CLIMATE MEDIATED SPATIAL AND TEMPORAL SEGREGATION OF SYMPATRIC

FELIDS AT A RANGE EDGE CONTACT ZONE

By

ARTHUR EDWARD SCULLY

A thesis submitted in partial fulfillment of

the requirements for degree of

MASTER OF SCIENCE IN NATURAL RESOURCE SCIENCES

WASHINGTON STATE UNIVERSITY

School of the Environment

DECEMBER 2016

© Copyright by ARTHUR EDWARD SCULLY, 2016

All Rights Reserved

© Copyright by ARTHUR EDWARD SCULLY, 2016

All Rights Reserved

ii

To the Faculty of Washington State University:

The members of the Committee appointed to examine the thesis of ARTHUR EDWARD

SCULLY find it satisfactory and recommend that it be accepted.

____________________________________

Daniel H. Thornton, Ph.D., Chair

____________________________________

Robert B. Wielgus, Ph.D.

____________________________________

Caren S. Goldberg, Ph.D.

iii

ACKNOWLEDGMENT

Funding for this research was provided by: Seattle City Light Wildlife Research Grant,

Washington Fish and Wildlife Aquatic Lands Enhancement Program Grant, NSF Catalyzing

New International Collaborations Grant (OISE # 1427542), and a Washington State University

New Faculty Seed Grant to D. Thornton. I especially thank the Washington Department of

Natural Resources for providing logistical and technical support for the field work. It is also

important to note: the data collection would not have been possible without the help and

expertise of Scott Fisher. I thank Dr. David Miller from Penn State University for assistance

with occupancy modeling. The data analysis would not have been possible without the

instruction and guidance from Dr. Daniel Thornton. This thesis would not have been possible

without the support of my lab mates, cohort, friends, and family.

iv

CLIMATE MEDIATED SPATIAL AND TEMPORAL SEGREGATION OF SYMPATRIC

FELIDS AT A RANGE EDGE CONTACT ZONE

Abstract

by Arthur Edward Scully, M.S.

Washington State University

December 2016

Chair: Daniel H. Thornton

Biotic Interactions may influence species distribution patterns, particularly at the edge of

their range, but remain poorly studied. In particular, we lack understanding of how climate may

mediate interactions between species. I studied spatial and temporal associations of Canada lynx

(Lynx canadensis) with their primary prey species (snowshoe hare; Lepus americanus) and

potential competitors (bobcats; Lynx rufus, and cougars; Puma concolor) along their southern

range edge in northern Washington. Although anecdotal evidence of these competitive

interaction exists, rigorous research remains absent from the literature. I collected

presence/absence data on these species from 2014-2016 across a 551 km2 landscape using

camera-traps along an elevation gradient in both snow-on and snow-off seasons. Then, using

single and two-species occupancy models, I used the data to examine three predictions: 1) lynx

occupancy is related to snowshoe hare availability, 2) distribution overlap between the sympatric

felids will be greatest in snow-off time periods, and 3) there will be evidence of spatial or

v

temporal avoidance of bobcats and cougars by lynx. Single-species models revealed probability

of occupancy was best modeled using abiotic and prey covariates for Canada lynx and abiotic

covariates for bobcats and cougar. The camera-trapping data revealed that spatial overlap of the

three felids species increased during snow-off time periods. Based on two-species occupancy

models Canada lynx showed a decrease in probability of occupancy when bobcats were present

at cameras, although this effect was scale dependent and did not vary between seasons. No

effect of cougar occupancy on lynx occupancy was seen during either season. All species had

high levels of overlap in activity throughout the year. Taken together, these results suggest that

biotic interactions are partly shaping large-scale lynx distribution patterns. The strong

relationship of lynx occupancy to snowshoe hare demonstrates the importance of this prey

species in shaping lynx habitat use. Observed negative interactions with bobcats at camera sites

provides empirical support for interactions between these species from modeling and anecdotal

evidence. Given these results, increasing temperatures and loss of snow may result in a

combination of habitat isolation and potential for increased competitive interactions for lynx at

their southern range edge.

vi

TABLE OF CONTENTS

Page

ACKNOWEDGEMENT............................................................................................................... iii

ABSTRACT .................................................................................................................................. iv

LIST OF TABLES ........................................................................................................................ vii

LIST OF FIGURES ................................................................................................................... .viii

CHAPTER 1

1. INTRODUCTION ...................................................................................................... 1

2. RESEARCH DESIGN AND METHODOLOGY ...................................................... 5

3. ANALYSIS AND DISCUSSION............................................................................... 15

BIBLIOGRAPHY ......................................................................................................................... 25

APPENDIX

A. Single Species Detection Models................................................................................ 45

B. Hourly Accumulation Curves .................................................................................... 46

C. Sunrise/Sunset Activity .............................................................................................. 47

vii

LIST OF TABLES

Page

1. Correlation Matrix .................................................................................................................... 33

2. Single Species Models ............................................................................................................. 34

3. Occupancy and Detection Estimates ........................................................................................ 35

4. 2-Species Occupancy Models .................................................................................................. 36

5. Daily Activity Overlap Estimates ............................................................................................ 37

viii

LIST OF FIGURES

Page

1. Study Area ................................................................................................................................ 38

2. Seasonal Detection Comparison Grid ...................................................................................... 39

3. Seasonal Detection Comparison Station .................................................................................. 40

4. Snowshoe Hare Influence on Lynx Occupancy ....................................................................... 41

5. Elevation Influence on Lynx, Bobcat, and Cougar Occupancy ............................................... 42

6. Slope Influence on Lynx and Bobcat Occupancy .................................................................... 43

7. Overlap of daily Activity Patterns ........................................................................................... 44

ix

Dedication

This thesis is dedicated to my parents, my advisor, my roommate, and my dog

who provided more support than they will ever know.

1

CHAPTER ONE

INTRODUCTION

Anthropogenic climate change is a critical environmental hazard that will dramatically alter

ecosystems worldwide. Species declines, local extinctions, and range shifts are already

occurring as a result of changing temperatures (Pounds et al. 1999; Thomas et al. 2004; Chen et

al. 2010; Wenger et al. 2011; Zhang et al. 2013), and these changes are expected to increase in

intensity over the coming decade. Despite climate change being one of the key conservation

challenges of the coming century, our ability to predict species’ responses to changing climates

remains inadequate (Araújo et al. 2005). To properly develop and implement effective

conservation strategies, a better understanding of the complex pressures that climate change will

impose upon each species is needed (Kissling et al. 2012).

A multitude of factors will impact species response to climate change (Martin 2001; Wenger et

al. 2011; HilleRisLambers et al. 2013); however, current climate-based niche models that have

formed the basis for distribution pattern forecasts and conservation policy guidelines (Franklin

2010; Early and Sax 2011; Schwartz 2012) mainly consider how temperature and precipitation

affect species distribution (Engler and Guisan 2009). Although these models provide a useful

starting point to understanding how climate change will influence species, they ignore a

potentially key factor influencing the nature of species response to climate change – their

interactions with other species. In fact, changing biotic interactions may be a more important

response to climate change than the effect of altered temperature and precipitation (Norberg et al.

2012; Schwartz 2012; Urban et al. 2012; Cahill et al. 2013). Accordingly, recent calls have been

made for improving our understanding of how biotic interactions change along climatic gradients

and shape range dynamics and large-scale distribution patterns (HilleRisLambers et al. 2013;

2

Alexander et al. 2016). Although experimental studies may ultimately be needed to address this

objective (Alexander et al. 2016), observational field studies of species interactions along

climatic gradients at the boundary of ranges may provide insight into how interactions shape use

of habitat at range limits, and thereby provide valuable information on the importance of biotic

interactions in influencing response to climate change.

Canada lynx (Lynx canadensis), at their southern range edge, are an excellent species for

assessing response to biotic interactions. Canada lynx are a cold-dependent feline specialist

whose southern range just enters the northern regions of the United States of America. Due to

their restricted range within the US borders, Canada lynx are federally listed as threatened and

are at greater risk of extirpation than their northern counterparts (Koehler and Aubry 1994).

Climate change is expected to negatively impact lynx and simple niche models predict

substantial range contraction for lynx as the climate warms (Peers et al. 2014). However, several

biotic interactions may be important for lynx persistence along the southern edge of their range.

Lynx are heavily dependent on snowshoe hare (Lepus americanus) as prey, with morphological

adaptations for pursuit of this prey in deep snow (Murray and Boutin 1991; Squires and

Ruggiero 2007). Although other prey may be utilized along the southern edge of lynx’s range

(Roth et al. 2007), snowshoe hare remain their most important prey. Consequently, the

distribution of snowshoe hare could influence lynx range limits and response to climate change

(Peers et al. 2014). In addition, several potential competitors may influence lynx distribution in

marginal environments. Bobcats (Lynx rufus) are closely related to Canada lynx, but lack the

adaptations for movement through deep snow. Bobcats are generalist predators who do not rely

on snowshoe hare but are still adept at hunting them (Dobson and Wigginton 1996). Bobcats

appear to be expanding northward (Lavoie et al. 2009), and current niche models predict that

3

bobcats will expand deep within the core Canada lynx habitat (Thornton, unpublished data).

