HABITAT AND COMMUNITY ECOLOGY OF CANADA LYNX ACROSS
INTENSIVELY MANAGED FOREST LANDSCAPES
by
Shannon Michael Crowley
B.Sc., University of Alaska Southeast, 1999
M.Sc., University of Northern British Columbia, 2010

THESIS SUBMITTED IN PARTIAL FULFILLMENT OF
THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
IN
NATURAL RESOURCES AND ENVIRONMENTAL STUDIES

UNIVERSITY OF NORTHERN BRITISH COLUMBIA
August 2024
© Shannon M. Crowley, 2024

Abstract
In the subboreal forests of central British Columbia, large-scale and rapid timber harvest
has resulted in fundamental changes to the age distribution of forests. Canada lynx (Lynx
canadensis) is a habitat and prey specialist on snowshoe hare (Lepus americanus) that is likely
sensitive to broad-scale habitat change. It is unclear how large-scale timber harvest may
influence community dynamics, including the much studied lynx-hare predator-prey cycle. The
density and associated population cycles of Canada lynx differ substantially between northern
and southern populations with few studies of the population or habitat ecology of lynx inhabiting
southern boreal forests. My research addressed gaps in our understanding of the population,
community, and habitat ecology of Canada lynx in subboreal forests of British Columbia.
Specifically, I investigated the ecological factors influencing the habitat use, community
interactions, and survey methods of Canada lynx during two contrasting periods of cyclic lynx
and hare abundance and across a landscape with widespread and rapid forest harvesting. In
Chapter 2, I compared habitat selection at two different movement scales using GPS-collared
lynx and American marten as well as camera data collected during two contrasting periods of
prey abundance. I found that camera traps, in general, reflected the habitat use of GPS-collared
lynx and marten. Mid-level and top-level vegetation cover were important predictors of habitat
use for both marten and lynx, but with opposite directional influences. Lynx and marten
demonstrated differential use of habitat defined by forest age and structure suggesting that each
species would serve as a unique indicator of forest condition and change. My objective in
Chapter 3 was to determine if a combination of camera traps, abundance estimates, and
behavioural cues could be used to monitor cyclic population trends. I found that lynx behaviours
and relative abundance were strongly correlated. Consistent with my predictions, years with

II

higher lynx and hare abundance were characterized by increases in cheek-rubbing, scentmarking, and grouping behaviours. Population indices and estimates, in combination with
behavioural observations for lynx, provided insights into the ecological drivers of population
trends. In Chapter 4, I used camera traps to investigate the influence of sympatric carnivores
(coyote, fisher, wolverine) and prey (snowshoe hare, red squirrel) on the habitat use and cooccurrence of Canada lynx. I found that lynx occurrences mirrored the cyclic change in hare
abundance, while the number of sympatric mustelid species and the occurrences of each species
increased during the low period. The co-occurrence of lynx with other sympatric carnivores
increased at a time of prey scarcity suggesting predator populations in subboreal forests may be
in a dynamic state of habitat overlap dependent on cyclic prey abundance. In total, my research
provides new insights on the habitat, behavioural, and community ecology of lynx found in
subboreal forests that are experiencing rapid change. The ecology of lynx in that system is a
dynamic response to not only change in forest structure, but also the abundance of their primary
prey, snowshoe hare. Also, my results provide guidance on the appropriate application and
possible biases of a range of methods for monitoring the distribution and abundance of lynx.

III

Table of Contents
ABSTRACT...................................................................................................................................................... II
LIST OF TABLES ....................................................................................................................................... VII
LIST OF FIGURES......................................................................................................................................... X
LIST OF APPENDICES ............................................................................................................................ XIII
ACKNOWLEDGEMENTS ............................................................................................................................ 1
CHAPTER 1 ...................................................................................................................................................... 2
INTRODUCTION ............................................................................................................................................ 2
BACKGROUND .................................................................................................................................................... 2
CANADA LYNX ECOLOGY .................................................................................................................................... 6
RESEARCH OBJECTIVES ..................................................................................................................................... 10

RESEARCH METHODS ............................................................................................................................. 12
STUDY AREA .................................................................................................................................................... 12
DATA COLLECTION ........................................................................................................................................... 15
Camera traps and hair snares ................................................................................................................. 15
GPS collars ............................................................................................................................................ 16
Habitat covariates .................................................................................................................................. 17

CHAPTER 2: LINKING ECOLOGICAL PROCESSES OF INDIVIDUAL-BASED DATA
AND OCCURRENCE SURVEYS: CANADA LYNX AND AMERICAN MARTEN
MOVEMENT AND HABITAT SELECTION USING GPS-COLLARS AND CAMERA
TRAPS ............................................................................................................................................................. 20
ABSTRACT..................................................................................................................................................... 20
INTRODUCTION .......................................................................................................................................... 21
MATERIALS AND METHODS ................................................................................................................ 25
STUDY AREA .................................................................................................................................................... 25
FIELD DATA COLLECTION ................................................................................................................................. 26
GPS-collared individuals ........................................................................................................................ 26
Camera traps.......................................................................................................................................... 27
Habitat covariates .................................................................................................................................. 28
DATA ANALYSIS ............................................................................................................................................... 30
Movement scales..................................................................................................................................... 30
Habitat selection..................................................................................................................................... 33

RESULTS......................................................................................................................................................... 36
MARTEN CAPTURES .......................................................................................................................................... 36
LYNX CAPTURES............................................................................................................................................... 36
CAMERA TRAPS ................................................................................................................................................ 37
MOVEMENT SCALES.......................................................................................................................................... 37
LYNX WINTER HABITAT MODELS...................................................................................................................... 37
LYNX SNOW-FREE HABITAT MODELS ................................................................................................................ 38

IV

MARTEN WINTER HABITAT MODELS .................................................................................................................. 38

DISCUSSION.................................................................................................................................................. 42
CHAPTER 3: BEHAVIOUR AS AN INDICATOR OF CYCLIC TRENDS IN ABUNDANCE
OF CANADA LYNX (LYNX CANADENSIS) AND SNOWSHOE HARE (LEPUS
AMERICANUS) ............................................................................................................................................. 46
ABSTRACT..................................................................................................................................................... 46
INTRODUCTION .......................................................................................................................................... 47
MATERIALS AND METHODS ................................................................................................................ 51
STUDY AREA .................................................................................................................................................... 51
FIELD DATA COLLECTION ................................................................................................................................. 51
SAMPLING AND ECOLOGICAL VARIABLES .......................................................................................................... 53
N-mixture models.................................................................................................................................... 53
Lynx behaviour ....................................................................................................................................... 54
DATA ANALYSIS ............................................................................................................................................... 55
Population estimates and indices ............................................................................................................ 55
Lynx behaviour ....................................................................................................................................... 58

RESULTS......................................................................................................................................................... 59
CAMERA TRAPS ................................................................................................................................................ 59
POPULATION ESTIMATES AND INDICES............................................................................................................... 60
BEHAVIOUR MODELS ........................................................................................................................................ 63

DISCUSSION.................................................................................................................................................. 69
CHAPTER 4: SHORT-TERM FLUCTUATIONS IN PREY AND ITS INFLUENCE ON A
CARNIVORE COMMUNITY: HABITAT CO-OCCURRENCE OF CANADA LYNX AND
SYMPATRIC MESOPREDATORS INCREASES FOLLOWING CYCLICAL REDUCTION
IN PRIMARY PREY .................................................................................................................................... 75
ABSTRACT..................................................................................................................................................... 75
INTRODUCTION .......................................................................................................................................... 76
MATERIALS AND METHODS ................................................................................................................ 79
STUDY AREA .................................................................................................................................................... 79
FIELD DATA COLLECTION ................................................................................................................................. 81
HABITAT COVARIATES ...................................................................................................................................... 82
DATA ANALYSIS ............................................................................................................................................... 84

RESULTS......................................................................................................................................................... 89
CAMERA TRAP OCCURRENCES .......................................................................................................................... 89
2015–2016 WINTER MODELS ............................................................................................................................ 92
2015–2016 SNOW-FREE MODELS ...................................................................................................................... 92
2020–2021 WINTER MODELS ............................................................................................................................ 93
2020–2021 SNOW-FREE MODELS ...................................................................................................................... 93

V

DISCUSSION.................................................................................................................................................. 98
CHAPTER 5: STUDY FINDINGS, MANAGEMENT IMPLICATIONS, AND
RECOMMENDATIONS FOR FUTURE RESEARCH ...................................................................... 105
SUMMARY OF RESULTS ................................................................................................................................... 105
ECOLOGICAL IMPLICATIONS ............................................................................................................................ 106
LANDSCAPE CHANGE AND LONG-TERM MONITORING ...................................................................................... 109
STRENGTHS, WEAKNESSES, AND FUTURE RESEARCH ....................................................................................... 110
MANAGEMENT AND CONSERVATION ............................................................................................................... 113

REFERENCES .............................................................................................................................................. 116
APPENDIX 1 ................................................................................................................................................ 142
APPENDIX 2 ................................................................................................................................................ 143
APPENDIX 3 ................................................................................................................................................ 145

VI

List of Tables
Table 1. Sampling methods and sample sizes for research on the habitat ecology of Canada lynx
(Lynx canadensis) in central British Columbia, Canada, 2015–2016 and 2020–
2021……………………………………………………………………………..………..17
Table 2. Variables used in the development of habitat selection models for Canada lynx (Lynx
canadensis) and American marten (Martes americana) using camera traps and GPScollared individuals in central British Columbia, Canada, 2015–2016 and 2020–2022...29
Table 3. A priori candidate models (multinomial logistic regression) representing habitat use of
Canada lynx (Lynx canadensis) and American marten (Martes americana) using camera
traps and GPS-collared individuals in central British Columbia, Canada, 2015–2016 and
2020–2022………………………………………………………………………………..35
Table 4. AICc scores, and AICc weights (w) for the three highest-ranked multinomial logistic
regression models representing habitat use of Canada lynx (Lynx canadensis) and
American marten (Martes americana) using GPS collars and camera traps in central
British Columbia, Canada, 2015–2016 and 2020–2021. Area under the curve (AUC) and
standard error (SE) for the receiver operating characteristic represents the predictive
accuracy of each model…………………………………………………………………..39
Table 5. Variables used in the development of models predicting abundance (N-mixture models)
or behaviours (cheek-rubbing, scent-marking, grouping) of Canada lynx (Lynx
canadensis) using camera traps in central British Columbia, Canada, 2015–2016 and
2020–2022………………………………………………………….…………………....55

VII

Table 6. A priori candidate models (logistic regression) representing three types of behaviour
(cheek-rubbing, scent-marking, grouping) of Canada lynx (Lynx canadensis) at remote
camera sites in central British Columbia, Canada, 2015–2016 and 2020–2022...…..….58
Table 7. Density estimates from spatial capture-mark-resight (SCMR) and N-mixture models for
Canada lynx (Lynx canadensis) and snowshoe hare (Lepus americanus) using camera
traps in central British Columbia, Canada, 2015–2016 and 2020–2022. RSE = relative
standard error………………………………………………………….…………………62
Table 8. AICc scores, and AICc weights (w) for logistic regression models predicting Canada
lynx (Lynx canadensis) behaviours at camera traps in central British Columbia, Canada,
2014–2016 and 2020–2022. Area under the curve (AUC) and standard error (SE) for the
receiver operating characteristic represents the predictive accuracy of each model…….65
Table 9. Spatial and temporal correlations (r) between occurrence rates (OR) generated with
camera images, N-mixture (N-mix) abundance estimates, and lynx behaviours for Canada
lynx (Lynx canadensis) and snowshoe hare (Lepus americanus) using camera traps in
central British Columbia, Canada, 2014–2016 and 2020–2022……………………..…..66
Table 10. Variables used in the development of habitat and co-occurrence models (multinomial
logistic regression) for Canada lynx (Lynx canadensis), sympatric mesopredators, and
prey at remote camera sites in central British Columbia, Canada, 2015-2016 and 20202021………………………………………………………………………………………84
Table 11. A priori candidate models (multinomial logistic regression) representing habitat use
and co-occurrence of Canada lynx (Lynx canadensis), sympatric mesopredators, and prey

VIII

at remote camera sites in central British Columbia, Canada, 2015–2016 and 2020–
2021…………………………………………………………………………………...….88
Table 12. Total occurrences and occurrences per 100 camera-days (rate) for sympatric
mesopredators and prey at camera traps during two contrasting periods of prey abundance
in central British Columbia, Canada, 2015–2016 (high) and 2020–2021 (low)…….…...90
Table 13. AICc scores, and AICc weights (w) for multinomial logistic regression models
representing habitat use and co-occurrence of Canada lynx (Lynx canadensis), sympatric
mesopredators, and prey at remote camera sites in central British Columbia, Canada,
2015–2016 and 2020–2021. Area under the curve (AUC) and standard error (SE) for the
receiver operating characteristic represents the predictive accuracy of each model.
Shaded value = AUC > .7………………….……………………………..…….………..94
Table 14. Coefficients for top-ranked models illustrating habitat use and co-occurrence by
Canada lynx (Lynx canadensis), sympatric mesopredators, and prey at camera sites
during two contrasting periods of prey abundance in central British Columbia, Canada,
2015–2016 (high) and 2020–2021 (low). Area under the curve (AUC) for the receiver
operating characteristic represents the predictive accuracy of each model. Dark grey =
negative association, Light grey = positive association, “●” = significant coefficient, “X”
= useful or high model accuracy. …..………………………………………………...…96

IX

List of Figures
Figure 1. Canada lynx (Lynx canadensis) study area with camera locations and hexagonal grid in
in John Prince Research Forest, British Columbia, Canada, 2015–2016 and 2020–
2022……………………………………………………………………………………...14
Figure 2. Camera trap and hair snare site for Canada lynx (Lynx canadensis) and other
mesopredators in John Prince Research Forest, British Columbia, Canada, 2015–2016
and 2020–2022.…………….……………………………………………………….……16
Figure 3. Map of study area illustrating camera locations as well as GPS-collar locations
representing fast and slow movements by Canada lynx (Lynx canadensis) during the
winter period in central British Columbia, Canada, 2015–2016 and 2020–2022. ……....31
Figure 4. Map of study area illustrating camera locations as well as GPS-collar locations
representing fast and slow movements by American marten (Martes americana) during
the winter period in central British Columbia, Canada, 2015–2016 and 2020–2022. ..…32
Figure 5. Significant coefficients for top-ranked models illustrating habitat use by Canada lynx
(Lynx canadensis) during the winter and snow-free periods using GPS collars and camera
traps in central British Columbia, Canada, 2015–2016 and 2020–2022. Blue = Camera
(2020–2022), Red = Camera (2015–2016), Green = GPS slow movement, Orange = GPS
fast movement……………………………………………………………………………40
Figure 6. Significant coefficients for top-ranked models illustrating habitat use by American
marten (Martes americana) using GPS collars and camera traps during the winter in
central British Columbia, Canada, 2015–2016 and 2020–2022. Blue = Camera data

X

(2020–2022), Red = Camera data (2015–2016), Green = GPS slow movement, Orange =
GPS fast movement……………………………………………….……………………..41
Figure 7. Comparison in trends of relative abundance indices (RAI) and N-mixture abundance
estimates with 95% confidence intervals for Canada lynx (Lynx canadensis) using camera
traps in central British Columbia, Canada, 2014–2016 and 2020–2022. Spatial capturemark-resight (SCMR) estimates for 2020 and 2022 provide an additional measure for
validating estimates of relative abundance from camera images and estimates of
abundance from N-mixture models……….……………………………………..………61
Figure 8. Coefficients for top-ranked models illustrating the presence of behaviours (cheekrubbing, scent-marking, and grouping) of Canada lynx (Lynx canadensis) at camera trap
sites during contrasting periods of prey abundance in central British Columbia, Canada,
2015–2016 (high) and 2020–2022 (low). MW = mid-winter, LW = late winter, ES = early
summer, LS = late summer…………………………………………………..…………..67
Figure 9. Canada lynx (Lynx canadensis) abundance and presence of behaviours (percentage of
visits with each behaviour recorded) in each year (cheek-rubbing, scent-marking, and
grouping) of Canada lynx (Lynx canadensis) at camera trap sites during contrasting
periods of prey abundance in central British Columbia, Canada, 2014–2016 (increasing high) and 2020–2022 (decreasing - low)………………………………………………..68
Figure 10. Example model development and outcomes (multinomial logistic regression)
investigating the habitat use and co-occurrence of Canada lynx (Lynx canadensis) and
sympatric mesopredators (coyote [Canis latrans], fisher [Pekania pennanti], and
wolverine [Gulo gulo]) at camera traps during two contrasting periods of prey abundance

XI

in central British Columbia, Canada, 2015–2016 (high hare) and 2020–2021 (low
hare).……………………………………………………………………………………87
Figure 11. Occurrences per 100 camera-days for sympatric mesopredators and prey at camera
traps during two contrasting periods of prey abundance in central British Columbia,
Canada, 2015–2016 (high hare) and 2020–2021 (low hare). Top = Winter; Bottom =
Snow-free…………………………………………………………………………….….91
Figure 12. Probability of lynx (Lynx canadensis) and wolverine (Gulo gulo) co-occurrence and
lynx and fisher (Pekania pennati) co-occurrence, with 95% confidence intervals, as a
function of canopy cover (3–10 m and >10 m) at cameras sites during a high and low in
snowshoe hare abundance in central British Columbia, Canada, February–April, 2015–
2016 and 2020–2021…………………………………………………………………….97

XII

List of Appendices
Appendix 1.1. Full model results, AICc scores, and AICc weights (w) for multinomial logistic
regression models representing habitat use of Canada lynx (Lynx canadensis) and
American marten (Martes americana) using GPS collars and camera traps in central
British Columbia, Canada, 2015–2016 and 2020–2021………………………………140
Appendix 2.1. Full model rankings, AICc scores, and AICc weights (w) for N-mixture models
predicting abundance of Canada lynx (Lynx canadensis) and snowshoe hare (Lepus
americanus) at camera traps in central British Columbia, Canada, 2015–2016 and 2020–
2022…………………………………………………………………………………….141
Appendix 2.2. AICc scores, and AICc weights (w) for spatial-capture mark-resight (SCMR)
models estimating Canada lynx (Lynx canadensis) density at camera traps in central
British Columbia, Canada, 2020 and 2022………………,…………………………….142
Appendix 3.1. Full model rankings, AICc scores, and AICc weights (w) for multinomial logistic
regression models representing habitat use and co-occurrence of Canada lynx (Lynx
canadensis), sympatric mesopredators, and prey at camera traps in central British
Columbia, Canada during a winter of low hare abundance (2020–2021)..……………143
Appendix 3.2. Full model rankings, AICc scores, and AICc weights (w) for multinomial logistic
regression models representing habitat use and co-occurrence of Canada lynx (Lynx
canadensis), sympatric mesopredators, and prey at camera traps in central British
Columbia, Canada during a winter of high hare abundance (2015–2016)……..………145
Appendix 3.3. Full model rankings, AICc scores, and AICc weights (w) for multinomial logistic
regression models representing habitat use and co-occurrence of Canada lynx (Lynx
XIII

canadensis), sympatric mesopredators, and prey at camera traps in central British
Columbia, Canada during a snow-free period of low hare abundance (2020–2021)….147
Appendix 3.4. Full model rankings, AICc scores, and AICc weights (w) for multinomial logistic
regression models representing habitat use and co-occurrence of Canada lynx (Lynx
canadensis), sympatric mesopredators, and prey at camera traps in central British
Columbia, Canada during a snow-free period of high hare abundance (2015–2016)….149

XIV

Acknowledgements
I would like to thank the members of my committee including Ken Otter, Heather Bryan, and
Dale Seip for their guidance and support at every stage of my dissertation. I am greatly indebted
to my advisor, Chris Johnson, for helping me grow as a scientist, analyst, and writer. I would like
to thank Dexter Hodder for his feedback and long discussions about hare and lynx ecology. My
work improved leaps and bounds throughout our chats, and I feel privileged to have him as a
colleague. I thank Sue Grainger and the entire JPRF Board of Directors for their support through
out this process. I would like to thank Tom Jung for being my external examiner and providing
thoughtful feedback.
I thank the Habitat Conservation Trust Foundation, Forest Enhancement Society of
British Columbia, John Prince Research Forest and University of Northern British Columbia for
providing funding to support this project. Tanizul Timber Ltd. provided important funds for
acquiring and processing LiDAR data. I thank Steven Murdoch, Lauren Wheelhouse, Morgan
Conrad, Amanda Carriere, Gabrielle Aubertin, Max Prince, Jason Mattes, Ronald Monk, and
Jonathon Tom for their hard work, helpful discussions, and companionship in the field. I thank
the Binche Whut’en, Nak'azdli Whut'en, and Tl’azt’en First Nations for their support of this
project within their traditional territory.
I cannot overstate how essential the love and support I receive from my family (Val,
Willa, and Bren) has been and continues to be in my life. It is inconceivable that I could have
made it this far without them.

