large distances result from distant points being less similar
than average, so that the semivariance is greater than the population variance (covariance
= population variance - semivariance). Aggregated data may exhibit positive
autocorrelation at distances corresponding to the gaps between patches or data clusters
(Legendre and Fortin 1989), and zero or negative autocorrelation in between.
Correlograms for variables containing only small-scale spatial autocorrelation will quickly
decrease and converge to or oscillate around zero.
– 69 –
With respect to Small-scale patterns, the evaluation of autocorrelation statistics at various
lag distances in a correlogram may be useful in determining the maximum scale of
significant spatial autocorrelation. This may indicate the size of the “ecological
neighborhood” for the species or phenomenon being investigated (Addicott et al. 1987), or
the range of interaction between sites (Cressie 1991). This information can also be
ascertained from a semivariogram, but not with any statistical certainty.
Recently, several ecological studies have made use of correlograms and spatial
autocorrelation statistics for examining large- and Small-scale spatial patterns (e.g.,
Legendre and Fortin 1989, Carroll and Pearson 1998, Koenig 1998, Koenig and Knops
1998).
2.1.3. Spatial Covariance Models
Ordinary least-squares regression analysis requires that model residuals are independent
(Neter et al. 1989). When this assumption is violated, model parameter estimates may be
biased (Robertson 1987, Anselin 1989), and confidence intervals too small (Legendre
1993). Upon detection of spatial dependence in regression model residuals using tests for
spatial autocorrelation (e.g., Moran’s I), the spatial covariance structure of the residuals
can be specified and modeled explicitly. Provided the residuals are stationary (i.e., no
trend), then semivariograms and/or covariograms may be used to define the nature of
spatial dependence in the residuals and estimate the appropriate spatial parameters.
One approach to spatial covariance modeling is based on geostatistical principles
integrated with a mixed model framework (Zimmerman and Harville 1991, Littell et al.
1996). The regression model is of the form yi = a + ei, where yi is the iº" observation and
ei is the corresponding error. If y; occurs at location si, then the covariance structure is
defined by assuming that the covariance of two locations, si and sj, is a function of the
distance between them (dii). The covariance model then takes on the general form:
Cov(e,e,) = cr” |f(d).] with Var(e) = o' = o,” + o,” (2.4)
- 70 -
Since the covariance and semivariance are directly related (under the assumption of
stationarity), these parameters can be expressed in terms of semivariogram parameters: d
is the range, oº is the sill, or” is the nugget, and ox2 is the remaining variance, or partial
sill (see Figure 2.1). The function f(diſ) depends on the shape of the semivariogram, and
is generally specified according to one of the models in Figure 2.2.
2.1.4. Objectives
In 1996, I initiated a study to assess the effects of landscape-level urbanization patterns On
breeding-bird abundance in the Santa Monica Mountains of southern California. In recent
decades, chaparral and other native communities within the Santa Monica Mountains have
been experiencing urban encroachment from the surrounding Los Angeles metropolitan
area. Currently, the area contains a mosaic of interspersed suburban development and
native chaparral vegetation (see Figure 1.1). It presents a unique opportunity to study the
effects of urban encroachment at its intermediate stage, before the native habitat is
severely fragmented.
In my study area, the abundances of several chaparral- and woodland-associated bird
Species were negatively associated with levels of surrounding urbanization, while the
abundances of two urban-associated species exhibited positive associations with
Surrounding urbanization (Chapter 1). Significant responses to large-scale landscape
characteristics (i.e., surrounding urbanization proportion within radii up to 4000m) were
observed in most of these species.
Because the bird census locations used in this study were distributed over a large (~17,000
ha) geographic area, they are unlikely to be completely independent. Small-scale spatial
heterogeneity may result in spatially autocorrelated abundance. In addition, urbanization
within the study area follows an east-west gradient. To better understand the relationship
between urbanization and bird abundance, I posed the following research questions:
- 71 -
1. What is the nature of the spatial structure that characterizes the abundances of
these chaparral birds?
2. Controlling for spatial structure, does surrounding urbanization affect breeding
bird abundance?
2.2. Field Methods and Previous Results
Bird abundance data were collected during the summers of 1996 and 1997. A stratified
random sample totaling 79 chaparral-vegetated sites was censused for birds in 1996 using
100-meter fixed radius point counts, and georeferenced with GPS readings (detailed field
methods are given in Chapter 1). Twenty-eight sites were added in 1997 (totaling 107)
and the same bird census procedures were repeated (Figure 1.1). Using this sample of 107
sites, I modeled the effect of surrounding urbanization on maximum-per-site abundance of
each species, using Poisson regression with a repeated measures covariance structure
(1997 data + 1996 data where available). Several measures of surrounding urbanization
were examined, including the proportion of the surrounding area (within a circle of a
specified radius) contained in urban land uses (including golf courses, urban parks and
other human-constructed green spaces). GIS land use data were obtained from a 1:24000
Arcſinfo polygon coverage, constructed from 1990 aerial photos by the Southern
California Association of Governments. Scales of urbanization ranged from 250m to
4000m radius buffer distances (Chapter 1). Sequential score tests (Tarone 1988) were
used to determine the largest measure of surrounding urbanization proportion contributing
significantly to each species abundance model.
2.3. Spatial Statistics Methods and Results
Of the thirteen species exhibiting significant responses to some measure of surrounding
urbanization in 1997 (Table 1.3), four were selected for further analysis in this study, due
to the relative robustness of their models, and the small departures from normality in their
model residuals: the Wrentit (Chamaea fasciata), Northern Mockingbird (Mimus
polyglottos), Black-headed Grosbeak (Pheucticus melanocephalus) and Blue-gray
- 72 -
Gnatcatcher (Polioptila caerulea). To confirm the validity of these non-spatial
urbanization models, I looked for spatial structure in the abundance of each of the species
and modeled it explicitly as needed.
This analysis included four major steps:
1. Investigation of large- and small-scale spatial dependence in bird abundance,
urbanization proportion and vegetation cover, using semivariograms and
correlograms.
2. Regression of bird abundance on surrounding urbanization levels using (non-spatial)
ordinary least squares (OLS) regression, and investigation of spatial dependence in
model residuals.
3. Removal of large-scale spatial trends in bird abundance (regression of bird abundance
on spatial coordinates), regression of trend model residuals on Surrounding
urbanization levels, and investigation of spatial structure in detrended model residuals.
4. Construction of spatial covariance models for bird abundance, using spatial covariance
structures suggested by semivariogram and correlogram analyses, in order to
statistically incorporate small-scale spatial dependence.
Semivariogram and correlogram generation were performed using S-Plus 4.5 (MathSoft
1997), S-SpatialStats (MathSoft 1996), and S-Plus for ArcView GIS (MathSoft 1998).
SAS Systems’ version 6.12 mixed models procedure (SAS 1997, Littell et al. 1996) was
used for spatial covariance modeling.
2.3.1. Investigating Spatial Structure in Bird Abundance
Upon visual inspection of distribution maps for each species examined, they all appear to
vary along the east-west urbanization gradient, to some degree (Figure 2.3). The Northern
Mockingbird (2.3b) seems to increase in abundance from west to east, while the other
three species decrease. The Wrentit (2.3a) and Blue-gray Gnatcatcher (2.3d) may also
- 73 -
exhibit north-south variation, with the Wrentit more abundant in the south, and the
gnatcatcher more abundant in the north. The latter trends are less clear, however, due
perhaps to the general lack of north-south variation among sampling locations, especially
in the eastern section of the study area.
Using two related structure functions, the semivariogram and the correlogram, I examined
the spatial covariance structure of the sampling locations with respect to bird abundance,
as well as urbanization proportion. For each variable, I examined large-scale trends, and
determined the effective range of spatial autocorrelation (if any).
For maximum 1997 abundance of each of the four species and urbanization proportion
within a 2000m radius, I constructed an omnidirectional semivariogram, as well as
directional semivariograms in the east-west (90° azimuth) and north-south (0° azimuth)
directions (using a 45° tolerance), based on the 107 sampled locations. Fourteen lag
distances, at 1000m intervals, were used for the omnidirectional and east/west
semivariograms; nine 1000m lag distances were examined in the north/south direction.
Semivariograms were generated using the variogram function in S-HSpatialStats (MathSoft
1996).
I also constructed an omnidirectional correlogram for each of the above variables using
Moran’s I statistic. Correlograms were constructed from Moran’s I autocorrelation values
individually calculated for each 1000m lag distance (e.g., 0-1000m, 1000-2000m), also
using S--SpatialStats (MathSoft 1996). In calculating Moran’s I values, point pairs within
the given interval were considered to be spatial neighbors, and assigned a spatial weight
(w) of 1; all other point pairs were assigned a spatial weight of 0. I tested for overall
significance of each correlogram using a Bonferroni-adjusted significance level (O. = 0.01)
to account for multiple comparisons, as well as for significance of individual correlation
values (O. = 0.05), as suggested by Legendre and Fortin (1989).
- 74-
Semivariogram Results — Original Variables
The overall trend apparent in the omnidirectional semivariogram for urbanization (Figure
24a) was driven primarily by east-west variation, as supported by the 90° directional
semivariogram (2.4b), in which the variation among points increases almost
monotonically with distance, particularly at lag distances of 7000m or larger. The north-
South (0°) semivariogram (2.4c) exhibited a similar trend between 0-8000m, but stops
there, constrained by the dimensions of the study area.
Three of the bird species examined—the Wrentit, Northern Mockingbird and Blue-gray
Gnatcatcher—exhibited distinct large-scale spatial trends (Figure 2.5a,d,j) that appear to
follow the urbanization trend (24a-c). These trends occurred primarily in the east-west
direction (2.5b,e,k), as suggested by their distribution maps (2.3a,b,c), and slightly in the
north-South direction (2.5c,fl). The semivariograms did not increase monotonically,
however, suggesting the presence of additional spatial structure not related to the east-
west or north-south gradients. The semivariograms for the Black-headed Grosbeak were
fairly constant, but with a small increase in the very largest distance classes, suggesting a
slight spatial trend (2.5g-i).
In general, the omnidirectional semivariograms for each of the species resembled the east-
west semivariograms, indicating that large-scale spatial trends were driven primarily by
the east-west gradient (perhaps resulting from urbanization). The north-south
semivariograms were most likely constrained by the configuration of the study area (long
