Utah State University DigitalCommons@USU
All Graduate Theses and Dissertations Graduate Studies
12-2017
Urban Bobcat (Lynx rufus) Ecology in the Dallas-Fort Worth, Texas Metroplex
Julie M. Golla Utah State University
Follow this and additional works at: https://digitalcommons.usu.edu/etd
Part of the Other Ecology and Evolutionary Biology Commons
Recommended Citation Golla, Julie M., "Urban Bobcat (Lynx rufus) Ecology in the Dallas-Fort Worth, Texas Metroplex" (2017). All Graduate Theses and Dissertations. 6857. https://digitalcommons.usu.edu/etd/6857
This Thesis is brought to you for free and open access by the Graduate Studies at DigitalCommons@USU. It has been accepted for inclusion in All Graduate Theses and Dissertations by an authorized administrator of DigitalCommons@USU. For more information, please contact [email protected]. URBAN BOBCAT (LYNX RUFUS) ECOLOGY IN THE
DALLAS-FORT WORTH, TEXAS METROPLEX
by
Julie M. Golla
A thesis submitted in partial fulfillment of the requirements for the degree
of
MASTER OF SCIENCE
in
Wildlife Biology
Approved:
______Julie K. Young Kimberly A. Sullivan Major Professor Committee Member
______Thomas C. Edwards Mark R. McLellan, Ph.D. Committee Member Vice President for Research and Dean of the School of Graduate Studies
UTAH STATE UNIVERSITY Logan, Utah
2017 ii
Copyright © Julie M. Golla 2017
All Rights Reserved
iii ABSTRACT
Urban Bobcat (Lynx rufus) Ecology in the Dallas-Fort Worth, Texas Metroplex
by
Julie M. Golla, Master of Science
Utah State University, 2017
Major Professor: Julie K. Young Department: Wildland Resources
Wildlife and people increasingly overlap in their use of space and resources as
cities rapidly expand, often increasing rates of human-wildlife conflict. Urban carnivores are of special concern because of their potential effects on human health, well-being, and livelihoods. Bobcats (Lynx rufus) are a top predator in several urban areas across the
United States and a potential contributor to human-carnivore conflicts. This study evaluated the movements and habitat utilization of bobcats in a highly urbanized area of the Dallas-Fort Worth (DFW), Texas metroplex. Spatial data were collected from 10 bobcats via Global Positioning Satellite (GPS) for approximately one year. Average home range size using kernel density estimators was 4.60 km2 (n=9, SE=0.99 km2) for all
resident bobcats, 3.48 km2 (n=5, SE=1.13 km2) for resident females, and 6.00 km2 (n=4,
SE=1.61 km2) for resident males. Resource selection function (RSF) models showed
bobcats vary their selection of habitats such as creeks, grass, and low-medium
development in a non-linear way. Bobcats avoided areas close to and far from grasslands
and low-medium development but selected for these areas at intermediate distances. iv Bobcats also selected areas closer to developed open space, agricultural areas, and railroads. In addition, camera trap data analyzed with maximum likelihood spatially
explicit capture-recapture (SECR) models informed by the RSF results estimated a
population density of 0.64 bobcats/km2 (SE = 0.22). Bobcats in DFW have significantly
smaller home ranges and occur at higher densities compared to rural bobcat populations.
Home ranges were also slightly smaller and densities higher than the most closely similar
peri-urban bobcat studies. These differences likely arise due to the abundant urban prey
species the DFW landscape provides despite limited space and habitat for bobcats. The
dense urban development surrounding this population of bobcats may also discourage
dispersing from the area, contributing to higher densities. These results provide
information to facilitate management of urban bobcats by providing new insight into how
bobcats live amidst people in an anthropogenic-driven ecosystem.
(67 pages) v PUBLIC ABSTRACT
Urban Bobcat (Lynx rufus) Ecology in the Dallas-Fort Worth, Texas Metroplex
Julie M. Golla
Urban landscapes are quickly replacing native habitat around the world. As wildlife and people increasingly overlap in their shared space and resources, so does the potential for human-wildlife conflict, especially with predators. Bobcats (Lynx rufus) are
a top predator in several urban areas across the United States and a potential contributor
to human-carnivore conflicts. This study evaluated the movements and habitat use of
bobcats in the Dallas-Fort Worth (DFW), Texas metroplex. Spatial data were collected
from 10 bobcats via Global Positioning Satellite (GPS) for approximately one year.
Average home range size was 4.60 km2 (n=9, SE=0.99 km2) for all resident bobcats, 3.48
km2 (n=5, SE=1.13 km2) for resident females, and 6.00 km2 (n=4, SE=1.61 km2) for
resident males. Resource selection function (RSF) models show that bobcats avoid areas
close to and far from grasslands and low-medium development, while selecting for these
areas at intermediate distances. Bobcats also selected areas closer to developed open
space, agricultural areas, and railroads. In addition, camera trap data analyzed with
spatially explicit capture-recapture (SECR) models informed by the RSF results
estimated a population density of 0.64 bobcats/km2 (SE = 0.22). Bobcats in DFW have
significantly smaller home ranges and occur at higher densities compared to rural bobcat
populations. Home ranges were also slightly smaller and densities higher than the most
closely similar peri-urban bobcat studies. These differences likely arise due to the
abundant urban prey species the DFW landscape provides despite limited space and vi habitat for bobcats. The dense urban development surrounding this population of bobcats may also discourage dispersing from the area, and contributing to higher densities. These results provide information to facilitate management of urban bobcats by providing new insight into how bobcats live amidst people in urban areas.
vii ACKNOWLEDGMENTS
I would like to acknowledge the Welder Wildlife Foundation and Terry
Blankenship for providing funding for this research and my graduate degree. I would also
like to thank Utah State University for providing coursework, travel funds, and academic
resources that assisted me in accomplishing my thesis goals. A great many thanks to my
major advisor, Julie K. Young, in providing essential guidance, patience, and funding in
organizing and executing this project. Thank you to my committee members, Thomas
Edwards, Jr. and Kimberly Sullivan for their support, input, and encouragement. Thank
you to the Texas Parks and Wildlife Department biologists Richard Heilbrun, Derek
Broman, and Brett Johnson who provided support and guidance during field work in
Dallas, Texas. Thank you to Rob Denkhaus and the Fort Worth Nature Center staff in
providing their time and resources in fixing traps and sharing many helpful conversations
regarding urban ecology. Thank you to all of my volunteers, both in the field and in the
office, in helping me with my enormous work load and sharing their time and good spirits with me.
Thank you to my friends and family who offered limitless love and support
throughout the ups and downs of completing this thesis. Above all of the skills and work
involved in completing this degree, their role in my success was of the upmost
importance.