Moreover, past modeling work and anecdotal evidence suggests the potential for competitive

interactions between these two species (Parker et al. 1983; Peers et al. 2013). Cougars (Puma

concolor) also compete with lynx through interference competition, but such competition may be

less intense during winter when deep snow prohibits cougar movement (Ruggiero et al. 1999).

Interactions between these three species remain poorly understood, and in particular, whether

and how these interactions may be mediated by environmental and seasonal variation.

Along the US/Canada Border, Canada lynx, cougars, and bobcats are sympatric in only a few

locations. These unique areas, which are at the southern range edge for Canada lynx, provide the

opportunity to increase our understanding of the degree to which these species overlap and

interact. Utilizing camera trapping across an expansive landscape in Northern Washington, and

single and conditional two-species occupancy models, I assessed broad-scale habitat and prey

associations of these three species, seasonal differences in spatial and temporal overlap, and

evaluted evidence for negative interactions between lynx and their potential competitors.

I hypothesized that if Canada lynx are avoiding otherwise suitable habitats due to exclusion from

sympatric felid competitors then they would partition their space and/or time use to avoid these

agonistic interactions and that this partitioning would be most prevalent in snow-off time periods

where deep snow does not limit use of the landscape by bobcats and cougars. Accordingly, I

tested several predicitons related to how biotic interactions influence lynx distribution patterns:

1) Broad-scale lynx occupancy patterns will be influenced by distribution of their primary

prey species, snowshoe hare.

2) Distributional overlap between lynx and their potential competitiors (bobcats and

cougars) will be greatest in snow-off time periods.

4

3) In areas of overlap, there will be evidence of lynx spatial or temporal avoidance of

competitors (bobcats and cougars), and that avoidance will be most pronounced in snow-

off time periods.

5

RESEARCH DESIGN AND METHODOLOGY

Study Area

I conducted the study in the Loomis State Forest within Okanogan County in north-central

Washington State, USA (Fig 1). Loomis contains one of the last remaining lynx populations in

the state (Washington Department of Wildlife 1993 – see Fig 2), where lynx are state-listed as

threatened, and are currently being considered for uplisting to endangered (J. Lewis, personal

communication). The study area consisted of the DNR managed lands within the forest which

has an area of 551 km2. Elevation in the study area ranged from 330-2520 m (average = 1266

m). The climate region of the area (the temperate steppe ecoregion) had hot dry summers, with

less than half of the precipiation falling as rain, and cold winters, where majority of the annual

precipitation falls as snow. The average annual precipitation accumulation was less than 50 cm

and the average temperature ranged from -9° to 30°C. As managed forests, the study area is

impacted by extractive forestry, cattle grazing (in summer and fall), and recreational activities.

The Loomis State Forest’s extensive topography and forestry practices created a diverse array of

overstory associations throughout the study area. The higher elevation areas were dominated by

lodgepole pine (Pinus contorta) in association with subalpine fir (Abies lasiocarpa) or

Englemann’s spruce (Picea englemannii) in areas of minimal disturbance. Ponserosa pine

(Pinus ponderosa) and Douglas fir (Pseudotsuga menziesii) dominated the mid to lower

elevations within the forest. The intermediate elevations consisted of a variety of associations of

the conifer species determined by a combination of elevation, slope, and aspect. Areas near

perennial streams often had quaking aspen (Populus tremuloides) as a dominant overstory

6

species; conversely, areas with low water retention and increased solar insolation (such as south

facing slopes) were dominated by grasses and sagebrush (Artemesia spp.).

Occupancy Data Collection

I collected data using remote infrared sensing cameras in the Loomis State Forest between July

2014 and August 2016. I divided the study area into 4 km2 grid cells (2 km by 2 km square cells)

over the study site. Within each grid cell, I randomly placed two cameras along roads and trails

with the restriction that cameras could be placed no closer than 500 m to each other. Cameras

were placed on the landscape in both snow-off (May-October) and snow-on (December-April)

time periods. I checked the cameras around 30 days after initial deployment to ensure there was

suffiecient battery and no malfunctions. Due to malfunction and theft, several grid cells did not

have two functional cameras during sampling. Cameras were left active for an average of 60

days at each site; however, there was some variability in how long cameras were left active at a

given site due to camera malfunction and accessibility, particularly in winter.

Data were downloaded from cameras and species (lynx, bobcat, cougar, and snowshoe hare)

identified by myself or trained volunteers. I automatically extracted the date/time of each photo

using ExifTool (Harvey 2015), and created spreadsheets containing this data for each photo for

each camera. From the spreadsheets of collected data, I created detection matrices for use in

occupancy modeling, and spreadsheets of the date and time of each detection of each species for

use in activity analysis. I created detection matrices of 0s (for not detected) and 1s (for detected)

for each species, for each camera, per 5-day sampling intervals within a deployment. Because

cameras were 500 m apart and often located on distinct trail system, and because I combined

detections into 5-day intervals, I considered detections at neighboring cameras to be independent

events. At the grid cell level, detection matrices were developed by denoting a 0 if none of the

7

cameras within a grid detected the focal species within the 5-day sampling interval and a 1 if at

least one detection occurred on any of the cameras in the grid during the 5-day sampling interval.

For the activity analysis each independent image of lynx, bobcat, and cougar was coded

according to 24 hour time of capture at each camera. I defined an independent image as captures

of a species on a camera greater than 60 minutes apart from a previous capture of that species on

the same camera.

Occupancy and Detection Covariates

Analysis of occupancy patterns occurred at two-scales: 1) the scale of the individual camera

location and 2) the scale of the 4km2 grid cell. I calculated several covariates that I hypothesized

would influence occupancy or detection in the immediate vicinity of each camera site (50m) and

at the grid cell level. Covariates for detection included trail type (primary roads that were

heavily used by vehicles vs. secondary roads and hiking trails), season (snow-off vs. snow-on),

and the number of cameras within a grid cell (grid-level analysis only). For occupancy

covariates, I selected a small number of variables that were known or suspected to influence lynx

distribution based on previous studies, and that also likely related strongly to occupancy of

bobcats and cougars. I purposely selected a small number of variables to reduce correlations

between predictors and decrease the number of parameters to be estimated in the single and two-

species occupancy models. At each camera site and grid cell I calculated several variables

related to the abiotic environment (‘abiotic’ model). These included elevation, slope, and aspect

(camera level only). Elevation and aspect relate strongly to temperature and snow

accumulation/retention as well as over story association on the landscape (Romano and Palladino

2002). Increasing elevation as well as north facing slopes (particularly in snow-off time periods)

has been found to correlate with increasing lynx habitat use (Koehler 1990; Ruggiero et al.

8

1999). Slope (the ratio of elevation change to horizontal distance) relates to ease of mobility

across a landscape, moisture retention, and over story association; lower levels of steepness in

slope has been found to be an important covariate to Canada lynx use (Koehler et al.2007).

Snow, temperature, and topographic conditions reflected in these abiotic variables also likely

exert strong effects on bobcat and cougar occupancy (Koehler and Hornocker 1991; Ruggiero et

al. 1999; Dickinson and Beier 2007; Peers et al. 2014). I used a digital elevation model from the

National Map Viewer (National Map 2015) to estimate average elevation within 50m of a

camera and within each grid cell. I used the Slope and Aspect tools in ArcMAP 10.4.1 (ESRI

2016) to create rasters of slope and aspect values for the study area. I calculated average slope

values within 50m of a camera and within each grid cell, and average values of aspect within

50m of the camera (aspect was ignored at the grid cell level because the variability within a grid

cell would not explain detection or use at the detectors).

I also calculated two variables related to the vegetative characteristics of the environment

(‘vegetation’ model). These included canopy cover and normalized difference vegetation index

(NDVI). Increasing canopy cover gives Canada lynx, their prey, and their competitors safety

while moving, resting, and hunting/foraging and lynx have been found to select for habitats with

increasing canopy cover (Squires et al. 2012). Normalized difference vegetation index gives a

measure of live green vegetation and its condition; NDVI is a ratio calculated by the near

infrared subtracted from the visible red sensed in a pixel divided by the near infrared added to

the visible red sensed in that pixel. Live green plants absorb the red wavelengths and reflect the

near infrared wavelengths. The greater the amount of healthy productive vegetation (high NDVI

values) the more browse and cover available, even in areas of low canopy cover. NDVI has been

found to be a positive indicator of lynx use (Carroll et al. 2001). Bobcats and cougars have also

9

been associated with heavier cover (Koehler and Hornocker 1991; Holmes and Laudré 2006;

Tucker et al. 2008; Thornton and Pekins 2015). I used the LANDFIRE (LANDFIRE 2012)

forest canopy cover layer to estimate the average percent tree canopy cover within 50m of a

camera and within each grid cell. I used multiband imagery from the National Agriculture

Imagery Program (NAIP 2014) and the Image Analysis Tools in ArcMAP 10.4.1 (ESRI 2016) to

create an NDVI raster for the study area and estimate the average NDVI within 50m of a camera

and within each grid cell. Finally, I calculated the ratio of snowshoe hare detections (no more

than one per hour per camera) to days the camera was working to get a variable reflecting the

availability of Canada lynx’s preferred prey at each camera station (‘prey’ model). For the grid

cell level, the coefficient of hare for each station with in the grid cell was averaged. Lynx are

heavily dependent on snowshoe hare across their range and therefore I expected this variable to

influence space use. Bobcats also predate on snowshoe hare (Knick et al. 1984), but I expected

this variable to be less influential because bobcats are not specialist predators on snowshoe hare.