and narrow); thus they were based on fewer point pairs within each lag interval, and may
be less reliable than the other semivariograms. The east-west semivariograms are most
suspect at the shorter lag distances (<1000m) due to fewer point pairs, while both
directional and omnidirectional semivariograms should be interpreted with caution at the
larger lag distances (>8000m) (see sample sizes for each lag distance in Figure 2.5).
- 75 -
Correlogram Results — Original Variables
The omnidirectional Moran’s I correlogram for urbanization proportion displayed a nearly
linear decrease in autocorrelation (covariance) with increasing lag distance (Figure 2.6). It
demonstrated overall significance (p<0.01), as well as significant autocorrelation values
for almost all distance classes. This indicates a clear, large-scale gradient, supporting the
conclusions of the semivariogram analysis (Figure 2.4a-c).
Correlograms also confirmed that all four species were distributed along a spatial gradient,
based on the steady (if not monotonic) decrease from positive to negative autocorrelation
and overall significance (p<0.01) of each correlogram (Figure 2.7). All species exhibited
significant positive autocorrelation (p < 0.05) in more than one of the closer distance
classes, while only the mockingbird and grosbeak also showed significant negative
autocorrelation (p < 0.05) at greater distance classes. Autocorrelation ranges suggested by
the correlograms are approximately 7000m for the Wrentit (2.7a), 5000m for the Northem
Mockingbird (2.7b), 6000m for the Black-headed Grosbeak (2.7c), and 4000m for the
Blue-gray Gnatcatcher (2.7d).
2.3.2. Non-Spatial Modeling of Bird Responses to Urbanization
Results from the previous section indicate large-scale trends in the abundances of all four
bird species examined, which could invalidate tests for spatial autocorrelation (Legendre
1993). That is, the similarity between sites that are close in spatial proximity may be due
to their position on a gradient, rather than some intrinsic heterogeneity-producing spatial
process. To obtain a reliable picture of small-scale autocorrelation, large-scale trends
must first be removed.
Because similar trends were also observed in landscape urbanization proportion (within a
2000m radius of each point), I suspected that these large-scale trends were actually
products of the urbanization gradient. Thus I modeled the effects of surrounding
urbanization levels on bird abundance and examined the resulting model residuals for
remaining spatial autocorrelation, using the same tools as in the previous Section.
- 76-
Previous analyses indicated that the Wrentit is sensitive to urbanization up to 500m away,
the Black-headed Grosbeak up to 1000m, Northern Mockingbird up to 2000m, and the
Blue-gray Gnatcatcher up to 4000m away (Chapter 1). Based on these results, and
ignoring other significant urbanization-related predictors of bird abundance, I regressed
maximum-per-point abundance on surrounding urbanization proportion at the appropriate
Scale, using ordinary least squares (OLS) regression.
OLS regression models (rather than Poisson maximum likelihood) were used to facilitate
comparison with the Gaussian mixed models used in spatial covariance modeling (Litell
et al. 1996). Model residuals did not exhibit major departures from normality, and non-
spatial OLS model results were not found to differ markedly from those of the Poisson-
based models.
Effects of Urbanization on Bird Abundance
As demonstrated previously, the level of surrounding urbanization was a significant
predictor of abundance for all four species examined (Table 2.1). Due to the high
variation in the system, R values are low, although regression coefficients (B) are
generally highly significant. The mockingbird exhibited a positive response to
surrounding urbanization proportion, while the other three species were negatively
influenced.
Semivariogram Results – Urbanization Model Residuals
One species, the Black-headed Grosbeak, did not appear to exhibit any spatial dependence
in its residuals, in that its omnidirectional and directional semivariograms were fairly flat,
fluctuating around a population variance of approximately 0.5 (Figure 2.8g-i). For the
three other species, omnidirectional semivariograms increased initially, but eventually
reached a maximum, indicating the possible presence of spatial autocorrelation in model
residuals up to the distance at which the sill is reached (range) (2.8a,d,j). Although the
sill and range differed, these species all exhibited several small peaks and troughs, with
maxima between 4000 and 10,000 meters. This type of semivariogram function may
- 77 –
indicate that several processes govern bird abundance, operating at different spatial scales.
A sharp discontinuity in the semivariogram may represent a critical scale threshold
(Turner et al. 1989), or change in the process driving variation. More likely, however, is
that the peaks and dips reflect patchiness or aggregation in the sampling locations
(Legendre and Fortin 1989, Meisel and Turner 1998). Particularly in the eastern portion
of the study area, there were large data gaps resulting from the inaccessibility of certain
chaparral areas, as well as urbanization-related habitat fragmentation (see Figure 2.3).
The omnidirectional semivariogram for Wrentit residuals reached a sill (~3) at 7000m
(Figure 2.8a), indicating that this rather large distance was the maximum range of
autocorrelation. This pattern appeared to be driven by north-south variation, as seen in the
corresponding directional semivariogram, where the peak at 7000m was quite
pronounced, suggesting possible outlier data (2.8c). The east-west semivariogram had a
much smaller sill and range, peaking at a distance of 2000m (2.8b). These discrepancies
suggest that autocorrelation in this species may be anisotropic (direction-dependent).
For Northern Mockingbird residuals, the omnidirectional semivariogram had an even
larger range, reaching a sill (~0.8) at 10,000m (Figure 2.8d). Thus this species still
appeared to exhibit a slight large-scale trend in its abundance, even after the effect of
urbanization was removed. As with the Wrentit, this large-scale trend appeared to be
driven by north-south variation (8f), while the east-west semivariogram (8e) was almost
flat, suggesting anisotropic conditions. This remaining north-south dependence may be a
product of topography, in that study area consists of many north-south running canyons.
Within these canyons, bird abundance may be more similar than between sites in different
canyons that are closer in a Euclidean sense. Both species may also vary with elevation,
which generally increases from south to north in the study area.
The omnidirectional semivariogram for the Blue-gray Gnatcatcher reached a sill (~0.4)
near 4000m, then decreased fairly rapidly, reaching a minimum near 10,000m (Figure
2.85). This suggests a large-scale pattern, or similarity between the extreme ends of the
- 78 -
local population. Directional semivariograms (2.8k-l) were fairly similar, indicating
isotropic (non direction-dependent) autocorrelation.
The presence of a sill in each semivariogram indicates that the residuals were stationary
over the study area, which in turn suggests that large-scale trends in bird abundance were
primarily explained by variations in urbanization (although causality cannot be
determined). However, the large autocorrelation ranges revealed by the semivariograms
(4000-10,000m) indicated that model residuals still contained large-scale spatial
dependence. This may have been due to the influence of other large-scale patterns
(Smaller than the size of the study area), such as patterns of variation in vegetation or
elevation, in determining bird abundance.
Correlogram Results – Urbanization Model Residuals
Examination of omnidirectional Moran’s I correlograms (Figure 2.9) provided a more
complete picture of spatial structure in urbanization model residuals. Correlograms were
overall significant (p < 0.01) for all species but the Grosbeak, with spatial autocorrelation
generally declining with distance and oscillating around zero (Figure 2.9). What the
correlograms suggested, beyond the semivariograms, was a maximum range of “Small-
scale” autocorrelation (when the correlogram first reaches zero). Three of the four species
exhibited significant small-scale spatial autocorrelation (p → 0.05), with maximum
autocorrelation ranges apparently near 2000m for the Wrentit (2.9a) and mockingbird
(2.9b), and 4000m for the gnatcatcher (2.9d). Additional autocorrelation beyond these
ranges may represent remaining large-scale trends and/or patchiness in the sampling sites.
Again, all species exhibited somewhat periodic dips and crests in the correlograms for
their model residuals. For the Wrentit, significant positive spatial autocorrelation
occurred at 1000m and 5000m; negative autocorrelation was present at the 7000m and
9000m lag distances (Figure 2.9a). The mockingbird was positively autocorrelated at
1000m and 12,000m, and negatively autocorrelated at 6000m and 10,000m (2.9b). The
grosbeak was positively autocorrelated at 11,000m (2.9c). The gnatcatcher exhibited
- 79 -
significant positive autocorrelation at 1000m and 2000m, and negative autocorrelation at
7000m (2.9d). Negative autocorrelation in the middle distances (i.e. 6000m-10,000m)
most likely indicates the influence of spatial patchiness in the sample. The positive
autocorrelation at large distances exhibited in grosbeak and mockingbird residuals may
represent similarities at extreme ends of the study area, caused by large-scale processes,
but more plausible is the influence of outliers, given the high variability and large gaps in
the data.
2.3.3. Modeling large-scale spatial trends
Because the east-west trend in urbanization was so prominent, teasing apart the effects of
urbanization and spatial location on bird abundance was difficult. Results from the
previous section suggested that large-scale trends in bird abundance are largely a
reflection of variations in urbanization, but that additional large-scale spatial dependencies
may have been present, primarily in the north-south direction. The question remained as
to whether these trends indicated non-stationarity over the study area (a "true gradient"),
or intrinsic autocorrelation among sampling locations separated by large distances (a
"false gradient") (Legendre 1993). If the former is true, then bird abundance should vary
as a function of spatial location, and residuals should be uncorrelated once the trend is
removed (Legendre 1993).
To statistically remove the large-scale trend, I separately regressed the abundance of each
species on the x-y coordinates of the sampling locations and the interaction between them
(Z ~ x + y + x*y). I then regressed the trend model residuals on urbanization proportion,
and compared the significance of the parameter estimates and overall predictive ability
with the original urbanization models for each species. Finally, I looked for remaining
spatial dependence in the residuals of the de-trended urbanization models for each species,
using semivariograms and correlograms.
- 80 -
Results of trend models
For the Blue-gray Gnatcatcher and Wrentit, spatial location explained a greater portion of
the variation in abundance than urbanization proportion—25% vs. 12% for the
gnatcatcher and 35% vs. 23% for the Wrentit (Table 2.1). For the Northern Mockingbird
and Black-headed Grosbeak, however, surrounding urbanization levels explained more of
the variance than spatial location (41% vs. 24% and 14% vs. 12%, respectively). Over
50% of the variation in urbanization (at the 2000m radius) was explained by spatial
location, which means that the two effects are difficult to separate.
After removing the large-scale spatial trend, urbanization proportion was still a significant
predictor of mockingbird, Wrentit and grosbeak abundance, although the explanatory
power of these models was generally low, compared with the initial OLS models (i.e.,
lower R and p-values) (Table 2.1). These results indicate that urbanization, in and of