Julie M. Golla
viii CONTENTS
Page
ABSTRACT ...... iii
PUBLIC ABSTRACT ...... v
ACKNOWLEDGMENTS ...... vii
LIST OF TABLES ...... x
LIST OF FIGURES ...... xi
INTRODUCTION ...... 1
METHODS ...... 4
Study Site ...... 4 Data Collection ...... 4
Captures ...... 4 GPS Collars ...... 5 Cameras ...... 6
Data Analyses ...... 7
Home Ranges ...... 7 Resource Selection Function Model ...... 8 Spatially Explicit Capture-Recapture ...... 11
RESULTS ...... 13
Data Collection ...... 13
Captures ...... 13 GPS Collars ...... 13 Cameras ...... 13
Data Analyses ...... 14
Home Ranges ...... 14 Resource Selection Function Models ...... 15 Spatially Explicit Capture-Recapture Model ...... 16 ix DISCUSSION ...... 17
LITERATURE CITED ...... 23
TABLES AND FIGURES ...... 30
APPENDICES ...... 45
Appendix 1 ...... 46 Appendix 2 ...... 49
x LIST OF TABLES
Table Page
1 Duration of camera station activity from 8 June – 15 November, 2014 ...... 30
2 Duration each bobcat was fitted with a GPS collar within the DFW study site between January 2014 and April 2015 ...... 31
3 Fix rate success of GPS collars fitted to bobcats in the DFW study site. DOP < 8 was used to define acceptable, accurate locations ...... 32
4 Variance inflation of initial habitat covariate classifications (minus forest) for RSF analysis as calculated by package ‘usdm’ in program R ...... 33
5 Variance inflation of second habitat covariate classifications (minus forest) for RSF analysis as calculated by package ‘usdm’ in program R ...... 34
6 Variance inflation of final habitat covariate classifications (minus forest and wetland, with low and medium development combined as LowMed Dev) for RSF analysis as calculated by package ‘usdm’ in program R ...... 35
7 Top models of <3 delta AICc from initial RSF analysis ...... 36
8 Final top models for DFW bobcat RSF with polynomials addressing non-linearities ...... 37
9 Parameter estimations for the effective survey area from the top SECR model l ...... 38
10 Top model coefficients for individual bobcat RSFs ...... 50 xi LIST OF FIGURES
Figure Page
1 The Dallas-Fort Worth Metroplex, Texas ...... 39
2 The DFW urban bobcat study site and capture locations during GPS collaring efforts from January 2014 to March 2015 ...... 40
3 DFW bobcat camera survey from June – November 2014 ...... 41
4 (a) 95% kernel density home range estimation and (b) 99% local convex hulls for 10 collared bobcats in DFW metroplex ...... 42
5 Population-based bobcat habitat selection probability throughout the DFW study site ...... 43
6 SECR surface density map defining bobcat density per habitat mask point in the DFW metroplex study site ...... 44
7 Pair-wise comparison chart of 11 initial habitat covariates considered for DFW bobcat RSF ...... 46
8 Pair-wise comparison chart second habitat covariate classification, excluding forest, considered for DFW bobcat RSF ...... 47
9 Pair-wise comparison chart for final habitat covariates considered for DFW bobcat RSF, excluding forest and with low and medium development condensed ...... 48
10 Individual bobcat heat maps from DFW ...... 51
11 Population heat map generated by averaging individual bobcat RSFs ...... 56
INTRODUCTION
Sixty percent of the 8.5 billion people on earth will reside in urban areas by 2030
(United Nations 2015). Widespread urbanization directly affects native wildlife species
as encroaching development removes native habitat, introduces invasive species, and
brings people and wildlife into direct competition for space and resources (Adams 2005,
Magle et al. 2012). Carnivores are particularly susceptible to challenges brought on by
rapidly changing urban landscapes, often resulting in increased human-carnivore conflict
and negative effects on human health, well-being, and livelihoods (Woodroffe 2000,
Treves and Karanth 2003, Gehrt et al. 2010, Athreya et al. 2013). As top predators utilize anthropogenic landscapes, unique challenges in management and conservation develop
(Batemen and Fleming 2012). There is a need to understand how carnivores persist among people in urban ecosystems to facilitate coexistence and manage the landscape to the benefit of both (Athreya et al. 2013).
The emergence of carnivores in urban areas has led to surprising animal behavior,
some perceived as optimistic for species survival, others foreboding for both wildlife and
human well-being (Woodroffe 2000). In an era where species conservation and recovery
is applauded, the social and functional carrying capacity for large carnivores is being
tested at the urban interface where conflict potential with livestock, pets, and people is
increasing (Woodroffe 2000). There is a need for improved understanding and adaptive
wildlife management and conservation amidst city landscapes.
Some of the most densely human populated places in the world are also home to
thriving populations of carnivores. In India, there are established populations of leopards
(Panthera pardus) and striped hyena (Hyanena hyaena) amidst densely populated rural 2 and urban areas (Singh et al. 2010, Athreya et al. 2013, Bhatia et al. 2013). Spotted hyenas benefit from increased human activity and presence (Boydston et al. 2003), and peri-urban spotted hyenas (Crocuta crocuta) live entirely on domestic prey species and
scavenge anthropogenic scraps in Ethiopia (Abay et al. 2011). Mountain lions (Puma
concolor) in the western United States persist in patches of highly fragmented urban
landscapes (Gehrt et al. 2010, Kertson et al. 2011, Knopff et al. 2014). Coyotes (Canis
latrans) are well documented in urban habitats across North America and are considered
one of the most urban-adaptable carnivore species (Quinn 1997, Farrar 2007, Gehrt et al.
2009, Grubbs and Krausman 2009, Gehrt et al. 2010, Gese et al. 2012). Bobcats (Lynx
rufus) have also been shown to persist in suburban and fragmented developed areas
(Tigas et al. 2002, Gehrt et al. 2010, Ordenana et al. 2010, Riley 2003, Alonso et al.
2012, Ruell et al. 2012), though few urban areas support this disturbance-sensitive
species (Crooks 2002).
Most carnivore species that are successful in urban environments are actually
omnivorous generalists, a life history that allows them to persist on a wide variety of food
sources and habitat, including abundant anthropogenic resources in urban areas (Riley
2003, Baker and Harris 2007). In North America, for instance, coyotes, black bears
(Ursus americanus), raccoons (Procyon lotor), striped skunks (Mephitis mephitis),
Virginia opossums (Didelphis virginiana), and red fox (Vulpes vulpes) are commonly
found in urban areas. Excluding black bears and coyotes, these species have relatively
smaller body size and smaller resource and space requirements compared to other
carnivores (Baker and Harris 2007). Black bears are omnivorous and capable of finding a
near constant food supply by scavenging garbage with little foraging effort to support 3 their larger body size in urban areas (Carlos et al 2009). Coyotes are also omnivorous, and typically prey on natural food sources such as small mammals and fruit in developed areas, but use twice as much space as neighboring coyotes in undeveloped areas (Morey et al. 2007, Gese et al. 2012). Obligate carnivores that require live prey are less common in urban settings because they have less plasticity in adapting to novel environments where garbage or other sedentary food sources are more abundant (Riley 2003, Baker and
Harris 2007). Foraging for live prey requires more space and resources, both of which are heavily competed for in fragmented habitat of urban areas.
Bobcats (Lynx rufus) are obligate carnivores and rely on a diet of live prey such as small mammals and birds (Hansen 2006, Gehrt et al. 2010). They are historically known to be elusive and sensitive to human disturbances and are not commonly found in areas densely populated by humans (Crooks 2002). However, some bobcat populations exist in a peri-urban setting within close proximity to or overlap with areas of high human densities (Riley 2003, Alonso et al. 2012). A population of bobcats resides in the Dallas-
Fort Worth, Texas metroplex (DFW). This population appears to differ from other urban bobcat populations in that it exists within the metroplex, not just on the expanding edges or suburbs. Little is known about this population, despite regular sightings and reports from residents, animal control officers, and wildlife biologists. This study seeks to learn more about bobcat ecology in the context of an established isolated urban population to inform local wildlife managers in managing and conserving carnivores on the urban landscape. The objective is to determine population densities and habitat selection of urban bobcats in the DFW metroplex and to what extent they utilize developed habitat.
4 METHODS
Study Site
DFW is the fourth largest, third fastest growing, and 19th densest populated
metropolis in the United States (2014 census). It covers more than 24,000 km2 of rolling
hills and large flood plains that require extensive networks of storm drainages and creeks
to divert flood water into the Trinity River. The metroplex lies within the cross timbers
area of Texas, formerly covered in habitat types such as oak trees (Quercus sp.) and
backland prairie (Gould 1975), and is now a mosaic of urban structures which includes
retail stores, residential areas, city parks, golf courses, and patches of undevelopable
wetland, connected by a network of roads, highways, and interstates (Fig. 1).