All continuous covariates were standardized prior to analysis. I found no evidence of correlation

among the predictors used in the analysis (all correlation coefficients were < |0.45|; Table 1).

Analyzing Data

Single-species occupancy models

Occupancy models are the preferred approach for dealing with large-scale data on species

presence and absence, because they can account for the fact that species will not be detected

100% of the time in sites where they are present (MacKenzie et al. 2002). Through repeated

surveys of a site (or subsampling the data to generate repeat measures, as I did), a detection

history for each site can be generated. These detection histories can then be used to estimate

10

probability of occupancy (ψ) and detection (p). For example, a detection history of 101 at site k

would be modeled as ψ*p*(1-p)*p, where ψ is the probability of occupancy and p is the

probability of detection, given occupancy. A detection history of 000 could be the result of true

absence or failure to detect the species, and would be expressed as (1- ψ)+ ψ*(1-p)*(1-p)*(1-p).

Because lynx in my study may have had more than one camera-trap or grid cell in their home

range, and thus moved between those sites during sampling, the occupancy estimator is best

interpreted as “probability of use of a camera location” rather than probability of occupancy

(MacKenzie 2006). For simplicity, I continue to use the term occupancy throughout the rest of

the thesis. Detection histories for lynx, bobcat, and cougar were the detection matrices derived

previously.

I combined data from snow-off and snow-on time periods into one dataset, and created a final

covariate indicating the status of the season (snow-off vs. snow-on) prior to analysis. Although

camera stations changed locations between seasons, many of the same grid cells were sampled in

both seasons. However, open models of explicit dynamics (extinction/colonization) would be

fairly uninformative given the short time interval between seasons and the wide-ranging nature

of nature of Canada lynx (Squires and Oakleaf 2005; Burdett et al. 2007), and thus my approach

of combining the two time periods can be considered as modeling implicit dynamics where

changes between time periods occur more or less at random (MacKenzie 2006). Because I

expect that relationships between occupancy and habitat covariates will change between time

periods, I include interaction between season and habitat covariates in all models (see below).

I built occupancy models in a sequential manner. I first determined the best-fitting detection

parameters in single species models. I fit detection models using none of the occupancy

covariates (null models). For each null model, I tested several different detection models, where

11

I included neither, one, or both detection covariates (season and trail at the camera scale or

season and number of cameras at the grid scale; see Appendix A for list of models). I used the

model that had the best AICc (Burnham and Anderson 2002) of these models to determine

whether season, trail type or number of cameras, both, or neither would be used as detection

covariates (Appendix A). Using the best-fitting detection parameters, I then fit a series of single

species occupancy models for each species to determine the most important covariates

influencing occupancy. I tested abiotic, vegetative, and prey models singly and in combination

(Table 2). I compared models using AICc and selected the lowest AICc model as the best-fitting

model. Model goodness-of-fit was calculated on global models using the MacKenzie-Bailey

goodness of fit test (Mackenzie and Bailey 2004) implemented in program PRESENCE 10.1

(Hines 2006). This test compares the observed number of sites with a particular detection history

to the expected number based on the fitted model, and calculates a chi-square statistic.

Parametric bootstrapping is used to determine if the observed chi-square statistic is unusually

large compared to the bootstrapped samples; p-values <0.05 indicate a potentially poorly fitting

model. If prediction 1 was correct, I expected to see a strong positive relationship between prey

and Canada lynx occupancy in the single species models.

Two-species occupancy models

As a last step in the model-building process, I fit conditional two-species occupancy models

(Richmond et al. 2010) for pairs of potential competitors with lynx: lynx-bobcat and lynx-

cougar. Conditional two species occupancy models are a recently developed extension of

occupancy models that allow for the assessment of positive or negative interactions between

species (Richmond et al. 2010; Robinson et al. 2014). This is accomplished by allowing the

probability of occupancy or detection of a subordinate competitor to be dependent on occupancy

12

or detection of a dominant competitor at that site. Importantly these models also allow for the

inclusion of habitat covariates of importance to each species for occupancy and detection, to

decrease the likelihood that avoidance of the dominant species is confused with differential

habitat selection. I included the habitat covariates that appeared in best single-species models for

each species in the two-species models. I used a model selection approach to determine if lynx

occupancy and/or detection were influenced by the dominant species. I ran six different models:

1) “Occupancy-Detection” where both occupancy and detection of Canada lynx are affected by

the occupancy and detection of the dominant species, 2) “Occupancy-Constant” where

occupancy is held constant (occupancy of Canada lynx is not affected by the presence of the

dominant species, but detection is allowed to vary), 3) “Detection-Constant” where detection is

held constant (detection of Canada lynx is not affected by presence of the dominant species but

occupancy is allowed to vary), 4) “Detection-In-Interval-Constant” where detection within a

sampling interval is held constant (occupancy and detection of Canada lynx is affected by

occupancy and detection of the dominant species, but not by recent detection), 5) “Occupancy-

Detection-In-Interval-Constant”, where occupancy and detection within a sampling interval is

held constant (occupancy and detection within a sampling interval of Canada lynx is not affected

by the dominant species but overall detection is allowed to vary), and 6) “Occupancy-

Detection-Constant”, where detection and occupancy are held constant (occupancy and

detection of Canada lynx are not affected by occupancy or detection of the dominant

competitor). All models were initially fit by allowing the effect of the competitor to vary

between seasons, and holding that effect constant. However, these models were not competitive

and are thus not considered further. These models were then compared using AIC to see which

model best predicted occupancy and detection of Canada lynx. Finally, conditional two species

13

occupancy models enable the calculation of a species interaction factor (SIF): (ψA ψBA)/( ψA(ψA

ψBA+(1- ψA) ψBa), where ψA is the probability of occupancy for the dominant species (bobcat or

cougar), ψBA is the probability of occupancy for the subordinate species (lynx) when species A is

present, and ψBa is the probability of occupancy of species B when species A is absent. An SIF

of 1 indicates that the two species occur independently, where as an SIF less than one indicates

spatial avoidance – i.e., that the subordinate species is less likely to co-occur with the dominant

species. I expected that if prediction 3 is correct I would see evidence of avoidance of bobcat or

cougar by lynx (i.e., negative effect of the competitor on lynx occupancy in a best-fitting model

and/or SIF less than 1), and that this would be most pronounced in snow-off time periods. All

two-species occupancy models were fit in program PRESENCE 10.1 (HINES 2006).

Spatial Overlap

I calculated the amount of seasonal spatial overlap between Canada lynx, bobcat, and cougar in

several ways. I determined the number of cameras and grid cells that were jointly occupied by

each species in each season, derived a minimum convex polygon around occupied camera

stations and overlapped those polygons for lynx bobcat and lynx cougar in each season. I

calculated the summed occupancy probabilities between species pairs by projecting the final

single season occupancy models across the landscape. If prediction 2 was correct, I expected to

see greater distributional overlap between lynx and competitors in snow-off time periods.

Daily Activity Patterns

I used the package “overlap” (Meredith and Ridout 2014) to estimate the activity patterns of

Canada lynx, bobcats, and cougars during snow-on and snow-off time periods. I converted the

time data into radians and created density plots of each species activity patterns in both snow-on

14

and snow-off seasons that show frequency of activity during different times of the day. I

estimated the degree of temporal overlap of each species during each season, graphed that

overlap, and estimated confidence intervals of overlap using 10,000 bootstrap samples (Meredith

and Ridout 2016). Due to the length of my study, and thus the change in sunrise, solar-noon, and

sunset in respect to the time of day, detections were also quantified in their relation to these

biologically important temporal markers. I did this to avoid misinterpretation of activity overlap

if, in actuality, activity was partitioned around sunrise or sunset, but had overlapping time of day

as the sun’s position changed throughout the year (Nouvellet et al. 2011). I used the package

“maptools” (Bivand and Lewin-Koh 2015) to calculate sunrise and sunset for each camera

location for each detection and created boxplots of the distribution of overlap to compare among

species pairs. If prediction 3 was correct, I expected to see evidence of temporal segregation of

activity periods, and that segregation would be greater in snow-on time periods.

To confirm that I had enough detections to sufficiently represent the activity patterns of the

species of interest (Tambling et al. 2015), I used hourly accumulation curves from the package

“vegan” (Oksanen et al. 2012) . Briefly, these hourly accumulation curves are similar to species

accumulation curves, but instead of modeling how the number of species increases as more sites

are sampled, hourly accumulation curves show how new hourly time periods of activity are

added as more pictures are analyzed. Based on the hourly accumulation curves, all three species

reached approximate asymptotes with activity data in all 24-hour time periods (Appendix B),

indicating that I had sufficient data for the analysis.

15

ANALYSIS AND DISCUSSION

I surveyed 107 grid cells with 205 camera stations during the snow-on time period (16,259

camera trapping days) and 95 grid cells with 192 camera stations during the snow-off time

period (10,940 camera trapping days). Obvious patterns of spatial segregation of the three

species in both seasons are apparent from the detection data (Fig 3-5). In accordance with my

second prediction, the number of cameras with dual occupancy of lynx-bobcat and lynx-cougar

increased from snow-on to snow-on time periods from 8 to 10 and 9 to 14 respectively, as did the

number of dual occupied grid cells for lynx-cougar from 10 to 17. Lynx-bobcat dual occupied

grid cells stayed the same at 13 for snow-on and snow-off time periods. Area of overlap of

minimum convex polygons around grid cells also increased in snow-off time periods for lynx-

bobcat and lynx-cougar comparisons (increase in overlap of 91 km2 and 56 km2 respectively).