itself, did explain a significant portion of the variation in abundance of these species, even
under the most conservative assumptions.
For the Blue-gray Gnatcatcher, urbanization proportion was no longer a significant
predictor of abundance when the spatial trend was removed. Thus the effects of spatial
location and urbanization were difficult to separate in this case, suggesting that
gnatcatcher abundance may have been distributed along a large-scale gradient determined
by something other than urbanization.
Semivariogram and Correlogram Results – Detrended Model Residuals
Semivariogram (Figure 2.10) and correlogram (Figure 2.11) analyses suggest that all
species but the Northern Mockingbird may have followed a “true” gradient, in that
detrended model residuals appeared to be uncorrelated. The extent to which these large-
scale trends in abundance were a product of urbanization remains unclear, however,
especially since surrounding urbanization proportion at a particular scale may not describe
all of the variation in urbanization. Because urbanization may have been driving the
observed trends, it is not necessarily helpful to conservatively attribute bird responses to
-81 -
Spatial location. It may be more appropriate to assume that all spatial variation not
attributable to urbanization is autocorrelation resulting from interactions between sites,
and incorporate this dependence into the urbanization model.
2.3.4. Modeling small-scale spatial autocorrelation
Spatial dependence detected in the residuals from non-spatial regression models means
that the models were improperly specified (Anselin 1989). If this autocorrelation
represents positive, Small-scale autocorrelation, rather than large-scale trends, then a
spatial covariance model can be used to adjust for the spatial dependence. By
constructing a spatial covariance model for the residuals, the autocorrelation structure can
be explicitly defined, so that independence assumptions are met and significance tests are
valid.
For spatial covariance models to be appropriate, the residuals should exhibit second order
stationarity (constant mean and variance over sampling area) and isotropy (covariance
independent of direction) (Littell et al. 1996). Based on the apparent convergence of the
semivariograms to a sill, I assumed that the stationarity assumption was met in all cases,
despite irregularities and large autocorrelation ranges in the semivariograms and
correlograms. I also assumed isotropic behavior, despite differences in east/west and
north/south semivariograms.
For each of the three species with autocorrelated residuals according to semivariograms
and Moran’s I correlograms (mockingbird, Wrentit and gnatcatcher), I fit a spatial
covariance model (Littell et al. 1996) corresponding to the non-spatial model. The model
was fit using mixed model regression in SAS version 6.12 with REML (restricted
maximum likelihood) estimation of model parameters (SAS 1997). Based on the shapes
of the semivariogram for non-spatial model residuals, I used an exponential model for the
covariance function:
f(d)= (*/ (2.5)
- 82 -
The range parameter (p) was fixed at a value consistent with the empirical semivariogram
for non-spatial model residuals, and the partial sill (op”) and nugget (ojº) values were
estimated simultaneous with estimation of regression coefficients (8). Using the
maximum values of the empirical semivariograms, the approximate ranges determined for
the Northern Mockingbird, Wrentit and Blue-Gray Gnatcatcher were 9000m, 7000m, and
4000m, respectively. These range estimates are rather large, however, and the spatial
covariance model is only intended for positive, small-scale dependence (Littell et al.
1996). Thus I also used correlograms to determine more appropriate ranges for
comparison. I used as a second range estimate the approximate distance at which the
correlogram for non-spatial model residuals first reached zero: 2000m for the Wrentit and
mockingbird and 4000m for the gnatcatcher (same as semivariogram estimate). For the
Wrentit, this smaller range also corresponded with the range suggested by the east-west
directional semivariogram. The use of two different ranges (small vs. large) also provided
a sensitivity analysis for the regression parameters.
Likelihood ratio x tests were used to determine the significance of the spatial (vs. non-
spatial) models. The "best" covariance model could not be determined statistically, but
was evaluated by comparing log likelihood values and significance of individual spatial
parameter estimates. Regression coefficients and standard errors were also compared
between spatial and non-spatial models to determine whether modeling of spatial structure
in the residuals changed the substantive interpretations of the regression analysis.
Results of spatial covariance models
For all species, the spatial models exhibited significantly better fit than the corresponding
non-spatial models (x likelihood ratio test, p < 0.05) (Table 2.2), indicating significance
of the spatial covariance parameters. Overall, the inclusion of spatial parameters caused
regression parameter estimates (AE) to decrease, although not always substantially.
The Wrentit exhibited the largest variation in regression parameter estimates between the
three models, with somewhat higher magnitude values estimated by the non-spatial model
– 83 –
than in either spatial covariance model, although confidence intervals for all parameter
estimates were overlapping (Table 2.2). Spatial parameter estimates (a12 and oz2) were
larger in the large range (7000m) model, but the nugget effect was only significantly
different from zero (p<0.05) in the small range (2000m) model. The log likelihood value
for the large-scale model was slightly higher, indicating that the incorporation of large-
Scale spatial structure (up to 7000m) improved model fit for this species, but with a
consequent reduction in the urbanization parameter estimate. This suggests that part of
the urbanization association may indeed be attributable to spatial autocorrelation. The
most appropriate model depends on the range of autocorrelation assumed.
The Northern Mockingbird exhibited a slightly smaller relative reduction in regression
parameter estimates with the inclusion of spatial parameters (Table 2.2). Spatial
covariance parameter estimates were also significantly different from zero (p<0.05) in
both spatial models, but the nugget estimate was quite a bit larger in the large range model
(9000m) than in the small range (2000m) model. For this species, the 2000m range
distance appeared to generate the best-fitting model, based on the higher log likelihood
value. Thus the incorporation of large-scale autocorrelation (above 2000m) was
unnecessary and inappropriate for this species.
In the Blue-gray Gnatcatcher models, the effect of urbanization proportion was reduced
enough that it is no longer significant in the spatial model (Table 2.2). The partial sill
estimate in the spatial model (4000m range) was significantly different from zero (p <
0.05), but not the nugget. This indicates, as does the trend model from the previous
section, that all of the urbanization-related variation in gnatcatcher abundance may also be
described by spatial location.
2.4. Discussion
The four species examined in this study have demonstrated significant population
responses to urbanization at large spatial scales. Abundance of the Wrentit, Black-headed
Grosbeak and Blue-gray Gnatcatcher were negatively associated with high levels of
– 84 -
Surrounding urbanization, while the Northern Mockingbird was positively associated with
urbanization. These results suggested that chaparral bird populations may be affected by
changes in landscape composition and pattern, even when apparently suitable habitat is
present.
Using a variety of spatial statistics tools, I was able to consider the influence of spatial
dependence in my analysis of landscape-level urbanization effects on chaparral birds, in
order to verify and/or qualify these results. Equally important, these techniques also
allowed a better understanding of the spatial structures of the bird populations under
study, and the landscape mosaic that they inhabit. Different insights were gained for each
of the species examined, while common spatial patterns were also illuminated.
All four species exhibited large-scale spatial trends in their abundance, primarily
coinciding with the east-west urbanization gradient along which they were sampled
(Figures 2.4-2.7). After accounting for urbanization effects, however, these trends were
largely absent from model residuals (Figures 2.8–2.9), suggesting that urbanization
explained most of the large-scale spatial structure in these species. Conversely, when the
spatial trends were removed, three of these species (Black-Headed Grosbeak, Wrentit and
Northern Mockingbird) still demonstrated significant responses to changes in urbanization
(Table 2.1).
Abundance patems for all four species exhibited additional spatial autocorrelation
unrelated to urbanization (Figures 2.8–2.9). The use of spatial covariance models to
incorporate this small-scale spatial dependence (where possible) allowed conclusions
about urbanization effects to be more robust. In general, the observed relationships
between urbanization and bird abundance hold, despite small-scale spatial structure
(autocorrelation) in the data. Furthermore, the spatial covariance models had better
predictive capabilities, indicated by the significant improvement in log likelihood
statistics. If desired, the spatial and regression parameter estimates obtained in these
analyses could be used to interpolate bird abundance at points between those sampled. º
- 85 -
statistics. If desired, the spatial and regression parameter estimates obtained in these
analyses could be used to interpolate bird abundance at points between those sampled.
2.4.1. Ranges of Spatial Autocorrelation
Ranges of Small-scale spatial autocorrelation, as suggested by the correlogram analysis,
can be interpreted as indications of the ecological neighborhoods for each species—that is,
the “regions of activity or influence during periods of time appropriate to particular
ecological processes” (Addicott et al. 1987, p.340). Depending on the process and time-
Scale, a mobile animal’s ecological neighborhood can be defined by factors like (in order
of increasing spatial scale) breeding territory size, home range size, natal/adult dispersal
distance, and spatial extent of a local population.
Territory sizes, home range sizes and dispersal distances vary widely among passerine
Species in general, as among the four species examined in this study. For the Wrentit, an
“extremely sedentary” resident chaparral species (Ehrlich et al. 1988, p. 420), breeding
territories average -0.3ha in size (Erickson 1938), which is equivalent to a 30m x 30m
area. Natal dispersal distance for this species averages 375m, with a maximum near 700m
(Baker et al. 1995). The Blue-gray Gnatcatcher, also a year-round resident in the Santa
Monica Mountains, has somewhat larger territories, averaging-1 .8ha, and may have
winter home ranges up to 9ha in size (Ellison 1992). Natal dispersal distances are
unknown for the gnatcatcher, but movements up to 75km between breeding sites have
been recorded (Ellison 1992). The Black-headed Grosbeak, a long-distance migrant
species, has similar territory sizes, averaging 0.7-2.7ha, and low natal philopatry, with no
individuals returning to breed in one New Mexico study area (Hill 1995). For the urban-
associated Northern Mockingbird, large movements (>300km) have also been recorded,
although natal dispersal distances are typically much shorter (Derrickson and Breitwisch
1992). Thus, while chaparral bird species, in general, tend to be sedentary (Soulé et al.
1988), three of the four species examined in this study are known to move over large
distances, possibly on the order of the autocorrelation ranges detected (2000-4000m).
-86 -
The large autocorrelation distances found in this study—on the order of dispersal ranges