The study area is in the center of DFW, bordered by state highway TX-183 to the
north, state highway TX- 360 to the east, the west fork of the Trinity River to the south,
and Interstate 820 to the west (Fig. 2). This area covers approximately 78.1 km2 and includes parts of Fort Worth, Arlington, Hurst, Bedford, Euless, and Grand Prairie municipalities. The site was selected based on its high frequency of bobcat sightings and complete immersion within dense urban development. The area includes varying levels of urbanization ranging from patches of natural habitat (i.e., Trinity River corridor), developed open space, residential areas, and industrial or retail zones.
Data Collection
Captures
Bobcats were captured and handled in accordance with the National Wildlife
Research Center’s Institutional Animal Care and Use Committee (IACUC) regulations 5 (QA-2211). Live cage traps were used to capture bobcats that were immobilized with a
weight appropriate dose of 5:1 ketamine (10 mg/kg) - xylazine (0.75 mg/kg) mixture sedative (Beltrán and Tewes 1995). Body measurements, blood, hair, parasites, and pelage photos (Heilbrun et al. 2006) were collected for physiological data on each captured bobcat. A tooth sample was collected from adult bobcats (n=9) for aging (Crow
1972). Bobcats weighing over 5 kg (n=12) were fitted with Lotek 3300s store-on-board
GPS tracking collars equipped with drop-off mechanisms. Initial trapping occurred from
January - April 2014, with targeted recapture attempts occurring September - December
2014 and March 2015 for bobcats with GPS collars that did not drop off as expected.
Trap sites were typically placed along natural corridors with probable bobcat traffic but limited human traffic (Fig. 2).
GPS Collars
Collars were programmed to record a GPS location every two hours from 18:00 to
06:00 (nocturnal activity) and every four hours from 06:00 to 18:00 (diurnal activity) with some variability in time. These settings were chosen in order to maximize data
collection in consideration of known bobcat activity (Crooks 2002, Gehrt et al. 2010,
Hansen 2006) and the expectations of collar battery life. In addition to this daily data
collection, collars were programmed to record locations at 20-minute intervals for a six-
hour time period on the 15th of each month. For each collar, this would occur across one
of four randomly chosen six-hour time periods. The drop-off mechanisms were
programmed at 56 weeks after collar deployment allowing for just over one year of data
collection. Location data within the initial 24 hours of data collected after an animal was 6 collared were removed to allow for a full recovery time from immobilization drugs
administered during captures. Failed fixes and locations with a dilution of precision
(DOP) value of >8 were removed. For spatial analysis, point files were then created in
ArcGIS (ESRI 2011) under the projected coordinate system of NAD 83 zone 14 North.
Cameras
A camera grid comprised of 41 double-camera stations documented individual bobcats throughout the study site (Fig. 3). Camera station locations were assigned using
ArcGIS 10.2.0 (ESRI 2011). Spacing of the stations was roughly based on the minimum expected home range observed with preliminary tracking of DFW collared bobcats (1.5 km2) and the minimum home range size of urban bobcats in previous studies (1.3 km2)
(Riley 2003, Alonso et al. 2012). For each station, a 300-m buffer was used to ensure
effective camera placement for detecting resident bobcats while avoiding theft or being
vandalized by people. This buffer is higher than the 200-m buffer that other camera
studies have used to detect bobcats (Alonso et al. 2012) to allow feasible camera
placement for as many stations as possible amidst a more variable and risky urbanized
landscape. Spacing between cameras averaged 1.05 km between each station A few
portions of the study site were not included in the camera grid due to limitations of access
and equipment.
For each station, two cameras were simultaneously placed on opposing sides of a
road or trail 2-5 m from the trail’s center and aimed 30-45 cm above the trail (Heilbrun et
al. 2006, Alonso et al. 2012). No lure was used with camera sets in order to avoid any
bias in detection (Heilbrun et al. 2006). Bushnell HD Trophy Camera model 119576, 7 Moultrie A-8 model MCG-12646, and one Reconyx PC900 HyperFire camera were used across the grid. The Bushnell and Reconyx cameras were programmed to take bursts of three photographs per trigger at their shortest trigger delay setting (≤ 1 second) to improve likelihood of photographing unique markings on passing bobcats. The Moultrie cameras could only take one photograph per capture with a one-minute trigger delay.
Moultrie cameras were only used when paired with either a Bushnell or Reconyx (n = 37
stations) to limit detection bias based on camera type. Remaining stations (n = 4) had two
Bushnell cameras. Stations were active for six to seven consecutive weeks across a single
trapping period from 8 June - 15 November 2014 (Table 1).
All photographs from the camera grid were sorted into folders by station and then
manually viewed and sorted into subfolders by species. Bobcat photos were then
compared across stations to document total number of individual bobcats captured and
recaptured across the survey grid. The Photo Warehouse database software, developed at
the Colorado Department of Wildlife (Newkirk 2014), was used to organize and assign
IDs to individual bobcats, both collared and uncollared. A 20-minute time interval was
used to determine independent photograph capture events. For an ID to be applied, three
physically unique features were used to match or differentiate individuals across
photograph capture events (Heilbrun et al. 2006). Photographs of bobcats that could not
be identified were labeled as unknowns.
Data Analyses
Home Ranges
Individual bobcat home ranges were calculated with 95% kernel density 8 estimators (KDE) for use in the resource selection analysis and comparison to other urban bobcat studies (Manly et al. 2002). To account for the hard boundaries and patchy nature of the urban landscape, 99% local convex hulls (LoCoH) were also calculated (Getz and
Wilmers 2004, Lichti et al. 2011). For KDEs, Geospatial Modeling Environment (GME) software’s (Beyer 2012) Kernel Density Estimation tool was used to generate raster datasets from cleaned GPS locations for each bobcat. Within the tool, the plug-in bandwidth estimator with a cell size of 42.7 was used with a Gaussian kernel type processed in a set extent of 3637168 top, 662151 left, 3626484 bottom, and 683263 right.
These rasters were then brought into the isopleth tool of GME with quantiles set to 0.95, creating the 95% KDE home ranges.
Local convex hulls were calculated using the rhr package in R (version 3.2.4, www.R-project.org, Signer and Balkenhol 2015). Individual bobcat GPS data were each imported under the projected coordinate system of NAD 83 zone 14 North. Within the configure > home range tab, the Local Convex Polygon tool was used to produce type
“a”, “k”, and “h” LoCoH polygons, considering both 0.99 and 0.95 quantiles in each. In evaluating outputs, the 0.99 type a local convex hull was most appropriate in representing bobcat home ranges amidst the patches and barriers of the DFW landscape.
Resource Selection Function Models
ArcGIS 10.2.0 (ESRI 2011) was used for spatial data preparation of bobcat GPS locations and spatial habitat data. All data were converted to the same coordinate system as the bobcat GPS data. The most current, freely accessible landscape data from Texas
Natural Resources Information System (TNRIS) for Tarrant County, Texas were used for 9 habitat analysis. Landscape layers include the 2012 creeks layer for Texas, 2014 railroads
layer for Texas, and 2011 roads layer for Texas all in vector format (TNRIS 2016). The
2011 National Land Cover Dataset (NLCD) was used to classify habitat types (Homer et
al. 2015, Xian et al 2011).