Single-species occupancy models

At the camera-level, best detection covariates were season for Canada lynx and trail and season

for bobcat and cougar (Appendix A). All three species had lower probabilities of detection in

snow-on time periods, where the odds of detection in snow-on time periods decreased by a factor

of 0.70, 0.49, and 0.35 at the grid cell level and 0.65, 0.39, and 0.37 at the camera station level

for lynx, bobcat, and cougar respectively. Bobcat and cougars were less likely to be detected at

cameras located on secondary vs. primary trails (odds of detection on secondary trails decreased

by a factor of 0.51 and 0.50 compared to primary trails for bobcat and cougars, respectively). At

the grid level, best detection covariates were season for Canada lynx, and number of cameras and

season for bobcats and cougars. Direction of the influence season was similar to the camera

level (i.e., greater detection probabilities in snow-off time periods), and grid cells with a single

16

camera decreased odds of detection by a factor of 0.35 and 0.90 compared to grid cells with

multiple cameras for bobcats and cougars, respectively

Best occupancy models for Canada lynx included the full model (abiotic + prey) at the grid cell

and camera station level (Table 2). Best occupancy models for bobcat included a model which

only included abiotic covariates at the grid cell and camera station level. Best occupancy models

for cougar only included a model with a season covariate at the grid cell level and only abiotic

covariates at the camera station level (Table 2).

Based on large parameter estimates relative to standard errors, prey was a highly influential

covariate of Canada lynx occupancy at the grid cell and camera level (Table 3; Figs 6 & 7). Odd

of occupancy increased by a factor of 1.93 for a standard deviation increase in hare detection

ratio. At the camera level elevation was highly influential (Table 3; Fig 8). Odds of occupancy

increased by 2.44 for a one standard deviation increase in elevation. There was also a significant

interaction between prey and season, and south facing slopes and season, with hare detection

ratio exerting a more positive effect on occupancy in snow-on time periods (Table 3; Fig 6 & 7).

In comparison to lynx, bobcats were responding less strongly to most covariates, and in

particular less strongly to both elevation and prey, with direction of influence opposite to lynx

(Figs 7 & 8). Seasonal interaction with slope was a highly influential predictor of bobcat

occupancy with a directional influence opposite to lynx at the grid cell level (Table 3; Fig 9).

For cougars, only season was influential at the grid cell level, with odds of occupancy in snow-

on time periods decreasing by a factor of 0.20 compared to snow-off time periods (Table 3). At

the camera level, elevation interacted strongly with season, exerting a positive influence on

occupancy in snow-off, and a slight negative effect in snow-on time periods (Table 3; Fig 8).

17

Two-species occupancy models

The best-fitting two species model for lynx-bobcat at the camera station level was a model in

which occupancy of Canada lynx depended on occupancy of bobcat, but detection of lynx is

unaffected by occupancy or detection of bobcats (Table 4). In accordance with my third

prediction, parameter estimates of this model indicated that lynx were negatively affected by the

presence of bobcats at a camera station. Parameter estimate of the effect (1.02) indicates that

odds of occupancy increased by a factor of 2.71 when bobcat were not present, though there was

substantial variability in this effect (95% CI of this effect just overlapped zero; -0.01 – 2.01).

However, contrary to expectations, avoidance wasn’t any more pronounced in snow-off vs.

snow-on, as models that allowed the effect of bobcat to vary per season were not competitive.

Moreover, SIF values calculated based on mean values of habitat covariates were both less than

1, indicating avoidance, but were similar between the two seasons (0.76, and 0.65 for snow-on

and snow-off, respectively). Although camera-level data indicated an effect of bobcats on lynx

occupancy, at the grid level, occupancy and detection of lynx was independent of occupancy and

detection of bobcats. A grid-level model with a slightly higher AIC than the best model which

allowed occupancy of lynx to vary according to bobcat presence (Detection-In-Interval-

Constant) did indicate a negative influence of bobcat on lynx as I found for the camera-level

models, but the effect (parameter estimate = 0.69) was smaller than the camera-level models and

more highly variable (95% CI on the effect; -0.49 – 1.88). Similar to the grid-level bobcat

analysis, best models for lynx-cougar indicated that occupancy and detection of lynx was

independent of occupancy and detection of cougar at both camera and grid levels (Table 4).

18

Daily activity patterns

Canada lynx had a mainly nocturnal activity pattern during the snow-off time periods and a

constant activity throughout the day and night in the snow-on time periods (Fig 10). Bobcat and

cougar had very similar nocturnal activity patterns to lynx during the snow-off time periods and

both displayed more activity throughout the day in snow-on time periods with had less mid-day

activity then Canada lynx (Fig 10). Estimates of activity overlap between lynx and bobcats

change from 81% during snow-on time periods to 86% during snow-off time periods, and

remained constant at 78% between lynx and cougar during both seasons (Table 5). Although the

estimate of activity overlap of bobcat and lynx during snow-off time periods does not fall within

the confidence interval for activity overlap during snow-on time periods, the reverse is not true.

This indicates that there may be a slight increase in temporal niche portioning between bobcats

and Canada lynx during snow-on time periods (Table 5), in accordance with my third prediction.

However, I found no significant partitioning of activity around sunrise, sunset, or solar noon

between Canada lynx and bobcats or cougars in snow-on or snow-off time periods (Appendix C).

Discussion

My analysis is one of the first to examine how biotic interactions influence large-scale

distribution patterns of a large vertebrate at their range limit. The results of my analysis show

that occupancy patterns of Canada lynx at the southern edge of their range are driven in part by

biotic interactions. Canada lynx responded strongly to prey availability even after accounting for

other potential habitat and topographic covariates. Moreover, the negative influence of bobcats

on use of camera sites by lynx also suggests that competitive interactions may play a role in

shaping lynx distribution on this landscape. While this influence appeared to be variable, having

19

a relatively high standard error, there was a marked decrease in probability of use of camera sites

by Canada lynx when bobcats were present. My results add to a growing literature regarding the

importance of biotic interaction in shaping distribution patterns and potentially range limits

(Urban et al. 2012; Wisz et al. 2013), which may be especially prevalent at the southern edge of

a species’ range where my study took place (MacArthur 1972; Normand et al. 2009). Of

particular note, in the context of climate change, is the fact that spatial overlap of lynx and

potential competitors was more pronounced in snow-off time periods. Although interaction

strength did not seem to vary greatly according to season, greater distributional overlap of the

three felids in snow-off time periods suggests the potential that interactions between these

species will become more common and perhaps heightened as the climate warms (Milazzo et al.

2013).

At both the grid cell and camera scale, the availability of snowshoe hare was positively

associated with Canada lynx occupancy, and that association was stronger in snow-on time

periods. Canada lynx’s dependence on snowshoe hare as well as their responses to declines in

hare is well documented (Brand et al. 1976; Ward and Krebs 1985; Koehler 1990; O’Donoghue

et al. 1997), and it appears, even at the southern edge of their range, where snowshoe hare

densities are comparatively low (Wirsing et al. 2002) and lynx diet is more diverse (Roth et al.

2007), hare can exert a strong influence on large-scale distribution patterns. The increased

influence of hare on lynx in the winter may be a function of lesser availability of alternative prey

or fewer intraguild competitors. Given that Canada lynx have a predation advantage in deep

snow (Murray and Boutin 1991), if snow-on seasons are reduced due to increasing temperatures

and decreasing precipitation, lynx’s ability to specialize on snowshoe hare may be jeopardized,

or exploitative competition with other carnivores may be more likely. Because bobcats/cougars

20

were not strongly associated with snowshoe hare on this landscape, the potential for food

competition between felids may be diminished. However, the degree to which exploitative

competition influence lynx remains a knowledge gap along their southern range edge (Murray et

al. 2008).

Evidence of negative interaction of lynx with bobcat supports research showing that bobcat may

be displacing lynx, or altering their niche, in areas of overlap (Parker et al. 1983; Peers et al.

2013). Lynx and bobcat are virtually the same size on this landscape, and bobcat densities

appear to be quite high in my study area (Scully, unpublished data), which could be a factor in

lynx avoidance of bobcats. However, avoidance on this landscape appears to be only manifest at

the smaller scale of the camera site, and does not scale up to the grid cell level, suggesting that

displacement may alter travel paths and fine scale use of the landscape but not necessarily

overall distribution patterns. Others have noted the potential scale dependent nature of biotic

interactions (Byers and Noonburg 2003; Sandel and Smith 2009), but I caution against

overinterpretation of these results given the smaller sample size of sites at the grid cell level

which reduced my power to detect interaction. In addition to spatial avoidance, lynx-bobcat

hybrids have been found along the southern range edge in Minnesota (Schwartz et al. 2004),

although this type of hybridization does not appear to be wide spread (Koen et al. 2014). The

marked topographic relief in my study area combined with the increased topographic separation

of bobcat and lynx during the breeding season (February-April), may reduce hybridization

potential on this landscape. This segregation could also be jeopardized by shorter and less

intense winter seasons. A failure to find any negative effect of cougar on lynx occupancy was

surprising; previous research has found that cougars will kill smaller felids when they are

sympatric (Koehler and Hornocker 1991), could easily scare a lynx off a recent kill (Ruggiero et

21

al. 1999), and have been found to be a significant source of lynx mortality (Squires and Laurion

2000). The experimental design of this research’s was meant to encompass multiple Canada

lynx home ranges, which has been found to be less than 50 km2 in north central Washington

(Aubry et al. 2000), not cougars, which can have home ranges over ten times that size in the

Pacific Northwest (Lambert et al. 2006). Given cougar’s large home ranges, it is possible that

the scale of this research was not broad enough to pick up on interference competition between

the two species. Finally, I acknowledge that camera-trapping, combined with two-species

occupancy modeling, allows some inference regarding association and avoidance of species

while controlling for habitat use, but my work could be strengthened by examining interactions

in more detail through telemetry or taking advantage of natural experiments like reintroductions

or novel range expansions (Alexander et al. 2016). At the landscape-scale over which I am

working, however, my approach provides an excellent starting point for exploring how

distribution patterns may be influenced by competitors in range-edge environments.