and population sizes, rather than territory and home range sizes—may indicate the
influence of population (vs. individual) dynamics, which is reasonable given the
intermediate temporal and spatial scales of the urbanization processes being studied.
Time lags may also result in a decoupling of ecological variables and responses in bird
abundance (Wiens 1989). Given the coarse resolution of this analysis, however,
autocorrelation at Smaller scales, resulting from interactions among individuals, could not
easily be detected.
Of the four species examined, the Northern Mockingbird demonstrated the strongest
small-scale autocorrelation, even after large-scale trends were removed (Figures 2.9 and
2.11). This indicated a strong interdependence among sampling sites, independent of the
location in the landscape, but may also have been due in part to the clustering of this
species near urban developments. Based on correlograms for non-spatial model residuals,
the range of Small-scale autocorrelation appeared to be 2000m (Figure 2.9), although
autocorrelation dropped off quickly, indicating that maximum autocorrelation occurred at
a somewhat smaller distance.
Similar results were obtained for the Wrentit, a species that was negatively associated
with surrounding urbanization, although for this species, residuals were longer correlated
after large-scale trends were removed (Figure 2.11). The Wrentit also seemed to exhibit
anisotropic variation in the shorter distance ranges (Figure 2.8), which likely explained the
weaker overall autocorrelation values. Autocorrelation in the north-south, but not east-
west, direction, after factoring out urbanization influences, may be due to the north-south
orientation of the canyons and trails along which sampling sites were located. Given the
sedentary nature of this species, a juvenile Wrentit may be more likely to disperse within a
canyon than across a ridgetop to the next canyon, especially since many of the ridges have
firebreaks, which this species may be reluctant to cross (Small 1974).
The comparatively large range of potential spatial autocorrelation (4000m) in the Blue-
gray Gnatcatcher (Figure 2.9) seemed to indicate the presence of a large-scale trend in
-87 -
abundance, rather than interaction among distant sites. This was also supported by the
non-significant urbanization parameter in both the spatial covariance model and trend
model for this species. Although the large-scale trend may have been urbanization- f
related, this could not be determined with certainty. In order to obtain more conclusive
results for this species, a larger area should be studied, given that 4000m is nearly 20% of
the length of the study area. Given the large movements documented in this species
(Ellison 1992), the 4000m range of autocorrelation may indeed reflect dispersal distance
and interaction among sites.
For the Northern Mockingbird and Blue-gray Gnatcatcher, the range of autocorrelation in
model residuals, as determined by correlogram analyses (Figure 2.9), coincided with the
maximum range of urbanization influence: 2000m and 4000m, respectively (Chapter 1).
These “scale agreements” suggest that either (1) urbanization patterns (e.g., clustering of
developments or recreational trails) were responsible for the observed spatial
dependencies; or (2) autocorrelation in bird abundance due to other factors (e.g., dispersal
limitations or social interactions) facilitated urbanization influences over larger distances.
Indeed, this chaparral system is so integrated with the surrounding urbanized area that
such interactions might be expected.
2.4.2. Large-scale Trends vs. Small-scale Autocorrelation
Spatial covariance modeling is most appropriate for Small-scale spatial dependence
resulting from the interactions among sites that are close in spatial proximity. Spatial
autocorrelation can also occur at larger scales, when the influence of one site carries a
long way through geographic space (Legendre 1993). But large-scale spatial dependence
commonly reflects a spatial gradient, and if so, it should be removed before spatial
autocorrelation is modeled, as it violates the stationarity assumption (Legendre 1993).
The distinction between the two is subtle, and also depends upon the temporal scale and
window of analysis (Wiens 1981). For example, a gradient at the landscape level may
represent a mere blip of heterogeneity at the regional level. For the purpose of
geostatistical analyses, a true gradient should encompass and extend beyond the study
- 88 -
area. Large-scale patterns that are smaller than this can be incorporated in a spatial
covariance model.
In this study, large-scale spatial dependence in bird abundance was certainly present, but
most of it was explained by the urbanization gradient. Remaining large-scale structure
appeared to be operating on a smaller scale than that of the study area (as demonstrated by
semivariograms in Figure 2.8) and may have followed patterns of variation in vegetation,
topography, or other geophysical factors. It may also have been a product of aggregated
Sampling locations (Legendre and Fortin 1989), resulting from a combination of habitat
fragmentation by urbanization, clustering of census sites along trails and fire roads, and
gaps in census sites due to restricted or privately-owned areas. Regardless of the source,
this large-scale spatial dependence appeared to be adequately described in the spatial
covariance models. The similarity between large- and small-scale covariance models,
however, in terms of regression parameter estimates and overall fit, indicated that small-
Scale autocorrelation is most important. The large-scale variations may have been too
unpredictable to model accurately with the type of model used, which is only intended for
positive, small-scale autocorrelation (Littell et al. 1996).
For all species but the Northern Mockingbird, residuals from the detrended models
exhibited no spatial autocorrelation (Figure 2.11), suggesting that spatial covariance
models may not be necessary to describe their abundance patterns if large-scale trends are
removed. Here emerges a possible trade-off between describing the distribution of bird
abundance and detecting urbanization-related changes. If the primary goal is to model the
abundance of these species in this particular landscape, then factoring in spatial location,
in addition to the effect of urbanization influence, may be adequate. The remaining
residuals were uncorrelated, so the models should be appropriately specified. Given the
distinct urbanization gradient, however, such a model diminishes the potential for
detection of urbanization influences. If autocorrelation is thought to occur over large
distances, than this spatial dependence can instead be explicitly incorporated in a spatial
covariance model.
- 89 -
The decision to de-trend a variable, in cases such as this when the trend is not obvious,
depends on the assumptions made about the system. Here, because the birds under study
are fairly widespread and common throughout the Santa Monica Mountains, and the
vegetation types in which I sampled were similar for all sites, I would expect urbanization
and landscape pattern to cause more variation in abundance than spatial location perse.
Directional semivariograms indicated that trends in bird abundance were primarily driven
- by east-west variation (Figure 2.6), but after factoring out urbanization, north-south
autocorrelation was responsible for remaining large-scale dependence in Wrentit,
mockingbird and snatcatcher abundance (Figure 2.8). These results suggest that east-west
trends were primarily urbanization-related, while north-south trends were likely due to
natural features. Anisotropic covariance models, explicitly incorporating direction, as
well as distance, may therefore be more appropriate for these species. Furthermore, the
use of local spatial statistics (Getis and Ord 1992, Anselin 1995, Bao and Henry 1996)
may help identify specific regions of autocorrelation.
2.5. Conclusion
Fundamentally, spatial autocorrelation in nature is caused by unknown factors that could
potentially be measured and modeled explicitly, as many have done, particularly in Small-
scale studies. Measuring these unknown variables is often infeasible or impractical,
however, especially when they involve things like animal behavior, indirect effects of
biotic interactions, and time lags. Furthermore, such “bottom-up” approaches may
sacrifice generalizability for detail (Wiens 1989, Root and Schneider 1995).
In this study, landscape-level urbanization influences on bird abundance were the primary
focus of interest, while habitat type (i.e., vegetation) were kept constant. Although other
potential explanatory variables could certainly be measured (and some were, with no
significant associations found), small-scale autocorrelation, which has a large Stochastic
component, is difficult to predict. Thus, in the absence of the elusive “right” variables,
– 90 -
This study provides an example of how spatial statistical methods can be used to control
for spatial structure in analyzing relationships between spatially distributed variables. The
methods used helped confirm my conclusions regarding the influence of urbanization on
bird abundance, while also providing further insight into the spatial structure of the bird
populations examined.
– 91 -
2.6. Literature Cited
Addicott, J.F., J.M. Aho, M.F. Antolin, D.K. Padilla, J.S. Richardson and D.A. Soluk.
1987. Ecological neighborhoods: scaling environmental patterns. Oikos 49:340-
346. -
Andrewartha, H.G. and L.C. Birch 1954. The Distribution and Abundance of Animals.
University of Chicago Press, Chicago. -
Anselin, Luc 1989. Spatial Econometrics: Methods and Models. Kluwer Academic,
Dordrecht, Netherlands.
Anselin, L. 1995. Local indicators of spatial association – LISA. Geographical Systems
27:93-115. *
Baker, M., N. Nur and G.R. Geupel. 1995. Correcting biased estimates of dispersal and
survival due to limited study area: theory and an application using Wrentits.
Condor 97:663-674.
Bao, S. and M. Henry. Heterogeneity Issues in Local Measurements of Spatial
Association. Geographical Systems 3:1-13. -
Bock, C.E. 1984. Geographical correlates of abundance vs. rarity in some North American
winter landbirds. Auk 101:266–273.
Bock, C.E. 1987. Distribution-abundance relationships of some Arizona landbirds: a
matter of scale? Ecology 68:124-129.
Brown, J.H. 1984. On the relationship between abundance and distribution of species.
American Naturalist 124:255-279.
Brown, J.H., D.W. Mehlman, and G.C. Stevens. 1995. Spatial Variation in Abundance.
Ecology 76:2028-2043.
Carlile, D.W., J.R. Skalski, J.E. Batker, J.M. Thomas and V.I. Cullinan. 1989.
Determination of ecological scale. Landscape Ecology 2:203-213.
Carroll, S.S. and D.L. Pearson. 1998. Spatial modeling of butterfly species richness using
tiger beetles (Cicindelidae) as a bioindicator taxon. Ecological Applications 8:531-
543.
- 92 -
Chemill, A.J. C. McClean, P. Watson, K. Tucker, S.P. Rushton and R. Sanderson. 1995.
Predicting the distributions of plant species at the regional scale: a hierarchical
matrix model. Landscape Ecology 10:197-207.
Cook, R.E. 1969. Variation in species density of North American birds. Systematic
Zoology. 18:63-84. -
Cressie, N.A.C. 1991. Statistics for Spatial Data. Wiley, New York.
Cullinan, V.I. and J.M. Thomas. 1992. A comparison of quantitative methods for
examining landscape pattern and scale. Landscape Ecology 7:211-227.
David, M. 1977. Geostatistical ore reserve estimation. Elsevier Scientific, New York.
Davis, M.B. and C. Zabinski. 1992. Changes in geographical range resulting from
greenhouse warming: effects on biodiversity in forests. Pp. 297-308 in Peters, R.L.
and T.E. Lovejoy, Eds. Global Warming and Biological Diversity. Yale University
Press, New Haven, CT.
DeCandolle, A.P.A. 1874. Constitution dans le regne vegetal de groupes physiologiques
applicables a la geographie ancienne et moderne. Archives des Science Physiques
et Naturelles, Geneva. -
Delcourt, H.R. and P.A. Delcourt. 1988. Quaternary landscape ecology: relevant scales in