Fourteen NLCD classes were present in the DFW study site, which were
condensed into eight categories: (1) forest, included three forest types; (2) wetland,
included two wetland types; (3) agriculture, included crops and pastures; (4) grassland, included grassland/herbaceous; and (5-8) four developed categories, represented by the
amount of impervious surface in NLCD. Open space included developed open space
(<20% impervious surface) and barren, while low development (20-49% impervious
surface), medium development (50-79% impervious surface), and high development
(>80% impervious surface) were not changed. One habitat category, open water, was not
included in analysis because it was more accurately represented by the 2012 creeks data.
These categorical habitat data and the vector data were rasterized by Euclidean distance
measures using the ArcGIS Euclidean Distance tool (ESRI 2011). All rasterized habitat variables were standardized.
A systematic approach was used to represent availability across the individual home range KDE isopleths. Available points were generated in R using the ‘sp’ package’s spsample command using a 30 m cell size and nonaligned systematic sampling
(Bivand et al. 2013, Pebesma and Bivand 2005). Generated available points were then combined with the bobcat GPS locations, or used points, to compile one complete dataset
(Manly et. al 2002). Distance to features such as land cover type, roads, and creeks were then extracted from rasterized habitat data to all points using the ArcGIS Extract Value to 10 Point tool (ESRI 2011).
Collinearity and variance of inflation (VIF) for all variables were checked in R
using the ‘usdm’ package and variables with greater than 0.7 collinearity and 4.0 VIF were lumped together or removed (Naimi 2015). Leniency for inclusion was used for variables with VIF< 4.0 despite having collinearity > 0.7 if it was deemed ecologically sensible to do so. General linear mixed effects models in the ‘lme4’ package (Bates et. al
2015) in R were used to determine how bobcat habitat usage is affected by the predictive landscape variables, with Bobcat ID as a random effect. Using a multi-model inference package called ‘MuMIn’ (Burnham and Anderson 2002), all possible general linear mixed effects models were run with the dredge command. Top models (< 3 delta AICc) were reviewed and evidence of irregular covariate variability between models was further investigated by incorporating second order polynomials and possible interactions. A step-
wise selection process was employed to determine which variables of the top models
were compromised. Once non-linearity and confoundedness problems were addressed,
top models were selected again by having delta AICc values of < 3.0 from the top model.
From the top model, a bobcat habitat selection probability map was generated
using the distance to habitat rasters and corresponding covariate coefficients plugged into
the logistic equation in Raster Calculator in ArcGIS 10.2 (ESRI 2011). Prior to raster
calculation, Euclidean distance to habitat rasters were truncated to their minimum and
maximum distance measurements found in the RSF analysis. The resulting heat map
visually represents probability of bobcat resource selection throughout and surrounding
the DFW study site. Values from this RSF raster were extracted to the habitat mask and
trap locations for inclusion in the SECR analysis to better inform models for bobcat 11 population density estimates in DFW (Efford 2011).
In addition to the population based RSF, these steps were repeated for each
bobcats’ dataset to determine individual resource selection. Individual RSF heat map
rasters were then averaged using Raster Calculator in ArcGIS 10.2 (ESRI 2011) to create
a second population heat map (Appendix 2).
Spatially Explicit Capture-Recapture (SECR)
Camera data were analyzed with the statistical program R (version 3.2.4, www.R-
project.org; R Core Team 2013). A spatially explicit capture-recapture (SECR) package
was used to run maximum likelihood models for bobcat population estimates (Efford
2011). To include multiple captures of individuals over a single occasion, a capture
history was developed using a “count” detector type with a capture file of 265 capture
events across 23 occasions and 41 stations. The RSF heat map was used to assign each
station a habitat selection probability to inform detection probabilities.
Active dates for camera traps were binned into “weeks” for occasion ID in the
SECR capture history. In order to maximize capture inclusion, a camera was considered
active in a week as long as it was active for one night. Each bobcat capture event was
assigned an occasion along with the station of detection.
Following methods detailed by Bashir et al. (2013), a habitat buffer for the camera grid was developed by creating a minimum convex polygon of the camera grid and then adding a buffer of ½ mean maximum distance moved (MMDM) as calculated in
SECR from the capture history object (Karanth and Nichols 1998, Bashir et al. 2013)
(Fig. 3). The Trinity River was established as the hard southern boundary of the effective 12 area surveyed, as resident collared bobcats were not observed crossing this natural
obstacle. Using the buffer, a habitat mask was created by systematically sampling points
every 106.82 m, as generated by SECR’s default mask settings for the DFW camera dataset. The RSF heat map was used to assign each point of the habitat mask a bobcat relative selection probability. The NLCD habitat categories were used to assign cover type to each mask point. Points designated “water” from NLCD were removed from the
mask (Homer et al. 2015, Xian et al 2011).
For the SECR model, density (D) was predicted by the RSF values from the mask,
while lambda0 (λ) and sigma (σ) varied by occasion. A hazard halfnormal detection
function was selected in use with the “count” detector type as described in the SECR
package (Efford 2011).
Hazard halfnormal equation:
2 2 λ(d) = λ0 exp ( −d / 2σ ) ; g(d) = 1 − exp(−λ(d))
( ) = ; ( ) = 1 ( ) 2 2 −𝑑𝑑 λ 𝑑𝑑 λ0𝑒𝑒𝑒𝑒𝑒𝑒 � 2 � 𝑔𝑔 𝑑𝑑 − 𝑒𝑒𝑒𝑒𝑒𝑒�−λ 𝑑𝑑 � 𝜎𝜎
13 RESULTS
Data Collection
Captures
Seventeen individual bobcats were captured, tissue samples and pelage photos
were collected on 16, and 12 met weight requirements for collaring (male = 7, female =
5). Of the 12 collared bobcats, 10 collars were successfully collected from the field.
Three of those collars were deployed for the full 56 weeks, while the other seven bobcat
collars were deployed 7- 46 weeks (Table 2).
GPS Collars
A total of 17,918 GPS locations were used for home range analyses of 10 bobcats
(Average =1791.7, SE= 348.9). Fix rate success averaged 77% (SE = 2.4%) of points
being usable for analysis (Table 3).
Cameras
Six cameras were stolen from 4 camera stations. If one camera was stolen from a
station (n=2), the remaining camera and a new camera were re-set close by with
additional camouflage on the same trail until a total of six weeks of data could be
collected from at least one camera at that location. If both cameras were stolen from a
station (n=2), an entirely new trail was selected within the initial 300-m camera
placement buffer. The six-week interval would start over at the new location. No camera
stations were stolen a second time using this technique. Stations with temporary camera
failure (n=3) were still considered active as long as one camera remained functional. 14 The camera survey documented 279 bobcat capture events from 1003 bobcat
photographs. Forty-two individuals were identified, including nine collared bobcats and
two previously trapped but non-collared individuals. Of the collared bobcats, two, B009
and B013, were not captured via camera traps although their GPS location data indicated
they were present in the area of the camera trap grid, and one was caught on camera,
B016, even though it was not collared until after camera trapping ceased. Fourteen
capture events were left as unknown, leaving 265 capture events to be used for SECR
analysis.
Other carnivore species documented included coyote (Canis latrans), raccoon
(Procyon lotor), grey fox (Urocyon cinereoargenteus), striped skunk (Mephitis mephitis),
Virginia opossum (Didephis virginiana), domestic cat (Felis catus), and domestic dog
(Canis familiaris). Other species photographed included fox squirrel (Sciurus niger), lagomorphs (Sylvilagus sp.), passerine and waterfowl species, nine-banded armadillo
(Dasypus novemcinctus), beaver (Castor canadensis), domestic cows (Bos sp.), and feral pigs (Sus scrofa).