Single-species models supported many of my initial assumptions about the environmental

variables driving lynx occupancy. The topographic variables of slope, aspect and elevation have

all been found to be important factors in Canada lynx habitat selection (Koehler 1990; Ruggiero

et al. 1999; Koehler et al. 2007). High elevations, northern aspects, and gentle slopes are abiotic

factors tied to moisture retention (Romano and Palladino 2002) and were important positive

influences on Canada lynx occupancy during the snow off time periods. These same covariates,

though in the opposite direction of influence, were important to bobcat occupancy (Table 3).

Interestingly, the vegetative model did not come out as important in my analysis; this initially

appears to be at odds with Squires et al. (2010) who found that horizontal cover within age-

diverse stands of Engelmann spruce (Picea engelmannii) and subalpine fir (Abies lasiocarpa)

22

were important in lynx habitat selection. The covariates of canopy cover and NDVI I used to

assess vegetation’s influence on Canada lynx’s occupancy was likely too broad to pick up this

selection by Canada lynx. While both metrics are directly correlated with vegetative growth,

health and available cover, they would also include the dry forest types which Koehler (1990)

found that lynx selected against. Additionally, Stage and Salas (2007) found that the metrics of

aspect, slope, and elevation (the abiotic model) are predictors of forest associations. Therefore,

it is possible that the vegetative model was not included in the best occupancy models because

the abiotic model was better at predicting the overstory associations that lynx selected for and

against while the vegetative model included both without differentiating. The single species

models and topographic patterns of use in snow-on and snow-off seasons for all three species do

suggest that as climate change progresses and increasing temperatures and decreasing

precipitation (especially snow) intensifies Canada lynx’s available habitat will be reduced to high

elevation enclaves with increasing encroachment from their congeneric competitors.

My study also has important implications for large-scale monitoring and research on felids in the

Pacific Northwest. Although bobcats and pumas have been the subject of camera-based work

(Negrões et al. 2010; Clare et al. 2015; Thornton and Pekins 2015), lynx have been relatively

unstudied in this regard. Large scale surveys such as the one I conducted can be used to study

key environmental drivers of species distribution, inform the development of better predictive

models of response to anthropogenic threats, and aid in adaptive management decisions.

Camera-trapping has emerged as a powerful non-invasive method for the generation of large-

scale and long-term datasets of carnivore distribution (e.g., Ahumada et al. 2013; Steenweg et al.

2015). My results show that camera trapping was effective at detecting these three species in

boreal forest environments. Detection probabilities were fairly high per 5-day trapping session

23

in both snow-off and snow-on time periods for all three species at the grid cell level for Canada

lynx (0.21 and 0.08, respectively) as well as bobcats and cougars in cells with more than one

camera (0.17 and 0,09 for bobcats and 0.13 and 0.09 for cougars). Detection probabilities at the

camera level per 5-day trapping session in both time periods were also fairly high for Canada

lynx (0.063 and 0.161) as well as bobcats, and cougars on primary trails (0.079 and 0.124 for

bobcats and 0.064 and 0.099 for cougars, respectively). This suggests that 60-90 day periods

were sufficient for the survey. The drop in probability of detection for all three species during

snow on time periods was likely due to the fact that the roads and trails where I placed cameras

became less of a draw as movement off road was no longer contrastingly more difficult.

Regardless, my methodology generated ample data on all three species across a large landscape

which suggests this methodology could be applied more widely in this region.

Compelling evidence for temporal niche partitioning was not suggested during any

season. All three species exhibited high amounts of activity overlap throughout the year

(ranging from 78-86%), with all species switching from nocturnal activity patterns during the

snow-off time period to activity throughout the day and night during the snow-on time periods.

Similarly, no species partitioned their activity around important sun zeniths (sunrise or sunset).

While temporal niche partitioning does not appear to be a method lynx are using to avoid conflict

with sympatric felids, my methodology generated enough data to explore temporal overlap with

the confidence that activity patterns of each species in each hour were adequately represented.

Thus, this methodology could be applied to a wide range of species to explore activity overlap or

segregation.

My results also have management implications for lynx, which are threatened in the US. Canada

lynx, at their southern range, are at increased vulnerability to the predicted effects of climate

24

change. Canada lynx, in this area, occupy environments with abiotic factors that lead to

increased moisture retention, decreased temperatures (Romano and Palladino 2002; Pohl et al.

2014), and specific over story associations (Koehler 1990; Stage and Salas 2007; Squires et al.

2010). These habitats are at risk of becoming increasingly isolated as climate change progresses,

like the sky island populations of American pika (Ochotona princeps) (Galbreath et al. 2009)

which has led to extinctions of isolated populations since 1999 (Beever et al. 2011). Protecting

habitat in high elevation environments that will be the most resistant to climate change, as well

as paths of connectivity, between high elevation environments, is imperative to Canada lynx’s

persistence at their range edge. Encouraging robust snowshoe hare populations will also be

highly beneficial, as found elsewhere (Stenseth et al. 1997). Congeneric competition appears

potentially important biotic factor in lynx habitat use, as I have found avoidance of sympatric

bobcats by lynx (Table 4), which could increase as climate change progresses and Canada lynx’s

morphological advantages are no longer superior to their competitors. Further monitoring of

lynx and their interactions with bobcats at their southern range should be considered; it is likely

that the areas which are both optimal lynx habitat and poor bobcat habitat will be reduced

leading to increased spatial overlap and competition. If further work supports the supposition of

negative interactions with bobcats and increased contact due to climate change, increased take of

bobcat, where sympatric with lynx, may reduce pressures on lynx populations and improve

survival of this federally threatened species.

25

BIBLIOGRAPHY

Ahumada, J. A., J. Hurtado, and D. Lizcano. 2013. Monitoring the status and trends of tropical

Forest terrestrial vertebrate communities from camera trap data: a tool for conservation.

Plos ONE 8:e73707.

Alexander, J. M., J. M. Diez, S. P. Hart, and J. M. Levine. 2016. When climate reshuffles

competitors: a call for experimental macroecology. Trends in Ecology and Evolution 31:

831-841.

Araújo, M. B., R. G. Pearson, W. Thuiller, M. Erhard. 2005. Validation of species-climate

impact models under climate change. Global Change Biology 11:1504-1513.

Aubry, K. B., G. M. Koehler, and J. R. Squires. 2000. Ecology of Canada Lynx in Southern

Boreal Forests. Chapter 13 In L. F. Ruggiero, K. B. Aubry, S.W. Buskirk, G. M. Koehler,

C. J. Krebs, K. S. McKelvey, and J. R. Squires. Ecology and conservation of lynx in the

United States. University Press of Colorado. Boulder, CO. 480

Beever, E. A., C. Ray, J. L. Wilkening, P. F. Brussard, and P. W. Mote. 2011. Contemporary

climate change alters the pace and drivers of extinction. Global Change Biology

17:2054-2070.

Bivand, R. and N. Lewin-Koh. 2015. maptools: Tools for Reading and Handling Spatial

Objects. R package version 0.8-37. http://CRAN.R-project.org/package=maptools

Brand, C. J., L. B. Keith, and C. A. Fischer. 1976. Lynx responses to changing snowshoe hare

densities in central Alberta. Journal of Wildlife Management 40:416-428.

Burnham, K. P., and D. R. Anderson. Model Selection and Multimodel Inference: A Practical

Information-theoretical Approach. New York: Springer, 2002. Print.

Burdett, C. L., R. A. Moen, G. J. Niemi, and L. D. Mech. 2007. Defining space use and

movements of Canada lynx with global positioning system telemetry. Journal of

Mammalogy 88:457-467.

Byers, J. E. and E. G. Noonburg. 2003. Scaledependent effects of biotic resistance to biolotical

invasion. Ecology 84::1428-1433.

Cahill, A. E., M. E. Aiello-Lammens, M. C. Fisher-Ried, X. Hua, C. J. Karanewsky, and H. Y.

Ryu. 2013. How does climate change cause extinction? Proceedingsof the Royal

Society of B 280:1-10.

Carroll, C., R. F. Noss, and P. C. Paquet. 2001. Carnivores as focal species for conservation

planning in the Rocky Mountain region. Ecological Applications 11:961-980.