space and time. Landscape Ecology 2:23–44.
Derrickson, K.C. and R. Breitwisch. 1992. Northern Mockingbird (Mimus polyglottos).
Birds of North America 7. A. Poole, P. Stettenheim and F. Gill, Eds. Philadelphia,
PA.
Dunning, J.B., D.J. Stewart, B.J. Danielson, B.R. Noon, TL. Root, R.H. Lamberson and
E.E. Stevens. 1995. Spatially explicit population models: current forms and future
uses. Ecological Applications 5:3-11.
Ellison, W.G. 1992. Blue-gray Gnatcatcher (Polioptila caerulea). Birds of North America
23. A. Poole, P. Stettenheim and F. Gill, Eds. Philadelphia, PA.
Erickson, M.M. 1938. Territory, annual cycle and numbers in a population of wren-tits
(Chamaea fasciata). University of California Publications in Zoology 42:247-334.
Fortin, M-J, P. Drapeau and GM Jacquez. 1996. Quantification of the spatial co-
occurrences of ecological boundaries. Oikos 77:51-60.
– 93 -
Getis, A. and K. Ord. 1992. The analysis of spatial association by the use of distance
statistics. Geographical Analysis 24:189–206.
Gleason, H.A. 1917. The structure and development of the plant association. Bulletin of
the Torrey Botanical Club 43:463-81. -
Grieg-Smith, P. 1983. Quantitative plant ecology. University of California Press,
Berkeley, CA.
Hill, G.E. 1995. Black-headed Grosbeak (Pheuticus melanocephalus). Birds of North
America 143. A. Poole and F. Gill, Eds. Philadelphia, PA.
Hill, M.O. 1973. The intensity of spatial pattern in plant communities. Journal of Ecology
61:225-235.
Holling, C.S. 1992. Cross-scale morphology, geometry and dynamics of ecosystems.
Ecological Monographs 62:447-502.
Holt, R.D., S.W. Pacala, T.W. Smith and J. Liu. 1995. Linking contemporary vegetation
models with spatially explicit animal population models. Ecological Applications
5:20-27.
Journel, A.G. and C.J. Huijbregts. 1978. Mining Geostatistics. Academic Press, London.
Kareiva, P. and U. Wennergren. 1995. Connecting landscape patterns to ecosystem and
population processes. Nature 373:299-302.
Koenig, W.D. 1998. Spatial Autocorrelation in California Land Birds. Conservation
Biology 12(3):612-620.
Koenig, W.D. and J.M.H. Knops. 1998. Testing for spatial autocorrelation in ecological
studies. 1998. Ecography 21:423-429.
Lawton, J.H., S. Nee, A.J. Letcher and P.H. Harvey. 1994. Animal distributions: patterns
and processes. Pp. 41-58 in Edwards, P.J., R.M. May and N.R. Webb, eds. Large
Scale Ecology and Conservation Biology. Blackwell, Oxford.
Legendre, P. 1993. Spatial autocorrelation: trouble or new paradigm? Ecology 74:1659–
1673.
Legendre, P. and M.-J. Fortin. 1989. Spatial pattern and ecological analysis. Vegetatio
80:107-138. “I
- 94 -
Levin, S.A. 1976. Spatial patterning and the structure of ecological communities. Lectures
on Mathematics in the Life Sciences 81:1-35.
Levin, S.A. 1978. Pattern formation in ecological communities. Pp. 433-465 in J.H.
Steele, ed. Spatial Pattern in Plankton Communities. Plenum, New York.
Levin, S.A. 1992. The problem of pattern and scale in ecology. The Robert H. MacArthur
Award Lecture. Ecology 73:1943-1967.
Levin, S.A., D. Cohen and A. Hastings. 1984. Dispersal strategies in patchy
environments. Theoretical Population Biology 26:165-191.
Levins, R. 1969. Some demographic and genetic consequences of environmental
heterogeneity for biological control. Bulletin of the Entomological Society of
America 15:237-240. -
Littell, R.C., G.A. Milliken, W.W. Stroup, and R. D. Wolfinger. 1996. SAS System for
Mixed Models. Chapter 9: Spatial Variability. SAS Institute Inc., Cary, NC.
MacArthur, R.H. 1972. Geographical Ecology. Harper and Row, New York.
Matheron, G. 1963. Principles of geostatistics. Economic Geology 58:1246-1266.
MathSoft. 1996. S-SpatialStats User's Manual, Version 1.0. MathSoft, Inc., Seattle, WA.
MathSoft. 1997. S-Plus User’s Guide. MathSoft, Inc., Seattle, WA.
MathSoft. 1998. S-Plus for ArcView User’s Guide. MathSoft, Inc., Seattle, WA.
Maurer, B.A. 1990. The relationship between distribution and abundance in a patchy
environment. Oikos 58:181-189.
Maurer, B.A. and S.G. Heywood. 1993. Geographic range fragmentation and abundance
in neotropical migratory birds. Conservation Biology 7:501-509. -
Meisel, J.E. and M.G. Turner. 1998. Scale detection in real and artificial landscapes using
semivariance analysis. Landscape Ecology 13:347-362.
Merriam, C.H. 1898. Life zones and crop zones of the United States. USDA Bulletin 10.
Washington, DC.
Moloney, K.A., A. Morin and S.A. Levin. 1991. Interpreting ecological patterns generated
through simple stochastic processes. Landscape Ecology 5:163–174.
Moran, P.A.P. 1950. Notes on continuous stochastic phenomena. Biometrika 37:17-23.
- 95 -
Neter, J., W. Wasserman and M.H. Kutner. 1989. Applied Linear Regression Models.
R.D. Irwin, Homewood, IL.
O'Neill, R.V., S.J. Turner, V.I. Cullinan, D.P. Coffin, T. Cook, W. Conley, J. Brunt, J.M.
Thomas, M.R. Conley and J. Gosz. 1991. Multiple landscape scales: an intersite
comparison. Landscape Ecology 5:137-144. -
Paine, R.T. and S.A. Levin. 1981. Intertidal landscapes: disturbance and the dynamics of
patches. Ecological Monographs 51:145-178.
Peet, R.K. and O.L. Loucks. 1977. A gradient analysis of southern Wisconsin forests.
Ecology 58:485-499.
Pulliam, H.R., J.B. Dunning and J. Liu. 1992. Population dynamics in complex
landscapes: a case study. Ecological Applications 2:165-177.
Ripley, B.D. 1978. Spectral analysis and the analysis of pattern in plant communities.
Journal of Ecology 66:965–981.
Robertson, G.P. 1987. Geostats in ecology: interpolating with known variance. Ecology
68:744–748. a'
Robertson, G.P., K.M. Klingensmith, M.J. Klug, E.A. Paul, J.C. Crum, and B.G. Ellis.
1997. Soil resources, microbial activity, and primary production across an
agricultural ecosystem. Ecological Applications 7: 158-170.
Root, T.L. and S.H. Schneider. 1995. Ecology and climate: research strategies and
implications. Science 269:334-341.
Root, T.L. 1988. Environmental factors associated with avian distributional boundaries.
Journal of Biogeography 15:489-505.
Rossi, R.E., D.J. Mulla, A.G. Journel and E.H. Franz. 1992. Geostatistical tools for
modeling and interpreting ecological spatial dependence. Ecological Monographs
62:277-314.
SAS. 1997. SAS/STAT” Software: Changes and Enhancements through Release 6.12.
SAS Institute Inc., Cary, NC.
Simpson, G.G. 1964. Species density of North American recent mammals. Systematic
Zoology 13:57-73. -
Small, A. 1974. The Birds of California. Winchester Press, New York, NY.
– 96 -
Soulé, M.E., D.T. Bolger, A.C. Alberts, R. Sauvajot, J. Wright, M. Sorice, and S. Hill.
1988. Reconstructed dynamics of rapid extinctions of chaparral-requiring birds in
urban habitat islands. Conservation Biology 2:75-92.
Sugihara, G. and R.M. May. 1990. Applications of fractals in ecology. Trends in Ecology
and Evolution 5:79-86. -
Swanson, F.J. T.K. Kratz, N. Caine and R.G. Woodmansee. 1988. Landform effects on
ecosystem patterns and processes. BioScience 38:92-98. -
Tarone, R.E. 1988. Score Statistics. Pp. 304-308 in Encyclopedia of Statistical Sciences,
Volume 8. John Wiley & Sons, New York. -
Taylor, L.R. 1984. Assessing and interpreting the spatial distributions of insect
populations. Annual Review of Entomology 20:321-357.
Terborgh, J. 1971. Distribution of environmental gradients: theory and a preliminary
interpretation of distributional patterns in the avifauna of the Cordillera
Vilcabamba, Peru. Ecology 52:23–40.
Tilman, D. 1994. Competition and biodiversity in spatially-structured habitats. Ecology
75:2–16.
Turner, M.G. 1989. Landscape ecology: the effect of pattern on process. Annual Review
of Ecology and Systematics 20:171-197.
Turner, M.G., V.H. Dale and R.H. Gardner. 1989. Predicting across scales: theory
development and testing. Landscape Ecology 3:245-252.
Turner, M.G., S.M. Pearson, W.H. Romme and L.L. Wallace. 1997. Landscape
heterogeneity and ungulate dynamics: what spatial scales are important? Pp. 331-
348 in J.A. Bissonette, ed. Landscape Ecology: a Primer for Wildlife Biologists.
Springer-Verlag, New York.
Turner, M.G., Y. Wu, W.H. Romme, and L.L. Wallace. 1993. A landscape simulation
model of winter foraging by large ungulates. Ecological Modelling 69:163–184.
Villard, M-A. and B.A. Maurer. 1996. Geostatistics as a tool for examining hypothesized
declines in migratory songbirds. Ecology 77:59-68.
Von Humboldt, A. 1807. Ideenzu einer geographie der pflangen nebat einem
naturgemalde der tropenlander. Tubingen.
– 97 -
Ward, D. and D. Saltz. 1994. Foraging at different spatial scales: dorcas gazelles foraging
for lilies in the Negev desert. Ecology 75:45-58. .
Warming, E. 1895. Plantesamfund grundtrak av den okologiska plantegeographi.
Philipsen, Copenhagen.
Watt, A.S. 1947. Pattern and process in the plant community. Journal of Ecology 35:1-22.
Weise, C.M. and J.R. Meyer. 1979. Juvenile dispersal and development of site-fidelity in
the Black-Capped Chickadee. Auk 96:40–55.
Whittaker, R.H. 1956. Vegetation of the Great Smokey Mountains. Ecological
Monographs 26:1-80.
Wiens, J.A. 1981. Scale problems in avian censusing. Studies in Avian Biology No.
- 6:513-521.
Wiens, J.A. 1989. Spatial scaling in ecology. Fünctional Ecology 3:385-3 97. -
Yaglom, A.M. 1957. Some classes of random fields in n-dimensional space, related to
stationary random processes. Theory of Probability and its Applications 2:273-
320.
– 98 -
Table 2.1. Non-spatial and detrended models for bird abundance. 3 is the regression
coefficient for the effect of urbanization proportion within a surrounding radius (500m for
Wrentit; 1000m for Black-headed Grosbeak; 2000m for Northern Mockingbird; 4000m
for Blue-gray Gnatcatcher). -
OLS model: Trend model Trend model: bird abundance /
bird abundance residuals - urbanization measures -
~ urbanization urbanization x,y coordinates
InCaSUITCS InCaSUTCS (x = east/west, y =north/south)
Wrentit
B+ S.E. –7.83 + 1.41 -3.82 + 1.24 Significant effects x, y, x*y
(p-value) (<0.0001) (0.003) (p<0.05) -
F-statistic 30.61 9.43 F-statistic 18.36
(p-value) (<0.0001) (0.003) (p-value) (<0.0001)
R2 0.23 0.08 R2 0.35
Northern
Mockingbird
3 + S.E. 3.42 + 0.40 1.65 + 0.43 Significant effects x, y, x*y
(p-value) (<0.0001) (<0.0001) (p<0.05)
F-statistic 72.05 14.89 F-statistic 10.67
(p-value) (<0.0001) (<0.0001) (p-value) (<0.0001)
R” 0.41 0.12 R2 0.24
Black-headed
Grosbeak
B+ S.E. -1.98 + 0.48 –0.97+ 0.47 Significant effects tº gº
(p-value) (<0.0001) (0.042) (p<0.05)
F-statistic 17.33 4.23 F-statistic 4.57
(p-value) (<0.0001) (0.042) (p-value) (0.005)
R2 0.14 0.04 R2 0.12
Blue-gray
Gnatcatcher
B+ S.E. –0.89 + 0.24 –0.04 + 0.22 Significant effects x, y, x*y
(p-value) (<0.0001) (0.86) (p<0.05)
F-statistic 14.31 0.03 F-statistic 11.49
(p-value) (<0.0001) (0.86) (p-value) (<0.0001)
R” 0.12 0.00 R2 0.25
- 99 -
Table 2.2. Non-spatial vs. spatial covariance models for Bird Abundance. 3 is the
regression coefficient for the effect of urbanization proportion within a surrounding radius
(500m for Wrentit; 2000m for Northern Mockingbird; 4000m for Blue-gray Gnatcatcher).
Non-Spatial Model Spatial Covariance Models
Wrentit -
B+ S.E. –7.83 + 1.41 –6.16 = 1.78 –4.93 + 1.82
(p-value) (<0.0001) (0.0008) (0.0078)
-2 Log Likelihood 400.93 384.34 382.59
Range (p) + S.E. 2000m 7000m
Nugget (olº) + S.E. 1.28 + 0.51 1.77 -- 0.81
Partial Sill (o,”) + S.E. 1.45 + 0.29 1.65 + 0.27
Model Likelihood Ratio 16.59 18.34
Chi-Square (p-value) (0.0009) (0.0004)
Northern Mockingbird
3 + S.E. 3.43 + 0.41 3.23 + 0.74 2.96 + 0.90
(p-value) (<0.0001) (<0.0001) (0.0014)
-2 Log Likelihood 241.85 204.84 210.39
Range (p) + S.E. 2000m 9000m
Nugget (o,”) + S.E. 0.69 + 0.22 2.30 + 0.87
Partial Sill (o,”) + S.E. 0.10 + 0.06 0.14 + 0.06
Model Likelihood Ratio 37.01 31.45
Chi-Square (p-value) (<0.0001) (<0.0001)
Blue-gray Gnatcatcher
3 + S.E. –0.89 + 0.23 -0.59 + 0.44
(p-value) (0.0003) (0.19)
–2 Log Likelihood 169.94 161.68
Range (p) + S.E. 4000m
Nugget (o,”) + S.E. 0.09 + 0.06
Partial Sill (o,”) + S.E. 0.22 + 0.04
Model Likelihood Ratio 8.27
Chi-Square (p-value) (0.016)
- 100 -
semivar
Q9.
Cº. -
: sill
Ç9.
49.
$0. * range
2 º'
10.
eſnugget
Ö Fo- —F.— —r
distance (h)