Data Analysis
Home Ranges
The 95% kernel density estimator resulted in an average home range size of 4.60
km2 (n=9, SE=0.99 km2) for all resident bobcats, 3.48 km2 (n=5, SE=1.13 km2) for
resident females, and 6.00 km2 (n=4, SE=1.61 km2) for resident males. One dispersing
male bobcat had a home range of 25.71 km2 and was not included in calculating home
range averages (Fig. 4a). 15 The local convex hull estimator resulted in an average home range size of 3.83
km2 (n=9, SE=0.81 km2) for all resident bobcats, 2.29 km2 (n=5, SE=1.03 km2) for
resident females, and 5.05 km2 (n=4, SE=1.13 km2) for resident males. The dispersing
male bobcat had a home range of 25.97 km2 and was not included in calculating home
range averages (Fig. 4b).
Resource Selection Function Models
The systematic sampling resulted in 74,563 availability points from across all individual home ranges. When deducting for home range overlap, the bobcat use area covered 46.26 km2. However, RSFs were modelled by individual bobcat, therefore the
actual sum of individual home ranges was 67.13 km2.
Variable collinearity (Appendix, Figs. 7-9) and VIF checks (Tables 4-6) resulted in condensing of the initial land cover classes from eleven to eight variables for the RSF models. Remaining habitat variables were: (1) agriculture, (2) creeks, (3) grassland, (4) high development, (5) low-medium development, (6) open space, (7) roads, and (8) railroads. Forest and wetland were removed due to extremely high VIF and collinearity
(Tables 4-6; Appendix, Figs. 7-8). Low and medium development were lumped together in order to maintain an urban gradient of the landscape.
Initial top models included variations of the global model interchanging high development and low-medium development (Table 7). From these models, step-wise selection for testing polynomial and interaction effects resulted in a final top model that included seven variables and three corresponding polynomials to resolve non-linear trends among creeks, grass, and low-medium development (Table 8). Positive 16 coefficients represent a positive slope in the linear relationship between increasing
distance and selection, thus avoidance. Negative coefficients represent selection.
Variables with accompanying polynomials (variable2) represent non-linear relationships
between distance and selection, meaning intermediate distances are either selected or
avoided. In applying these coefficients to their corresponding habitats, bobcat resource
selection probabilities ranged from 0.730 – 0.086 across the study site and surrounding
areas (Fig. 5). The top models for individual RSF analysis are reported in Appendix 2.
Spatially Explicit Capture-Recapture Model
The “count” detector used 265 capture events across 23 trapping occasions for the
SECR analysis. Using the ½ MMDM method, a habitat mask using a buffer of 710 m
with the Trinity River as a hard boundary yielded an effective camera survey area of
63.53 km2 (Fig. 3). The habitat mask included 5,569 points across this area. Once water
points were removed, the final habitat mask contained 5,144 sampled points across 58.70
km2.
The SECR model predicted bobcat densities ranging from 2.06 – 0.46/km2 across the habitat mask (Fig. 6). The overall density for the camera survey area was 0.55/km2
(SE = 0.17), which calculates to 26.83 (SE = 9.98) bobcats across the entire camera grid
(Table 9).
17 DISCUSSION
Bobcats in DFW have much smaller home ranges than those in rural areas, where
home ranges average as large as 57.3 km2 (Donovan et al. 2011). Home ranges were comparable to but slightly smaller than those of bobcats in other urban areas, where averages are reported to be 8.83 km2 in Orange County, California (Alonso et al. 2012)
and 6.74 km2 for males and 1.51 km2 for females near San Francisco, California (Riley
2003). In our study, females typically had smaller home ranges than males and younger
male bobcats utilized developed open space and low levels of development more than
females and older males.
The differences in home range size between urban and rural bobcats may be
attributed to a high density of prey items supported by anthropogenic food and habitat
(e.g., garbage, bird feeders, landscaping) within a much smaller amount of available space and habitat. Previous bobcat studies show that bobcats have larger home ranges with lower prey densities (Blankenship 2000). Urban bobcats maintain smaller home ranges due to lack of space and abundance of prey. The differences between DFW bobcat home ranges and other peri-urban bobcat populations may be reflective of the size of habitat patches available. For example, many urban areas of southern California occur in mountainous terrain, where areas that are too steep or unstable to develop remain as habitat oases for urban wildlife (Alonso et al. 2012). DFW has fewer of these large and
undeveloped spaces, aside from floodplains and wetlands interspersed through the urban
matrix. This further compacts the natural habitat available to and selected for by DFW’s
bobcats, creating greater overlap of and smaller of home ranges. 18 Bobcats in DFW predictably selected for areas closer to more natural habitat
cover, such as developed open space, agriculture, railroads, and creeks, and avoided areas
close to higher levels of development, such as low-medium development and roads
(Table 8). However, several variables showed signs of non-linearity indicating that bobcats may reverse selection or avoidance to certain habitat types at intermediate distances. Intermediate distances, as represented by polynomial Grass2, showed the
highest level of selection among all variables, despite closer distances to grass being the
highest level of avoidance (Table 8). This is likely due to the use of Euclidean distance to
measure habitat for the RSF. Euclidean distance measures to cover type provide
continuous variables that general linear mixed effects models work well with. However,
the habitat patches that bobcats used in DFW are completely surrounded by various
levels of urbanization, thus any given area in natural habitat cover is always within a
certain distance of development, as opposed to an infinite distance that a linear model
assumes. Due to the complex landscape structure of urban areas, linear distance measures
can be misleading in their assumption of a continuous linear relationship between
selection and distance. For example, greater distances from grassland coincide with
closer distances to developed space, so, even though a linear relationship shows that as
distance to grass increases, selection increases, selection will not continue to increase as
habitats get closer to avoided cover, such as roads. This is why second order polynomials
must be considered in order to capture the true relationship of selection and distance to habitats. The non-linear trend of grassland (Grass2) shows that grass was less selected for
or avoided at very close and very far distances, but highly selected for at intermediate
distances. This also ecologically makes sense with known bobcat behavior in utilizing 19 edge habitats for travel and foraging (Crooks 2002, Hansen 2006). Bobcats may not
commonly use areas in the center of a grassland, but will select areas of habitat
neighboring grassland.
A similar trend was observed for low-medium development. Bobcats showed avoidance of low-medium development on a linear scale, but showed slight selection for low-medium development on a non-linear scale (Table 8). Again, this corresponds to
natural habitat cover types that occur at intermediate distances to low-medium development. This is similar to habitat use from similar bobcat studies near urban areas
(Riley et al. 2003).
Creeks were an important variable in this model; they are abundant on the landscape both in natural and developed regions within the DFW study site. This is likely why creeks were selected for on a linear scale but avoided on a non-linear scale (Table
8). When bobcats use creeks in urban areas, they act as narrow corridors amidst low- medium and high development areas. Even intermediate distances from most of the creeks on the landscape fall within highly developed areas that are otherwise avoided.
While these results suggest bobcat habitat use is similar to that found in other urban areas, the small sample size of bobcats and limited number of juvenile animals that were fitted with GPS collars make it difficult to detect spatial and resource selection differences between age and sex classes. When considering the DFW population estimate of 27 ± 10 bobcats, this study only sampled at most 59% of the resident population for
GPS data collection, and only half of those bobcats collected >8 months of data to cover multiple seasons. Additional sampling for longer time periods could address questions such as seasonal, sex, and age differences in home ranges and resource selection, as well 20 as broader population dynamics such as age structure, reproductive rates, and mortality.
Further, carnivores may use different habitat types depending on behavioral state, such as
when they are foraging versus resting (Abrahms et al. 2015). In such a mosaic of urban
landscape cover, it would be informative to perform analyses, such as for resource
selection, based on “active” or “inactive” periods. Thus, future habitat analyses should
consider bobcat activity patterns. GPS collar fix rate success may also be biased
depending on habitat, anthropogenic signal interference, weather, bobcat activity
patterns, and general quality of the collar (Table 3).