1

CHAPTER 1
Introduction
Background
Habitat loss and fragmentation are leading causes of the global decline in biodiversity (Fischer
and Lindenmeyer 2007). There is abundant empirical evidence documenting a relationship
between habitat change and declines in population abundance, range contraction, reductions in
genetic diversity, and altered community dynamics (Lino et al. 2019). However, those
relationships can vary greatly depending on species and ecosystem context. Species with larger
body mass, terrestrial or arboreal niches, and dependency on forested habitats are often most
sensitive to habitat loss and fragmentation (Lino et al. 2019). Habitat specialists and species that
range across large areas can be especially vulnerable (Crooks 2002, Fischer and Lindenmayer
2007, Lino et al. 2019).
Canada lynx (Lynx canadensis) is a habitat and prey specialist that is likely sensitive to
broad-scale habitat change. The influence of large-scale habitat loss and fragmentation on harelynx population cycles is unknown. The synchrony, period, and amplitude of the lynx-hare cycle
in the boreal forest is thought to be dependent on large tracts of contiguous forest that facilitate
dispersal by Canada lynx and other predators (Krebs et al. 2001, Krebs et al. 2014). Habitat
loss/fragmentation and an increase in generalist predators (e.g., coyotes [Canis latrans]) are two
hypotheses to explain dampened hare-lynx cycles (Keith 1963, Wolf et al. 1980, Buskirk et al.
2000, McKelvey et al. 2000, Strohm and Tyson 2009). Using a theoretical modeling approach,
Vitense et al. (2016) found that both generalist predation and habitat loss can result in stabilizing
effects and loss of hare cycles. However, both causes are likely to occur simultaneously in cases

2

where the abundance of generalist predators increase in disturbed landscapes (Buskirk et al.
2000).
Landscape disturbance that increases edge habitat may negatively affect hare and lynx
populations. The exploration and development of natural resources, including forest harvesting
and the exploitation of hydrocarbons, typically increases habitat fragmentation and the density of
linear features (i.e., roads, pipelines, seismic lines). In a large mammal system, increases in
linear features reduced the efficacy of prey refugia and increased the spatial overlap between
predator and prey (DeMars and Boutin 2017). Refugia in dense understory further from edge
may also be important for hares during cyclic lows (Wolff 1980, O’Donoghue et al. 1998b).
Edge habitat may increase hare exposure to predation by both specialist and generalist predators
(Vitense et al. 2016). However, empirical data on the influence of intensive habitat loss and
fragmentation on cyclic lynx-hare populations is lacking. Most studies of the hare-lynx cycle
have occurred in areas with less human disturbance (Mowat et al. 2000, Krebs et al. 2001, Krebs
et al. 2014).
Approximately 18 million hectares of British Columbia’s (BC) forests have been affected
by a mountain pine beetle (Dendroctonus ponderosae; MPB) epidemic that began in the mid1990s (BC FLNRO 2012). Since 2001, the Allowable Annual Cut (AAC) has been increased to
salvage dead and dying lodgepole pine (Pinus contorta) trees. This has resulted in fundamental
changes to the age distribution of forests across large portions of central BC with implications
for wildlife habitat (FPB 2009). Across central BC, trappers and biologists are concerned that
recent accelerated forest harvesting has negatively influenced the abundance and persistence of
furbearers, including Canada lynx (Bridger 2015). For example, Bridger et al. (2016) found an
83% decline in the availability of optimal habitat for lynx in areas with rapid salvage harvesting,
3

yet little relationship between the abundance of lynx, as indicated by lynx harvest, and habitat
change. That contradiction in findings highlights the need for a better understanding of the
relationship between forest change and the habitat needs and population dynamics of lynx.
Non-invasive surveys at point locations are used with increasing frequency to monitor
wildlife populations. Examples include camera traps, hair snares, and surveys of scat, tracks or
other sign (Long et al. 2008). These non-invasive surveys are often used as the basis for
management decisions. However, the biases and accuracy of many of these methods are not well
understood. Furthermore, bias and accuracy may differ according to species, abundance, or
variation in behaviour among populations (e.g., Long et al. 2010, Crowley et al. 2013,
Monterroso et al. 2014). This is especially true for forest carnivores due to their behaviour (i.e.,
elusiveness, low densities) and habitats (vegetation that obscures visibility) that have historically
limited survey options (Long et al. 2008).
Historically, Canada lynx have been monitored with hair snares or snow tracks, and only
more recently with camera traps (McDaniel et al. 2000, Squires et al. 2012, Crowley et al. 2017,
Doran-Myers et al. 2021). During the last two decades there has been an increase in the use of
camera traps to measure population-level patterns of occurrence, distribution, and abundance of
terrestrial wildlife (Burton et al. 2015, Steenweg et al. 2017) including Canada lynx (Crowley et
al. 2017, Thomas et al. 2019, Doran-Myers et al. 2021, Anderson et al. 2023, Kenney et al.
2024). Widespread use of increasingly more reliable camera traps has allowed monitoring at
larger spatial scales. However, there are still many assumptions and questions focused on the
ecological mechanisms that influence the precision and bias of this survey method, including the
assessment of community dynamics (Burton et al. 2015, Steenweg et al. 2017, Blanchet et al.
2020). In addition, snowshoe hare and lynx populations cycle with an 8–11-year frequency that
4

results in years of contrasting lynx and hare densities (Keith 1963, O’Donoghue et al. 1997,
Slough and Mowat 1996, Krebs et al. 2014, Krebs et al. 2017). Estimates or indices of population
abundance will likely be influenced by these changing ecological conditions and need to be
placed in the context of regional fluctuations in hare and lynx populations.
The vast majority of lynx studies have occurred in northern boreal forests (e.g., Mowat et
al. 2000, Poole et al. 1996, Slough and Mowat 1996) or on the southern periphery of their range
in the northern United States and southern Canada (e.g., Apps 2000, Koehler 1990, Vashon et al.
2008a, b). The density of lynx and associated population cycles differ substantially between
northern and southern populations with little known about Canada lynx occupying areas in
southern boreal forests. In British Columbia, there have been no studies of lynx populations
within the subboreal forest of the central interior.
Canada lynx have consistently been one of the top three harvested furbearers in BC
(American beaver [Castor canadensis], American marten [Martes americana], and Canada lynx)
and are considered sensitive to harvest (Class 2 species; Hatler and Beal 2003). Government
biologists have been asked in recent years by members of the trapper community to extend
harvest seasons for lynx (M. Anderson, pers. comm). However, the status and nature of cyclic
populations in many areas of BC is unknown. Increases in fur harvest during a hare low could
have a disproportionate effect on lynx populations. In addition to being an important fur
resource, Canada lynx may be an indicator of community dynamics in early seral and mixed-age
forests. In combination with other species in the mesopredator guild, lynx could provide
important information on forest change across landscapes. However, the influence of large-scale
intensive timber harvest on populations cycles and lynx persistence in sub-boreal forests is
unknown. This is of concern for the conservation of Canada lynx across its range, and especially
5

relevant in areas experiencing rapid and cumulative landscape change, such as central British
Columbia.

Canada Lynx Ecology
Canada lynx are morphologically adapted for living in cold climates with deep snow (thick fur,
long legs, large and well-furred feet for low foot loading) typical of boreal and subboreal forests
of Alaska and Canada (Anderson and Lovallo 2003). Canada lynx are considered a prey
specialist on snowshoe hare. Across lynx range, snowshoe hare can comprise 43–100% of lynx
diet (Nellis et al. 1972, Brand et al. 1976, O'Donoghue et al. 1998). Other prey items include red
squirrels (Tamiasciurus hudsonicus), northern flying squirrels (Glaucomys sabrinus), grouse
(Bonasa spp.), small mammals, small birds, and ungulate carrion (Brand et al. 1976, Parker et al.
1983, O'Donoghue et al. 1998, Aubry et al. 1999, Mowat et al. 1999, Jung 2022). Grouse are the
most frequently observed birds eaten by lynx (Staples 1995, O’Donoghue et al. 1998a, Jung
2022). Red squirrels are often the second most dominant prey item, typically serving as alternate
prey and contributing more during lows in snowshoe hare abundance (Koehler 1990, Staples
1995, O’Donoghue et al. 1998b, Aubry et al. 2000, Apps 2000, Mowat et al. 2000). The
dominance of snowshoe hare in lynx diet is especially evident during the winter with secondary
prey becoming more common during the snow-free season (Saunders 1963, Parker et al. 1983,
Mowat et al. 2000). At the southern extent of their range, lynx may exhibit a more generalist diet
and use of habitat compared to more northern lynx populations (Roth et al. 2007, Murray et al.
2008).
The population dynamics of lynx and snowshoe hare are closely linked (Elton and
Nicholson 1942, Keith 1963, O’Donoghue et al. 1997, Slough and Mowat 1996). These two

6

species cycle with an 8–11-year period in the northern boreal forests of Alaska and Canada (high
to high or low to low; Elton and Nicholson 1942, Keith 1963, O’Donoghue et al. 1997, Slough
and Mowat 1996, Krebs et al. 2014, Krebs et al. 2017). The peak in the amplitude of cycling lynx
and hare populations are asynchronous, with lynx demonstrating a 1–2-year lag (Elton and
Nicholson 1942, Keith 1963, O’Donoghue et al. 1997, Krebs et al. 2014, Krebs et al. 2017). That
cycling dynamic may differ greatly across the range of lynx. For example, there is a 4-fold
difference in the high-to-low points in the cycle for populations in central Canada (Keith et al.
1977) and up to 17-fold difference in northern Canada (Slough and Mowat 1996, Poole 1994). A
decrease in productivity and kitten survival (Nellis et al. 1972, Brand and Keith 1979, Parker et
al. 1983, Mowat et al. 1996), and an increase in mortality, emigration, and home-range size
(Slough and Mowat 1996, Ward and Krebs 1985, Poole 1997) typically characterize the low or
declining population phase of the cycle. The amplitude and period of the cycle has not been well
described for populations in BC.
There are several proposed mechanisms to explain the cycling dynamics of lynx and hare,
including climate (Yan et al. 2013), weather (Krebs et al. 2014), forest succession (Krebs et al.
2014), food availability and quality (Krebs et al. 1986, Sinclair et al. 1988, Krebs et al. 2017),
and predation (Tyson et al. 2010, Krebs et al. 2014, Krebs et al. 2017). The majority of evidence
suggests that predation is the most important driver of the hare-lynx cycle (Krebs et al. 2017).
However, many of those factors may work synergistically to influence population fluctuations. In
particular, food availability and quality for hares, in combination with predation, may strongly
contribute to cyclic fluctuations (Krebs et al. 2017). In addition to limitations on food abundance,
there is evidence that severe browsing by hares can also cause plants to produce secondary
compounds that affect food palatability and quality (Sinclair et al. 1988, Rodgers and Sinclair
7

1997). Such changes in the quality of forage can result in reduced reproductive productivity of
hares. Krebs et al. (2017) stated that more research was needed to better understand the complex
effects of plant secondary compounds relative to the distribution and abundance of snowshoe
hares and other herbivores.
Canada lynx, a specialist predator, are generally assumed to have a type-II functional
response characterized by a decreasing proportion of prey eaten as prey density increases (Tyson
et al. 2010, Fryxell et al. 2014). A type-II functional response can lead to destabilization of hare
populations when hares increase above carrying capacity (Fryxell et al. 2014). Hare
overcompensation combined with a time lag in the numerical response of predators to hare
abundance can result in increased depensatory predation (increasing proportion of hare killed
with declining hare abundance; O’Donoghue et al. 1998b). The combination of functional and
numerical responses can contribute to density-dependent cycles (Fryxell et al. 2014). In addition
to direct mortality, lynx and other predators may also induce chronic stress in hares, resulting in
decreased reproduction that further drives the population decline phase (Sheriff et al. 2009, 2010,
2011). Stress hormone levels can remain elevated for a prolonged period and be transmitted from
mothers to offspring (Sheriff et al. 2009, 2010, 2011; Krebs et al. 2017). Physiological responses
by hares likely have a time lag that also contributes to population destabilization and cyclic
fluctuations.
Lynx breed in March to early April, and male-female pairs may travel together for
several days (Quinn and Parker 1987, Poole 2003). After a gestation period of 61–70 days,
kittens are born in late May to early June (Saunders 1963, Poole 1994, Slough 1999, Lavoie et al.
2019). Litter sizes can range from 1–8 but more commonly consist of 2–4 kittens (Brand et al.
1976, Mowat and Slough 1998, Moen et al. 2008, Lavoie et al. 2019). Litter sizes fluctuate
8

considerably in response to snowshoe hare densities (Lavoie et al. 2019). For example, litters
range from 4 or 5 kittens at population highs to single kitten litters or no reproduction during
population lows (Lavoie et al. 2019). Kittens begin making movements away from the den with
their mother 6–8 weeks after parturition and remain with their mother for 9–10 months traveling
and hunting as a group through their first winter (Lavoie et al. 2019).
Lynx typically select habitats where snowshoe hares are abundant (as summarized by
Aubry et al. 2000). Throughout their range, both lynx and snowshoe hares are strongly
associated with forest that is characterised by dense understory (Adams 1959, Brocke 1975,
Wolff 1980, Wolfe et al. 1982, Litvaitis et al. 1985, Homyack et al. 2007). That understory may
include dense shrub cover or early successional conifer growth. Snowshoe hare use such forest
types for cover and food (Hodges 2000a, b). Cover from predators, precipitation, and
temperature extremes may be especially important during the winter (Whittaker and Thomas
1983, Hodges 2000a, b). For this reason, optimal hare and thus lynx habitat has often been
associated with older regenerating stands aged 20–35 years.
The size of home ranges for lynx can vary considerably by year and area depending on
the abundance of snowshoe hare. During high hare densities in northern boreal forests, the home
ranges of male and female lynx typically cover 20–45 and 13–21 km2, respectively (Quinn and
Parker 1987, Poole 1994, Lavoie et al. 2013). During low hare densities, home ranges for both
sexes can cover hundreds of kilometres (Quinn and Parker 1987, Poole 1994, Lavoie et al. 2013).
Home ranges at the southern periphery of the distribution of lynx tend to be more stable and
intermediate in size relative to populations found in the boreal north (Aubry et al. 2000, Mowat
et al. 2000).

9

Coyotes, bobcats (Lynx rufus), wolverines (Gulo gulo), fishers (Pekania pennanti),
cougars (Felis concolor), and wolves (Canis lupus) are sympatric with lynx populations across
parts of their range. Coyote and bobcat are the most common exploitive competitors of Canada
lynx (Buskirk et al. 2000). Wolverine may be the dominant species in interference competition
(Jung et al. 2023). Wolf, coyote, cougar, wolverine, and fisher have been observed to directly kill
lynx (Poole 1994, Slough and Mowat 1996, Squires and Laurion 2000, O’Donoghue et al. 1995,
1997, McClellan et al. 2018). In Maine, predation was the leading cause of mortality for Canada
lynx with at least 18 of 65 lynx moralities over a 12-year period attributed to predation
(McLellan et al. 2018). Of these mortalities, 14 of 18 were caused by fisher.

Research Objectives
My research addressed gaps in our understanding of the population, community, and
habitat ecology of Canada lynx in subboreal forests of British Columbia. Specifically, I
investigated the ecological factors influencing the habitat use, behaviours, community
interactions, and survey methods of Canada lynx during two contrasting periods of cyclic lynx
and hare abundance and across a landscape with widespread and rapid forest harvesting. The
general research objectives were:
1) Determine and contrast the habitat selection of Canada lynx and American marten (Martes
americana) at two behavioural scales of movement using GPS collars and camera traps.
2) Determine if N-mixture models, camera trap occurrence rates, and behaviours can be used to
monitor trends in the abundance of Canada lynx.
3) Determine the influence of sympatric and intraguild mesopredators on the distribution and
habitat use of lynx.
10

My dissertation includes three interrelated chapters. In Chapter 2, I used data from
collared individuals and camera traps to determine the effects of vegetation characteristics, prey
density, and disturbance on habitat selection by lynx and marten at the scale of intra- and interpatch movement. The addition of American marten to this analysis allowed me to expand the
research question to include a species with contrasting ecology and habitat relationships to
Canada lynx and widen monitoring and management applications. The identification of habitat
selection at multiple scales provides important information on the ecological processes
influencing lynx distribution that is necessary for interpreting monitoring surveys and largerscale population processes. In Chapter 3, I compared multiple methods for generating abundance
estimates for Canada lynx. I then used the observed behaviour of lynx at camera traps to
investigate the ecological factors influencing abundance as well as the efficacy of using
behaviours as a co-indicator of population trends. This chapter provides insights that allow us to
interpret biases in estimates of abundance, especially in the context of cyclical dynamics of lynx
and hare. In Chapter 4, I investigated the co-occurrence of lynx and sympatric mesopredators and
the implications for habitat use by lynx. My results provide new insights on interspecific
competition in subboreal forests and the possible implications of competition for the habitat use
and population dynamics of lynx in the context of a cyclic prey base. My analyses represented
two time periods with contrasting periods of hare abundance typical of the observed hare-lynx
cycle. In total, my dissertation addresses key knowledge gaps on the habitat and population
ecology of Canada lynx and provides new and important information on the ecological processes
that influence monitoring surveys.

11

Research Methods
Study Area
The research was conducted in and adjacent to the John Prince Research Forest (JPRF) in northcentral BC, Canada (Figure 1). The JPRF is a 16 500-ha portion of forested provincial land 45
km northwest of the town of Fort St. James that is co-managed by the University of Northern
British Columbia and Binche Whut’en, and Tl’azt’en First Nations. The study area is dominated
by the Sub-Boreal Spruce biogeoclimatic zone and represents the northern extent of contiguous
Rocky Mountain Douglas-fir (Pseudotsuga menziesii var. glauca) in the interior of British
Columbia (Delong et al. 1993). The JPRF is the focus of a long-term monitoring program
investigating the influences of climate and landscape change on a relatively large number of
carnivores and ungulates. There has not been any significant trapping in the study area for almost
two decades.
The JPRF has experienced a wide variety of logging activities over the past 75 years and
contains a mosaic of old and young forest (continuum from new harvest to old growth >250
years old) with interspersed deciduous stands. The JPRF harvested approximately 20,000 m3 on
an annual basis during the study period. More intensive industrial forest harvest occurred in the
study area outside the Research Forest boundaries in the 15 years before this study (2000-2015).
Although salvage harvest continued, relatively fewer stands were actively harvested in between
the two data collection periods (e.g. 2015–2016 to 2020–2022). The study area is within a
broader region that is experiencing rapid and widespread harvest of dead and dying lodge pole
pine affected by a mountain pine beetle epidemic.

12

The research forest is characterized by rolling terrain with low mountains (700 m to 1500
m above sea level) and is traversed by many small streams that flow into either Tezzeron or
Pinchi Lakes. Mean daily average temperatures (1981–2010) were 3.5 ºC and ranged from a
monthly mean daily average of -9.5 ºC in January to 15.4 ºC in July. Mean annual precipitation
was 487.2 mm, with an average of 172.7 cm of snow (Environment and Climate Change Canada
2018).
American marten, Canada lynx, short-tailed weasel (Mustela erminea), American mink
(Neovison vison), river otter (Lontra canadensis), coyotes (Canis latrans), red fox (Vulpes
vulpes) fisher (Pekania pennanti), and wolverine (Gulo gulo) are mesopredators that occur in the
study area. American beaver (Castor canadensis), muskrat (Ondatra zibethicus), snowshoe hare
(Lepus americanus), red squirrels (Tamiascurus hudsonicus), flying squirrels (Glaucomys
sabrinus), ruffed grouse (Bonasa umbellus), deer mice (Peromyscus maniculatus), and voles
(Myodes spp.) are some of the most common prey species throughout the study area.

13

Figure 1. Canada lynx (Lynx canadensis) study area with camera locations and hexagonal grid in
John Prince Research Forest, British Columbia, Canada, 2015–2016 and 2020–2022.

14

Data Collection
Table 1 summarizes the years, data types, and sample sizes used for each of the research
objectives and data chapters.
Camera Traps and Hair Snares
I deployed and monitored 66 trail cameras across a hexagonal grid (2.5 km apart; 5.41 km2 cell
size) from January–March in 2015 and 2016 and from April–October in 2016 (Figure 1). These
same camera sites were set from February 2020 to May 2022 (Table 1). At each site, a rub
station was established 2–3 m directly in front of a camera trap (located between 0.5 and 1 m
above the snow on a tree; Figure 2). Each rub station consisted of a small diameter log (<15 cm)
that was secured in the snow with one end above the ground (45 cm) and pointing directly at the
camera. Snow was packed down around the log to maintain a consistent height above ground
throughout the survey. The log acted as a protruding solid object that Canada lynx could sniff,
scent mark, and rub their faces against. A wire gun-cleaning brush (30-caliber) was placed at the
end of the log to attract and collect hair from lynx. The gun brush was attached in a vertical
orientation flush against the butt end of the log with approximately one-quarter of the brush
length protruding above the top edge. A local commercial lure containing beaver castor and
catnip oil, as the two primary ingredients, was placed on either side of the gun brush. A small
piece of hanging American beaver meat (~5 cm diameter) was hung by wire directly above the
end of the log (~60 cm) to serve as an additional attractant. Cameras were checked, lure and bait
added, hair collected, and brushes replaced every two weeks in the winter and every four weeks
in the snow-free months. Beaver bait was not added to the sites during snow-free months due to
bear activity.

15

Figure 2. Camera trap and hair snare site for Canada lynx (Lynx canadensis) and other
mesopredators in John Prince Research Forest, British Columbia, Canada, 2015–2016 and 2020–
2022.

GPS Collars
I used GPS collars (Lotek LiteTrack 250; Lotek Engineering Inc., Newmarket, Ontario) to
monitor the movements and habitat use of lynx in each of 3 years, 2020–2022. Lynx were
captured in cage traps (Tomahawk 210A; 122 cm x 51 cm x 66 cm; Tomahawk Live Trap,
Hazelhurst, WI, U.S.A.) from February–April. Collars weighed 250–280g and were placed on
both adult male and female lynx. Marten were captured in cage traps (Havahart; 61 cm x 18 cm
x18 cm; Woodstream Corporation, Lancaster, PA, U.S.A.) from January–February of 2015 and
16

2016. I equipped marten weighting >600 g with miniature GPS collars (ATS G10 ultralite GPS
logger; Advance Telemetry Systems, Isanti, MN, U.S.A.). Research protocols for lynx were
approved by the UNBC Animal Care and Use Committee (Protocol # 2020-3) and British
Columbia Government (BC Wildlife Act, Permit PG19-596814). Research protocols for marten
were approved by the UNBC Animal Care and Use Committee (Protocol # 2013-20) and British
Columbia Government (BC Wildlife Act, Permit PG13-92070).

Table 1. Sampling methods and sample sizes for research on the habitat ecology of Canada lynx

Chapter 2

Chapter 3

Chapter 4

(Lynx canadensis) in central British Columbia, Canada, 2015–2016 and 2020–2021.

Camera traps (n = 66): 2015–2016

X

X

X

Camera traps (n = 66): 2020–2022

X

X

X

GPS collars (n = 17): Lynx 2020–2021

X

X

GPS collars (n = 9): Marten 2015–2016

X

Sampling method

Habitat Covariates
The study area had a forest inventory derived from high-density light detection and ranging
(LiDAR; 8–10 pulses/m2) data obtained in August and September 2015. Independent covariates
were represented by three broad categories that included forest cover and structure, disturbance,
17

and topography. Variables representing forest cover and structure from LiDAR data included
canopy closure at 3 vertical layers calculated at a 50m radius from a camera or GPS location
(bottom [0–3m], mid [3–10m], and top [>10m]) and cover calculated at the stand polygon scale
(ground [0–1m], low [1–3m], mid [3–10m], and top [>10m]).
The LiDAR data were collected during summer when deciduous vegetation was in leaf.
Thus, I used empirical data and values from the literature to adjust the LiDAR data for canopy
conditions typical of the leaf-off season (Davison et al., 2020, Wasser et al., 2013). I measured
canopy cover in 4 cardinal directions at 3 m from each camera location and >3 m above ground
and averaged those values during the leaf-on and leaf-off seasons (Canopeo Application;
www.canopeo.com; Alamo Software Foundation). I applied a correction factor of -25% and 52% to the leaf-off season for sites with few (26–50%) and a large (51–100%) proportion of
deciduous trees, respectively.
I used provincial forest inventory data (Parminter 2000) to represent a number of
ecological factors that were likely of relevance to the distribution of lynx, hare, and
mesopredators. That included distance of each camera site to riparian areas, including lakes,
wetlands and streams; the proportion of area (250m radius adjacent to each camera) containing
recent cutblocks (<15 years post harvest), density of edge between forest and recent forest
openings (<15 years post disturbance), elevation, and percent slope.
I used camera occurrences to measure the relative abundance of the typical prey of lynx
(red squirrel, snowshoe hare, grouse). Cameras have been compared to live trapping methods and
used to estimate population densities of red squirrels, red-backed voles, deer mice, and snowshoe
hares (Villette et al. 2016, Villette et al. 2017, Kenney et al. 2024). I used pellet transects

18

monitored from 2017 to 2022 as a secondary measure of the density of snowshoe hares
(Chisholm 2023, Krebs et al. 2023).