Figure 2.1. Idealized Semivariogram Function (from Littell et al. 1996).
exponential spherical
o
so *
© :
e
©
13 g
©
c
ă. ă
©
•
©
cN .
G-> GN
o
2 4. 6 8 10 12 14 2 4. 6 8
distance distance
gaussian linear
e
sº
co
©
º co
# s g
ă •. ă º
C
§ -
o
tº —w- ~~ -v- —w- wr
2 4 6 8 10 12 $4 2 4. 6 8
distance distance




Figure 2.2. Theoretical Semivariogram Models (from MathSoft 1996).
s
Max. Abundance
1 - 5 N 3000 0 3000 6000
a 6 - 7 ====
• 8 - 10 Meters
Figure 2.3a. Wrentit abundance. Maximum of two counts (1997). 107 sampling locations.
º
º º
º rºtº- º
§: --
- º º
#. I. º
º
d
º rº-tº-
sº
º
-
-
-
º º
- - -
- - --. - -
--- Fºſº -
ºś
Residential
Commercial/Industrial
Open Space and Recreation
Water












s
º º S. - -
Sºº-
º:
-
-
º:
Nº.
-------
º º
ºf ººº-ºº:
ºś º º
Lº-
º º
Fº
º
:
º,
º
S㺠-
º
º:
º
* …ºl lºº
zºº.
_º. sº - º
sº º º:
| º º
º
º |- º
tºº-ºº:
- º -º-º: -
- ! ºft .º --- T
lº. - -
º º
º
- º
º
º —-ºº. -
º Fºº-ºº-ºº: Hºſº º
--~~~~ ºlºtººs
T Commercial/Industrial
Open Space and Recreation
Meters | Water
3000 O 3000 6000 Residential
----
Figure 2.3b. Northern Mockingbird Abundance. Maximum of two counts (1997). 107 Sampling locations.










































