The bobcat population density in DFW is higher than those reported in other
bobcat studies. Heilbrun et al. (2006) reported 0.48 bobcats/km2 in a rural wildlife area of
Texas. In this study, there was a range of densities across the camera survey grid from
2.06 – 0.46 bobcats/km2 (Fig. 6) with much of the study area having densities higher than
Heilbrun et al. (2006). The overall density estimate for DFW was 0.55/km2 (SE = 0.17),
which was at the highest range reported from peri-urban areas (Alonso et al. 2012). GPS
data also show a trend of high home range overlap, indicating high densities in urban
areas, as shown in similar urban bobcat studies (Alonso et al. 2012, Riley 2003). This
trend of high densities in urban areas likely relates to the limited amount of space and
dispersal opportunities for a resident bobcat population that is also supplemented by a
higher density of urban prey items. Indeed, GPS collared bobcats in this study were
observed hunting squirrels and waterfowl from uniquely urban features such as golf
courses and landscaped apartment complexes.
Even so, estimates may be artificially inflated as potential bias in SECR estimates
could exist due to several field and analysis methods: (1) cameras were specifically 21 placed on trails and in areas that would improve likelihood of detecting bobcats in the area, thus potentially overestimating density and detection probabilities due to non- random placement within the 300-m buffer; (2) only one camera survey session was conducted, thus limiting the assumption that bobcat detections were representative of the year-round population; and (3) some GPS collared bobcats known to travel near cameras were not detected (n = 2), indicating that we did not achieve 100% detection with our camera grid, thus likely underestimating actual bobcat population densities. To improve
SECR estimates, future camera surveys should incorporate randomized camera placement, more than one camera trapping session conducted between different seasons and years, and cameras should be set at a higher frequency on the landscape to improve chance of complete detection of the population.
SECR models could also be improved by incorporating non-euclidean distance measures (Royle et al. 2013). Bobcats rarely move in straight lines, especially in an urban landscape where there are many avoided areas and barriers affecting their movements.
The analysis used the default Euclidean distance measure to estimate home range centers and maximum distances moved for individual bobcats. Correcting this assumption of straight-line movements with “ecological distances” could improve the accuracy of the overall size of the effective camera survey area (Fig. 3) and also improve density estimates based on more accurate bobcat home ranges produced within the SECR package.
Understanding the habitat usage and potential behaviors of urban bobcats will aid wildlife managers and conservationists in continuing efforts to improve human-carnivore coexistence in urban areas globally. Urbanization and loss of habitat is a leading threat to 22 many large mammalian species, especially carnivores that often directly compete with humans for resources (Crooks et al. 2011). Findings from this DFW bobcat study revealed optimistic possibilities in the potential for carnivores to thrive in an urban landscape with minimal conflict. Citizens recreating or living near areas associated with high probability of bobcat occurrence can be informed accordingly to enhance current preventative methods in minimizing bobcat conflict with pets or domestic animals.
Observing such a high density of bobcats in an urban setting is informative for urban areas anticipating other predator species that may adapt and thrive amidst high densities of people. These results have promising implications for species preservation in areas where urbanization is advancing and threatening wildlife. This and other urban carnivore studies focus on areas that host a unique urban-wildlife interface where appropriate management of can improve the landscape in which both people and wildlife share.
23 LITERATURE CITED
Abay, G. Y., H. Bauer, K. Gebrihiwot, and J. Deckers. 2011. Peri-urban spotted hyena
(Crocuta crocuta) in Northern Ethiopia: diet, economic impact, and abundance.
European Journal of Wildlife Research 57:759–765.
Abrahms, B., N.R. Jordan, K.A. Golabek, J.W. McNutt, A.M. Wilson, and J.S. Brashares.
2015. Lessons from integrating behaviour and resource selection: activity-specific
responses of African wild dogs to roads. Animal Conservation 19:247-255.
Adams, L.W. 2005. Urban wildlife ecology and conservation : A brief history of the
discipline. Urban Ecosystems 8:139–156.
Alonso, R.S., E.E. Boydston, B.R. Noon, and D.M. Theobald. 2012. The effects of
urbanization and road development on carnivores in Southern California.
Colorado State University.
Athreya, V., M. Odden, J.D.C. Linnell, J. Krishnaswamy, and U. Karanth. 2013. Big cats
in our backyards: persistence of large carnivores in a human dominated landscape
in India. PloS One 8:e57872.
Baker, P.J., and S. Harris. 2007. Urban mammals: what does the future hold? An analysis
of the factors affecting patterns of use of residential gardens in Great Britain.
Mammal Review 37:297–315.
Bashir, T., T. Bhattacharya, K. Poudyal, S. Sathyakumar, and Q. Qureshi. 2013.
Estimating leopard cat Prionailurus bengalensis densities using photographic
captures and recaptures. Wildlife Biology 19:462–472.
Bateman, P.W., and P.A. Fleming. 2012. Big city life: carnivores in urban environments. 24 Journal of Zoology 287:1–23.
Bates, D., M. Maechler, B. Bolker, and S. Walker 2015. Fitting linear mixed-effects
models using lme4. Journal of Statistical Software 67:1-48.
Beltrán, J.F., and M.E. Tewes. 1995. Immobilization of ocelots and bobcats with
ketamine hydrochloride and xylazine hydrochloride. Journal of Wildlife Diseases
31:43–48.
Beyer, H.L. 2012. Geospatial Modelling Environment (Version 0.7.3.0). (software).
URL: http://www.spatialecology.com/gme. Accessed 5 May 2015.
Bhatia, S., V. Athreya, R. Grenyer, and D.W. MacDonald. 2013. Understanding the role
of representations of human-leopard conflict in Mumbai through media-content
analysis. Conservation Biology 27:588–94.
Bivand, R.S., E. Pebesma, and V. Gomez-Rubio. 2013. Applied spatial data analysis with
R, second edition. Springer, New York, New York, USA.
Blankenship, T.L. 2000. Ecological response of bobcats to fluctuating prey populations
on the Welder Wildlife Foundation Refuge. Dissertation, Texas A&M University-
Kingsville.
Boydston, E.E., K.M. Kapheim, H.E. Watts, M. Szykman, and Kay E. Holekamp. 2003.
Altered behaviour in spotted hyenas associated with increased human activity.
Animal Conservation 6:207-219.
Burnham, K.P. and D.R. Anderson. 2002. Model selection and multimodel inference: a
practical information-theoretic approach. 2nd ed. Springer-Verlag, New York,
New York, USA.
Carlos, D., A.D. Bright, T.L. Teel, and J.J. Vaske. 2009. Human–black bear conflict in 25 urban areas: an integrated approach to management response. Human Dimensions
of Wildlife 14:174–184.
Crooks, K.R. 2002. Relative sensitivities of mammalian carnivores to habitat
fragmentation. Conservation Biology 16:488–502.
Crooks, K.R., C.L. Burdett, D.M. Theobald, C. Rondinini,, and L. Boitani. 2011. Global
patterns of fragmentation and connectivity of mammalian carnivore habitat.
Philosophical Transactions of the Royal Society of London B: Biological
Sciences 366:2642-2651.
Crowe, D.M. 1972. The presence of annuli in bobcat tooth cementum layers. The Journal
of Wildlife Management 36:1330–1332.
Donovan, T.M., M. Freeman, H. Abouelezz, K. Royar, A. Howard, and R. Mickey. 2011.
Quantifying home range habitat requirements for bobcats (Lynx rufus) in
Vermont, USA. Biological Conservation 144:2799–2809.
Efford, M.G. 2011. Estimation of population density by spatially explicit capture–
recapture analysis of data from area searches. Ecology 92:2202-2207.