26

Clare, J. D. J., E. M. Anderson, D. M. MacFarland, and B. L. Sloss. 2015. Comparing the costs

and detectability of bobcat using scat-detecting dog and remote camera surveys in central

Wisconsin. Wildlife Society Bulletin 39:210-217.

Chen, P-Y, C. Welsh, and A. Hamann. 2010. Geographic variation in growth response of

Douglas-fir to interannual climate change variability and projected climate change.

Global Change Biology 16:3374-3385

Dickson, B. G. and P. Beier. 2007. Quantifiying the influence of topographic position on cougar

(Puma concolor) movement in southern California USA. Journal of Zoology 271:270-

277.

Dobson, F. S. and J. D. Wigginton. 1996. Environmental influences on the sexual dimorphism

in body size of western bobcats. Oecologia 108:610-616.

Early, R. and D. F. Sax. 2011. Analysis of climate paths reveals potential limitations on species

range shifts. Ecology Letters 14: 1125-1133.

Engler, R. and A. Guisan. 2009. MIGCLIM: predicting plant distribution and dispersal in a

changing climate. Diversity and Distributions 15:590-601.

ESRI, 2016. ArcGIS Desktop: Release 10.4.1. Redlands, CA: Environmental Systems

Research Institute.

Franklin, J. 2010. Moving beyond static distribution models in support of conservation

biogeography. Diversity and Distributions 16:321-330.

Galbreth, K. E., D. J. Hafner, and K. R. Zamudio. 2009. When cold is better: climate driven

elevation shifts yield complex patterns of diversification and demography in an alpine

specialist (American pika, Ochotona princeps). Evolution 63:2848-2863.

Harvey, P. 2015. ExifTool Kingston, Ontario: Queen’s University.

HilleRisLambers, J., M. A. Harsch, A. K. Ettinger, K. R. Ford, and E. J. Theobald. 2013. How

will biotic interactions influence climate change-induced range shifts? Annals of the

New York Academy of Sciences 1297:112-125.

Hines, J. E. 2006. PRESENCE2: software to estimate patch occupancy and related parameters.

U.S. Geological Survey, Patuxent Wildlife Research Center. Laurel, Maryland, USA.

http://www.mbr-pwrc.usgs.gov/software/presence.html

Holmes, B. R. and J. W. Laundré. 2006. Use of open, edge, and forest areas by pumas Puma

concolor in winter: are pumas foraging optimally? Wildlife Biology 12:201-209.

27

Kissling, W. D., C. F. Dormann, J. Groeneveld, T. Hickler, I. Kühn, G. J. McInerny, J. M.

Montoya, C. Römermann, K. Schiffers, F. M. Schurr, A. Singer, J. Svenning, N. E.

Zimmerman, and R. B. O’Hara. 2012. Towards novel approaches to modelling biotic

interactions in multispecies assemblages at large spatial extents. Journal of

Biogeography 39:2163-2178.

Knick, S. T., S. J. Sweeney, J. R. Alldredge, and J. D. Brittell. 1984. Autumn and Winter food

habits of bobcats in Washington State. The Great Basin Naturalist 44:70-74.

Koehler, G. M. 1990. Population and habitat characteristics of lynx and snowshoe hare in north

central Washington. Canadian Journal of Zoology 68:845-851.

Koehler, G. M. and M. G. Hornocker. 1991. Seasonal resource use among mountain lions,

bobcats, and coyotes. Journal of Mammalogy 72:391-396.

Koehler, Gary M.; Aubry, Keith B. 1994. Chapter 4: Lynx. In: Ruggiero, Leonard F.; Aubry,

Keith B.; Buskirk, Steven W.; Lyon, L. Jack; Zielinski, William J., tech. eds. The

scientific basis for conserving forest carnivores: American marten, fisher, lynx, and

wolverine in the western United States. Gen. Tech. Rep. RM-254. Fort Collins, CO: U.S.

Department of Agriculture, Forest Service, Rocky Mountain Forest and Range

Experiment Station. p. 74-98

Koehler, G. M., B. T. Maletzke, J. A. Von Kienast, K. B. Aubry, R. B. Wielgus, and R. H.

Naney. 2007. Habitat fragmentation and the persistence of lynx populaitons in

Washington state. The Journal of Wildlife Management 72:1518-1524.

Koen, E.L., J. Bowman, J. L. Lalor, P. J. Wilson. 2014. Continental-scale assessment of the

hybrid zone between bobcat and Canada lynx. Biological Conservation 178:107-115.

Lambert, C. M. S., R. B. Wielgus, H. S. Robinson, D. D. Katnik, H. S. Cruickshank, R. Clarke,

and J. Almack. 2006. Cougar population dynamics and viability in the Pacific

Northwest. Journal of Wildlife Management 70:246-254.

LANDFIRE, 2012, Existing Vegetation Type Layer, LANDFIRE 1.3.0, U.S. Department of the

Interior, Geological Survey. Accessed 10 December 2015

at http://landfire.cr.usgs.gov/viewer/.

Lavoie, M., P. Collin, F. Lemieux, H. Jolicoeur, P. Canac-marquis, and S. Lariviére. 2009.

Understanding fluctuations in bobcat harvest at the northern limit of their range. The

Journal of Wildlife Management 73:870-875.

MacArthur R. 1972. Geographical ecology: patterns in the distribution of species. New York,

NY: Harper & Row.

28

MacKenzie, D. I, J. D. Nichols, G. B. Lachman, S. Droege, J. A. Royle, and C. A. Langtimm.

2002. Estimating site occupancy rates when detection probabilities are less than one.

Ecology 83:2248-2255.

MacKenzie, D. I. and L. L. Bailey. 2004. Assessing the fit of site-occupancy models. Journal

of Agricultural, Biological, and Environmental Statistics 9:300-318.

MacKenzie, D. I. 2006. Modeling the probability of resource use: the effect of, and dealing

with, detecting a species imperfectly. Journal of Wildlife Management 70:367-374.

Martin, T. E. 2001. Abiotic vs. biotic influences on habitat selection of coexisting species:

climate change impacts? Ecology 82:175-188.

McCutchan, M. H. and D. G. Fox. 1986. Effect of elevation and aspect on wind, temperature,

and humidity. Journal of Applied Meterology and Climatology 25:1996-2013.

Meredith, M. and M. Ridout. 2014. Overview of the overlap package.

https://cran.r-project.org/web/packages/overlap/vignettes/overlap.pdf

Meredith, M. and M. Ridout. 2016. overlap: Estimates of coefficient of overlapping for

animal activity patterns. R package version 0.2.4

Milazzo, M. S. Mirto, P. Domenici, and M. Gristina. 2013. Climate change exacerbates

interspecific interactions in sympatric coastal fishes. Journal of Animal Ecology 82:468-

477.

Murray, D. L. and S. Boutin. 1991. The influence of snow on lynx and coyote movement: does

morphology affect behavior? Oecologia 88:463-469.

Murray, D. L., T. D. Steury, J. D. Roth. 2008. Assessment of Canada lynx research and

conservation needs in the southern range: another kick at the cat. The Journal of

Wildlife Management 72:1463-1472.

NAIP, 2014. USGS. National Agriculture Imagery Program (NAIP). URL:

https://lta.cr.usgs.gov/NAIP

National Map, 2015. U.S. Department of the Interior, Geological Survey. Accessed 2 December

2015 at http://nationalmap.gov/index.html

Negrões, N, P. Sarmento, J. Cruz, C. Eira, E. Revilla, C. Fonseca, R. Sollmann, N. M. Tŏrres, M.

M. Furtado, A. T. A. Jácomo, L. Silveira. 2010. Use of camera-trapping to estimate

puma density and influencing factors in central Brazil. The Journal of Wildlife

Management 74:1195-1203.

29

Nielsen, C. K. and M. A. McCollough. 2009. Considerations on the use of remote cameras to

detect Canada lynx in northern Maine. Northeastern Naturalist 16:153-157.

Norberg, J., M. C. Urban, M. Vellend, C. A. Klausmeier, and N. Loeuille. 2012. Eco-

evolutionary responses of biodiversity to climate change. Nature Climate Change 2:747-

751.

Normand, S., U. A. Treier, C. Randin, P. Vittoz, A. Guisan, J. Svenning. 2009. Importance of

abiotic stress as a range-limit determinant for European plants: insights from species

responses to climatic gradients. Global Ecology Biogeography 18:437-449.

Nouvellet, P. , G. S. A. Rasmussen, D. W. Macdonald, F. Courchamp. 2011. Noisy clocks and

silent sunrises: measurement methods of daily activity pattern. Journal of Zoology

286:179-184.

O’Donoghue, M., S. Boutin, C. J. Krebs, and E. J. Hofer. 1997. Numerical responses of coyotes

and lynx to the snowshoe hare cycle. OIKOS 80:150-162.

Oksanen, J., F. G. Blanchet, R. Kindt, P. Legendre, D. McGlinn, P. R. Minxhin, R. B. O’Hara,

G. L. Simpson, P. Solymos, M. H. M. Stevens, E. Szoecs, H. Wagner. 2016. Vegan:

community ecology package. R package version 2.4-1 edn, https://cran.r-

project.org/web/packages/vegan/vegan.pdf

Parker, G. R., J. W. Maxwell, and L. D. Morton. 1983. The ecology of the lynx (Lynx

canadensis) on Cape Breton Island. Canadian Journal of Zoology 61:770-786.