º w º
º: º
º º - º
- sº. - º
Rºsº -----
º *
º
* * *
iłſº º : º: -- º
º º
º
3 L º º -
- º º
- - - º º
- K º º -- º -
Max. abundance - -
0 N 3000 0 3000 6000 Residential
º 1
• 2 - 3
Commercial/Industrial
Open Space and Recreation
Meters Water
Figure 2.3c. Black-headed Grosbeak Abundance. Maximum of two counts (1997). 107 sampling locations




#
Max. Abundance
3000 0 3000 º
O
e 1-2
Residential
6000 Commercial/Industrial
Open Space and Recreation
Meters Water
Figure 2.3d. Blue-gray Gnatcatcher Abundance. Maximum of two counts (1997). 107 Sampling locations.


- 105-

0.
1
(a) Urbanization – Omnidirectional
0.
O8
0.
06
0.
O4
O
02
Jº
O
|al
o o P
n D
O
0 2000 4000 6000 axo
10000 120CO 14000
lag distance in meters(h)
(b) Urbanization - East/West (c) Urbanization - North/South
O. 1 O.1
* 0.08 a * * 0.08
£ E
s E
§ 0.06 gº § 0.06
© © -
# º #
# 0.04 lºſ = 0.04
& º sº
2 º # D D
ă 0.02 P—o-c 3. 0.02 Taſ U
tº D º O
º (º,
0.12–5– —I- & s s I 0 le—t--- ſ T
O 2000 4000 6000 8000 10000 12000 14000 O 2000 4000 6000 8000 10000 12000 14000
lag distance in meters(h) lag distance in meters (h)
Figure 2.4. Empirical Semivariograms, Y(h), for Urbanization Proportion (2000m radius).
East/West = 90° azimuth, 45° tolerance; North/South = 0° azimuth, 45° tolerance. Lag
distance values represent averages distances between pairs of points within each 1000m
distance class."
* Sample sizes for each distance class in omnidirectional semivariograms:
Avg.dist. 361 1044 2038 3010 4024 5011 5980 7013 7950 8987 9989 10939 1 1972 13041
# Pairs 72 282 357 567 580 582 504 388 287 158 158 151 110 137
East/West:
Avg.dist. 379 1060 2033 3002 4024 5010 5994 7020 7973 9021 9989 10939 11972 13041
# Pairs 31 164 166 297 322 332 292 256 199 130 158 151 110 137
North/South: -
Avg.dist. 348 1021 2043 3018 4025 5012 5960 6999 7898 8832
# Pairs 41 118 191 270 258 250 212 132 88 28
- 106 - .
(a) Wrentit-Omnidirectional (d) Northern Mockingbird – Omnidirectional
2
gºe gº D
65 63
E
E # 15
§ § º
8 8 1 -->
5 5
F # D D []
2 ..? O.5 ſº O Os,
E E a * D
8 3 tº
O I——r— y r T r
O 2000 4000 6000 8000 10000 12000 14000 O 2000 4000 6000 8000 10000 12000 14000
lag distance in meters(h) lag distance in meters(h)
(b) Wrentit—East/West (e) Northern Mockingbird-EastWest
2
º
º 7
5
# i 1.5
© g 1 _ſº []
: F . º ºr
# F C, O
ºme .2 [] ſº f = }
E E 0.5 D D ſºl º lºſt Tºſ
§ §
O —r- r —I- r ¥ T
O 2000 4000 6000 8000 10000 12000 14000 O 2000 4000 6000 8000 10000 12000 14000
lag distance in meters (h) lag distance in meters (h)
(c) Wrentit--NorthSouth (f) Northern Mockingbird-North/South
2
g g
E # 15
§ §
[]
g § 1 º,
º &
E E 0.5 [º]
8 8 O [s]
o |P
O 2000 4000 6000 8000 10000 12000 14000 O 2OOO 4000 6000 8000 10000 12000 14000
lag distance in meters (h) lag distance in meters(h)
Figure 2.5. Empirical Semivariograms, Y(h), for Maximum 1997 Bird Abundance.
East/West = 90° azimuth, 45° tolerance; North/South = 0° azimuth, 45° tolerance. Lag
distance values represent averages distances between pairs of points within each 1000m
distance class."



- 107 -
(g) Black-headed Grosbeak--Omnidirectional
lag distance in meters (h)
1 0.5
# 0.8 # 04
E E
§ 0.6 § 03
© 8
# ſº
3 0.4 # 02
g §
E. O.2 E 0.1
§ §
O O
O 2000 4000 6000 8000 10000 12000 14000 O
(j) Blue-gray Gnatcatcher — Omnidirectional
2000 4000 6000 8000 10000 12000
lag distance in meters(h)
14000
(h) Black-headed Grosbeak – EastWest
Tº 'º
E E
E 5
§ Se
3 #
# #
tº; &
-> 2
E E
8 §
O 2000 4000 6000 8000 10000 12000 14000 0
lag distance in meters(h)
(k) Blue-gray Gnatcatcher - East/West
2000 4000 6000 8000 10000 12000
lag distance in meters (h)
14000
(i) Black-headed Grosbeak-North/South
(I) Blue-gray Gnatcatcher-North/South
º Tº
E V. E
E E
§ 0. §
3 §
É #
83 º
O 2000 4000 6000 8000 10000 120CO 14000 O 2000 4000 6000 8000 10000 12000 14000
lag distance in meters(h) lag distance in meters (h)
Figure 2.5. (continued)
* Sample sizes for each distance class in omnidirectional semivariograms:
Avg.dist. 361 1044 2038 3010 4024 5011 5980 7013 7950 8987 9989 10939 11972 13041
# Pairs 72 282 357 567 580 582 504 388 287 158 158 151 110 137
East/West:
Avg.dist. 379 1060 2033 3002 4024 5010 5994 7020 7973 9021 9989 10939 11972 13041
# Pairs 31 164 166 297 322 332 292 256 199 130 158 151 1 10 137
North/South:
Avg.dist. 348 1021 2043 3018 4025 5012 5960 6999 7898 8832
# Pairs 41 118 191 270 258 250 212 132 88 28






- 108 -

0.8 -
0.6 .
0.4 -
0.2 -
i
O
–0.4 -
-0.6 -
-0.8 -
-1
r —r- I G TT n I-
-0.29 2000 4000 6000 8300 ° 10000 - 12000 14000
lag distance in meters (h)
Figure 2.6. Moran’s I correlogram for Urbanization Proportion (2000m radius). Shaded
squares represent significant (p<0.05) autocorrelation. Lag distance values represent
maximum distances between pairs of points within each 1000m distance class.”
** Sample sizes for each distance class:
Maximum
distance 1000 2000 3000 4000
Number of 198 317 480 556
5000 6000 7000 8000 9000
10000 11000 12000 13000
Pairs
583 577 403 384 211
148 162 120 120
- 109 -

lag distance in meters (h)
(a) Wrentit (b) Northern Mockingbird
0.8 0.8
0.6 - os.
0.4 - O.4 -
º *
§ 02 | " . . . . § 02:
C O C [.
# O I —r —º-e g [...] —r |É O T *—P—a-º. I I
§ 2000 4000 sooo sooo 10000 1300, 14:00 b 2000 4000 effo sooo P.10800° 1300 tºo
22. -02:
-0.4 -0.4
lag distance in meters (h) lag distance in meters (h)
(c) Black-headed Grosbeak (d) Blue-Gray Gnatcatcher
0.8 O.8
06. 0.6 -
0.4 – 0.4 -
º º
; 0.2 - 5 0.2 . * =
> O tº tº D D D $: O —fºl - * o al
I w T -I 3. T I- *—U-UI —I I O
* 2000 4000 sooo &o * 1900 12000 1400 $ 2000 4000 sooo - sºoo toooo 12:00° 14000
-02 . -0.2 -
–0.4 -0.4
lag distance in meters (h)
Figure 2.7. Moran’s I correlograms for 1997 Maximum Bird Abundance. Shaded squares
represent significant (p<0.05) autocorrelation. Lag distance values represent maximum
distances between pairs of points within each 1000m distance class."
** Sample sizes for each distance class:
Maximum
distance 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 1 1000 12000 13000
Number of 198 317 480 556 583 577 403 384 211 148 162 120 - 120
Pairs
- 110 -