19

Chapter 2: Linking ecological processes of individual-based data and
occurrence surveys: Canada lynx (Lynx canadensis) and American marten
(Martes americana) movement and habitat selection using GPS-collars and
camera traps
Abstract
There has been a rapid increase in the use of camera traps to measure population-level
patterns of occurrence, distribution, and abundance of terrestrial wildlife. Individual patterns of
movement and behaviour may influence camera trap surveys. However, there has been little
study of the relationship between animal behaviour and occurrence at camera traps. Different
scales of behaviour, as represented by movement, by collared animals (slow vs. fast) can aid in
the interpretation of data collected during camera trap surveys. Habitat associations derived from
camera trap surveys may reflect most habitat used by an animal or only a small portion
depending on the influence of movement scale and associated behaviour. I used GPS-collar
locations and occurrence data at camera traps to compare habitat selection by Canada lynx (Lynx
canadensis) and American marten (Martes americana) at two movement scales (slow vs. fast)
during two periods of contrasting prey abundance. I found that camera traps, in general, reflected
habitat use measured with GPS-collars for both species. Mid-level and top-level vegetation cover
derived from LiDAR data were important predictors of habitat use for both marten and lynx, but
with opposite directional influences. For lynx, the direction and significance of habitat covariates
were nearly identical for camera traps and slow and fast movements, but with GPS-collar models
also being negatively associated with ground-strata (0–1m) cover. For marten, camera trap
locations were differentiated from GPS-collar locations (fast and slow) by being positively

20

associated with closer distances to riparian edge. Differential use of habitat defined by forest age
and structure by lynx and marten suggest that each species would serve as a unique indicator of
forest condition and change. Although I found that camera traps and GPS collars provided
similar assessments of habitat use for lynx and marten, it is unclear if this is the case for other
wildlife species. Estimates of distribution, abundance, and habitat use derived from camera-trap
surveys may be misinterpreted without knowledge of behavioural influences that likely vary by
species, time period, and geographic area.

Introduction
During the last two decades there has been an increase in the use of camera traps to measure
population-level patterns of occurrence, distribution, and abundance of terrestrial wildlife
(Burton et al. 2015, Steenweg et al. 2017). Habitat use often is derived from these surveys as a
core objective or as an integral component of population estimates. However, there is a
considerable gap in our understanding of the relationship between species occurrence at camera
traps and underlying ecological processes (Burton et al. 2015). For example, density estimates
from camera trap surveys are often dependent on species’ movement rates and home range
characteristics that are not well understood (Kjellander et al. 2004, Burton et al. 2015).
Individual patterns of movement (i.e., movement rate and path) and habitat use are two
mechanisms that explain population-level processes (i.e., abundance, distribution, occurrence) on
a landscape (Wiens et al. 1993). The movement rate and path provide insights on the decisions
animals make in moving from one location to another and the choice of patches determines the
responses of animals to variation in habitat availability (Wiens et al. 1993). Thus, environmental
21

heterogeneity influences both the movement and habitat choice of individual animals, and that
may strongly influence estimates derived from population-level surveys (Popescu et al. 2014,
Neilson et al. 2018, Stewart et al. 2018, Chandler et al. 2022). For this reason, the pairing of
GPS-collars and camera traps could provide valuable insights into the interpretation of point
surveys.
Most studies using camera traps assume that site detections fully represent animal space
use (Burton et al. 2015, Neilson et al. 2018, Stewart et al. 2018). A few studies have investigated
the influence of space use on estimates of occupancy or abundance by simultaneously collecting
species occurrence and animal movement data (e.g., Sollman et al. 2013, Popescu et al. 2014).
For fishers (Pekania pennanti), Popescu et al. (2014) found that telemetry and camera trap data
provided consistent information on space use. Stewart et al. (2018) studied a different population
of fishers and found that the magnitude of animal movements better reflected occurrence than the
proximate space use at camera trap locations.
The relationships between animal movement, habitat use, and camera trap locations are
unknown for many species. An understanding of habitat is an important component of
conservation and management plans for wildlife, yet it is unclear what types of habitats are
represented by camera data. Camera trap surveys may reflect most habitat used by an animal or
only a small portion. In addition, different behaviours (i.e., foraging, resting, travel, etc.) may
influence habitat associations. For example, animals may use different resources or habitat
features when hunting as compared to those resources used when moving to patches of habitat or
different portions of a seasonal range (Nams and Bourgeois 2004). Thus, camera trap locations
may represent a foraging patch or a travel corridor connecting foraging patches. Information may
be lost if animal locations are grouped into one behavioural category since variation in the
22

movement patterns and use of different habitats define interpretations of resource selection
(Johnson et al. 2002, Nams and Bourgeois 2004). A better understanding of the ecological
processes influencing camera trap surveys is needed if we are to have confidence that cameraderived habitat associations are effective for conservation and management decision making.
Canada lynx (Lynx canadensis) and American marten (Martes americana) are sympatric
at the landscape scale but often have contrasting habitat requirements at the scale of the habitat
patch. Marten are positively associated with late successional forests or the structural attributes
found in old forests (Buskirk and Powell 1994, Powell et al. 2003, Thompson et al. 2012), and
lynx are associated with forests exhibiting early successional characteristics that support
snowshoe hares (Lepus americanus; Mowat et al. 2000, Anderson and Lovallo 2003, Poole
2003). Selection for cover by lynx may be related to hare availability, but areas with extensive
forest cover may also provide thermal refugia during the summer (Vashon et al. 2008) and
protection from sympatric carnivores (McLellan et al. 2018). Horizontal cover can be an
important constituent element of den sites selected by lynx (Organ et al. 2008, Squires et al.
2008). The consistent but contrasting patterns of habitat associations between lynx and marten
are useful characteristics of indicator species (Lambeck 1997, McLaren et al. 1998, Noss 1999,
Carignan and Villard 2000), that may also lend well to studies investigating ecological processes
and surveys. The contrasting ecologies of lynx and marten can provide insights into the
effectiveness of camera traps for monitoring habitat use that expand monitoring and management
applications.
I used GPS-collar data (Global Positioning Systems) and camera trap surveys to
investigate behavioural and interspecific differences in habitat selection of sympatric populations
of Canada lynx and American marten. GPS collars can provide the fine-scale data necessary for
23

investigating behavioural indicators, including movement and habitat selection. To better
interpret resource selection, it helps to not only know what habitat animals use, but to also
understand how and why they are using these habitats. Different scales of movement by collared
animals (slow vs. fast) can be indicators of behaviours (i.e., foraging or resting vs. travel; Sibly
et al. 1990, Johnson et al. 2002a, b) and aid in the interpretation of data collected during point
surveys. For example, if camera locations represent a single movement scale (i.e., slow
movements) and associated habitat, survey results may not be representative of all habitat types
used by a species. Typically, location or “use” data are gathered with GPS or VHF collars.
However, similar data can be gathered with camera traps. To-date, there has been no comparison
of model outputs and resulting ecological inference generated with both collar and camera data
collected for marten or lynx. I compared habitat selection at two different movement scales to
camera data collected in the same area and time period. To better understand the consistency in
patterns of habitat use temporally, I also compared these three measures (e.g., camera trap, slowmovement GPS collar, and fast-movement GPS collar) to camera trap data collected outside of
this window during a contrasting period of prey abundance.
I predicted that lynx and marten would have two scales of movement: foraging patch
(slow; intrapatch) and corridor movements between foraging patches (fast; interpatch). For
lynx, I predicted that slow, hunting-type movements would occur in snowshoe hare habitat.
These are typically early successional forests with canopy heights ranging from 3–10m (Mowat
et al. 2000, Anderson and Lovallo 2003, Poole 2003). For marten, I predicted that slow
movements would occur in closed-canopy mature forest (>10 m canopy height). Marten are
known to select such habitats as they provide prey and vertical and horizontal cover (e.g.,
coarse woody debris) from predators (Buskirk and Powell 1994, Powell et al. 2003, Thompson
24

et al. 2012). I predicted that habitat selection using intrapatch movement data (more frequent,
slower rate of movement) collected from collared animals would be most similar to data
collected at camera-trap locations.

Materials and Methods
Study Area
The research was conducted in and adjacent to the John Prince Research Forest (JPRF) in northcentral BC, Canada, which is co-managed by the University of Northern British Columbia,
Binche Whut’en, and Tl’azt’en First Nations. The study area is ~390 km2 and characterized by
rolling terrain with low mountains (700 m to 1500 m above sea level). The region represents the
northern extent of contiguous Douglas-fir (Pseudotsuga menziesii var. glauca) forests in the
interior of British Columbia and is dominated by the Sub-Boreal Spruce biogeoclimatic zone
(Delong et al. 1993). The area has experienced a wide variety of logging activities over the past
75 years and contains a mosaic of old and young forest (continuum from new harvest to old
growth >250 years old) with interspersed deciduous stands. The study area had relatively little
forest harvesting during the time of data collection and has not had any significant trapping for
almost two decades. The study area is part of a long-term monitoring program investigating the
influences of climate and landscape change focused on multi-species detections across the JPRF
as well as an adjacent area managed by a major forest company.

25

Field Data Collection
GPS-collared Individuals
I used GPS collars (Lotek LiteTrack 250; Lotek Engineering Inc., Newmarket, Ontario) to
monitor the movements and habitat use of lynx in each of 3 years, 2020–2022. Lynx were
captured in cage traps (Tomahawk 210A; 122 cm x 51 cm x 66 cm; Tomahawk Live Trap,
Hazelhurst, WI, U.S.A.) from February–April. Collars weighed 250–280g and were placed on
both adult male and female lynx. Captured lynx were immobilized with a syringe pole using a
ketamine (5 mg/kg)/medetomidine (0.05 mg/kg) combination as an immobilizing agent. Each
lynx was equipped with a uniquely numbered (between animals) and colored ear tag in each ear.
Collars were placed on lynx weighing >7.5kg and programmed to transmit 6 locations per day
for the last 10 days of a month and 2 locations per day for the remainder of the month. Collars
automatically released from the lynx one year after the capture date. Research protocols for lynx
were approved by the UNBC Animal Care and Use Committee (Protocol # 2020-3) and British
Columbia Government (BC Wildlife Act, Permit PG19-596814).
Marten were captured in cage traps (Havahart; 61 cm x 18 cm x18 cm; Woodstream
Corporation, Lancaster, PA, U.S.A.) from January–February of 2015 and 2016. I guided
captured marten into a handling cone and immobilized them with isoflurane administered
through a portable diffuser (DesMarcheliers et al. 2007). I equipped marten weighting >600 g
with miniature GPS collars (ATS G10 ultralite GPS logger; Advance Telemetry Systems, Isanti,
MN, U.S.A.). Collars were programmed to record a location every five minutes and marten were
recaptured during the same winter to retrieve data (6–8 weeks). All successful GPS fixes were
calculated with four or more satellites (3D fixes). A passive integrated transponder tag (PIT) was

26

injected subcutaneously between the shoulder blades for individual identification. Research
protocols for marten were approved by the UNBC Animal Care and Use Committee (Protocol #
2013-20) and British Columbia Government (BC Wildlife Act, Permit PG13-92070).
For both species, I checked traps once or twice daily, depending on weather conditions
and location. Cage trapping was suspended if conditions were wet or temperatures were < -20˚
C. Cage traps were baited with beaver meat (Castor canadensis) and commercial lures. A
hanging grouse feather/wing was added to lynx traps. I placed conifer boughs and corrugated
plastic over the trap to provide protection from precipitation. Hair and tissue samples were
collected from anesthetized animals. Standard morphological measurements were taken (i.e.,
weight, body length, chest girth, tail length, shoulder length, zygomatic arch). I assessed age by
weight and an examination of tooth wear and coloration.
Camera Traps
I used data collected from camera traps during two contrasting periods (2015–2016 and 2020–
2022) of hare abundance (Chisholm 2023). Occurrence rates (# occurrences/# camera days) of
snowshoe hares at camera traps decreased 73% between the winter of 2015–2016 (i.e., high hare
abundance) and 2020–2022 (i.e., low hare abundance). This trend was supported by a 40%
decrease in hare densities measured at pellet plots established 3 years after the estimated hare
peak (2018–2021; John Prince Research Forest, unpublished data). I deployed 66 camera traps
(Bushnell Trophy Trail Cameras models 119467 and 11947; Bushnell Outdoor Products,
Missouri, USA) on a hexagonal grid (2.5 km apart; 5.41 km2 cell size) from January–March in
2015 (23 Jan–03 Apr) and 2016 (01 Feb–10 Apr). In 2020–2022, trail cameras (Browning Dark
Ops HD Pro Trail Cameras model BTC-6HDP; Browning, Utah, USA) were set at the same sites
as 2015–2016. Camera data used from 2020–2022 included: 01 February–11 April and 24 June–
27

14 October. Camera models used during the two time periods had similar trigger speeds (0.2
seconds). I also used a consistent measured distance between camera traps and scent posts to
standardize detection distance and the field of view (see paragraph below). Camera traps were
randomly placed at or near the center of each hexagonal cell.
At each site, a scent post was established 2.5–3 m directly in front of a remote camera
between 0.5 and 1 m above the ground on a tree. Each scent post consisted of a small diameter
log (<15 cm) secured with one end above the ground (45 cm) and pointing directly at the camera.
A local commercial lure containing beaver castor and catnip oil was placed at the end of the log.
A small piece of hanging beaver meat (~5 cm diameter) was hung by wire directly above the end
of the log (~60 cm) to serve as an additional attractant. Bait was consumed quickly by most
carnivores (~2–3 days), but the remnant scent continued to serve as an additional lure to
encourage animals in the vicinity of the camera trap to move into view. Cameras were checked
and lure and bait added every two weeks in the winter and every four weeks in the snow-free
months. Only scent lure was added to sites during snow-free months due to bear activity. In
2015–2016, cameras were set to take 30s of video with a 1s delay between video-recordings. In
2020–2022, cameras were set to take 10s of video with a 1s delay between video-recordings.
These schedules allowed for nearly continuous recording of the time an animal was in view.

Habitat Covariates
The study area had a forest inventory derived from high-density light detection and ranging
(LiDAR; 8–10 pulses/m2) data obtained in August and September 2015. Independent covariates
were represented by three broad ecological and environmental categories that included forest
cover and structure, disturbance, and topography (Table 2). Variables representing forest cover

28

and structure were derived from LiDAR data and included canopy closure at 3 vertical layers
calculated at a 50m radius from a camera or GPS location (bottom [0–3m], mid [3–10m], and top
[>10m]) and cover calculated at the stand polygon scale (ground [0–1m], low [1–3m], mid [3–
10m], and top [>10m]). I used GIS data to measure the distance of each camera site to riparian
areas, the proportion of area (250m radius) containing recent cutblocks (<15 years in age), edge
density, elevation, and slope.

Table 2. Variables used in the development of habitat selection models for Canada lynx (Lynx
canadensis) and American marten (Martes americana) using camera traps and GPS-collared
individuals in central British Columbia, Canada, 2015–2016 and 2020–2022.

Parameter
cover (0–3)
cover (3–10)
cover (10+)
Stand cover (0–1)
Stand cover (1–3)
Stand cover (3–10)
Stand cover (10+)
elevation
slope
dist.riparian
dist.edge
prop. forest age (<20 yrs)
edge density

Description
Average canopy cover 0–3 m (LiDAR, 50-m radius)
Average canopy cover 3–10 m (LiDAR, 50-m radius)
Average canopy cover >10m (LiDAR, 50-m radius)
Average canopy cover (LiDAR; 0–1 m) stand polygon scale
Average canopy cover (LiDAR; 1–3 m) stand polygon scale
Average canopy cover (LiDAR; 3–10 m) stand polygon scale
Average canopy cover (LiDAR; 10+ m) at the stand polygon scale
Digital elevation model (LiDAR; m)
Digital elevation model (LiDAR; percent)
Distance to stream and lake edge (m)
Distance to forest stand edge (m)
Proportion of area (250-m radius) with recent cutblocks (<20 yrs)
Edge density (250-m radius)

Variable
type
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous

29

Data Analysis
Movement Scales
I used a non-linear curve fitting procedure to determine if observed movement patterns
represented two scales (i.e., slow vs. fast; Sibly et al. 1990, Johnson et al. 2002a, b). I first
calculated the movement rate (distance/time) between successive GPS-collar locations using a 4hr interval for both lynx and marten. For lynx, this represented a 10-day period each month. A
two-process non-linear curve was then fit to the frequency distribution of movement rates for
each collared animal by season and species to identify movement scales (i.e., large-scale vs.
small-scale processes). Specifically, I used a piece-wise “hockey-stick” regression to determine a
break point in movement rates for individuals of each species (Bacon and Watts 1971). This
model assumes the process can be fit by two straight lines, with different slopes, and calculates
the two slopes and movement rate value (breakpoint) at which the slope changes. I grouped all
individual-based movement scales (fast vs. slow) for use in further analyses (Figure 3 and 4).

30

Figure 3. Map of study area illustrating camera locations as well as GPS-collar locations
representing fast and slow movements by Canada lynx (Lynx canadensis) during the winter
period in central British Columbia, Canada, 2015–2016 and 2020–2022.

31

Figure 4. Map of study area illustrating camera locations as well as GPS-collar locations
representing fast and slow movements by American marten (Martes americana) during the
winter period in central British Columbia, Canada, 2015–2016 and 2020–2022.

32

Habitat Selection
I used multinomial logistic regression to investigate habitat characteristics that influenced the
occurrence of lynx and marten. Occupancy modelling was considered but determined not
necessary given our target species (common with high detection rates), survey effort and timing
(continuous monitoring during period of interest), camera spacing (< 2.5km between sites), use
of bait and lure that were designed to provide a high likelihood of detection, and focus on
occurrence throughout the entire survey duration (Mackenzie and Royle 2005, Tigner et al.
2015). Furthermore, home ranges of lynx and marten were large enough to encompass multiple
camera stations violating the assumption of spatial independence for occupancy models (Burton
et al. 2015, MacKenzie et al. 2004, Parsons et al. 2019). Multinomial regression allowed me to
investigate differences in resource selection between camera and collar data sets using the same
set of random locations.
For the dependent variable in the camera trap portion of the multinomial models, I first
summed all daily occurrences (interval = 24 hour) at a site during each season and biological
period: winter low hare abundance (2020–2022), winter high hare abundance (2015–2016),
snow-free low hare abundance (2020–2021), and snow-free high hare abundance (2016). I then
calculated the occurrence rate (#occurrence days/#operational camera days) for lynx and marten
at each site. I classified the frequency of lynx or marten into two categories (moderate/high (1)
vs no/low use (0)) based on the 50th percentile of the occurrence rate at a site.
The binary dependent variable for occurrence rate allowed for consistent analyses
between years (i.e., much greater number of lynx and marten occurrences in 2015–2016), and an
accounting of variation in sampling effort between seasons and sites. The response variable
served as a coarse measure of relative abundance (no/low use vs moderate/high use) rather than
33

presence or absence. For lynx in the winter low hare season, a “0” represented a occurrence rate
of < 1.4 lynx occurrences/site/day. For snow-free low hare, a “0” represented 0.6 lynx
occurrences/day. Lastly, for lynx winter high hare and snow-free high hare, a “0” equated to <
2.9 and < 1.2 occurrences/site/day, respectively. For marten in the winter low hare and high hare
years, a “0” represented an occurrence rate of < 1.06 and < 2.17 marten occurrences/site/day,
respectively.
There were 5 possible outcomes in each multinomial logistic model that represented the
occurrence of lynx or marten: GPS fast or slow movements, 2020–2022 camera data, and 2015–
2016 camera data. All data sets were compared to a fifth set of random locations selected from
across the camera grid study area at a minimum 1:1 ratio for each multinomial data set. Models
for understanding resource selection with cameras included camera trap sites with moderate/high
use by lynx or marten in 2020–2022 (Outcome 1) and camera trap sites with moderate/high use
by lynx or marten in 2015–2016 (Outcome 2). Models for understanding resource selection using
GPS collars were stratified into two scales of behaviour that represented slow movement GPScollar locations (Outcome 3), and fast movement GPS-collar locations (Outcome 4; Figure 3 and
4). Models for lynx were fit for 2 seasons (winter and snow-free), while models for marten were
fit for winter only, the only season of monitoring with GPS collars.
I used an Information Theoretic Model Comparison (ITMC) approach to develop a set of
10 multinomial regression models that contrasted resource selection as quantified with camera
traps and GPS-collars for each species and season (i.e., lynx winter, lynx snow-free, marten
winter; Burnham and Anderson 2004; Table 3). I used Akaike’s information criterion for small
sample sizes (AICc) to identify the most parsimonious model explaining habitat association and

34

differentiation between camera occurrence and GPS-collar data sets (Burnham and Anderson
2004). I used both ΔAICc and Akaike weights (AICcw) to rank and compare models.

Table 3. A priori candidate models (multinomial logistic regression) representing habitat use of
Canada lynx (Lynx canadensis) and American marten (Martes americana) using camera traps
and GPS-collared individuals in central British Columbia, Canada, 2015–2016 and 2020–2022.

Model Name
Canopy Cover (50-m buffer)
Topography
Disturbance
Stand Canopy Cover
Riparian + Cover
Disturbance + Topography 1
Disturbance + Topography 2
Topography + Cover
Disturbance + Cover
Null

Model
cover (0–3) + cover (3–10) + cover (10+)
elevation + slope + dist. riparian
prop. forest age (<20 yrs) + edge density
cover (0–1) + cover (1–3) + cover (3–10) + cover (10+)
dist. riparian + cover (3–10 m) + cover (10+ m)
prop. forest age (<20 yrs) + edge density + elevation + slope
prop. forest age (<20 yrs) + dist.riparian + elevation
elev + slope + dist.riparian + cover (0 –3 m) + cover (3–10 m) +
cover (10+ m)
edge density + cover (3–10 m) + cover (10+ m)
no independent covariates

K
4
4
3
5
4
5
4
7
4
1

I used the receiver operating characteristics and resulting area under the curve (AUC) to
assess the predictive ability of the best model (Pearce and Ferrier 2000). I used a one-fold crossvalidation routine to withhold each record sequentially from the model building process and then
calculated the independent probability of that withheld record being a species location. I
considered a model with an AUC score of 0.7 to 0.9 to be a useful application and a model with a

35

score >0.9 as highly accurate (Boyce et al. 2002). I used the AUC scores to compare the relative
predictive ability of the 4 outcomes for each species’ model set.
I used 95% confidence intervals to assess the strength of effect of each predictor
covariate on the dependent variable. Poor power and inconclusive statistical inference are
expected from covariates with confidence intervals that approach or overlap 0. I used tolerance
scores to assess variables within each model for excessive collinearity (Menard 2001). All data
analyses were performed using Stata (version 17.0; Statacorp, College Station, Texas).

Results
Marten Captures
I captured and monitored a total of 9 marten (5 male, 4 female) with GPS collars in 2015 (3
marten) and 2016 (6 marten). I obtained only 157 locations over 31 days for one collared female.
The 5-minute fix success rate for the remaining 8 collars (<10m DOP) was 13.0%. Collars
successfully recorded between 748 and 2996 (‫ = ̅ݔ‬1680.8, SE = 278.7) locations over periods of
35 to 52 days (‫ = ̅ݔ‬46.0, SE = 2.3). This fix schedule corresponded to an average of 36.3 (SE =
5.38) locations per day (Range = 16.6–58.9). Location data were standardised to an interval of 3–
4 hours (446 marten locations) to ensure consistency with the collar schedule used for GPS
collars deployed on lynx. A total of 446 random locations were used in this analysis.