Peers, M. J. L., D. H. Thornton, and D. L. Murray. 2013. Evidence for large-scale effects of

competition: niche displacement in Canada lynx and bobcat. Proceedings of the Royal

Society London B 280:1-10.

Peers, M. J. L., M. Wehtje, D. H. Thornton, and D. L. Murray. 2014. Prey switching as a means

of enhancing persistence in predators at the trailing southern edge. Global Change

Biology 20:1126-1135.

Pohl, S., J. Garvelmann, J. Wawerla, and M. Weiler. 2014. Potential of a low-cost sensor

network to understand the spatial and temporal dynamics of a mountain snow cover.

Water Resources Research 50:2533-2550.

Pounds, J. A., M. P. L. Fogden, and J. H. Campbell. 1999. Biological response to climate

change on a tropical mountain. Nature 398:611-615.

Richmond, O. M. W., J. E. Hines, and S. R. Beissinger. 2010. Two-species occupancy models:

a new parameterization applied to co-occurrence of secretive rails. Ecological

Applications 20:2036-2046

30

Robinson, Q. H., D. Bustos, G. W. Roemer. 2014. The application of occupancy modeling to

evaluate intraguild predation in a model carnivore system. Ecology 95:3112-3123.

Roth, J. D., J. D. Marshall, D. L. Murray, D. M. Nickerson, T. D. Steury. 2007. Geographic

gradients in diet affect population dynamics of Canada lynx. Ecology 88:2736-2743.

Romano, N. and M. Palladino. 2002. Prediction of soil water retention using soil physical data

and terrain attributes. Journal of Hydrology 265:56-75.

Ruggiero, L. F., K. B. Aubry, S.W. Buskirk, G. M. Koehler, C. J. Krebs, K. S. McKelvey, J. R.

Squires. 1999. Ecology and conservation of Lynx in the United States. United States

Department of Agriculture.

Sandel, B. and A. B. Smith. 2009. Scale as a lurking factor: incorporating scale-dependence in

experimental ecology. OIKOS 118:1284-1291.

Schwartz, M. K., K. L. Pilgrim, K. S. McKelvey, E. L. Lindquist, J. J. Claar, S. Loch, L. F.

Ruggiero. 2004. Hybridization between Canada lynx and bobcats: genetic results and

management implications. Conservation Genetics 5:349-355.

Schwartz, M. W. 2012. Using niche models with climate projections to inform conservation

management decisions. Biological Conservation 155:149-156.

Squires, J. R. and T. Laurion. 2000. Lynx home range and movement in Montana and

Wyoming -preliminary results. Chapter 11 In L. F. Ruggiero, K. B. Aubry, S.W.

Buskirk, G. M. Koehler, C. J. Krebs, K. S. McKelvey, and J. R. Squires. Ecology and

conservation of lynx in the United States. University Press of Colorado. Boulder, CO.

480

Squires, J. R. and R. Oakleaf. 2005. Movements of a male Canada lynx crossing the greater

Yellowstone Area, including highways. Northwest Science 79:196-201.

Squires, J. R. and L. F. Ruggiero. 2007. Winter prey selection of Canada lynx in northwestern

Montana. The Journal of Wildlife Management 71:310-315.

Squires, J. R., N. J. Decesare, J. A. Kolbe, and L. F. Rugiero. 2010. Seasonal resource selection

of Canada lynx in managed forests of the Northern Rocky Mountains. The Journal of

Wildlife Management 74:1648-1660.

Squires, J. R., N. J. DeCesare, L. E. Olson, J. A. Kolbe, M. Hebblewhite, and S. A. Parks. 2012.

Combining resource selection and movement behavior to predict corridors for Canada

lynx at their southern range periphery. Biological Conservation 157:187-195.

31

Stage, A. R. and C. Salas. 2007. Interactions of elevation, aspect, and slope in models of forest

species composition and productivity. Forest Science 53:486-492.

Steenweg, R., J. Whittington, and M. Hebblewhite. 2015. Canadian Rockies remote camera

multi-species occupancy project: examining trends in carnivore populations and their

prey using remote cameras in Banff, Jasper, Kootenay, Yoho and Waterton Lakes

National Parks. Final report. Parks Canada.

Stenseth, N. C., W. Falck, O. N. Bjornstad, and C. J. Krebs. 1997. Population regulation in

snowshoe hare and Canadian lynx: asymmetric food web configurations between hare

and lynx. Proceedings of the National Academy of Sciences 94:5147-5152.

Tambling, C. J., L. Minnie, J. Meyer, E. W. Freeman, R. M. Santymire, J. Adendorff, and G. I.

H. Kerley. 2015. Temporal shifts in activity of prey following large predator

reintroductions. Behavioral Ecology and Sociobiology 7:1153-1161.

Thomas, C. D., A. Cameron, R. E. Green, M. Bakkenes, L. J. Beaumont, Y. C. Collingham, B. F.

N. Erasmus, M. F. de Siqueira, A. Grainger, L. Hannah, L. Hughes, B. Huntley, A. S. van

Jaarsveld, G. F. Midgley, L. Miles, M. A. Ortega-Huerta, A. T. Peterson, O. L. Phillips,

and S. E. Williams. 2004. Extinction risk from climate change. Nature 427:145-148.

Thornton, D. H. and C. E. Pekins. 2015. Spatially explicit capture-recapture analysis of bobcat

(Lynx rufus) density: implications for mesocarnivore monitoring. Wildlife Research

42:394-404.

Tucker, S. A., W. R. Clark, and T. E. Gosselink. 2008. Space use and habitat selection by

bobcats in the fragmented landscape of south-central Iowa. The Journal of Wildlife

Management 72:1114-1124.

Urban, M. C., J. J. Tewksbury, and K. S. Sheldon. 2012. On a collision course: competition

and dispersal differnces create no-analogue communitites and cause extincitons during

climate change. Proceedings of the Royal Society Series B 279:2072-2080.

Ward, M. P. and C. J. Krebs. 1985. Behavioural responses of lynx to declining snowshoe hare

abundance. Canadian Journal of Zoology 63:2817-2824.

Washington Department of Wildlife. 1993. Status of the North American lynx (Lynx

canadensis) in Washington. Unpubl. Rep. Wash. Dept. Wildl., Olympia.

Wenger, S. J., D. J. Isaak, C. H. Luce, H. M. Neville, K. D. Fausch, J. B. Dunham, D. C.

Dauwalter, M. K. Young, M. M. Elsner, B. E. Rieman, A. F. Hamlet, and J. E. Williams.

2011. Flow regime, temperature, and biotic interactions drive differential declines of

trout species under climate change. Proceedings of the National Academy of Sciences of

the United States of America 108:14175-14180.

32

Wirsing, A. J., T. D. Steury, D. L. Murray. 2002. A demographic analysis of a southern

snowshoe hare population in a fragmented habitat: evalusting the refugium model.

Canadian Journal of Zoology 80:169-177.

Wisz, M. J., J. Pottier, D. W. Kissling, L. Pellissier, J. Lenoir, C. F. Damgaard, C. F. Dormann,

M. C. Forchhammer, J. Grytnes, A. Guisan, R. K. Heikkinen, T. T. Hove, I. Kuhn, M.

Luoto, L. Maiorano, M. Nilsson, S. Norman, E. Ockinger, N. M. Schmidt, M.

Termansen, A. Timmermann, D. A. Wardle, P. Aastrup, J. Svenning. 2013. The role of

biotic interactions in shaping distributions and realized assemblages of species:

implications for species distribution modelling. Biological Reviews 88:15-30.

Zhang, M.. Z. Zhou, W. Chen, C. H. Cannon, N. Raes, and J. W. F. Slik. 2013. Major declines

in woody plant species ranges under climate change in Yunnan, China. Diversity and

Distributions 20:405-415.

33

TABLE 1

Grid Cell Scale

NDVI CAN ELE SLO PREY

NDVI - 0.175867 -0.44841 0.174882 0.060159

CAN 0.175867 - -0.1225 0.131573 0.101305

ELE -0.44841 -0.1225 - -0.20303 0.262409

SLO 0.174882 0.131573 -0.20303 - -0.18229

PREY 0.060159 0.101305 0.262409 -0.18229 -

Camera Station Scale

NDVI CAN ELE SLO PREY

NDVI - -0.00434 -0.34464 0.113153 0.078806

CAN -0.00434 - -0.09967 0.115497 0.021217

ELE -0.34464 -0.09967 - -0.17042 0.223065

SLO 0.113153 0.115497 -0.17042 - -0.13419

PREY 0.078806 0.021217 0.223065 -0.13419 -

Correlation matrix of occupancy covariates at the grid and camera station level

34

TABLE 2

Canada Lynx Grid Level Bobcat Grid Level Cougar Grid Level

Model AICc AICc Weight Model AICc

AICc

Weight Model AICc AICc Weight

abiotic + prey 872.0996 0.8886 abiotic 1470.968 0.4746 season 942.7284 0.7369

Abiotic + vegetative + prey 876.7986 0.0848 abiotic + prey 1471.008 0.4652 prey 946.5063 0.1114

prey 880.6757 0.0122 vegetative + prey 1477.114 0.0220 abiotic 946.6288 0.1048

abiotic 881.3909 0.0085 null 1478.334 0.0119 abiotic + prey 949.747 0.0220

vegetative + prey 882.3011 0.0054 abiotic + vegetative 1479.192 0.0078 vegetation 950.5334 0.0149

abiotic + vegetative 887.2789 0.0004 abiotic + vegetative + prey 1479.356 0.0072 abiotic + vegetative 954.5358 0.0060

null 900.1059 0.0000 season 1480.244 0.0046 vegetative + prey 954.6263 0.0020

season 900.5034 0.0000 prey 1480.512 0.0040 abiotic + vegetative + prey 956.9486 0.0019

vegetative 906.2989 0.0000 vegetative 1481.253 0.0028 null 965.158 0.0000

Canada Lynx Camera Station Level Bobcat Camera Station Level Cougar Camera Station Level