(a) Wrentit-Omnidirectional (d) Northern Mockingbird - Omnidirectional
1.
5
O
5
ſel
tº
C
_ſal
CJ
O
[J-
ſº
O
O 2000 4000 6000 8000 10000 12000 14000 O 2000 4000 6000 8000 10000 12000 14000
lag distance in meters (h) lag distance in meters(h)
(b)wrentit—Eastwest (e) Northern Mockingbird-East/West
1.5
0.5 as ſº {e}–
[s]
al
tº
[s]
º
1
o 2000 4000 6000 8000 10000 12000 14000 O 2000 4000 6000 8000 10000 12000 14000
lag distance in meters(h) lag distance in meters
(c) Wrentit-North/South (f) Northern Mockingbird - North/South
º
o 2000 4000 sooo sooo too 12000 14000 O 2000 4000 6000 8000 10000 12000 14000
lag distance in meters(h) lag distance in meters(h)
Figure 2.8. Empirical Semivariograms, Y(h), for Non-Spatial Urbanization Model
Residuals. East/West = 90° azimuth, 45° tolerance; North/South = 0° azimuth, 45°
tolerance. Lag distance values represent averages distances between pairs of points within
each 1000m distance class."
- 111 -
(g) Black-headed Grosbeak-Omnidirectional ()
2000 4000 6000 8000 10000
lag distance in meters (h)
12000 14000
Blue-gray Gnatcatcher - Omnidirectional
2000 4000 6000 8000 10000 14000
lag distance in meters (h)
12000
(h) Black-headed Grosbeak – EastWest
2000 4000 6000 8000 10000 14000
lag distance in meters(h)
12000
(k) Blue-gray Gnatcatcher — EastWest
2000 4000 6000 8000 10000
lag distance in meters (h)
12000 14000
(i) Black-headed Grosbeak - North/South
(ſ) Blue-gray Gnatcatcher-North/South
O 2000 4000 6000 8000 10000 12000 14000 O 2000 4000 6000 8000 10000 12000 14000
lag distance in meters (h) lag distance in meters (h)
Figure 2.8. (continued)
* Sample sizes for each distance class in omnidirectional semivariograms:
Avg.dist. 361 1044 2038 3010 4024 5011 5980 7013 7950 8987 9.989 10939 1 1972 13041
# Pairs 72 282 357 S67 580 582 504 388 287 158 158 151 1 10 137
East/West:
Avg.dist. 379 1060 2033 3002 4024 5010 5994 7020 7973 9021 9989 10939 1 1972 13041
# Pairs 31 164 166 297 322 332 292 256 199 130 158 151 110 137
North/South:
Avg.dist. 348 1021 2043 3018 4025 5012 5960 6999 7898 8832
# Pairs 41 118 191 258 250 212 132 88 28
270






- 112 -

(a) Wrentit (b) Northern Mockingbird
0.8 O.8
0.6 - 0.6 -
0.4 . 0.4 -
.* so
§ 02: 5 0.2 -
# O * e o * o O D # O tº ſº O.
I O —r —I- i º I n g I I– y I-
© 2000 4000 6000 = 8&O 10000 12000 o 14000 b 2000 ° 4000 P 6000 °sho o 10000 12000 a 14boo
-0.2 Q -0.2 -
[D
-0.4 -0.4
lag distance in meters (h) lag distance in meters (h)
(c) Black-headed Grosbeak (d) Blue-gray Gnatcatcher
O.8 O.8
0.6 - 0.6 -
0.4 – 0.4 -
º, g
5 0.2 - 5 0.2 + s =
# O ſº º O # O ſº [...]
tº-w—o-r—tº-ºr- –3– I N - f Tej I *—e—E-E
() 2000 4000 600 ° sooo “ logo 12000 G 14000 0 2000 4500 P 6000 s 8óbo 10000 12000 14000
-0.2 - - -0.2 :
-0.4 -0.4
lag distance in meters (h) lag distance in meters (h)
Figure 2.9. Moran’s I correlograms for Non-Spatial Urbanization Model Residuals.
Shaded squares represent significant (p<0.05) autocorrelation. Lag distance values
represent maximum distances between pairs of points within each 1000m distance class."
** Sample sizes for each distance class:
Maximum
distance 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000 12000 13000
Number of 198 317 480 556 583 577 403 384 211 148 162 120
120
Pairs -
- 113 -

(a) Wrentit-Omnidirectional
(d) Northern Mockingbird-Omnidirectional
2
# = 1.5
§ # [s]
g § 1
5 - a P - P - * ~ * 5 G tº tº
+ ** } * > Twº y wºme, T ~ Tºmº Tº J
§ º º 2 G p P D
E º tº # 0.5 ºl
§ O G D
—I T I g I T - O J g I- º T I
2000 4000 6000 8000 10000 12000 14000 2000 4000 6000 8000 10000 12000 14000
lag distance in meters (h) lag distance in meters(h)
(b) Wrentit—EastWest (e) Northern Mockingbird-EastWest
2
g i 1.5
ă Se [...]
g [...] 3 1 Uſ
5 Ç— ºl ſel C [...] [º G 5 º D
§ º º [s] § ſº º ſº º ſº º
E D º E 0.5 DTD & O
§ §
I —I- i t w n O T —T I † I —I
2000 4000 6000 8000 10000 12000 14000 2000 4000 6000 8000 10000 12000 14000
lag distance in meters(h) lag distance in meters(h)
(c) Wrentit-North/South (f) Northern Mockingbird – North/South
2
† º
E E 1.5
ă E
Se § [s]
© º º 1
t £3 “E D
g d g º
E D E 0.5 n p B
§ § o a 5
I- —r n w I I O —I- r —, y I I
2000 4000 6000 8000 10000 12000 14000 2000 4000 6000 8000 10000 12000 14000
lag distance in meters (h) lag distance in meters (h)
Figure 2.10. Empirical Semivariograms, Y(h), for Detrended Urbanization Model
Residuals. East/West = 90° azimuth, 45° tolerance; North/South = 0° azimuth, 45°
tolerance. Lag distance values represent averages distances between pairs of points within
each 1000m distance class."
- 114 -

(g) Black-headed Grosbeak-Omnidirectional
(j) Blue-gray Gnatcatcher-Omnidirectional
1 O.5
* 0.8 Tº 0.4
E E
# os [...] # os
co V. E} - Cº V. ſºn — ſº —iº
* º, D d TD * Tº Tº Tºy
is." º ... • * # , " : * c q c
‘E ſº t sº
s s º Ul
# 02 ă O.1 e—a–
O —T- i I- n I s O T t —I- I r I
O 2000 4000 6000 8000 10000 12000 14000 O 2000 4000 6000 8000 10000 12000 14000
lag distance in meters (h) lag distance in meters(h)
(h) Black-headed Grosbeak-EastWest (k) Blue-gray Gnatcatcher-EastWest
1 0.5
* 0.8 * 0.4
E E
5 [s]
# 0.6 U TUTU- tal § 0.3 R—u—
g ſº d P º o P 3 D D D
ă o4 |P 5 D 5 O2 [º] al
E ſº
# 02 # 0.1 e—P
O I T -T- I i I O i t T I —H i
O 2000 4000 6000 8000 10000 12000 14000 O 2000 4000 6000 8000 10000 12000 14000
lag distance in meters(h) lag distance in meters(h)
(i) Black-headed Grosbeak – North/South (l) Blue-gray Gnatcatcher - North/South
1 0.5
79 0.8 Tº 0.4
£ E
# D 5 D
tº 0.6 * —D— Sº 0.3 UJ ſº {=} -E}
o O D go [...]
# D D D # o P D O
# 0.4 ſº # O.2
2 .2
ă 0.2 ă 0.1
O I I T I I y O y i —I- f I- I
O 2000 4000 6000 8000 10000 12000 14000 O 2000 4000 6000 8000 10000 12000 14000
lag distance in meters (h) lag distance in meters(h)
Figure 2.10. (continued)
* Sample sizes for each distance class in omnidirectional semivariograms:
Avg.dist. 361 1044 2038 3010 4024 5011 5980 7013 7950 8987 9989 10939 i 1972 13041
# Pairs 72 282 357 567 580 582 504 388 287 158 158 151 110 137
East/West:
Avg.dist. 379 1060 2033 3002 4024 5010 5994 7020 7973 9021 9989 10939 1 1972 13041
# Pairs 31 164 166 297 322 332 292 256 199 130 158 151 110 137
North/South:
Avg.dist. 348 1021 2043 3018 4025 5012 5960 6999 7898 8832
# Pairs 41 118 191 270 258 250 212 132 88 28
- 115 -

(a) Wrentit (b) Northern Mockingbird
0.8 0.8
0.6 - 0.6 -
[…]
0.4 – 0.4 -
º p
5 0.2 - § 02:
º C
# 0 –F–F–a–s—º-s—a-º-º-o- # O —: r—e—ſº T W
() 2000 D 4000 6000 8600 10000 12OOO 14000 2000 ° 4000 ° spo 8000 ° 10000° (2000 1400
–0.2 – –0.2 - O
-0.4 -0.4
lag distance in meters (h) lag distance in meters (h)
(d) Blue-gray Gnatcatcher (c) Black-headed Grosbeak
0.8 0.8
0.6 - 0.6 -
0.4 - 0.4 -
..so º,
5 0.2 - # 02.
S # Dr.
-E º : º
O O fººl ºl tº –º- O ſº G fº º Jºl.
i º —r [s] n U —r- D º lºsſ \ºſT T n O ki T T
0 2000 4000 ° 6000 8000 10000 12000 14000 ( 2000 4000 6000 a 8000 G 10000 12000 14too
-0.2 – –0.2 - -
O
-0.4 -0.4
lag distance in meters (h) - lag distance in meters (h)
Figure 2.11. Moran’s I correlograms for Detrended Urbanization Model Residuals.
Shaded squares represent significant (p<0.05) autocorrelation. Lag distance values
represent maximum distances between pairs of points within each 1000m distance class."
** Sample sizes for each distance class:
Maximum
distance 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000 12000 13000
Number of 198 317 480 556 583 577 403 384 211 148 162 120 120
Pairs
§ - N
º， º
∞
ºſ
Millin
|
*：2.
1. º~
Z J. （...
º ：



º
* * * * * * *
* * * * * * * * * * *
* * tº . * * *
* * * * * * * º
* : * * * * * * * * * * *
a º * * * * d
at we w w w tº
tº a tº a tº
s & º £ tº
* * * * : * : * : * * * * * * *
º, wº wºº a .
A : * * * *
tº a w
as , ,
º
* * * * * * * * * ,
* * * * * * * B = < *
A tº
* , t , a w tº a , i.
tº it. : * : * ~ * R", , yº."
A, R & sº t, sº
ſ *** * :
|
:
º
8
.
i
i
i
:
:
:
.
|
º
|
:
:
º
M :
:
º i
:
{
:
Hº:
:
#
º
f
D
ºf a , wº - a
* - U º
#
º
. d
cº-3-. wº
* { * {
tº
:
Vº
:
sº