Lynx captures
In total, I captured and collared 17 Canada lynx (11 male, 6 female) from 2020–2022. I captured
7, 4, and 6 lynx in 2020, 2021, and 2022, respectively. Overall fix success rate (<10m DOP) was
70.0%. Collars successfully recorded between 140 and 1338 (‫ = ̅ݔ‬827.3, SE = 82.4) locations
36

over periods of 32 to 382 days (‫ = ̅ݔ‬262.7, SE = 26.8). After filtering the data set for 4-hour fixes
and locations within the LiDAR area, I used a total of 5,335 lynx locations and 5,335 random
locations (winter = 3,193, snow-free = 2,142) for the habitat analyses.

Camera Traps
In 2015–2016, cameras functioned properly an average of 96.0% (SE = 0.9; 8,866 camera-days)
and 90.3% (SE = 2.4; 5,008 camera-days) of potential camera days in the winter and snow-free
periods, respectively. In 2020–2022, cameras functioned properly an average of 95.1% of
potential camera days in the winter (SE = 0.8; 13,181 camera-days) and 91.1% (SE = 1.4; 10,101
camera-days) of potential camera days in the snow-free periods. There was a 76.2% decrease in
lynx occurrence rates at camera traps between 2015–2016 (8.26 occurrences/day) and 2020–
2021 (1.97 occurrences/day) during the winter, and a 65.9% decrease (3.17 to 1.08
occurrences/day) in the snow-free season. There was a 39.4% decrease in marten occurrence
rates between 2015–2016 (5.81 occurrences/day) and 2020–2021 (3.52 occurrences/day) during
the winter.

Movement Scales
For both lynx and marten, the two-process non-linear model fit the loge frequency distribution of
movement rates (P < 0·001) for most individuals (lynx winter = 12 of 17; lynx snow-free = 12 of
15; marten = 6 of 9). Data from individuals that fit the two-process non-linear model were used
in the subsequent habitat analyses. For lynx in the winter and snow-free period, the break point
between slow and fast movements was 7.42 m/min (11.9% fast) and 6.58 m/min (14.4% fast),
respectively. For marten in winter, the break point was 6.9 m/min (18.6% fast).

37

Lynx Winter Habitat Models
Canopy cover at the stand scale was the top-ranked model explaining habitat selection by lynx
(AICcw = 1.0); Table 4). Lynx were positively associated with low-strata (1–3m) and mid-strata
vegetation cover (3–10m) during winter, but negatively associated with top-strata cover (10m +)
(Figure 5). The direction and significance of covariates were identical for both camera trap
periods, and nearly identical for both movement scales. GPS-collar models were differentiated
from camera traps by also being negatively associated with ground-strata (0–1m) cover. All
models had useful predictive accuracy (AUC > or overlapping 0.7).

Lynx Snow-free Habitat Models
Similar to winter, the stand-level cover model was best at representing the resource selection of
lynx (Table 4; AICcw = 1.00). Lynx selected for sites with more mid-strata vegetation cover and
less cover >10 meters for both years of camera occurrence data and GPS-collar locations
measured at both movement scales (Figure 5). Although lynx were positively associated with
low-strata vegetation (1–3m) for all models, this covariate was not significant for the 2015–2016
camera data set. GPS-collar locations (fast and slow) were differentiated from camera trap
locations by being negatively associated with ground-strata (0–1m) cover. Only the 2020–2022
camera model had a useful predictive accuracy (AUC = 0.69, SE = 0.042).

Marten Winter Habitat Models
The regression model with topography and cover covariates provided the most parsimonious
representation of resource selection by marten during winter (Table 4; AICw = 0.99). Marten
monitored with GPS collars and cameras were positively associated with greater amounts of
38

forest cover above 10 meters (Figure 6). At camera traps, marten selected for sites closer to
riparian edge. Although not significant, slow and fast movements were associated with further
and closer distances to riparian features, respectively. Marten selected for lower elevation
portions of the study area, but this covariate was only significant when included in the GPS
collar models for both scales of movement. All model outcomes had useful predictive accuracy
(AUC > 0.7).

Table 4. AICc scores, and AICc weights (w) for the three highest-ranked multinomial logistic
regression models representing habitat use of Canada lynx (Lynx canadensis) and American
marten (Martes americana) using GPS collars and camera traps in central British Columbia,
Canada, 2015–2016 and 2020–2021. Area under the curve (AUC) and standard error (SE) for the
receiver operating characteristic represents the predictive accuracy of each model.
Model
LYNX WINTER
Cover Stand Scale
Disturbance + Cover
Topography + Cover
LYNX SNOW-FREE
Cover Stand Scale
Topography + Cover
Riparian + Cover
MARTEN WINTER
Topography + Cover
Disturbance + Topo. 1
Disturbance + Topo. 2

K

AICc

ΔAICc

w

AUC1(SE)

AUC2(SE)

AUC3(SE)

AUC4(SE)

5
4
7

8489.4
8741.5
8756.5

0.0
252.1
267.1

1.00
0.00
0.00

0.72 (0.03)

0.74 (0.01)

0.68 (0.02)

0.70 (0.04)

5
7
4

8248.4
8265.8
8324.9

0.0
17.5
76.5

1.00
0.00
0.00

0.69 (0.04)

0.67 (0.01)

0.64 (0.02)

0.62 (0.04)

7
6
4

1789.3
1803.7
1807.8

0.0
14.4
18.5

1.00
0.00
0.00

0.70 (0.04)

0.75 (0.02)

0.77 (0.03)

0.73(0.04)

39

Figure 5. Significant coefficients for top-ranked models illustrating habitat use by Canada lynx
(Lynx canadensis) during the winter and snow-free periods using GPS collars and camera traps
in central British Columbia, Canada, 2015–2016 and 2020–2022. Blue = Camera (2020–2022),
Red = Camera (2015–2016), Green = GPS slow movement, Orange = GPS fast movement.
40

Figure 6. Significant coefficients for top-ranked models illustrating habitat use by American
marten (Martes americana) using GPS collars and camera traps during the winter in central
British Columbia, Canada, 2015–2016 and 2020–2022. Blue = Camera data (2020–2022), Red =
Camera data (2015–2016), Green = GPS slow movement, Orange = GPS fast movement.

41

Discussion
This is the first study to investigate ecological processes influencing camera trap surveys of
sympatric Canada lynx and marten. I found that camera traps, in general, provided similar
inferences about habitat selection when compared to GPS-collar data. This relationship held for
two species with different ecological niches and contrasting habitat needs as well as between
survey periods with differences in the abundance of snowshoe hare. My data suggest that camera
traps can be a reliable method for measuring habitat relationships of lynx and marten.
I found minor differences in habitat selection between the two scales of movement. Fast
movements by GPS-collared marten were associated with steeper slopes, more mid-level
vegetation cover, and adjacency to riparian features. Although these covariates were nonsignificant, the relationships suggested that marten moved faster in riparian habitats or older
regenerating stands. Marten may be using these habitats for inter-patch travel between foraging
or resting habitat. Habitat selection by lynx was consistent for both slow- and fast-scale
movements. Canada lynx are habitat specialists and may spend the majority of their time
foraging and traveling in similar habitats at the 4-hr movement intervals used in this study. The
location interval for GPS collars on lynx limited my analyses to only two movement scales. Data
representing shorter interval movements could reveal finer scales of movement and differences
in habitat use at additional spatiotemporal scales (Nams and Bourgeois 2004).
The influence of habitat covariates was similar between time periods with contrasting
prey abundance. In addition to a decrease in hare, lynx, and marten occurrence, there was an
increase in red squirrel, fisher, and wolverine from 2015–2016 to 2020–2022 (See Chapter 4.). I
predicted that GPS-collar locations would best match camera-trap data that were collected
concurrently, when ecological conditions across the study area were the same. My results
42

generally supported that prediction, but I also found similar habitat selection using camera-trap
data collected outside my GPS-collar periods. Although the abundance of lynx and marten, their
prey, and sympatric carnivores changed between monitoring periods, vegetation cover remained
a consistent and important predictor of the occurrence of these two habitat specialists.
Mid-level and top-level vegetation cover were important predictors of habitat use for both
marten and lynx, but with some opposite directional influences. For example, lynx selected for
less top canopy cover, while the inverse relationship was found for marten. For lynx, the topranked model included cover covariates measured at the stand polygon scale, while the top
marten model had cover covariates measured at the area surrounding the camera (i.e., 50-m
radius). Although measured at different spatial scales, cover covariates were often significant and
clearly an important component of the habitat of lynx and marten. For marten, the selection for
habitat closer to riparian features was the covariate that differed most between the two methods.
Distance to riparian edge significantly influenced the occurrence of marten at camera locations
during both time periods (2015–2016 and 2020–2022) but not at GPS-collar locations. Riparian
edge is a relatively narrow and small proportion of the total land area, and marten may be using
riparian corridors frequently to travel between foraging habitats. Although marten may spend a
relatively small proportion of their time in riparian areas, they were captured on camera
consistently when moving in these corridors.
My habitat selection results for both marten and lynx were consistent with other habitat
studies in North America. Both species are generally considered habitat specialists, but with
contrasting habitat requirements. Although American marten can persist in habitats such as
regenerating or deciduous forests (Poole et al. 1994, Payer and Harrison 2003), they are
generally associated with mature spruce/fir forests containing sufficient canopy cover and
43

structure (Buskirk and Powell 1994, Chapin et al 1998, Hargis et al 1999). Large open areas are
often avoided by marten and may limit dispersal (Buskirk and Powell 1994, Moriarty et al 2015).
Habitat fragmentation and loss of interior forests may result in reduced abundance or even local
extirpation of marten populations (Hargis et al 1999, Moriarty et al 2011). Consistent with these
studies, I found that canopy cover above 10m was a key predictor of habitat selection by marten.
In contrast, Canada lynx are generally found in boreal, sub-boreal and montane forests with finer
scale habitat selection matching the distribution of snowshoe hare (Mowat et al. 2000, Poole
2003, Thomas et al. 2019). At the stand scale, lynx often select for older regenerating stands
(>20 years old) with dense understory that contains horizontal and vertical cover, while selecting
against younger regenerating stands (Mowat et al. 2000). Canada lynx in my study also selected
for stands with greater canopy cover (1–10m) typically found in older regenerating forests.
Lynx and marten both display consistent and identifiable patterns of habitat association, a
key characteristic of indicator species. Indicator species are used to measure the overall status
and condition of ecosystems often in relation to human disturbance (Caro 2010). The indicator
species concept is attractive because it allows sampling of the occurrence or abundance of a
small set of species rather than the entire community (Caro 2010, De Caceres et al. 2010). The
sensitivity of marten and lynx to forest age and structure (i.e., cover) as well as variation in the
abundance of prey suggests that both species are good indicators of forest condition and change.
Maintaining sufficient habitat for these two species on the same landscape would require a
mosaic of stands with varying cover characteristics likely beneficial to many other wildlife
species. However, more research is required to confirm the extent (i.e., number of species or
species guilds) and strength of that relationship relative to varying forest conditions.

44

Many ecological patterns and processes have their mechanistic roots in individual
behaviours of movement and habitat use (Wiens et al. 1993). Historically, many assessments of
habitat use by wildlife were based on location data collected with radio transmitters or GPS
collars. Although this technique can provide accurate individual-based location data that
represent movement and habitat use, these studies are expensive, labor intensive, and expose
animals to the risk of capture and immobilisation. Given those costs, studies of mesopredators
often have small sample sizes of collared animals limiting the extrapolation of results across
larger landscapes. In the last two decades, new technologies focused on DNA and camera
sampling have offered alternatives for researching and monitoring the distribution of wildlife
populations across larger areas. However, we still have little understanding of the ecological
processes influencing these surveys and how they may affect the assumptions and biases
incorporated into survey design and analyses. Although I found that camera traps and GPS
collars provided similar assessments of habitat use for two sympatric species that are relatively
common across much of North America’s boreal and subboreal forests, I recommend more
paired comparisons of GPS-telemetry and non-invasive survey methods. Estimates of
distribution, abundance, and habitat use derived from camera-trap surveys may be misinterpreted
without knowledge of behavioural influences that likely vary by species, time period, and
geographic area (Burton et al. 2015).

45

Chapter 3: Behaviour as an indicator of cyclic trends in abundance of Canada
lynx (Lynx canadensis) and snowshoe hare (Lepus americanus)
Abstract
The behaviour of individual animals influences reproduction and survival. Thus, variation
in animal behaviour can result in changes in the distribution and abundance of populations as
well as broader community dynamics. The pairing of behavioural cues with abundance indices
can provide additional insights into the ecological processes influencing population and
community trends. I used a combination of occurrence and behavioural data from camera traps to
measure variation in the abundance of Canada lynx populations during two time periods with
contrasting abundance of their primary prey, snowshoe hare. My objective was to determine if
N-mixture models, camera occurrence rates, and behavioural cues could be used to monitor
trends in cyclic abundance. Behaviours were measured to both interpret abundance estimates and
serve as a potential co-indicator of population trends. I found that lynx behaviours and relative
abundance were correlated among years with varying abundances of snowshoe hare. Consistent
with my predictions, years with greater abundance of lynx and hare were characterized by
increases in cheek-rubbing, scent-marking, and grouping behaviours. Variation in behaviour and
indices or estimates of abundance proved to be strong co-occurring indicators of the cyclic
population dynamics of Canada lynx. Behaviour can serve as an indicator of reproduction,
demographics, and survival that both influence and are influenced by population density. Thus,
measures of behaviour from individual animals can provide insights into changes in populationlevel occurrence rates at camera traps.

46

Introduction
The behaviour of individual animals both influences and is influenced by reproduction and
survival. Thus, variation in animal behaviour can result in changes in the distribution and
abundance of populations as well as broader community dynamics (Bro-Jorgensen et al. 2019,
Green et al. 2019, Higashide et al. 2021). A better understanding of the ecological factors
influencing variation in behaviour could aid in the design or interpretation of population surveys.
Also, where behavioural changes are related to environmental or ecological conditions, including
interspecific interactions (e.g., predator-prey dynamics), they could complement abundance
estimates and be used as additional or alternative indicators of population status and trends.
In the northern boreal forests of Alaska and Canada, snowshoe hare (Lepus americanus)
populations cycle over an 8–11-year period (O’Donoghue et al. 1997, Krebs et al. 2014, Krebs et
al. 2017). This change in hare abundance influences the body condition, reproduction, and
associated behaviour of predators, such as Canada lynx (Lynx canadensis), and could have
implications for methods designed to quantify changes in population abundance of prey and
predator. For example, barbed and scented rub pads that rely on cheek-rubbing are among the
most commonly used DNA-based method for surveying Canada lynx (McDaniel et al. 2000,
Mills et al. 2000, Kendall and McKelvey 2008). Demographic or temporal variation in cheekrubbing behaviour may influence the accuracy, precision, or bias of that technique. Variability in
cheek-rubbing or scent-marking is not well understood, but may be influenced by sex, age,
habitat, prey and predator density, season, lure type and timing, and previous visits by
conspecifics. For example, there can be a distinct seasonal peak in the rubbing behaviour of
Canada lynx, coinciding with the breeding season in late winter (Crowley et al. 2013, Crowley et
al. 2017). Similarly, the efficacy of hair-snare surveys for Eurasian lynx (Lynx lynx) can also
47

vary by season, being more successful during the breeding period (Schmidt and Kowalczyk
2006). Also, scent-marking can vary by habitat type and prey density. For example, the Sunda
clouded leopard (Neofelis diardi) and the Iberian lynx (Lynx pardinus) were more likely to scent
mark habitats with a high density of prey (Allen et al. 2016a, Burgos et al. 2018). The frequency
and duration of scent-marking by puma (Puma concolor) was influenced by previous visitation
of conspecifics (Allen et al. 2016b). Vogt et al. (2016) suggested that Eurasian lynx faced a
trade-off between enhancing detection of scent-marks by conspecifics and avoiding detection by
prey. These ecological and behavioural factors undoubtedly affected sampling bias and resulting
estimates of occupancy and abundance used for the conservation and management of felid
species.
Non-invasive methods for estimating the abundance of elusive, forest-dwelling
carnivores include camera traps, sampling of fecal material or hair for DNA, and surveys of
tracks or other sign (Long et al. 2008). Due to their relatively low costs and applications across
large landscapes, these surveys are often employed as an alternative or complement to studies
that require invasive marking or collaring of individuals (e.g., GPS collars, ear tags). Advances
in laboratory techniques and more widespread use of recreational game cameras has reduced per
unit costs of sampling and increased reliability allowing for greater sample sizes and more
rigorous data collection.
In addition to new methods for collecting non-invasive data for population monitoring,
there has been considerable innovation in methods for analysing those data. Spatial mark-resight
models and count models are used with increasing frequency to generate estimates of population
abundance from camera images. The N-mixture count model is one of the more common
methods to estimate abundance of completely unmarked wildlife populations (Royle 2004, Priol
48

et al. 2014, Kidwai et al. 2019). This model is attractive because it uses count data at repeatedly
sampled sites to estimate abundance and detection probability simultaneously without the need to
identify unique individuals (Royle 2004). A latent state N represents the absolute abundance and
probability, p, of the detection process (Royle 2004, Royle and Dorazio 2006). The variation in
N across sites follows a Poisson distribution where the expected value is logarithmically linked
to linear covariates and is assumed to remain constant across multiple observations within each
site (Royle 2004). In these models, detection-driven and abundance-driven variation are assumed
to occur within sites and between sites, respectively. However, predictions of population
abundance can be sensitive to underlying assumptions, including: 1) abundance at each site is
random and independent of other sites, 2) population remains closed between surveys, 3)
individuals are not double counted, and 4) every individual in the population has an equal
detection probability (Royle 2004, Priol et al. 2014, Burgar et al. 2018, Ficetola et al. 2018, Link
et al. 2018, Sun et al. 2022). Other challenges include data separation for the abundance and
detection components of the N-mixture model and choosing the most accurate and precise model
from a set that represent variation in the environmental conditions at each sampling site (Koetke
et al. 2024).
Canada lynx are generally not recognizable by pelage patterns in the winter and require
DNA-based surveys to identify individuals (but see Anderson et al. 2022). Camera trap data and
N-mixture models could be useful for monitoring temporal and spatial trends in abundance, but
the underlying behavioral processes influencing the precision and reliability of this method are
unknown. I used a combination of behavioural responses, abundance estimates, and abundance
indices to measure temporal trends in the abundance of Canada lynx. I conducted the analysis for

49

two time periods when the primary prey of lynx, snowshoe hare, was relatively abundant and
scarce, according to the well-documented population cycle.
I estimated the relative abundance of lynx for a total of six years using N-mixture models
and occurrence rates ((lynx occurrence/operational camera days) *100) that were derived from
observations at camera traps. For each image, I recorded three behavioural responses by lynx:
cheek-rubbing, scent-marking, and grouping. All of these behaviours are associated with
reproduction and most likely influenced by population density. My objective was to determine if
a combination of abundance and behavioural responses could be used to monitor and interpret
cyclic population trends. Also, I estimated lynx abundance using spatial capture-mark-resight
(SCMR) to aid in the interpretation of N-mixture abundance estimates. SCMR models are
commonly used to estimate the abundance or density of wildlife populations where some
individuals are marked (Rich et al. 2014, Efford and Hunter 2017, Doran-Myers et al. 2021).
My second objective was to investigate the underlying environmental and behavioural
factors influencing cheek-rubbing, scent-marking, and grouping behavior of Canada lynx.
Variation in behaviour could result in a spatial or temporal bias in the number of observed lynx
and the resulting population estimate or index. I developed a suite of covariates to explain the
presence or absence of each behaviour during a lynx occurrence at a camera trap including
vertical and horizontal cover, prey and conspecific abundance, season, lure timing, group size,
and previous visits by conspecifics. I predicted that cheek-rubbing, scent-marking, and grouping
behaviours would be positively influenced by greater hare and lynx densities at sites, prior visits
by conspecifics, luring by researchers, and time period relative to the snowshoe hare cycle.
Lastly, I predicted that the frequency of scent-marking and cheek-rubbing behaviour would
increase with the lynx breeding season (mid-winter; February–mid-March).
50

Materials and Methods
Study Area
The research was conducted in and adjacent to the John Prince Research Forest (JPRF) in northcentral British Columbia (BC), Canada, which is co-managed by the University of Northern
British Columbia, Binche Whut’en, and Tl’azt’en First Nations. The study area is ~390 km2 and
characterized by rolling terrain with low mountains (700 m to 1500 m above sea level). The
region represents the northern extent of contiguous Douglas-fir (Pseudotsuga menziesii var.
glauca) forests in the interior of British Columbia and is dominated by the Sub-Boreal Spruce
biogeoclimatic zone (Delong et al. 1993). The area has experienced a wide variety of logging
activities over the past 75 years and contains a mosaic of old and young forest (continuum from
new harvest to old growth >250 years old) with interspersed deciduous stands. The study area
had relatively little forest harvesting during the time of data collection and has not had any
significant trapping for almost two decades. The study area is part of a long-term monitoring
program investigating the influences of climate and landscape change on the distribution and
abundance of carnivore communities in central BC.

Field Data Collection
I deployed 66 camera traps (Bushnell Trophy Trail Cameras models 119467 and 11947;
Bushnell Outdoor Products, Missouri, USA) on a hexagonal grid (2.5 km apart; 5.41 km 2 cell
size) from January–March in 2015 (23 Jan–03 Apr) and 2016 (01 Feb–10 Apr) and from June–
October (22 Jun–19 Jul; 12 Aug–08 Sep; 22 Sep–19 Oct) in 2016. In 2014, I deployed cameras
at 41 of these same locations from 12 February–08 April. Data from January–April represented
the winter period, and June–October the snow-free period. In 2020–2022, camera traps
51

(Browning Dark Ops HD Pro Trail Cameras model BTC-6HDP; Browning, Utah, USA) were set
at the same sites as 2015–2016. Although the same camera sites were monitored continuously
from January 2020–April 2022, I used time periods in which the highest proportion of cameras
were active, survey starts coincided with camera check and lure days, and survey timing
overlapped with dates in 2014–2016. I only included sites in the analyses in which the remote
camera was active >50% of the survey period. A total of four cameras were removed from the
snow-free high period as well as one camera from the snow-free low period due to camera
malfunction.
The 2014–2016 and 2020–2022 data sets represented periods of high and low hare
abundance, respectively. Although the exact timing of the peak and trough in hare populations
was uncertain, the data represented two time periods of contrasting abundance. Occurrence rates
of snowshoe hares at camera traps decreased 73% between the two winter periods. This trend
was supported by a 40% decrease in hare densities measured at pellet plots established 3 years
after the estimated hare peak (2018–2021; John Prince Research Forest, unpublished data).
The hexagonal grid was randomly placed on the study area and cameras set near the
center of each 5.41-km2 cell. Although some camera traps were set near roads and trails, none
were set directly along these linear features. At each camera site, a scent post was established
2.5–3 m directly in front of a camera trap between 0.5 and 1 m above the ground on a tree. This
set-up maintained a consistent height and angle of view to the scent post. Each scent post
consisted of a small diameter log (<15 cm) secured with one end above the ground (45 cm) and
pointing directly at the camera. A local commercial lure containing beaver (Castor canadensis)
castor and catnip oil was placed at the end of the log. A small piece of American beaver meat
(~5-cm diameter) was hung by wire directly above the end of the log (~60 cm) to serve as an
52

additional attractant during winter. Bait was consumed quickly by most carnivores, but its scent
continued to serve as an additional lure to encourage animals in the vicinity of the camera trap to
move into view. Cameras were checked and lure and bait added every two weeks in the winter
and every four weeks in the snow-free months. Only lure was added to sites during snow-free
months due to bear activity. In 2014–2016, cameras were set to take 30s of video with a 1s delay
between video-recordings. In 2020–2022, cameras were set to take 10s of video with a 1s delay
between video-recordings. These schedules allowed for nearly continuous recording of the time
an animal was in view.