ESRI 2011. ArcGIS desktop: release 10. Redlands, CA: Environmental Systems
Research Institute.
Farrar, R.O. 2007. Assessing the Impact of Urban Coyote on People and Pets in Austin,
Travis County, Texas. Wildlife Damage Management Conferences --
Proceedings, 334–341. Retrieved from
http://digitalcommons.unl.edu/icwdm_wdmconfproc/62. Accesssed 5 May 2015.
Gehrt, S.D., C. Anchor, and L.A. White. 2009. Home range and landscape use of coyotes
in a metropolitan landscape: conflict or coexistence? Journal of Mammalogy 26 90:1045–1057.
Gehrt, S.D., S.P. Riley, and B.L. Cypher. 2010. Urban carnivores: ecology, conflict, and
conservation. Johns Hopkins University Press, Baltimore, Maryland, USA.
Gese, E.M., P.S. Morey and S.D. Gehrt. 2012. Influence of the urban matrix on space use
of coyotes in the Chicago metropolitan area. Journal of Ethology 30:413–425.
Getz, W.M., and C.C. Wilmers. 2004. A local nearest neighbor convex-hull construction
of home ranges and utilization distribution. Ecography 27:1–17.
Gould, F.W. 1975. The Grasses of Texas. Texas A&M University Press, College Station,
Texas.
Grubbs, S.E., and P.R. Krausman. 2009. Use of urban landscape by coyotes.
Southwestern Association of Naturalists 54:1–12.
Hansen, K. 2006. Bobcat: master of survival. Oxford University Press, New York, New
York, USA.
Heilbrun, R.D., N.J. Silvy, M.J. Peterson, and M.E. Tewes. 2006. Estimating bobcat
abundance using automatically triggered cameras. Wildlife Society Bulletin
34:69–73.
Homer, C.G., J.A. Dewitz, L. Yang, S. Jin, P. Danielson, G. Xian, J. Coulston, N.D.
Herold, J.D. Wickham, and K. Megown. 2015. Completion of the 2011 National
Land Cover Database for the conterminous United States-Representing a decade
of land cover change information. Photogrammetric Engineering and Remote
Sensing 81:345-354.
Karanth, K.U., and J.D. Nichols. 1998. Estimation of tiger densities in India using
photographic captures and recaptures. Ecology 79:2852–2862. 27 Kertson, B.N., R.D. Spencer, J.M Marzluff, J. Hepinstall-Cymerman, and C.E. Grue.
2011. Cougar space use and movements in the wildland – urban landscape of
western Washington. Ecological Applications 21:2866–2881.
Knopff, A.A., K.H. Knopff, M.S. Boyce, and C.C. St. Clair. 2014. Flexible habitat
selection by cougars in response to anthropogenic development. Biological
Conservation 178:136–145.
Lichti, N.I., and R.K. Swihart. 2011. Estimating utilization distributions with kernel
versus local convex hull methods. Journal of Wildlife Management 75:413–422.
Magle, S.B., V.M. Hunt, M. Vernon, and K.R. Crooks. 2012. Urban wildlife research:
past, present, and future. Biological Conservation 155:23–32.
Manly, B.F.J., L.L. McDonald, D.L. Thomas, T.L. McDonald, and W.P. Erickson. 2002.
Resource selection by animals: statistical design and analysis for field studies, 2nd
Edition. Kluwer Academic Publishing, London, England.
Morey, P.S., E.M. Gese, and S. Gehrt. 2007. Spatial and temporal variation in the diet of
coyotes in the Chicago metropolitan area. The American Midland Naturalist
158:147-161.
Naimi, B. 2015. usdm: Uncertainty analysis for species distribution models. R package
version 1.1-15. https://CRAN.R-project.org/package=usdm. Accessed 5 May
2015.
Newkirk, E.S. 2014. CPW Photo Warehouse. Colorado Parks and Wildlife, Fort Collins,
Colorado, USA.
http://cpw.state.co.us/learn/Pages/ResearchMammalsSoftware.aspx. Accessed 5
May 2015. 28 Ordeñana, M.A., K.R. Crooks, E.E. Boydston, R.N. Fisher, …and D.H. Van Vuren.
2010. Effects of urbanization on carnivore species distribution and richness.
Journal of Mammalogy 91:1322–1331.
Pebesma, E.J., and R.S. Bivand, 2005. Classes and methods for spatial data in R. R News
5 (2), http://cran.r-project.org/doc/Rnews/. Accessed 5 May 2015.
Quinn, T. 1997. Coyote (Canis latrans) habitat selection in urban areas of western
Washington via analysis of routine movements. Northwest Science 71:289–297.
R Core Team. 2013. R: A language and environment for statistical computing. R
Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-
project.org/. Accessed 5 May 2015.
Riley, S.P.D. 2003. Spatial ecology of bobcats and gray foxes in urban and rural zones of
a national park. Journal of Wildlife Management 70:1425–1435.
Riley, S.P.D., R.M. Sauvajot, T.K. Fuller, E.C. York, D.A. Kamradt, C. Bromley, and
R.K. Wayne. 2003. Effects of urbanization and habitat fragmentation on bobcats
and coyotes in Southern California. Conservation Biology 17:566–576.
Royle, J.A., R.B. Chandler, K.D. Gazenksi, and T.A. Graves. 2013. Spatial capture-
recapture models for jointly estimating population density and landscape
connectivity. Ecology 94:287–294.
Ruell, A.E.W., S.P.D. Riley, M.R. Douglas, M.F. Antolin, J.R. Pollinger, J.A. Tracey,
L.M. Lyren, E.E. Boydsten, R.N. Fisher, and K.R. Crooks. 2012. Urban habitat
fragmentation and genetic population structure of bobcats in coastal Southern
California. The American Midland Naturalist 168:265–280.
Singh, P., A.M. Gopalaswamy, and K.U. Karanth. 2010. Factors influencing densities of 29 striped hyenas (Hyaena hyaena) in arid regions of India. Journal of Mammalogy
91:1152–1159.
Tigas, L.A., D.H. Van Vuren, and R.M. Sauvajot. 2002. Behavioral responses of bobcats
and coyotes to habitat fragmentation and corridors in an urban environment.
Biological Conservation 108:299–306.
TNRIS. 2016. Texas Natural Resources Information System. Texas Water Development
Board, Austin, Texas. URL http://www.tnris.org/. Accessed 5 May 2015.
Treves, A., and K.U. Karanth. 2003. Human-carnivore conflict and perspectives on
carnivore management worldwide. Conservation Biology 17:1491–1499.
United Nations, Department of Economic and Social Affairs, Population Division. 2015.
World Urbanization Prospects: The 2014 Revision, (ST/ESA/SER.A/366).
Accessed 5 May 2015.
Woodroffe, R. 2000. Predators and people: using human densities to interpret declines of
large carnivores. Animal Conservation 3:165–173.
Xian, G., C. Homer, J. Dewitz, J. Fry, N. Hossain, and J. Wickham. 2011. The change of
impervious surface area between 2001 and 2006 in the conterminous United
States. Photogrammetric Engineering and Remote Sensing 77: 758-762.
30 TABLES AND FIGURES
Table 1. Duration of camera station activity from 8 June – 15 November, 2014.
ID Jun Jul Aug Sep Oct Nov 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 31 Table 2. Duration each bobcat was fitted with a GPS collar within the DFW study site between January 2014 and April 2015.
ID Age Sex Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr B002 Adult F B004 Adult M B006 Juvenile F B007 Adult F B008 Adult M B009 Adult M B011 Adult F B013 Adult M B014 Juvenile M B016 Adult F 32 Table 3. Fix rate success of GPS collars fitted to bobcats in the DFW study site. DOP< 8 was used to define acceptable, accurate locations. Usable fixes are from all fixes obtained for each GPS collar.