Model AICc AICc Weight Model AICc AICc

Weight Model AICc AICc Weight

abiotic + prey 1138.577 0.9100 abiotic 1894.266 0.4352 abiotic 1288.79 0.4796

abiotic + vegetative + prey 1143.215 0.0895 abiotic + prey 1895.375 0.2500 abiotic + prey 1289.827 0.2855

abiotic 1155.249 0.0002 abiotic + vegetative 1895.708 0.2116 season 1291.272 0.1386

abiotic + vegetative 1155.416 0.0002 abiotic + vegetative + prey 1897.836 0.0730 prey 1294.333 0.0300

vegetative + prey 1162.836 0.0000 vegetative 1901.157 0.0139 abiotic + vegetative 1294.462 0.0281

prey 1163.9 0.0000 null 1902.699 0.0064 abiotic + vegetative + prey 1294.521 0.0273

null 1190.917 0.0000 season 1903.188 0.0050 vegetative 1297.828 0.0052

season 1192.201 0.0000 vegetative + prey 1904.048 0.0033 null 1298.218 0.0043

vegetative 1195.506 0.0000 prey 1905.436 0.0016 vegetative + prey 1300.649 0.0013

AICc values of single species occupancy models for Canada lynx, bobcat, and cougar at the grid cell and camera station level

35

TABLE 3

Estimates

Species

Snow-

On Asp (S) Elevation Slope NDVI Canopy Prey Snow:Elevation S:Slope S:Asp(S) S:NDVI S:Canopy S:Prey S(Det) Camera #

Trail

(2nd)

CL.g

1.071

(0.595)

-

0.523

(0.414)

-0.276

(0.362)

-

-

0.662

(0.231)

0.629

(0.617)

-0.251

(0.493)

-

-

-

0.740

(1.098)

-1.20

(0.245)

-

-

CL.50

-0.705

(0.650)

-0.504

(0.432)

0.891

(0.348)

-0.233

(0.239)

-

-

0.457*

(0.186)

0.106

(0.519)

-0.33

(0.408)

2.426*

(0.882)

-

-

2.06

(0.902)

-1.04

(0.236)

-

-

B.G

-0.211

(0.551)

-

0.1498

(0.301)

0.0773

(0.274)

-

-

-

-0.2425

(0.466)

1.5168*

(0.666)

-

-

-

-

-0.671

(0.179)

-0.432

(0.270)

-

B.50

-0.875

(0.651)

-0.616

(0.4523)

-0.413

(0.406)

0.152

(0.241

-

-

-

0.121

(0.492)

0.875

(0.539)

0.724

(0.665)

-

-

-

-0.491

(0.194)

-

-0.681

(0.196)

C.G

-1.619

(0.472)

-

-

-

-

-

-

-

-

-

-

-

-

-0.436

(0.273)

-2.312

(0.614)

-

C.50

-1.9713*

(0.616)

0.354

(0.487)

0.7163

(0.385)

0.3911

(0.294)

-

-

-

-1.0344*

(0.470)

-0.3531

(0.379)

1.0847

(0.729)

-

-

-

-0.468

(0.288)

-

-0.702

(0.241)

Occupancy and Detection Estimates from the best models for Canada lynx, bobcats, and cougars at the scale of the grid cell and

camera station. Standard errors for each estimate is shown below in parentheses.

36

TABLE 4

Bobcat-Lynx Grid Level

Model AIC deltaAIC AIC weight Model Likelihood # of Parameters -2*LogLike

Occupancy-Detection-Constant 2331.35 0.00 0.3070 1.0000 20 2291.35

Occupancy-Detection-In-Interval-Constant 2331.89 0.54 0.2344 0.7634 21 2289.89

Detection-In-Interval-Constant 2332.63 1.28 0.1619 0.5273 22 2288.63

Detection-Constant 2332.89 1.54 0.1422 0.4630 21 2290.89

Occupancy-Constant 2333.79 2.44 0.0906 0.2952 22 2289.79

Occupancy-Detection 2334.49 3.14 0.0639 0.2080 23 2288.49

Null 2364.68 33.33 0.0000 0.0000 11 2342.68

Cougar-Lynx Grid Level

Model AIC deltaAIC AIC weight Model Likelihood # of Parameters -2*LogLike

Occupancy-Detection-Constant 1817.29 0.00 0.9983 1.0000 16 1785.29

Null 1830.02 12.73 0.0017 0.0017 11 1808.02

Bobcat-Lynx Camera Station Level

Model AIC deltaAIC AIC weight Model Likelihood # of Parameters -2*LogLike

Detection-Constant 3031.05 0.00 0.3714 1.0000 25 2981.05

Occupancy-Detection-Constant 3032.28 1.23 0.2008 0.5406 24 2984.28

Detection-In-Interval-Constant 3032.68 1.63 0.1644 0.4426 26 2980.68

Occupancy-Constant 3033.18 2.13 0.1280 0.3447 26 2981.18

Occupancy-Detection 3033.94 2.89 0.0876 0.2357 27 2979.94

Occupancy-Detection-In-Interval-Constant 3035.15 4.10 0.0478 0.1287 25 2985.15

Null 3089.23 58.18 0.000 0.0000 11 3067.23

Cougar-Lynx Camera Station Level

Model AIC deltaAIC AIC weight Model Likelihood # of Parameters -2*LogLike

Occupancy-Detection-Constant 2396.31 0.00 0.5195 1.0000 24 2348.31

Occupancy-Constant 2397.48 1.17 0.2894 0.5571 26 2345.48

Occupancy-Detection-In-Interval-Constant 2398.31 2.00 0.1911 0.3679 25 2348.31

Null 2456.31 60.00 0.0000 0.0000 11 2434.31

Viable 2-species occupancy models for Canada lynx and bobcat/cougar at the camera station and grid cell level.

37

TABLE 5

Activity overlap

estimate

95% Confidence

interval

Canada lynx-bobcat snow on 0.81 0.63 - 0.84

Canada lynx-bobcat snow off 0.86 0.76 - 0.93

Canada lynx-cougar snow on 0.78 0.59 - 0.84

Canada lynx-cougar snow off 0.78 0.63 - 0.83

Estimate and confidence interval of overlap between Canada lynx and bobcats, and cougars

during snow-on and snow-off time periods.

38

FIGURE 1

Inset shows location of the study area in Washington. Main Figure shows the location of 2x2 km

grid cells and cameras (blue and yellow circles, for snow-off and snow-on camera stations).

Yellow line at the top of figure shows the US-Canada border.

39

FIGURE 2

Canada lynx and cougar or bobcat detections during snow-on and snow-off seasons at the camera

station level. Height of the bar represents the amount of detections of each species.

40

FIGURE 3

Canada lynx and cougar or bobcat detections during snow-on and snow-off seasons at the grid

cell level. Height of the bar represents the amount of detections of each species.

41

FIGURE 4

Hare detection ratio’s influence on Canada lynx probability of use in both snow-on and snow-off

time periods at the scale of the camera station.

42

FIGURE 5

Elevation’s influence on Canada lynx, bobcat and cougar probability of use in both snow-off and

snow-on time periods at the scale of the camera station.

43

FIGURE 6

Slope’s influence on Canada lynx and bobcat probability of use in both snow-off and snow-on

time periods at the scale of the grid cell.

44

FIGURE 7

Overlap of daily activity patterns of Canada lynx with bobcat and cougar during snow-on and

snow-off time periods.

45

APPENDIX A

Canada Lynx Grid Level Bobcat Grid Level

Model AICc AICc

Weight Model AICc

AICc

Weight

Season 900.1059 0.6058

Cams+Season 1478.334 0.5040

Cams+Season 900.9656 0.3942

Season 1478.371 0.4947

null 922.6499 0.0000

null 1491.101 0.0009

Cams 924.5419 0.0000

Cams 1492.476 0.0004

Canada Lynx Camera Station Level

Bobcat Camera Station Level

Model AICc AICc

Weight Model AICc

AICc

Weight

Season 1190.917 0.7333

Trail + Season 1902.699 0.9901

Trail + Season 1192.941 0.2666

Trail 1913.29 0.0050

null 1210.464 0.0000

Season 1913.346 0.0048

Trail 1210.328 0.0000

Null 1921.606 0.0001

AICc values of single species detectoin models for Canada lynx, bobcat, and cougar at the grid cell and camera station level

46

APPENDIX B

Hourly accumulation curves showing that the asymptote of number of the number of

observations needed to represent activity in all hours was reached.

47

APPENDIX C

Activity of Canada lynx in comparison to cougars and bobcats around sunrise and sunset in snow-on and snow-off time periods.

There is significant overlap in their activity and it does not appear that they are partitioning their activity around either solar zenith