Sampling and Ecological Variables
N-mixture Models
The study area had a forest inventory derived from high-density light detection and
ranging (LiDAR; 8–10 pulses/m2) data obtained in August and September 2015. Independent
covariates for N-mixture models were represented by three broad categories that included forest
cover and structure, disturbance, and prey abundance (Table 5). The number of days since lure
was added to a site by a researcher was included as a detection covariate. Variables representing
forest cover and structure from LiDAR data included canopy closure at 4 vertical layers
calculated at the stand polygon scale (ground (0–1m), low (1–3m), mid (3–10m), and top
(>10m)). I used GIS data to measure the distance of each camera site to riparian edge, proportion
of area (1000m radius) containing recent cutblocks (<20 years in age), and proportion of area
(1000m radius) containing potential hare habitat (forest 21–50 years in age). I included a prey
covariate that represented hare abundance. Relative hare abundance was calculated using N-

53

mixture models and camera data (24-hour independent visit). I calculated a hare abundance
estimate for each site and season.
Lynx Behaviour
I developed a number of ecological and temporal variables that were hypothesised to explain the
behaviour of lynx at each camera trap. I used LiDAR data to calculate measures of vegetation
structure at three vertical heights within a 50-m radius of a camera location: ground-low (0–3m),
mid (3–10m), and top (>10m; Table 5). Temporal covariates included time of day (day or night),
season ((mid-winter (23 January–15 March), late winter (16 March–30 April), early summer (23
June–30 July), late summer (01 August–15 September), fall (16 September–31 October)) and
hare abundance relative to an assumed peak and trough in the cycle (2015–2016 = high hare,
2020–2022 = low hare). I related behaviour to the abundance of lynx and hare at each site, as
calculated using N-mixture models and occurrence rates (see previous section). Other covariates
included the number of days since bait and lure were added to a site by researchers, the number
of days since a site was last visited by a lynx, and number of individuals in a group of lynx
(binary; 1 or >1 individual).

54

Table 5. Variables used in the development of models predicting abundance (N-mixture models)
or behaviours (cheek-rubbing, scent-marking, grouping) of Canada lynx (Lynx canadensis) using
camera traps in central British Columbia, Canada, 2014–16 and 2020–22.

Parameter
cover (1–3)*
cover (3–10)*
cover (10+)*
#hare*
#lynx
period
season
daily
lure days
lynx days
group size
new harvest age**
hare age**
riparian**

Description
Average canopy cover 0–3 m (LiDAR, stand polygon)
Average canopy cover 3–10 m (LiDAR, stand polygon)
Average canopy cover >10m (LiDAR, stand polygon)
Site abundance estimates of snowshoe hare (N-mixture models)
Site abundance estimates of lynx (N-mixture models)
Hare abundance period (2015–2016 = high, 2020–2022 = low)
Mid-winter, late winter, early summer, late summer, fall
Behavior occurred at night or day
Number of days since bait and lured added to a site
Number of days since last visited by a lynx
1 individual or >1 individual
Proportion of 1000-m buffer with forest age >20 years old
Proportion of 1000-m buffer with forest age 21–50 years old
Distance to riparian edge

Variable
type
Continuous
Continuous
Continuous
Continuous
Continuous
Categorical
Categorical
Categorical
Continuous
Continuous
Categorical
Continuous
Continuous
Continuous

*N-mixture and behaviour models, **N-mixture models only

Data Analysis
Population Estimates and Indices
I used the “unmarked” package (version 1. 4. 1) in R (version 4. 3. 2) to fit N-mixture models
(pcount) to estimate abundance of Canada lynx and snowshoe hare at camera trap sites in six
winters (Fiske and Chandler 2011). To minimize double counting at the same site and detection
of the same individual at multiple sites within a single occasion, I used an 8-hr image interval to
define an independent visit to avoid double counting of individuals. I used time periods that did
not overlap with major changes in ecological conditions to lessen potential violations of
55

population closure. For example, I did not include data beyond mid-April for N-mixture models
to avoid spring break up and dispersing juveniles. Although there is potential for individual- and
sex-biased detection, I was not able to test this assumption with my sample size of collared
individuals. I used a 2-step model fitting process where I first fit detection models with and
without a single covariate of lure days and no independent variables for abundance. This
covariate represented the number of days since lure and bait were added to a site by researchers.
I then selected the best fitting detection model (with or without lure variable) to use with the
abundance model. Each model contained a single covariate for predicting abundance resulting in
7 models: cover at three vertical strata, proportion young forest, proportion older regenerating
forest, distance to riparian, and site-specific hare abundance.
I fitted the 8-model set (including a null model) for each winter season and year (2014,
2015, 2016, 2020, 2021, 2022). I used linear regression to estimate correlation (r) and compare
trends in abundance estimated by N-mixture models with relative abundance indices of camera
occurrence rates (number of camera occurrences/number of operational camera days). This
comparison was conducted temporally over the 6-year period (average camera occurrences per
day/operational camera days) and spatially among camera sites in each year (camera occurrences
per site/operational camera days).
I used spatial capture–mark–resight (SCMR; “secr” package (version 4. 6. 4) in R
(version 4. 3. 2)) to estimate Canada lynx density within the study area during the winter and
hare low period (2020–2002). That analysis allowed me to investigate the reliability and
precision of estimates with unmarked animals (i.e., N-mixture, indices of occurrence rates). This
method uses a combination of detection data of marked individuals at survey sites and counts of
unmarked individuals to estimate activity centers and model detection probabilities based on the
56

distance of survey sites from these activity centers (Efford et al. 2009, Efford and Hunter 2017).
I parameterised the SCMR model with images of individual lynx that were identified with GPS
(Global Positioning Systems) collars (Lotek LiteTrack 250) and uniquely colored ear tags at
camera traps from 2020–2022. In 2021, I had insufficient sample size of marked individuals to
conduct the SCMR analysis. Capture and handling protocols can be found in in Chapter 2.
Research protocols for lynx were approved by the UNBC Animal Care and Use Committee
(Protocol # 2020-3) and British Columbia Government (BC Wildlife Act, Permit PG19-596814).
Observations of lynx for the SCMR models included capture data when collaring
individual animals, capture histories of marked individuals at camera sites, and counts of
unmarked or unknown lynx generated with camera data. For SCMR models in 2020, there were
4, 1-week marking occasions and 4, 1-week sighting occasions. In 2022, there were 3, 1-week
marking occasions and 5, 1-week sighting occasions. I tested three SCMR models for lynx
density: 1) model with no independent covariates (null), 2) model with detection probability
varying between occasion type (marking or sighting), and 3) model with a habitat variable for
forest age. I developed a relatively simple index of lynx habitat that included non-habitat (water
bodies > 50 hectares) and categorical forest age (young = <20 yrs., mid = 21–50 yrs., and old =
>51 yrs.). The sampling area was buffered by 6 and 8 km in 2020 and 2022, respectively. That
buffered area, representing the range of collared lynx within and outside the study area,
prevented the use of LiDAR data. All models were adjusted for overdispersion of count data. I
compared density estimates generated from SCMR and N-mixture models. I used confidence
intervals and relative standard error (RSE) to assess the precision of estimates.

57

Lynx Behaviour
I used an Information Theoretic Model Comparison (ITMC) approach to develop a set of 9
logistic regression models that represented hypotheses explaining the display of lynx behaviour
(Burnham and Anderson 2004; Table 6). Camera images revealed the presence (1) or absence (0)
of a behavior at each site. I developed and compared the models independently for each
behavioural type: cheek-rubbing, scent-marking, and grouping. The group size covariate was not
included in any models used to explain lynx grouping behaviour. I used Akaike’s information
criterion for small sample sizes (AICc) to identify the most parsimonious model (Burnham and
Anderson 2004). I used both ΔAICc and Akaike weights (AICcw) to rank and compare models.

Table 6. A priori candidate models (logistic regression) representing three types of behaviour
(cheek-rubbing, scent-marking, grouping) of Canada lynx at remote camera sites in central
British Columbia, Canada, 2015–2016 and 2020–2022.

Model Name
Canopy cover
Hare abundance
Activity + scent
Lynx abundance
Temporal
Combined 1
Combined 2
Combined 3
Null

Model
cover (0–3 m) + cover (3–10 m) + cover (10+ m)
#hare
daily + lure days + lynx days
#lynx + group size
period + season
hare + period + season + lure days
season + group size + lure days + cover (0–3 m)
period + season + group size + lure days + cover (0–3 m) + #lynx
no independent covariates

58

K
4
2
4
3
6
8
8
10
1

Receiver operating characteristics and resulting area under the curve (AUC) were used to
assess the predictive ability of the best model (Pearce and Ferrier 2000). I used a onefold cross
validation routine to withhold each record sequentially from the model building process and then
calculated the independent probability of that withheld record being the presence of a behaviour.
I considered a model with an AUC score of 0.7 to 0.9 to be a useful application and a model with
a score >0.9 as highly accurate (Boyce et al. 2002). I used 95% confidence intervals to assess the
strength of effect of each predictor covariate on the dependent variable. Poor power and
inconclusive statistical inference was expected from covariates with confidence intervals that
approach or overlap 0. I used tolerance scores to assess variables within each model for
excessive collinearity (Menard 2001). All data analyses were performed using Stata (version
17.0; Statacorp, College Station, Texas).
I compared trends in the proportion of each behaviour in each year to both camera
occurrence rates and N-mixture abundance estimates using standard linear regression and
correlation (r). Lastly, I tested for correlation between lynx and hare abundance estimates
generated with N-mixture models over the 6-year period and spatially within camera sites in each
year.

Results
Camera Traps
In 2015–2016, cameras functioned properly an average of 96.0% (SE = 0.9; 8,866 camera-days)
and 90.3% (SE = 2.4; 5,008 camera-days) of potential camera days in the winter and snow-free
periods, respectively. In 2020–2022, cameras functioned properly an average of 95.1% of
potential camera days in the winter (SE = 0.8; 13,181 camera-days) and 91.1% (SE = 1.4; 10,101
59

camera-days) of potential camera days in the snow-free periods. There was a 76.2% and 65.9%
decrease in lynx occurrence rates between 2015–2016 and 2020–2021 during the winter and
snow-free seasons, respectively. There was a similar steep decrease in snowshoe hares (72.7% =
winter, 79.0% = snow-free).

Population Estimates and Indices
The top N-mixture models varied between years but included models with canopy cover 3–10 m
(2014 and 2015), canopy cover 1–3 m (2021 and 2022), and proportion of forest 21–50 years old
(2016; Appendix 2.1). For all top models, the single covariate positively influenced the
abundance of lynx. The detection model including the number of days since lure and bait were
added to a site ranked higher than the null model only in 2016. Estimates of density ranged from
15 lynx/100km2 (CI = 11–21) in 2016 to 7 lynx/100km2 (CI = 4–12) in 2021 (Table 7; Figure 7).
Corresponding camera occurrence rates were 11.1 lynx/100 camera days in 2016 and 1.9
lynx/100 camera days in 2021. N-mixture estimates and camera hits for both lynx and hare were
strongly correlated spatially by site and temporally over the 6-year period of study. The
population trend for lynx showed an increasing population from 2014–2016 with a decrease four
years later from 2020–2021. Abundance estimates were still low in 2022 but with some evidence
of a slight increase. Although the overall trend was similar for snowshoe hare compared to lynx,
there was a decrease in hare abundance rather than an increase from 2015 to 2016. Relative
standard errors ranged from 17% in 2016 to 37% in 2021 (Table 7).

60

Figure 7. Comparison in trends of relative abundance indices (RAI; number of captures/number
of camera days x 100) and N-mixture abundance estimates with 95% confidence intervals for
Canada lynx (Lynx canadensis) using camera traps in central British Columbia, Canada, 2014–
2016 and 2020–2022. Spatial capture-mark-resight (SCMR) estimates for 2020 and 2022 provide
an additional measure for validating estimates of relative abundance from camera images and
estimates of abundance from N-mixture models.

61

Table 7. Density estimates from spatial capture-mark-resight (SCMR) and N-mixture models for
Canada lynx (Lynx canadensis) and snowshoe hare (Lepus americanus) using camera traps in
central British Columbia, Canada, 2015–16 and 2020–2022. RSE = relative standard error.
Estimate method
Year Species
SCMR
2020
Lynx
SCMR
2022
Lynx
N-mixture
2014
Lynx
N-mixture
2015
Lynx
N-mixture
2016
Lynx
N-mixture
2020
Lynx
N-mixture
2021
Lynx
N-mixture
2022
Lynx
N-mixture
2014
Hare
N-mixture
2015
Hare
N-mixture
2016
Hare
N-mixture
2020
Hare
N-mixture
2021
Hare
N-mixture
2022
Hare
2
* = lynx/100km , hare/hectare

Density*
2.64
2.89
9.16
12.72
14.80
11.31
6.63
7.92
0.23
0.24
0.19
0.15
0.14
0.18

SE
0.71
0.71
2.38
2.30
2.52
3.28
2.45
2.14
0.07
0.04
0.04
0.04
0.05
0.04

LCI
1.57
1.80
4.50
8.21
9.86
4.88
1.83
3.73
0.09
0.16
0.12
0.08
0.04
0.10

UCI
4.44
4.63
13.82
17.23
19.74
17.74
11.43
12.11
0.37
0.32
0.26
0.22
0.24
0.26

RSE
26.9%
24.6%
26.0%
18.1%
17.0%
29.0%
37.0%
27.0%
30.4%
16.7%
20.0%
23.3%
35.7%
21.7%

There were 35 resights of seven GPS-collared lynx in 2020, and 41 resights of eight
collared lynx in 2022. Collared lynx represented 49% and 45% of all lynx occurrences at camera
traps in 2020 and 2022, respectively. Population density estimates using SCMR models were
similar between 2020 (3 lynx/100km2 [CI = 2–4]) and 2022 (3 lynx/100km2 [2–5]) with
reasonable precision (RSE of estimate = 26.9% and 24.6%, respectively). The top model for both
years allowed detection probability to vary between marking and resighting occasions. SCMR
abundance estimates were less than N-mixture models and had non-overlapping confidence
intervals.

62

Behaviour Models
The top-ranked regression model explaining cheek-rubbing behaviour included covariates
representing temporal factors, lynx abundance, vegetation cover, and the diminishing effect of
bait and lure (model = Combined 3; AICcw = 1.00; Table 8). Those covariates were effective at
predicting cheek-rubbing (AUC = 0.75, SE = 0.01). Rubbing behaviour was positively associated
with a period of hare abundance (2015–2016), mid-winter and early summer seasons, groups of
lynx, more ground cover, and lynx abundance at camera sites (Figure 8). Covariates with
negative associations included the fall and late winter seasons as well as the number of days
since bait and lure were added to a site by a researcher (i.e., rubbing behaviour increased closer
to lure dates).
The highest ranked model explaining scent-marking by lynx included covariates for hare
abundance, lure days, period, and season (model = Combined 1; AICcw = 0.998; Table 8). Scentmarking was positively associated with hare abundance both temporally (2015–2016 period) and
spatially at camera trap sites, and negatively associated with the mid-winter season (Figure 8).
However, that model had poor predictive accuracy (AUC = 0.61, SE = 0.01).
Two of the combined models for lynx grouping behaviour were considered equivalent
(ΔAICc < 2; >95% of AICcw). For both models, lynx grouping was positively associated with a
period of hare abundance (2015–2016) and the mid-winter season, and negatively associated
with the early summer season and increasing days since lure was added to a site (Figure 8). Both
models had reasonable predictive accuracy (AUC = 0.68–0.69, SE = 0.02).
There was a strong, positive correlation (r) between abundance estimates for lynx during
winter and each type of behaviour (Table 9; Figure 9), but that relationship was especially
63

pronounced for cheek-rubbing behaviour (r = 0.96). Although there was a strong positive
correlation between lynx and hare abundance annually from 2014–2022 (r = 0.95), there was no
or very weak correlation between site-specific hare and lynx abundance estimates within each
year (r mean = 0.23, SE = 0.04). Coefficients

64

Table 8. AICc scores, and AICc weights (w) for logistic regression models predicting Canada
lynx (Lynx canadensis) behaviours at camera traps in central British Columbia, Canada, 2014–
2016 and 2020–2022. Area under the curve (AUC) and standard error (SE) for the receiver
operating characteristic represents the predictive accuracy of each model (top 95%).
Model

Cheek-rubbing
Combined 3
Combined 1
Temporal
Combined 2
Lynx abundance
Activity + scent
Canopy cover
Hare abundance
Null
Scent-marking
Combined 1
Combined 3
Temporal
Hare abundance
Lynx abundance
Activity + scent
Combined 2
Canopy cover
Null
Lynx grouping
Combined 1
Combined 3
Temporal
Combined 2
Activity + scent
Canopy cover
Hare abundance
Lynx abundance
Null

K

AICc

ΔAICc

w

AUC (SE)

10
8
6
8
3
4
4
2
4

1910.6
1981.9
2011.9
2052.5
2089.6
2200.7
2229.4
2240.7
2263.5

0.0
71.3
101.3
141.9
179.0
290.1
318.8
330.1
352.9

1.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00

0.75 (0.01)

8
10
6
2
3
4
8
4
4

2090.2
2104.0
2105.3
2119.6
2134.6
2140.3
2144.8
2152.1
2155.7

0.0
13.8
15.1
29.4
44.4
50.1
54.6
61.9
65.5

1.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00

0.61 (0.01)

8
10
6
8
4
4
2
3
4

763.3
764.0
768.4
799.3
809.8
815.7
818.2
819.1
822.3

0.0
0.7
5.1
36.0
46.5
52.4
54.9
55.8
59.0

0.55
0.40
0.04
0.00
0.00
0.00
0.00
0.00
0.00

0.68 (0.02)
0.69 (0.02)

65

Table 9. Spatial and temporal correlations (r) between occurrence rates (OR) generated with
camera images, N-mixture (N-mix) abundance estimates, and lynx behaviours for Canada lynx
(Lynx canadensis) and snowshoe hare (Lepus americanus) using camera traps in central British
Columbia, Canada, 2014–16 and 2020–22.

Comparison
Species
Year
Correlation (r)
Spatial (camera trap sites, n = 198 (hare low), n = 173 (hare high)
OR and N-mix
lynx
2014–2016
0.95
OR and N-mix
lynx
2020–2022
0.88
Lynx and Hare (N-mix)
lynx/hare
2014–2016
0.23
Lynx and Hare (N-mix)
lynx/hare
2020–2022
0.21
Lynx and Hare (OR)
lynx/hare
2014–2016
0.21
Lynx and Hare (OR)
lynx/hare
2020–2022
0.19
Temporal (Annual winter, n = 6)
OR and N-mix
lynx
2014–2022
0.91
OR and N-mix
hare
2014–2022
0.77
Lynx and Hare (N-mix)
lynx/hare
2014–2022
0.95
Lynx and Hare (OR)
lynx/hare
2014–2022
0.69
Cheek-rubbing and N-mix
lynx
2014–2022
0.96
Cheek-rubbing and OR
lynx
2014–2022
0.91
Scent-marking and N-mix
lynx
2014–2022
0.94
Scent-marking and OR
lynx
2014–2022
0.83
Lynx group and N-mix
lynx
2014–2022
0.92
Lynx group and OR
lynx
2014–2022
0.87

66

Figure 8. Coefficients for top-ranked models illustrating the presence of behaviours (cheekrubbing, scent-marking, and grouping) of Canada lynx (Lynx canadensis) at camera trap sites
during contrasting periods of prey abundance in central British Columbia, Canada, 2015–2016
(high) and 2020–2022 (low). MW = mid-winter, LW = late winter, ES = early summer, LS = late
summer.
67

Figure 9. Canada lynx (Lynx canadensis) abundance and presence of behaviours (percentage of
visits with each behaviour recorded) in each year (cheek-rubbing, scent-marking, and grouping)
at camera trap sites during contrasting periods of prey abundance in central British Columbia,
Canada, 2014–2016 (increasing - high) and 2020–2022 (decreasing - low).