ID Raw data Fixes Success DOP<8 Accurate Usable rate (%) fixes (%) fixes (%) B002 2335 1512 64.8 1441 95.3 61.7 B004 3908 3390 86.7 3182 93.9 81.4 B006 651 604 92.8 587 97.2 90.2 B007 3783 3120 82.5 3027 97.0 80.0 B008 1760 1619 92.0 1351 83.4 76.8 B009 3734 3168 84.8 3067 96.8 82.1 B011 499 405 81.2 349 86.2 69.9 B013 1713 1490 87.0 1339 89.9 78.2 B014 3879 2908 75.0 2799 96.3 72.2 B016 1060 901 85.0 825 91.6 77.8
33 Table 4. Variance inflation of initial habitat covariate classifications (minus forest) for RSF analysis as calculated by package ‘usdm’ in program R. Variables with VIF > 4 and collinearity coefficients > 0.7 were removed.
Variable VIF Coef > 0.7 Ag 2.29 Creeks 1.31 Forest 15.32 X Grass 5.25 High Dev 2.62 Low Dev 4.18 X Med Dev 2.81 Open Space 2.36 Roads 2.54 Railroads 2.10 Wetlands 13.27 X
34 Table 5. Variance inflation of second habitat covariate classifications (minus forest) for RSF analysis as calculated by package ‘usdm’ in program R. Variables with VIF > 4 and collinearity coefficients > 0.7 were removed.
Variable VIF Coef > 0.7 Ag 2.253 Creeks 1.247 Grass 3.686 High Dev 2.631 Low Dev 4.367 X Med Dev 2.834 Open Space 2.357 Roads 2.592 Railroads 2.046 Wetlands 6.077 X
35 Table 6. Variance inflation of final habitat covariate classifications (minus forest and wetland, with low and medium development combined as LowMed Dev) for RSF analysis as calculated by package ‘usdm’ in program R. Variables with VIF > 4 and collinearity coefficients > 0.7 were removed.
Variable VIF Coef > 0.7 Ag 1.785 Creeks 1.228 Grass 2.112 High Dev 2.255 LowMed Dev 2.279 Open Space 2.108 Roads 2.269 Railroads 1.879
Table 7. Top models of <3 delta AICc from initial RSF analysis. These models provide guidance for stepwise selection considering non-linearities and interactions.
36
Table 8. Final top models for DFW bobcat RSF with polynomials addressing non-linearities.
37
38 Table 9. Parameter estimations for the effective survey area from the top SECR model. D represents density in km2, lamda0 represent average expected number of detections, and sigma represents average home range movements in m.
Parameter Link Estimate SE Lower CL Upper CL D log 0.55 0.17 0.30 1.00 Lambda0 log 0.24 0.05 0.16 0.35 Sigma log 1162.50 108.30 968.90 1394.80
39
Figure 1. The Dallas-Fort Worth Metroplex, Texas. The study site included the cities of Arlington, Bedford, Euless, Fort Worth, and Hurst. Bobcats were studied from January 2014 to April 2015.
40
Figure 2. The DFW urban bobcat study site and capture locations during GPS collaring efforts from January 2014 to March 2015. The Trinity River and three major roadways define the research area. Trapping was focused in or along edges of natural habitat amidst the urban landscape.
41
Figure 3. DFW bobcat camera survey from June – November 2014. The survey boundary represents the ½ MMDM buffer as calculated in SECR, with the Trinity River as a hard boundary to the south.
42 a
b
Figure 4. (a) 95% kernel density home range estimation and (b) 99% local convex hulls for 10 collared bobcats in DFW metroplex. 43
Figure 5. Population-based bobcat habitat selection probability throughout the DFW study site. Likelihood of bobcat habitat selection ranges from the highest at 0.73 (red) to the lowest at 0.086 (blue). Intermediate selection likelihoods are represented by yellow.
44
Figure 6. SECR surface density map defining bobcat density per habitat mask point in the DFW metroplex study site. Density ranges from 2.06 -0.46 bobcats/km2.
45
APPENDICES
46 APPENDIX 1
Figure 7. Pair-wise comparison chart of 11 initial habitat covariates considered for DFW bobcat RSF. Collinearity values are shown in the on the lower half of the chart, plotted points of covariate relationships are shown in the upper half.
47
Figure 8. Pair-wise comparison chart second habitat covariate classification, excluding forest, considered for DFW bobcat RSF. Collinearity values are shown in the on the lower half of the chart, plotted points of covariate relationships are shown in the upper half.
48
Figure 9. Pair-wise comparison chart for final habitat covariates considered for DFW bobcat RSF, excluding forest and with low and medium development condensed. Collinearity values are shown in the on the lower half of the chart, plotted points of covariate relationships are shown in the upper half.
49 APPENDIX 2
METHODS
An individual-based RSF analysis was conducted in addition to the population- based RSF to help determine individual variance. The same modelling methods were employed as described for the population RSF for each of the 10 bobcat datasets to develop individual relative probability of selection heat maps. However, since a random effect of individual was used for the population-based models, random effects were not used in the general linear mixed effects models for the individual-based RSF. Individual heat map rasters were averaged using Raster Calculator in ArcGIS (ESRI 2011) to create a population heat map that represents bobcat resource selection across the study area.
RESULTS
Individual-based RSFs vary between DFW bobcats. Top models demonstrate the differences in combinations and coefficient values of variables for each individual (Table
10). Individual heat maps also reveal variance in DFW bobcat habitat selection (Figure
4a-j). 50 2 − − − − − − 0.060 -0.317 -0.086 -1.948 Road − − − − 0.037 0.276 0.207 0.281 0.121 Road -0.072 − − − − − 0.855 -0.065 -0.368 -0.786 -0.020 Railroad 2 − − − − 0.038 0.651 0.002 1.960 -0.084 -0.057 OpenSpace − − 0.285 0.055 -0.294 -0.287 -0.153 -0.002 -1.563 -0.137 OpenSpace − − − − − − − − − -0.504 LowDe v 2 − − − − − − − − -0.077 -0.030 LowMe d − − − − − 0.020 0.065 0.244 -0.351 -0.102 LowMe d − − − − − − − − 0.405 -0.247 MedDev 2 − − − − − 0.028 0.164 0.006 0.529 -0.049 HighDev − − 0.313 0.172 0.036 -0.077 -0.844 -0.298 -0.016 -0.094 HighDev − − 1.578 0.218 0.774 0.055 0.655 -4.915 -1.092 -0.072 Grass − − − − 0.162 -0.908 -2.584 -0.231 -0.692 -2.274 We tland − − − − -10.20 -1.947 -3.752 -4.745 -2.993 -4.769 Fore st 2 − − − − − − − − − 0.829 Creeks − 0.219 0.017 0.505 0.024 0.468 -1.048 -0.503 -2.049 -0.061 Creeks − − Ag 0.268 0.131 0.725 -0.650 -0.553 -2.714 -0.021 -0.151 0.070 -12.47 -7.077 -12.22 -5.619 -1.526 -0.102 -5.270 -4.030 -6.428 Intercept Table 10. Top coefficients model bobcat for individual RSFs. Bobcat ID B002 B004 B006 B007 B008 B009 B011 B013 B014 B016 51 10a.
10b.
52 10c.
10d.
53 10e.
10f.
54 10g.
10h.
55 10i.
10j.
Figure 10. Individual bobcat heat maps from DFW. Red indicates higher relative probability of selection, blue indicates low relative probability of selection. 56
Figure 11. Population heat map generated by averaging individual bobcat RSFs.