68

Discussion
Few studies have quantified individual- or population-level behaviours to changes in population
abundance (O’Donoghue et al. 1998a, Green et al. 2019, Higashide et al. 2021). This is one of
the first studies to combine co-occurring indicators of behaviour and abundance to measure
cyclic trends. Lynx behaviours, occurrence rates, and relative abundance estimates followed
similar patterns among years and between hare abundance periods. Consistent with my
predictions, years with greater lynx and hare abundance were characterized by increases in
cheek-rubbing, scent-marking, and grouping behaviours. The combination of abundance and
behaviours indicative of body condition and reproduction provided insights into the ecological
drivers of population change. My results provide support for the use of camera traps to measure
population and behavioural trends for Canada lynx. Similar relationships may hold for other
felids that are known to demonstrate repeated scent-making behaviours (e.g., Allen et al. 2016a,
b).
All behaviours by lynx were positively associated with the period when snowshoe hares
were most abundant in my study area (i.e., 2015–2016). Also, scent-marking varied spatially,
increasing at camera sites with greater hare abundance. I developed an accurate model
representing cheek-rubbing by lynx, with a positive relationship between that behaviour and sitespecific estimates of lynx abundance, lynx groups, and increased ground cover. The proportion
of lynx occurrences with cheek-rubbing differed among years (8–59%) but was strongly
associated with years of greater hare abundance. The combination of fewer lynx and lower
propensity for cheek-rubbing during the low in hare abundance have implications for monitoring
or research reliant on sampling of DNA contained in hair. Although hair sampling has been used
with success during a winter with high lynx densities (Doran-Myers et al. 2021), I found it to be
69

unreliable during a low in the lynx-hare cycle. Variable cheek-rubbing among years can result in
inconsistent DNA collection and imprecise population estimates that may mask inter-year or
cyclical variation in lynx abundance. For example, I used DNA from rub-stations and spatially
explicit capture-recapture (SECR) models that were not included in this study to estimate lynx
abundance during three years of the population low, but few recaptures and the non-detection of
multiple collared animals resulted in imprecise estimates (RSE = 60–120%).
Previous studies have documented increased cheek-rubbing during late winter that
coincided with lynx breeding (Crowley et al. 2013, Crowley et al. 2017). I found a similar
increase in cheek-rubbing during the breeding season, but the presence of cheek-rubbing was
also positively associated with early summer. Canada lynx typically give birth in May and June
and the early summer coincides with rearing of lynx kittens at den sites (Anderson and Lovallo
2003). However, if an increase in cheek-rubbing were due to kitten rearing and associated
territoriality, I would expect it to occur only in years with lynx litters and this was not the case.
This same pattern of behaviour occurred in years with no apparent lynx reproduction in my study
area. From February 2020–May 2022, no lynx kittens were observed on my camera traps.
The application of catnip to scent poles at camera sites may partially explain increased
cheek-rubbing in early summer (Uenoyama et al. 2021). Most felid species have a positive
neurophysiological response to catnip (nepetalactone; Todd 1962, Hill et al. 1976, Tucker and
Tucker 1987), and its chemical compounds have been found to repel mosquitoes following
rubbing behaviour in domestic cats (Uenoyama et al. 2021). The early summer period coincides
with the most intense insect activity of the year. Further study is required to determine if
chemical defense and insect avoidance are a potential explanatory mechanism of cheek-rubbing
in response to catnip during this time of year.
70

Kenney et al. (2024) found that camera occurrence rates were strongly correlated with
density estimates or other population indices for several boreal forest mammals. For Canada
lynx, indices of relative abundance based on camera trap and snow track surveys produced
similar population trends (Kenney et al. 2024). I found a strong positive correlation between
occurrence rates and N-mixture abundance estimates. In agreement with Kenney et al. (2024),
my results provide additional support for the use of camera traps to monitor trends in lynx and
hare populations. I urge caution, however, in using N-mixture models to estimate absolute
abundance. Although N-mixture estimates and occurrence rates followed similar temporal and
spatial patterns of abundance (see Figure 7; Table 7), confidence intervals for N-mixture
estimates were wide and overlapping between years. Additionally, in the two years I had markresight estimates for comparison, N-mixture estimates were 3–4 times greater than estimates
generated with the SCMR model. The 3 lynx/100 km2 density estimate produced by SCMR
methods was ecologically plausible, matching estimates of lynx density during hare lows in the
boreal forest (Brand et al. 1976, Poole 1994, Slough and Mowat 1996, O’Donoghue et al. 1997).
SCMR models were parameterised using a large number of resighted individuals (35 resights of
7 lynx in 2020; 41 resights of 8 lynx in 2022) with collared lynx representing 49% and 45% of
all lynx occurrences at camera traps in 2020 and 2022, respectively. For these reasons, I consider
SCMR estimates to be reliable and reasonably precise. Although I adopted a conservative
measure of independent camera observations (i.e., 1-day counts with 8-hr intervals to define an
independent visit), N-mixture models were likely sensitive to this model violation, possibly
biasing the results (Barker et al. 2018, Link et al. 2018). Although N-mixture models may not be
suitable for estimates of absolute abundance they proved useful for measuring relative abundance
and population trends in my study. Strong positive correlations between annual abundance

71

estimates and behavioral cues (r = 0.92–0.96) further support their use as a measure of
population trends. However, occurrence rates were also strongly correlated with both N-mixture
estimates and behaviours providing a reliable, yet simpler, method for monitoring population
trends.
Abundance of hare and lynx were strongly correlated between years, but not spatially at
camera traps within each year. This density-habitat relationship may explain why the abundance
of lynx and hare tracked temporally but not spatially among camera sites that varied considerably
in vegetation cover. Several studies have suggested that lynx may not be able to hunt hare
sufficiently in dense forest (Mowat et al. 2000, Fuller et al. 2007) and that habitat with dense
understory may provide a refuge for hares during cyclic lows (Wolff 1980, O’Donoghue et al.
1998b). For this reason, lynx may be more numerous at sites with intermediate cover (Mowat et
al. 2000, Fuller et al. 2007).
I found that time since bait and lure were added to a site was not a significant influence
on the detection probability of lynx in N-mixture models. This suggests that lure and bait had
little influence on the number of images recorded by cameras and estimates of abundance for
lynx in my study area. Once a lynx was at a site, however, lure positively influenced cheekrubbing and scent-marking behaviours. Bait and lure primarily encouraged animals already in the
vicinity of the camera trap to engage in cheek-rubbing and scent-marking. Contrary to my
predictions and in contrast to pumas (Puma concolor; Allen et al. 2016b), time since a previous
visit by a conspecific did not influence these behaviours.
Differences in behaviour between the low and high in hare abundance may be explained
by a combination of lynx density and reproduction. During highs in hare abundance, lynx
movements decrease as individual animals spend less time and effort searching for prey (Nellis
72

and Keith 1968, Brand et al. 1976, Wards and Krebs 1985). Less energy devoted to hunting may
provide more energy for reproduction and associated behaviours such as cheek-rubbing, scentmarking, and grouping. Breeding pairs of male and female lynx were commonly recorded on
camera traps displaying these behaviours during the hare high. Additionally, greater lynx
densities may increase breeding competition and further influence the frequency of these
behaviours that can be related to intraspecific interactions. Conversely, during lows in hare
abundance, lynx increase movements but consume less biomass of prey (Brand et al. 1976).
Lynx in my study area did not have kittens or form breeding pairs, and engaged less frequently in
cheek-rubbing and scent-marking behaviours during the apparent low in the hare cycle.
Other behaviours besides those associated with reproduction may change throughout the
lynx-hare cycle. O’Donoghue et al. (1998a) identified changes in lynx behaviour that included
differences in hunting methods, kill rates, and prey caching. Their study relied on backtracking
that can be labor intensive, dependent on snow conditions, and typically limited to smaller areas.
Camera traps, especially video footage, provided unique insight into behaviours over a large area
that has not been available with other survey methods. Variation in behaviour and indices or
estimates of abundance proved to be strong co-occurring indicators of the cyclic population
dynamics of Canada lynx. Behaviour is an indicator of reproduction, demographics, and survival
that both influence and is influenced by population density; thus, providing insights into the why
for changes in occurrence rates at camera traps. In addition, changes in some behaviours may be
indicators of future trends in population abundance. For example, increases in cheek-rubbing
related to improvements in ecological conditions (i.e., prey abundance or body condition) could
signal changes in breeding activity that may not affect population abundance for another year. A
smaller number of sites could be used to monitor behaviours and inform the timing of more
73

intensive monitoring efforts required for estimating abundance. Accurate assessment of trends
over time will ultimately improve management and conservation efforts for Canada lynx and
other species where behaviour and occurrence can be monitored simultaneously.

74

Chapter 4: Short-term fluctuations in prey and its influence on a carnivore
community: Habitat co-occurrence of Canada lynx and sympatric
mesopredators increases following cyclical reduction in primary prey
Abstract
In multi-predator systems, niche expansion may increase habitat overlap among species and the
potential for competitive or predatory interactions. The strength of those interactions is likely
mediated by prey availability. Although interspecific competition may drive habitat
specialization and niche partitioning, that differentiation may not evolve when prey populations
cycle in large amplitudes and over short durations. In the northern boreal and subboreal forests of
Alaska and Canada, snowshoe hare populations cycle over an 8–11-year period and that variation
can strongly influence the distribution and population dynamics of carnivores. I used camera
traps to investigate the influence of sympatric carnivores (coyote, fisher, wolverine) and prey
(snowshoe hare, red squirrel) on the habitat use and co-occurrence of Canada lynx during two
contrasting periods of hare abundance. Given optimal foraging theory and the relatively short
interval between changes in prey abundance, I predicted that habitat overlap and co-occurrence
would increase during a low in hare abundance when Canada lynx hunt alternative prey in varied
habitats. I found that lynx occurrences mirrored the decrease in hares, while the number of
sympatric carnivore species as well as the combined occurrences of these species increased
during the low in hare abundance. The co-occurrence of lynx with other sympatric carnivores
increased at a time of prey scarcity. Predator populations in subboreal forests may be in a
dynamic state of habitat overlap dependent on cyclic prey abundance. Canada lynx may lose

75

their competitive advantage during hare lows and be especially vulnerable to cumulative forest
change and future climate conditions that are beneficial to many of their sympatric competitors.

Introduction
When resources are limiting, optimal foraging theory states that a species expands the breadth of
habitat and diet (Emlen 1968, MacArthur and Pianka 1966). ,In a multi-predator system, niche
expansion by each individual species could then increase habitat overlap among species and the
potential for competitive or predatory interactions. During long-term resource limitation,
interspecific competition may reduce niche overlap through habitat specialization and niche
partitioning (Hardin 1960, Rosenzwieg 1981). Short-term resource fluctuations, however, are
more likely to result in an increase in the breadth a species’ niche rather than specialization as
the latter is often the result of longer-term processes such as competitive exclusion, adaptation,
and natural selection (Hardin 1960). Wiens (1977) suggested that co-existing species may follow
a high-low-high sequence of niche overlap in response to fluctuating resources; responding
opportunistically to the same abundant resources but with lessened competition, specializing as
resources become more limiting, and increasing their niche overlap and competitive interactions
as resources become scarce. In systems where prey populations cycle in large amplitudes and
over short durations, there may be insufficient time for the development of behaviours or
resource use that facilitates niche partitioning. In this scenario, populations would be in a
dynamic state of niche overlap dependent on prey-predator densities at a particular stage in a
cycle.

76

In the northern boreal forests of Alaska and Canada, snowshoe hare (Lepus americanus)
populations cycle over an 8–11-year period that results in contrasting years of prey abundance
for the carnivore community (high to high or low to low; (Krebs et al. 2014, Krebs et al. 2017,
O’Donoghue et al. 1997, Slough and Mowat 1996). Canada lynx (Lynx canadensis; hereafter
lynx), in particular, cycle asynchronously with snowshoe hare and are considered a specialist on
this prey species (Krebs et al. 2014, Krebs et al. 2017, O’Donoghue et al. 1997). Red squirrels
(Tamiascurus hudsonicus) are often the second most dominant prey item in lynx diet, typically
contributing more during lows in snowshoe hare abundance (Aubry et al. 2000, Koehler 1990,
O’Donoghue et al. 1998b). The dominance of snowshoe hare in lynx diet is especially evident
during the winter with secondary prey becoming more common during the snow-free seasons
(Mowat et al. 2000, Parker et al. 1983, Saunders 1963).
Coyote (Canis latrans) and great-horned owl (Bubo virginianus) can also demonstrate
asynchronous cycling dynamics in response to changes in the abundance of hare in the boreal
forest (O’Donoghue et al. 2001, Rohner et al. 2001, Tyson et al. 2010). Although less
pronounced, fisher (Pekania pennanti) populations in Ontario have shown a delayed and positive
numerical response to increases in hare abundance (Bowman et al. 2006). Unlike lynx, coyotes
and fishers are generalists that may be better able to adapt to low hare numbers by supplementing
their diet with other prey. Although large fluctuations in the abundance of hare affect different
predators to varying degrees, the hare cycle results in contrasting periods of prey biomass that
likely has strong influences on community dynamics.
Generalist predators are thought to have a stabilizing effect on predator-prey cycles. The
influence of generalist predators on lynx populations, especially during cyclic lows in hare
numbers, is unclear. Previous studies have focused on potential coyote and lynx interactions by
77

measuring numerical, functional, and behavioral responses to changes in hare abundance of each
species individually (Murray et al. 1994, O’Donoghue et al. 1997, O’Donoghue et al. 1998a, b).
Coyotes and fishers can be both exploitative and interference competitors as well as occasionally
preying on lynx (McClellan et al. 2018, O’Donoghue et al. 1995, O’Donoghue et al. 1998b). In
Maine, predation was the leading cause of mortality for Canada lynx (28%) with 77% of
predation events (14 of 18) over a 12-year period caused by fishers (McLellan et al. 2018).
Although wolverine (Gulo gulo) typically occupy a different ecological niche than lynx, their
diets overlap, wolverine may be the dominant species in interference competition (Jung et al.
2023), and snowshoe hare can be a high proportion of wolverine diet (Banci 1987, Robitaille et
al. Unpubl. Data ). Lynx in central British Columbia (BC) are sympatric with coyotes, fisher, and
wolverine and no studies have investigated the habitat overlap and co-occurrence of lynx with
this assemblage of other predators.
Habitat loss and fragmentation can have a significant influence on species distribution,
but these changes can also affect interspecific interactions, making it difficult to disentangle the
influence of habitat and interspecific interactions on species distribution (Estevo et al. 2017,
Fisher et al. 2012). Approximately 18 million hectares of BC’s forests have been affected by a
mountain pine beetle (Dendroctonus ponderosae; MPB) epidemic that began in the mid-1990s
(BC FLNRO, 2012). From 2001–2017, the Allowable Annual Cut (AAC) was increased to
salvage dead and dying lodgepole pine (Pinus contorta) trees. This resulted in fundamental
changes to the age distribution of forests across large portions of central BC with implications
for wildlife habitat (FPB, 2009). It is currently unknown how large-scale changes in habitat loss
and fragmentation will affect the cyclic dynamics of snowshoe hares and their predators.
Historically, most studies of species distribution have focused on the relationship between
78

individual species and habitat covariates one at a time, partly due to the difficulties of accurately
measuring and implementing multi-species studies. The development of more inexpensive and
reliable trail cameras for multi-species detection have opened the possibility of investigating a
fuller range of community interactions.
I investigated the influence of sympatric mesopredators (coyote, fisher, wolverine) and
prey (snowshoe hare, red squirrel) on the habitat use and co-occurrence of lynx during two
contrasting periods of hare abundance. Natural cycles in hare populations provide an opportunity
to investigate the influence of short-term food abundance and scarcity on the habitat overlap and
co-occurrence patterns of sympatric predators. Although not sufficient to assess the full range of
competitive interactions (spatial and temporal interactions, diet; Murray et al. 2023), differences
in habitat co-occurrence across temporal variation in resource availability can provide important
insights into potential community interactions. Habitat overlap may be an indicator of potential
competition between species (Murray et al. 2023). However, my intention was not to directly
assess competition but to measure habitat overlap during two contrasting periods of prey
abundance. Given optimal foraging theory and the relatively short interval between changes in
prey abundance, I predicted that habitat overlap and co-occurrence would increase during a low
in hare abundance when Canada lynx hunt alternative prey in varied habitats.

Materials and Methods
Study Area
The research was conducted in and adjacent to the John Prince Research Forest (JPRF) in northcentral BC, Canada, which is co-managed by the University of Northern British Columbia,
Binche Whut’en, and Tl’azt’en First Nations. The study area is ~390 km2 and characterized by
79

rolling terrain with low mountains (700 m to 1500 m above sea level). The region represents the
northern extent of contiguous Douglas-fir (Pseudotsuga menziesii var. glauca) forests in the
interior of British Columbia and is dominated by the Sub-Boreal Spruce biogeoclimatic zone
(Delong et al. 1993). The area has experienced a wide variety of logging activities over the past
75 years and contains a mosaic of old and young forest (continuum from new harvest to old
growth >250 years old) with interspersed deciduous stands. The JPRF harvested approximately
20,000 m3 on an annual basis during the study period. More intensive industrial forest harvest
occurred in the study area outside the Research Forest boundaries in the 15 years before this
study (2000-2015). Although salvage harvest continued, relatively fewer stands were actively
harvested in between the two data collection periods (e.g. 2015–2016 to 2020–2022). The study
area is within a broader region that is experiencing rapid and widespread harvest of dead and
dying lodge pole pine affected by a mountain pine beetle epidemic. The study area is part of a
long-term monitoring program investigating the influences of climate and landscape change
focused on multi-species detections across the JPRF as well as an adjacent area managed by a
major forest company. There has not been any significant trapping in the study area for almost
two decades.
American marten (Martes americana), Canada lynx, short-tailed weasel (Mustela
erminea), American mink (Neogale vison), river otter (Lontra canadensis), fisher, wolverine,
coyote, and red fox (Vulpes vulpes) are small- to medium-size carnivores present throughout the
study area. American beaver (Castor canadensis), muskrat (Ondatra zibethicus), snowshoe hare,
red squirrels, flying squirrels (Glaucomys sabrinus), ruffed grouse (Bonasa umbellus), deer mice
(Peromyscus maniculatus), and voles (Clethrionomys gapperi, Myodes spp. and Microtus spp.)
are some of the most common prey species throughout the study area.
80

Field Data Collection
I deployed sixty-six camera traps (Bushnell Trophy Trail Cameras models 119467 and 11947;
Bushnell Outdoor Products, Missouri, USA) on a hexagonal grid (2.5 km apart; 5.41 km 2) from
January–March in 2015 (23 Jan–03 Apr) and 2016 (01 Feb–10 Apr) and from April–October (22
Jun–19 Jul; 12 Aug–08 Sep; 22 Sep–19 Oct) in 2016. The total study area was 357 km2. In
2020–2021, trail cameras (Browning Dark Ops HD Pro Trail Cameras model BTC-6HDP;
Browning, Utah, USA) were set at the same sites as 2015–2016. Although the same camera sites
were monitored continuously from January 2020 to October 2021, I used time periods in which
the highest proportion of cameras were active, survey starts coincided with camera check and
lure days, and survey timing overlapped with dates in 2015–2016. Camera data used in both
2020 and 2021 included: 01 February–11 April and 24 June–14 October (24 Jun–21 Jul; 05
Aug–01 Sep; 16 Sep–13 Oct). The 2015–2016 and 2020–2021 data sets represented a time of
high and low hare abundance, respectively. Although the exact timing of a peak and trough in
hare populations was uncertain, my two data sets (2015–2016 and 2020–2021) represented two
time periods of contrasting hare abundance. Occurrence rates of snowshoe hares at camera traps
decreased 73% between the two winter periods. This trend was supported by a 40% decrease in
hare densities measured at pellet plots established 3 years after the estimated hare peak (2018–
2021; John Prince Research Forest, unpublished data). I only included sites in the analyses in
which the remote camera was active >50% of the survey period. A total of four cameras were
removed from the snow-free high period as well as one camera from the snow-free low period
due to camera failure.
Although random placements of some sites were in the vicinity of roads or trails, no
cameras were directly placed on these habitat features. At each site, a scent post was established
81

2.5–3 m directly in front of a camera trap between 0.5 and 1 m above the ground on a tree. This
set-up maintained a consistent height and angle of view to the scent post. Each scent post
consisted of a small diameter log (<15 cm) secured with one end above the ground (45 cm) and
pointing directly at the camera. A local commercial lure containing beaver castor and catnip oil
was placed at the end of the log. A small piece of beaver meat (~5 cm diameter) was hung by
wire directly above the end of the log (~60 cm) to serve as an additional attractant. Bait was
consumed quickly by most carnivores but continued to serve as an additional lure to encourage
animals in the vicinity of the camera trap to move into view. Cameras were checked and lure and
bait added every two weeks in the winter and every four weeks in the snow-free months. Only
lure was added to sites during snow-free months due to bear activity. In 2015–2016, cameras
were set to take 30s of video with a 1s delay between video-recordings. In 2020–2021, cameras
were set to take 10s of video with a 1s delay between video-recordings. These schedules allowed
for nearly continuous recording of the time an animal was in view.

Habitat Covariates
The study area had a forest inventory derived from high-density light detection and ranging
(LiDAR; 8-10 pulses/m2) data obtained in August and September 2015. Independent covariates
were represented by three broad categories that included forest cover and structure, disturbance,
and prey abundance (Table 10). Variables representing forest cover and structure from LiDAR
data included canopy closure at 3 vertical layers (bottom [0-3 m], mid [3-10 m], and top [>10
m]) and canopy height. Because LiDAR data were measured during the leaf-on season, I used a
coarse correction factor for the winter based on empirical data collected at sites using the digital
Canopeo Application (www.canopeo.com; Alamo Software Foundation) and a literature search
(Davison et al. 2020, Wasser et al. 2013). Canopy cover measurements were taken in 4-cardinal
82

directions at 3m from the camera location and >3m above ground and averaged during the leafon and leaf-off seasons. I applied a correction factor of -25% and -52% change to the leaf-off
season for sites with few (26–50%) and a large (51–100%) proportion of deciduous trees,
respectively.
I used field measurements to assess tree type (binary; conifer leading = >75% conf,
mixed/deciduous = <75% conf) and volume of coarse woody debris. Field data for tree type were
based on a count of trees (>12.5 cm dbh) within 4 plots (11.28 m at camera location and three
3.99 m plots at the end of 50m transects). Volume of coarse woody debris (>7.5cm dbh) was
measured along three 50m transects spaced equally and radiating out from the camera location. I
used GIS data to measure the distance of each camera site to riparian areas, forest edge, and road
as well as proportion of area (250m radius) containing recent cutblocks (<15 years in age), and
edge density. Prey covariates included hare abundance, red squirrel abundance, and grouse
abundance. I measured relative prey abundance using camera occurrences. I calculated the
occurrence rate per site (# of days with a prey occurrence/active camera survey days) for each
prey species and season.

83

Table 10. Variables used in the development of habitat and co-occurrence models (multinomial
logistic regression) for Canada lynx (Lynx canadensis), sympatric mesopredators, and prey at
remote camera sites in central British Columbia, Canada, 2015-16 and 2020-21.
Parameter
cover (0-3 m)
cover (3-10 m)
cover (10+ m)
canopy height
coarse woody debris
conf/decid
dist.riparian
dist.edge
prop. Forest age (<15 yrs)
edge density
dist.road
hare
grouse
red squirrel

Description
Average canopy cover 0-3m (LiDAR, 50-m radius)
Average canopy cover 3-10m (LiDAR, 50-m radius)
Average canopy cover >10m (LiDAR, 50-m radius)
Average canopy height (LiDAR, 50-m radius)
Volume coarse woody debris (field data; 3, 50-m transects)
Count of trees; conifer = >75% conf, mixed/decid = <75% conf
Distance to stream and lake edge (m)
Distance to forest stand age (m)
Proportion of area (250-m radius) with recent cutblocks (<15 yrs)
Edge density (250-m radius)
Distance to truck driveable road (m)
Occurrence rate per site for hare (# days detected/survey days)
Occurrence rate per site for grouse (# days detected/survey days)
Occurrence rate per site for squirrel (# days detected/survey days)

Variable type
Continuous
Continuous
Continuous
Continuous
Continuous
Categorical
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous
Continuous

Data Analysis
I used multinomial logistic regression to investigate habitat characteristics that influenced the cooccurrence of lynx, sympatric mesopredators (fisher, wolverine, coyote), and their prey
(snowshoe hare, red squirrel). I considered the application of occupancy-based co-occurrence
models (Mackenzie et al. 2004), but decided that approach was unsuitable given the survey effort
and timing (continuous monitoring throughout season of interest), camera spacing, and the use of
bait and lure that were designed to provide a high likelihood of detection (Mackenzie and Royle
2005, Tigner et al. 2015). Furthermore, home ranges of all mesopredators in this study were
large enough to encompass multiple camera stations violating the assumption of spatial

84

independence for occupancy models (Burton et al. 2015, MacKenzie et al. 2004, Parsons et al.
2019). By using a multinomial model approach, I was able to investigate habitat association and
differentiation of lynx relative to other sympatric mesopredators and their prey across all data
sets.
For the dependent variable in the camera models, I first summed all daily occurrences
(independent visit = 24 hour) at a site during each season and biological period (winter low,
winter high, snow-free low, snow-free high). I then calculated the occurrence rate (#occurrence
days/#operational camera days) for each site. To create a binary variable, I separated lynx
occurrences at each site into two categories (high (1) vs no/low use (0)) based on the 50th
percentile of the occurrence rate at a site. I used the occurrence rate and 50th percentile for all
model sets because of the large number of sites (82%) with lynx occurrences in 2015–2016, to
keep analyses consistent between years, and to account for variation in sampling effort between
seasons and sites. Occurrence rates of lynx were strongly correlated to site abundance estimates
from N-mixture models in my study area (r = 0.88–0.95; Crowley et al. in prep). In the boreal
forest of Canada, camera data have also been highly correlated with independent density
estimates of snowshoe hares, lynx, and coyotes (Villette et al. 2016, Kenney et al. 2024). By
using the 50th percentile, the binary response variable represented a coarse measure of relative
abundance (no/low use vs moderate/high use) rather than presence or absence. This approach
provided an ecologically plausible response variable likely more representative of common and
widespread species. Except for the winter high period, this delineation generally combined sites
with no or single lynx occurrences during each time period. For the snow-free high and winter
low season, a “0” represented an occurrence rate of <1.2 and <1.4 lynx occurrences/site/day,
respectively. For snow-free low, a “0” represented <0.6 lynx occurrences/day. Lastly, for winter
85

high, a “0” equated to < 2.9 occurrences/site/day. For each prey (hare, squirrel), the 50th
percentile was also divided into two categories (high (1) vs low use (0)) based on the 50th
percentile of the occurrence rate at a site. For all other sympatric carnivores, the 50th percentile
reflected a true binary outcome (e.g,, 0 = 0 fisher occurrences/site/day).
There were four possible outcomes in each model according to the occurrence of lynx
and a competitor or prey–lynx, fisher, wolverine, coyote, hare, red squirrel–at a camera site. For
example, in models investigating habitat association between lynx and fisher, the base outcome
(0) represented sites with low use by lynx and no fisher. The contrast outcomes included sites
with high use by lynx with no fisher (1), sites with low use by lynx and fisher was present (2),
and sites with high use by lynx and fisher was present (3; Figure 10). I fit models for all pairs of
potential competitors of lynx (i.e., lynx-fisher, lynx-wolverine, and lynx-coyote) as well as the
two most common prey species of lynx (lynx-hare, lynx-red squirrel). All two-species models
(e.g., lynx-fisher) were fit for two seasons (winter and snow-free) and during two contrasting
periods of prey abundance within the hare cycle (hare low and hare high) for a total of four
potential model sets.

86

Figure 10. Example model development and outcomes (multinomial logistic regression)
investigating the habitat use and co-occurrence of Canada lynx (Lynx canadensis) and sympatric
mesopredators (coyote [Canis latrans], fisher [Pekania pennanti], and wolverine [Gulo gulo]) at
camera traps during two contrasting periods of prey abundance in central British Columbia,
Canada, 2015–2016 (high hare) and 2020–2021 (low hare).

87

I used an Information Theoretic Model Comparison (ITMC) approach to develop a set of
11 multinomial regression models to explain habitat associations and differentiations between
lynx and other sympatric mesopredators and prey (Burnham and Anderson 2004; Table 11). I
used Akaike’s information criterion for small sample sizes (AICc) to identify the most
parsimonious model explaining habitat association and differentiation (Burnham and Anderson
2004). I used both ΔAICc and Akaike weights (AICcw) to rank and compare models.

Table 11. A priori candidate models (multinomial logistic regression) representing habitat use
and co-occurrence of Canada lynx (Lynx canadensis), sympatric mesopredators, and prey at
remote camera sites in central British Columbia, Canada, 2015–2016 and 2020–2021.

Model Name
Canopy Cover
Riparian + Cover
Canopy Height + Riparian
Canopy Height + Ground Cover
Disturbance
Disturbance + Edge
Edge + Cover
Disturbance + Canopy Height
Prey- 3 Species
Prey- 2 species
null

Model
cover (0–3 m) + cover (3–10 m) + cover (10+ m) + conf/decid
dist. riparian + cover (3–10 m) + cover (10+ m)
canopy height + dist. riparian + conf/decid
canopy height + coarse woody debris + cover (0–3 m)
prop. forest age (<15 yrs) + edge density + dist. Road
prop. forest age (<15 yrs) + dist. Edge
dist. Edge + cover (0–3 m) + cover (3–10 m) + cover (10+ m)
canopy height + prop. forest age (<15 yrs)
hare + grouse + red squirrel
hare + red squirrel
no independent covariates

K
5
4
4
4
4
3
5
3
4
3
1

I used the receiver operating characteristics and resulting area under the curve (AUC) to
assess the predictive ability of the best model (Pearce and Ferrier 2000). I used a onefold cross
validation routine to withhold each record sequentially from the model building process and then
88

calculated the independent probability of that withheld record being a species occurrence. I
considered a model with an AUC score of 0.7 to 0.9 to be a useful application and a model with a
score >0.9 as highly accurate (Boyce et al. 2002). I used the AUC scores to compare the relative
predictive ability of the three outcomes for each 2-species model set.
I used 95% confidence intervals to assess the strength of effect of each predictor
covariate on the dependent variable. Poor power and inconclusive statistical inference is
expected from covariates with confidence intervals that approach or overlap 0. I used tolerance
scores to assess variables within each model for excessive collinearity (Menard 2001). All data
analyses were performed using Stata (version 17.0; Statacorp, College Station, Texas).

Results
Camera Trap Occurrences
In 2015–2016, cameras functioned properly an average of 96.0% (SE = 0.9; 8,866 camera-days)
and 90.3% (SE = 2.4; 5,008 camera-days) of potential camera days in the winter and snow-free
periods, respectively. In 2020–2021, cameras functioned properly an average of 96.8% of
potential camera days in the winter (SE = 0.7; 8,947 camera-days) and 91.1% (SE = 1.4; 10,101
camera-days) of potential camera days in the snow-free periods.
There was a 76.2% and 65.9% decrease in lynx occurrence rates between 2015–2016 and
2020–2021 during the winter and snow-free seasons, respectively (Table 12; Figure 11). There
was a similar steep decrease in snowshoe hares (72.7% = winter, 79.0% = snow-free). Both
fisher and wolverine occurrence rates increased from 2015–2016 to 2020–2021 during the winter
and snow-free periods. Coyote occurrences were similar between the hare abundance periods in
the winter, but decreased from 2015–2016 to 2020–2021 during the snow-free period. Red
89

squirrels increased from 2015–2016 to 2020–2021 during both seasons (winter = 42.3%, snowfree = 83.0%). In general, the probability of co-occurrence of lynx and other mesopredators
increased with more mid-strata and less top-strata cover (Figure 12). Lynx were strongly
associated with hares and these cover characteristics during all time periods. This relationship
was most pronounced in winter for wolverine-lynx and fisher-lynx during the hare low, and for
coyote-lynx during snow-free periods. For differences in AICc scores and weights, null models
ranked low (9 to 11 of 11 models) for all species and time periods (ΔAICc >14.6, AICcw =
<0.001).

Table 12. Total occurrences and occurrences per 100 camera-days (rate) for sympatric
mesopredators and prey at camera traps during two contrasting periods of prey abundance in
central British Columbia, Canada, 2015–2016 (high) and 2020–2021 (low).
2015–2016 (Hare High)

2020–2021 (Hare Low)

Species
Lynx
Wolverine
Fisher
Coyote

Winter
Total
Rate
732
8.26
15
0.17
2
0.02
26
0.29

Snow-free
Total
Rate
159
3.17
0
0
0
0
27
0.54

Winter
Total
Rate
176
1.97
95
1.06
52
0.58
22
0.25

Snow-free
Total
Rate
109
1.08
8
0.08
22
0.22
12
0.12

Hare
Squirrel

708
1804

432
1302

195
2591

183
4807

7.99
20.35

8.63
26.00

2.18
28.96

1.81
47.59

90

1.2

9
8

1

7
6
5

0.6

4

0.4

3
2

0.2

1

0

0

HIGH

LOW

Cyclic hare abundance – Winter (a)
0.6

10
9

0.5

8

7

0.4

6
0.3

5
4

0.2

Occurrence/100 days (lynx, hare)

Occurrence/100 days (fisher, wolverine, coyote)

0.8

3

2

0.1

1
0

0
HIGH

LOW

Cyclic hare abundance – Snow-free (b)
WOLVERINE

FISHER

COYOTE

LYNX

HARE

Figure 11. Occurrences per 100 camera-days for sympatric mesopredators and prey at camera
traps during two contrasting periods of prey abundance in central British Columbia, Canada,
2015–2016 (high hare) and 2020–2021 (low hare). Top = Winter (a); Bottom = Snow-free (b).
91

2015–2016 Winter Models
In general, the canopy cover models ranked above disturbance and prey abundance models for
the 2015–2016 winter season. There was a clear top model for all species except for fisher in the
winter (AICcw > 0.833; Table 13). I was unable to run models for fisher with only two total
occurrences (one in each year) during this survey period. The model including all three canopy
cover strata and percent conifer was the highest ranked during the 2015–2016 winter for all other
species except for snowshoe hare. For lynx-coyote models, lynx high use was not associated with
the occurrence of coyotes, but had outcomes with useful predictive accuracy at sites with low use
by lynx and no coyotes (AUC > 0.77). Lynx high use was not associated with the occurrence of
wolverine. At sites with lynx and hare, hares selected for sites with greater canopy cover (3–
10m), but differentiated from lynx by being positively associated with sites further from riparian
edge (Table 14).

2015–2016 Snow-free Models
For the snow-free period in 2015–2016, I could only run lynx-hare, lynx-squirrel, and lynxcoyote models because fisher and wolverine did not occur at any sites (Table 13). Lynx were not
strongly associated with the occurrence of red squirrels during this period. Two models for
coyote-lynx–disturbance and cover–accounted for 95% of the AICc weight (ΔAICc < 4.0). For
the disturbance model, only sites with coyotes (low lynx use) had useful predictive accuracy.
Coyotes were associated with further distances from truck roads, lower edge density, and higher
amounts of recent disturbance. The cover model was predictive of sites where high use by lynx
was associated with the occurrence of coyotes (AUC = 0.78, SE = 0.09). When lynx and coyotes

92

occurred together, both species selected for sites that were closer to riparian habitat with more
mid-level cover and less cover above 10 meters (Table 14).

2020–2021 Winter Models
In winter 2020–2021, similar to 2015–2016, cover models ranked higher than disturbance and
prey abundance models in winter (Table 13). In contrast to 2015–2016, however, the highest
ranked model for all species contained covariates for distance to riparian edge and the mid- and
top-strata canopy cover. Lynx occurrence was strongly associated with hare or squirrel
occurrence (AUC > 0.9). In all winter 2020–2021 models, lynx were positively associated with
mid-strata cover (3–10m) and closer distances to riparian edge, but negatively associated with
top-strata cover (10m +) (Table 14). Except for hare, all other models regardless of species or
multinomial outcome were positively associated with closer distances to riparian edge. Although
there was some model uncertainty for lynx-wolverine and lynx-fisher models, sites with high use
by lynx were associated with the occurrence of both species. Lynx, fisher, and wolverine
selected for sites with greater mid-strata cover and less top-strata cover. Wolverine also selected
for sites closer to riparian edge, fully mirroring lynx high use sites. Lynx high use was not
associated with coyotes.

2020–2021 Snow-free Models
The first and second ranked models for the 2020–2021 snow-free season contained either edge
distance or riparian edge distance in combination with cover covariates (Table 13). Similar to all
other time periods, lynx and hare occurrence were strongly associated. Lynx high use was not
associated with fisher, but positively associated with coyote occurrence (AUC = 0.75, SE =
0.14). Similar to winter 2020-2021, lynx high use sites were associated with the occurrence of
93

wolverine. In 2-species models, lynx, wolverine, hare, coyote, and squirrel selected for greater
mid-strata cover and less top-strata cover (Table 14).

Table 13. AICc scores, and AICc weights (w) for multinomial logistic regression models
representing habitat use and co-occurrence of Canada lynx (Lynx canadensis), sympatric
mesopredators, and prey at remote camera sites in central British Columbia, Canada, 2015–2016
and 2020–2021. Area under the curve (AUC) and standard error (SE) for the receiver operating
characteristic represents the predictive accuracy of each model. Shaded value = AUC > 0.7.
HARE LOW WINTER
Model
COYOTE–LYNX
Riparian + Cover
WOLVERINE–LYNX
Riparian + Cover
Edge + Cover
Canopy Cover
FISHER–LYNX

K

ΔAICc

w

AUC1 (SE)

AUC2 (SE)

AUC3 (SE)

4

0.0

0.961

0.88 (0.05)

0.27 (0.09)

0.51 (0.18)

4
5
5

0.0
2.3
2.4

0.621
0.192
0.182

0.59 (0.10)
0.64 (0.09)
0.67 (0.10)

0.20 (0.08)
0.41 (0.10)
0.41 (0.10)

0.90 (0.05)
0.84 (0.07)
0.82 (0.08)

Riparian + Cover

4
4
5
4
5

0.0
0.9
2.3
2.6
2.9

0.406
0.255
0.128
0.109
0.096

0.74 (0.08)
0.68 (0.09)
0.71 (0.08)
0.75 (0.08)
0.66 (0.08)

0.27 (0.08)
0.62 (0.10)
0.44 (0.10)
0.45 (0.10)
0.46 (0.11)

0.87 (0.07)
0.74 (0.09)
0.78 (0.09)
0.83 (0.07)
0.76 (0.10)

4
4

0.0
5.4

0.874
0.060

0.57 (0.10)
0.69 (0.10)

0.55 (0.11)
0.60 (0.10)

0.9 (0.05)
0.87 (0.06)

4

0.0

0.961

0.74 (0.09)

0.57 (0.10)

0.95 (0.04)

K

ΔAICc

w

AUC1 (SE)

AUC2 (SE)

AUC3 (SE)

5

0.0

0.939

0.77 (0.07)

0.39 (0.12)

0.63 (0.11)

5

0.0

0.833

0.82 (0.06)

0.2 (0.20)

0.22 (0.11)

Canopy Height + Gr. Cover
Canopy Cover
Canopy Height + Riparian
Edge + Cover
HARE–LYNX
Riparian + Cover
Canopy Height + Riparian
RED SQUIRREL–LYNX
Riparian + Cover
HARE HIGH WINTER
Model
COYOTE–LYNX
Canopy Cover
WOLVERINE–LYNX
Canopy Cover

94

Table 13 (continued)
Edge + Cover
HARE–LYNX
Riparian + Cover
RED SQUIRREL–LYNX
Canopy Cover

5

3.6

0.138

0.79 (0.06)

0.43 (0.22)

0.16 (0.08)

4

0.0

0.956

0.28 (0.09)

0.80 (0.08)

0.82 (0.07)

5

0.0

0.966

0.62 (0.10)

0.36 (0.10)

0.86 (0.07)

K

ΔAICc

w

AUC1 (SE)

AUC2 (SE)

AUC3 (SE)

5

0.0

0.97

0.68 (0.07)

0.44 (0.18)

0.75 (0.14)

4
5

0.0
2.8

0.719
0.181

0.71 (.07)
0.67 (0.07)

0.61 (0.15)
0.23 (0.09)

0.73 (0.11)
0.65 (0.15)

5
4
5

0.0
0.8
1.6

0.451
0.310
0.201

0.78 (0.07)
0.81 (0.06)
0.77 (0.07)

0.24 (0.12)
0.28 (0.09)
0.33 (0.12)

0.48 (0.10)
0.44 (0.10)
0.37 (0.10)

4
5

0.0
2.4

0.703
0.205

0.52 (0.10)
0.37 (0.10)

0.44 (0.11)
0.47 (0.12)

0.81 (0.07)
0.76 (0.07)

5
4

0.0
1.8

0.620
0.253

0.66 (0.10)
0.64 (0.10)

0.46 (0.10)
0.40 (0.10)

0.69 (0.09)
0.76 (0.08)

K

ΔAICc

w

AUC1 (SE)

AUC2 (SE)

AUC3 (SE)

4
4

0.0
4.0

0.827
0.113

0.59 (0.09)
0.36 (0.09)

0.73 (0.10)
0.23 (0.11)

0.59 (0.11)
0.78 (0.09)

5

0.0

0.999

0.65 (0.11)

0.59 (0.12)

0.76 (0.08)

4

0.0

0.967

0.40 (0.11)

0.73 (0.10)

0.69 (0.09)

HARE LOW SNOW-FREE
Model
COYOTE–LYNX
Edge + Cover
WOLVERINE–LYNX
Riparian + Cover
Edge + Cover
FISHER–LYNX
Edge + Cover
Riparian + Cover
Canopy Cover
HARE–LYNX
Riparian + Cover
Edge + Cover
RED SQUIRREL–LYNX
Edge + Cover
Riparian + Cover
HARE HIGH SNOW-FREE
Model
COYOTE–LYNX
Disturbance
Riparian + Cover
HARE–LYNX
Edge + Cover
RED SQUIRREL–LYNX
Disturbance

95

Table 14. Coefficients for top-ranked models illustrating habitat use and co-occurrence by
Canada lynx (Lynx canadensis), sympatric mesopredators, and prey at camera sites during two
contrasting periods of prey abundance in central British Columbia, Canada, 2015–2016 (high)
and 2020–2021 (low). Area under the curve (AUC) for the receiver operating characteristic
represents the predictive accuracy of each model. Dark grey = negative association, Light gray =
positive association, “●” = significant coefficient, “X” = useful or high model accuracy.

Coyote
Coyote–Lynx
Lynx
Wolverine
Wolverine–lynx
Lynx
Fisher
Fisher–lynx
Lynx
Hare ●
Hare–lynx
Lynx
Squirrel
Squirrel–lynx

●

● ●

X
X

● ●
● ●

X

X

● ●
● ● ●
●
● ● ● X ● ●

● ● ● X
● ●
X**

●

● ● ●
● ● ●

X
X
X

cover (10+ m)
AUC > 0.7

● ●
*

● ●
● ● ● X ● ● ●

cover (0–3 m)
cover (3–10 m)

HARE LOW

AUC > 0.7
dist_riparian
dist_edge

edge_density
dist_roads

cover (3–10 m)
cover (10+ m)
AGE < 15

AUC > 0.7
X

SNOW-OFF
HARE HIGH

dist_riparian
dist_edge
cover (0–3 m)

cover (3–10 m)
cover (10+ m)

●

AUC > 0.7
dist_riparian

● X ● ● ●

cover (10+ m)
conifer

Model outcome
Lynx

dist_riparian
cover (0–3 m)
cover (3–10 m)

WINTER
HARE HIGH
HARE LOW

●

●
● ●

●
X
● ● X
●
● ● X
● ● X

X
● ●
●

X

● ● X
●
●
● ● X

2 model ouputs:*disturbance model, **cover model

96

Figure 12. Probability of lynx (Lynx canadensis) and wolverine (Gulo gulo) co-occurrence and
lynx and fisher (Pekania pennati) co-occurrence, with 95% confidence intervals, as a function of
canopy cover (3–10 m and >10 m) at cameras sites during a high and low in snowshoe hare
abundance in central British Columbia, Canada, February–April, 2015–2016 and 2020–2021.

97

Discussion
This is the first study to document changes in habitat overlap and co-occurrence between Canada
lynx and sympatric mustelid species during the snowshoe hare cycle. A significant change in the
mesopredator community occurred between two time periods separated by only four years. Lynx
occurrences mirrored the decrease in hares, while the number of sympatric carnivore species and
the combined occurrences of these species increased during the low in hare abundance. Overlap
in habitats used by lynx, wolverine, and fisher increased during a low in hare abundance.
Consistent with my prediction, habitat overlap increased at a time of prey scarcity. Conversely,
coyote occurrences decreased from the high to low in hare abundance during the snow-free
period. Unlike the mustelids, interspecific overlap in habitat used by lynx and coyote was
greatest during the snow-free seasons in all years.
Cover models were the most important for all time periods. Cover models were the most
important for all time periods. The importance of cover for mesocarnivores and their prey is well
documented (Feldhamer et al. 2003), and was a clear driver of habitat use in this study. Prey
models consistently ranked lower than cover models, but hares were often a positive and
significant influence on the use of sites by lynx. Although disturbance models ranked low and
decreases in cover could also be related to forest harvest, disturbance was not as important as
cover at the spatial scale of my camera grid and variables.
A key difference between the hare low and high periods in winter was the importance of
edge distance, especially riparian, during the low in hare abundance. Riparian edges can be
positively associated with species diversity and abundance including small mammals (Hamilton
et al. 2015, Larsen-Gray and Loehle 2022, Perault and Lomolino 2000). Riparian areas likely
provide a dual function for mesopredators, providing not only hunting opportunities, but also
98

serving as travel corridors that provide cover from sympatric competitors/predators. Although
disturbance models did not rank high, my results have implications for increased disturbance on
the landscape. Following from regulatory requirements, riparian corridors can often be the only
remnant forest left behind in areas of extensive timber harvest. If riparian habitats with
concentrated mesopredator activity are more limited or become the only available cover then
overlap and competition could increase.
Mid-level and top-level cover were important predictors of where lynx and sympatric
carnivores co-occurred. Ground level forest cover and shrub vegetation (0–3) were not included
or significant covariates in most top models. This could be the result of an imprecise correction
factor for the LiDAR data that were collected during the snow-free season. However, canopy
cover (0–3) also did not play a strong role in habitat use during the snow-free season. Although
canopy cover (0–3) had a lesser role in habitat use in my study, it was a significant covariate in
several lower-ranked models. This suggests that ground cover also played a role, but to a lesser
degree than canopy cover at the two upper strata. Past studies of lynx and hare did not have
accurate cover data at multiple strata, and often used one coarser measurement of upper canopy
cover (>3m). Differences in the direction of influence of the two separate upper strata (3–10m,
>10m) in this study would not have been apparent, and their importance to habitat use studies
may have been underestimated.
Hare abundance was the only major ecological condition that changed between time
periods. There was relatively little forest harvesting and no fire activity in or adjacent to the
study area during the time of data collection. Also, the study area has not had any significant
trapping for almost two decades. Deep snow should give lynx a competitive advantage over
other carnivores (except wolverine) and there was a difference in snow depth in February/March
99

(Fort St James, B.C.; https://climate.weather.gc.ca/historical_data) between 2015–2016 (‫= ̅ݔ‬
23.1cm, SE = 1.3) and 2020–2021 (‫ = ̅ݔ‬35.5cm, SE = 1.3). However, snow was deeper during
the hare low when there were less lynx and more sympatric carnivores and is an unlikely
explanatory factor for habitat overlap in my study.
The potential influence of bait on movement patterns is an important consideration when
designing, implementing, and interpreting wildlife field studies. A study of fisher, however,
found that any effects of bait on animal movements were eclipsed by habitat heterogeneity
(Stewart et al. 2019), and that bait may increase detection probability improving ecological
inferences. In another analysis in my study area, I found that time since bait and lure were added
to a site was not a significant influence on the detection probability of lynx (Chapter 3). Lastly,
bait was very small, consumed quickly by most carnivores, and often not present at sites for long
periods. For this reason, bait primarily served as an additional scent lure before and after
consumption that encouraged animals already in the vicinity of the site to move into camera
view.
Short-term fluctuations in prey abundance and scarcity may drive short-term fluctuations
in habitat overlap and potential niche expansion and competitive interactions. Consistent with
optimal foraging theory (i.e., niche expansion during prey scarcity; Emlen, 1968, MacArthur and
Pianka, 1966) and with short-term fluctuations in prey abundance (Wiens 1977, 1993), I
observed differences in the habitat use and overlap of sympatric carnivores between the two hare
abundance periods. Although I was not able to measure competition as it might relate to diel
variation in space use or differences in diet, habitat overlap is one key element in assessing the
potential for niche overlap and competitive interactions (Murray et al. 2023). The inclusion of
temporal variation in resource availability is another important component of investigations of
100

niche overlap (Murray et al. 2023) that was part of my study. Changes in habitat overlap between
lynx and other mesopredators was likely influenced by temporal differences in hare abundance.
Short-term cycles may not only drive fluctuations in habitat overlap, but also the identity
and roles of dominant and submissive species. Extreme changes over short durations in the
biomass and composition of the prey community will provide varying benefits to different
predator species. A competitive advantage during one stage of a cycle may not be as beneficial
during another. My research reveals two important potential outcomes for lynx and competitive
mesopredators. During the hare high period, lynx were the numerically dominant species on the
landscape. Canada lynx are morphologically adapted for living in cold climates with deep snow
(thick fur, long legs, large and well-furred feet for low foot loading) (Anderson and Lovallo,
2003). Lynx had a numerical advantage because of their specialized ability to use the most
abundant food source (i.e., hares). Their ability to hunt hares in deep snow provided an
advantage over conspecifics as an exploitative competitor (Peers et al. 2020). During the hare
low period, the specialized ability to hunt hares and the numerical advantage of lynx was
diminished. Data from camera traps and hare pellet plots showed a significant decrease in the
snowshoe hare population between time periods. Additionally, both camera traps and lynx
captures revealed that lynx were in poor body condition and without kittens during the low (John
Prince Research Forest, Unpub. Data). During this time, generalist predators may have both an
interference and exploitative competitive advantage over lynx.
Generalist mesopredators may derive some benefits from an increase in the availability of
non-hare prey. During the hare low, reduced exploitative competition as well as the ability to
hunt other prey resources may have increased the distribution and abundance of mustelids and
led to increased co-occurrence with lynx. Unlike the mustelids, overlap in habitat use between
101

lynx and coyotes occurred in both snow-free seasons regardless of hare abundance. Coyote and
snowshoe hare in the boreal forest often follow a similar pattern of abundance (O’Donoghue et
al. 2001, Tyson et al. 2010), that was consistent with my observations of coyote during the snowfree season. Habitat overlap and potential competition with coyotes may increase during the
summer when lynx lose their hunting advantage in snow. In contrast, coyote occurrences at
camera sites during the winter were similar between time periods. In the boreal forest of Yukon,
O’Donoghue et al. (1998b) found that coyotes switched to hunting voles in years with a
reduction in the abundance of hares and increased numbers of small mammals. Prey switching
by coyotes may partially explain the similar patterns of occurrence between 2015–2016 and
2020–2021 when I observed relatively fewer lynx and hare during the winter.
There are several proposed mechanisms to explain the cycling dynamics of lynx and hare,
including climate (Yan et al. 2013), weather (Krebs et al. 2014), forest succession (Krebs et al.
2014), food availability and quality (Krebs et al. 1986, Sinclair et al. 1988, Krebs et al. 2017),
and predation (Tyson et al. 2010, Krebs et al. 2014, Krebs et al. 2017). Although many of these
factors may work synergistically, predation is likely the most important driver of the hare-lynx
cycle (Krebs et al. 2017). The pattern and scale of the hare cycle has long been a function of
large, undisturbed, and contiguous forests and it is unclear how a fragmented landscape, like in
many parts of central BC, may affect the cyclic dynamics and interactions of hare, lynx, and
other carnivores in the long-term. My research suggests that habitat overlap among conspecifics
in disturbed landscapes could play a role in the persistence of this cyclic dynamic.
The potential influence of competing predators on lynx populations becomes especially
relevant given the cumulative effects of landscape disturbance and climate change (Scully et al.
2018). Accumulated snow cover over most of Canada has decreased over the last several decades
102

and is projected to follow that trend in the future (Bush and Lemmen 2019). Decreases in snow
may lessen the competitive advantage lynx have over other predators potentially resulting in
niche displacement (Buskirk et al. 2000, Peers et al. 2013, Scully et al. 2018). Besides being
generalists and likely gaining some competitive advantage during hare scarcity, species such as
fisher and coyote will also benefit from reduced snowfall (Murray and Boutin 1991, Krohn et al.
1995, Pozzanghera et al. 2016, Peers et al. 2020, Pauli et al. 2022). Coyotes have gone through
an extensive range expansion in the north over many decades (Hody and Kays 2018). Fisher
populations also have expanded their range likely due to reduced snowfall and increased forest
harvesting where they often outcompete marten populations (Krohn et al. 1995, Jung and Slough
2011, Suffice et al. 2020, Kupferman et al. 2021, Pauli et al. 2022). Bobcats (Lynx rufus) are
another generalist predator not in my study area, but that are expanding northward in other parts
of North America (Roberts and Crimmins 2010) and may have a competitive advantage over
lynx in areas with less snow cover (Peers et al. 2012,2013). The existing and future novel
mesopredators likely have an interference and exploitative advantage over lynx during hare lows
or in areas of recent forest disturbance. Except for possibly wolverine, lynx also lose much of
their competitive advantage in years or areas with lesser snow cover. The current state of
cumulative forest change and projected future conditions may be detrimental to lynx populations,
but beneficial to many of its sympatric competitors. However, the scale and pace of
environmental change and its effects on natural cycles and community interactions is unclear and
requires additional long-term study.
Wiens (1977, 1993) suggested that unless populations are in a resource-defined
equilibrium, short-term studies representative of most grant or thesis projects will provide only
snapshots in time that may be difficult to interpret. My data spans a 7-year period and represents
103

contrasting environmental conditions. Nonetheless, there is uncertainty in the mechanisms
underlying the observed patterns of lynx and their competitors. Clearly, longer term monitoring
is required as has been achieved in a small number of other study areas (Einoder et al. 2018,
Krebs et al. 2001, 2023; Lindenmayer et al. 2022, Kenney et al. 2024). Extreme changes in my
mesopredator community occurred over a very short time period, thus, a frequent monitoring
interval is required. Without a basic understanding of competition among sympatric carnivores,
we cannot begin to address broader scale questions regarding the influence of climate change,
forest disturbance, and mesopredator persistence on the landscape.

104

Chapter 5: Study Findings, Management Implications, and
Recommendations for Future Research
Summary of Results
In Chapter 2, I compared the habitat selection of GPS-collared lynx and American marten at two
behavioural scales. Also, I compared differences in habitat selection for the two species
measured using GPS data and camera images. I found that habitat use measured using camera
traps, in general, was similar to the habitat use of GPS-collared lynx and marten. Mid-level and
top-level vegetation cover were important predictors of habitat use for both marten and lynx, but
with opposite directional influences. Lynx selected for more mid-level cover and less top-level
canopy cover, while the inverse relationship was found for marten. Lynx and marten
demonstrated differential use of habitat defined by forest age and structure suggesting that each
species would serve as a unique indicator of forest condition and change. For Canada lynx, the
direction and significance of habitat covariates were nearly identical using camera trap and GPScollar data and consistent between survey periods with differences in the abundance of their
primary prey species, snowshoe hare. My data suggested that camera traps can be a reliable
method for measuring habitat relationships of lynx and marten.
I found that lynx behaviours and relative abundance were correlated between years and
hare abundance periods (Chapter 3). Consistent with my predictions, years with higher lynx and
hare abundance were characterized by increases in cheek-rubbing, scent-marking, and grouping
behaviours. These behaviours proved to be good indicators of cyclic fluctuations in lynx and
hare abundance. Measures of abundance and behaviours provided insights into the ecological
processes influencing hare and lynx population change.
105

In Chapter 4, lynx occurrences mirrored the decrease in hares that I observed during the
2020–2021 monitoring period, while the occurrences of sympatric fisher and wolverine species
increased during that period. I found that the co-occurrence of lynx with other sympatric
carnivores increased at a time of prey scarcity suggesting predator populations in subboreal
forests may be in a dynamic state of habitat overlap dependent on cyclic abundance of snowshoe
hare.

Ecological Implications
My research considered a broad range of ecological relationships that were focused on
the spatial and temporal variation in the habitat, behaviour, and abundance of lynx. That
variation included a rapid decrease in the abundance of the primary prey of lynx, snowshoe hare.
Despite that variation, I observed some consistent responses of lynx to components of the
subboreal forest. In particular, vegetation cover was a strong and consistent factor that explained
habitat use, carnivore co-occurrence, abundance, and behaviours. Specifically, the co-occurrence
of lynx with other mesopredators, habitat selection measured with GPS collars and camera traps,
and abundance estimates were explained by mid- and top-strata cover. Abundance, GPS-collar
locations, and cheek-rubbing behaviours were also positively associated with greater vegetative
ground or low cover. The accuracy of LiDAR-derived cover measures relative to coarser VRI
data likely aided in revealing these relationships. Past studies have found that ground cover is an
important predictor of lynx habitat (Adams 1959, Brocke 1975, Wolff 1980, Wolfe et al. 1982,
Litvaitis et al. 1985, Homyack et al. 2007). Ground cover was important in my study, but midlevel cover and upper canopy cover were also associated with lynx habitat. Past studies of lynx
and hare did not have accurate cover data at multiple strata, and often used one coarser
measurement of upper canopy cover (>3m). Differences in the direction of influence of the two
106

separate upper strata (3–10m, >10m) in this study would not have been apparent, and their
importance to habitat use studies may have been underestimated.
Lynx were positively associated with areas closer to riparian features in several analyses
of my dissertation. Lynx co-occurrence with sympatric carnivores increased in areas closer to
riparian edge during the low in hare abundance. There can be both positive and negative effects
of edge habitat. Species diversity has been positively associated with habitat edges, while
increases in nest predation and barriers to dispersal are potential negative effects (Yahner 1988,
Lidicker Jr. 1999, Kremsater and Bunnell 1999, and Baldi 2004, Hamilton et al. 2015, Willmer
et al. 2022). Riparian zones, in particular, can be positively associated with species diversity, and
serve as movement corridors for wildlife (Bunnell and Dupuis 1993, Perault and Lomolino 2000,
Hamilton et al. 2015, Larsen-Gray and Loehle 2022).
Riparian edge may be important to mesopredators in my study area, particularly during
times of prey scarcity. Greater prey abundance and diversity relative to nearby non-riparian
habitat may provide more hunting opportunities. Also, riparian areas likely have a dual function
for mesopredators, providing not only foraging habitat, but also serving as travel corridors
between other habitats that provides cover from sympatric competitors, predators, and
environmental conditions. Lynx increase movement rates during hare lows potentially increasing
the use of riparian corridors for both foraging and travel (Nellis and Keith 1968, Brand et al.
1976, Wards and Krebs 1985, Poole 1994).
The habitat ecology of lynx, and potential competitors, were strongly related to the
temporal and spatial variation in the abundance of snowshoe hare. My study included two
contrasting periods of hare abundance, providing the opportunity to investigate the interaction
between prey abundance and the habitat selection, community dynamics, and survey methods for
107

lynx. I found that hare abundance influenced the abundance of lynx, co-occurrence with
competitive mesopredators, and the efficacy of survey methods. The absence of hares available
to carnivores likely created a ripple effect throughout the entire community. Increases in fisher
and wolverine between the high and low in hare abundance, were paralleled by decreases in
marten, lynx, and hare. My results suggest that the carnivore community in subboreal forests
may be in a dynamic state of habitat overlap dependent on cyclic prey and predator abundance.
These changes in ecological conditions can strongly influence how we monitor this community.
Behavioral research can help explain population and community models as well as
identify proximate mechanisms underlying behavioural responses (Bro-Jorgensen et al. 2019).
Similarly, variation in individual behaviour can influence the precision or even bias the results of
population estimates. For example, Crowley et al. (2012) found that movement behaviours and
track-sign heterogeneity among individual river otters could bias surveys used to assess and
monitor changes in population distribution and abundance. My research provides new insights on
the relationship between behaviour of Canada lynx and the design and interpretation of studies
focused on documenting habitat use and abundance. Surveys were influenced by behaviours
including movement, habitat choice, and cheek-rubbing. My research supports the pairing of
behavioural responses with other monitoring efforts to provide additional insights into factors
influencing population and community trends. This approach is critical to the interpretation of
population-level surveys and will ultimately improve the conservation and management of
wildlife populations.

108

Landscape Change and Long-term Monitoring
My results highlight the need for long-term monitoring of wildlife populations and their habitat.
Such monitoring is especially important when resources are of concern or have high social or
economic value, as part of a legal mandate or judicial decision, or in reaction to a crisis
(McComb et al. 2010). When wildlife monitoring is implemented as a reaction to a crisis, we
often do not have “before” data which limits our ability to understand and interpret population
changes as well as predict future conditions. McComb et al. (2010) state that a rigorous, unbiased
monitoring program is needed to determine if a population is changing and to inform
management decisions. Only through long-term monitoring can we understand long-term
dynamics that are a product of cyclic behaviour. A typical 2–3-year research project could
produce very different results and management recommendations, depending on the phase of the
cycle that was studied. These differences could be caused by ecological changes as well as their
effects on the efficacy of survey methods.
The lynx-snowshoe hare cycle is a keystone relationship in northern forests. In this case,
a species/relationship is considered keystone if removal of a species or fundamental change in
the relationship would disproportionately affect the broader ecosystem dynamic (Caro 2010).
Continued long-term monitoring is critical to understanding the persistence of this relationship
across disturbed landscapes. At the very least, short-term studies need to be placed in the context
of regional hare populations. For example, habitat selection, movement dynamics, and
competitive interactions of lynx may differ with varying density of snowshoe hare. Survey
methods that are successful during high population densities may differ from those during
population lows. Without this knowledge, data could be misinterpreted and result in biased or
incomplete management recommendations and practices.
109

Indicator species are used to measure the status and conditions of ecosystems (Caro
2010). Marten is often recommended as a management indicator and umbrella species for mature
and old growth forest conservation and monitoring initiatives (Hepinstall and Harrison 2001,
Mortelliti et al. 2022, Woollard et al. 2024). The close association among lynx, hare, and lateregenerating forests also holds promise for the use of lynx as an indicator and umbrella species
of younger forest types. My research has demonstrated that marten and lynx are sensitive to
forest cover and age. Those relationships support the suitability of these two mesopredators as
co-indicators of forest condition and change. Maintaining sufficient habitat for marten and lynx
on the same landscape would require a mosaic of stands with varying cover likely beneficial to
many other wildlife species.
Canada lynx may be particularly suited as an indicator species of environmental change
because of their suspected sensitivity to variation in climate (King et al. 2020). Lynx is a
specialist predator of snowshoe hare and is uniquely adapted to hunting prey in deep snow.
Accumulated snow cover over most of Canada has decreased over the last several decades and is
projected to follow that trend in the future (Bush and Lemmen 2019). Decreases in snowfall may
lessen the competitive advantage lynx have over other predators (Buskirk 2000, Peers et al.
2013). Monitoring changes in lynx populations across a latitudinal gradient may provide an
indicator of how environmental change and community dynamics of this pan-North American
predator-prey system is affected by climate change.

Strengths, Weaknesses, and Future Research
A clear strength of my research was the combined availability and use of individual movement
(GPS collars), non-invasive survey (camera trap and hair snare), and fine-scale habitat (LiDAR)

110

data. Comparable data from GPS-collars and non-invasive surveys often are not available for the
same land base and time period. LiDAR provided habitat data that have not been available for
lynx studies in the past. The combined data sets provided novel insights into the ecological and
behavioural processes influencing the habitat ecology of lynx. My research also benefited from a
longer-term data set that spanned different cyclic periods of lynx and hare abundance. My
research incorporated these differences in abundance, and provided additional insights into how
ecological conditions influence lynx habitat use, carnivore co-occurrence, and survey methods.
The annual sample size of GPS-collared lynx in my study was small. Although 17 lynx
were used in my analyses (Chapter 2), lynx capture spanned several years with only 4–7 lynx
monitored in each year. Low densities of lynx in the study area limited capture success and
sample size. Although my annual sample size was small, density estimates produced from SECR
models (Chapter 3), suggested that approximately 50% of the lynx population were outfitted with
GPS collars during each year of the study. Raw detection data corroborate this estimate with
collared lynx representing 49% and 45% of all lynx occurrences at camera traps in 2020 and
2022, respectively. Data from GPS collars on lynx, however, represented a period of low hare
abundance only and it is unclear how mark-resight and N-mixture estimates would compare
during a hare high. If there is 4-fold difference in density between low and high periods as has
been documented in Alberta (Keith et al. 1977), my SCMR density estimate would extrapolate to
12 lynx/100 km2 during a hare high. Although reasonable, extreme variation in the amplitude of
cycles within the same region or between regions make this estimate difficult to assess for our
study area.
The use of LiDAR to measure fine-scale habitat characteristics was a clear benefit to my
study. However, LiDAR data are costly to acquire and process, limiting the overall availability
111

of these data. I was unable to use location data from several collared lynx that ranged beyond the
LiDAR area. In addition, predictive habitat mapping beyond the boundaries of my study area
was not possible. Further research needs to connect LiDAR based metrics to past and current
harvest practices. An understanding of this relationship can help predict future habitat conditions
and inform best management practices for maintaining forest structure required for lynx and
other forest carnivores.
Each of my chapters provided direction for future research. In Chapter 2, I highlighted
the need to investigate the influences of movement behaviour on point surveys for wildlife
species. It is unclear if different types of point surveys can provide similar assessments of habitat
use when compared to individual-based movement data. This could be investigated by pairing
GPS-collars or backtracking studies with camera trap or hair snare surveys. Finer-scale GPS fix
rates (i.e., 10 minutes to 1 hour) could also provide more detailed assessments of sinuosity and
associated behaviours (i.e., resting, hunting, traveling). Backtracking during the winter also
provides this information as well as specific behaviors such as hunting, prey chases and kills,
scent-marking, and resting beds (Parker et al. 1983, O’Donoghue et al 1998a).
In Chapter 3, I did not have sufficient data to generate mark-resight estimates of lynx
abundance during a high in hare abundance. Although I found similar patterns of habitat
selection using both GPS collars and camera trap data, GPS collars deployed during a hare high
could provide additional insights into the habitat ecology of lynx and the efficacy of N-mixture
models. In addition, the influence of disturbance on lynx abundance was difficult to assess at the
spatial scale of my study. The influence of regional forest disturbance on cyclic fluctuations
needs to be assessed across a much larger area (i.e., Northern BC). Long-term monitoring of
multiple camera grids across varying levels of disturbance would be a first step in addressing that
112

knowledge gap. Although this would require considerable effort, there is an opportunity for
cross-jurisdictional collaboration in a study of this type. For Chapter 4, future research should
include data on carnivore diet to assess the full range of potential competitive interactions
(spatial, temporal, diet; Murray et al. 2023). Although I had estimates of prey availability, diet
data collected across time periods of varying prey availability could provide additional insights
into competitive interactions within the mesopredator community.

Management and Conservation
Late-regenerating forests are important to hare and lynx populations (Mowat et al. 2000,
Hodges et al. 2000a, b, Anderson et al. 2003, Murray et al. 2003). Greater ground- and mid-level
cover, often a characteristic of late regenerating stands (20–35 years old), were an important
predictor of lynx habitat in my study. Representation of late-regenerating stands in a mosaic of
multi-age forest is likely critical to maintaining hare populations and the cyclicity of their
populations. The biomass represented by hares is important to a variety of other species that goes
beyond the lynx-hare relationship. Representation and maintenance of these older regenerating
stands should be an important component of forest and landscape management planning. In
addition, cover and structure were not only key predictors of lynx, but also the habitat use of
marten, wolverine, fisher, and coyote. My results suggest that mesopredator communities in
subboreal forests are dependent on landscapes that contain a mosaic of stand types of various
ages that represent vegetation cover at multiple strata (Holbrook et al. 2019). Large areas of
even-aged forest will likely limit species diversity. Harvest practices that retain stand cover or
maintain connectivity corridors with cover will benefit a variety of mesopredator species.

113

The importance of riparian habitat to fish and other aquatic or semi-aquatic wildlife is
well known. However, the integrity of riparian systems is also important to lynx and other
terrestrial carnivore species. My results provide some insights on the ecological effects of
increased forest disturbance at landscape scales. Riparian habitats, including adjacent forests,
account for a small proportion of the landscape, but they have a disproportionately high
ecological value (Morgan and Lashmar 1993, Naiman et al. 1993, Moore and Richardson 2003,
Li at al. 2024). Human modification and loss of riparian habitat is a continuing trend globally
(Riis et al. 2020, Urbanic 2022) and in British Columbia (Li et al. 2024). In some parts of my
study area, riparian corridors were the only remnant forest left behind due to regulatory
requirements that limit harvest close to streams and lakeshores. If riparian forest cover becomes
more limited, species overlap will likely increase in these habitats.
Accurate monitoring of population abundance is critical for determining and assessing
trapping regulations. Canada lynx is one of the top three harvested furbearers in BC and are
considered sensitive to harvest (Class 2 species; Hatler and Beal 2003). Government biologists
have been asked on multiple occasions by members of the trapper community to extend harvest
seasons for lynx (S. Marshall, pers. comm.). However, the absence of past studies on the habitat
ecology and abundance of lynx in BC have made it difficult to make informed management
decisions. During years when lynx are less abundant, non-compensatory trapping could prolong
the low phase of the lynx-hare population cycle. An understanding of lynx populations in the
context of regional and continent-wide fluctuations becomes especially relevant to decisions on
trapping quotas and the length of the trapping season. Future monitoring using camera traps will
allow us to estimate the abundance of lynx populations, including documenting the increasing or
declining phase of the cycle, and adjust trapping regulations accordingly.
114

Although there are large differences in the ecology and behaviour of felid species
throughout the world, there are also many similarities and behavioural consistencies. For
example, the rubbing behaviour of lynx on scented objects is generally universal among felid
species (Allen et al. 2016a). The survey techniques and design as well as statistical methods of
my research will likely influence research on other cat species. This is especially important in
areas with rare or threatened species that receive little funding for research and conservation.
According to the IUCN Red List, 13.2% of 36 felid species are endangered, 34.2% are
vulnerable, and another 18.4% are near threatened (IUCN 2020). Research on many of these
species is limited and underfunded but of critical and urgent importance for conservation. My
research can inform the research and conservation efforts of these less studied felid species.

115

References
Adams, L. 1959. An analysis of a population of snowshoe hares in northwestern Montana.
Ecological Monographs 29:141–170.
Allen, M.L., H.U. Wittmer, E. Setiawan, S. Jeffe, and A.J. Marshall. 2016a. Scent marking
Sunda clouded leopards (Neofelis diardi): novel observations close a key gap in
understanding felid communication behaviours. Scientific Reports 6:35433.
Allen, M.L., V. Yovovich, and C.C. Wilmers. 2016b. Evaluating the responses of a territorial
solitary carnivore to potential mates and competitors. Scientific Reports 6:27257.
Anderson, E.M., and M.J. Lovallo. 2003. Bobcat and lynx (Lynx rufus and Lynx canadensis). In
G. Feldhamer, B. Thompson, and J. Chapman (Eds.), Wild Mammals of North America:
Biology, Management, and Conservation. 2nd ed (pp. 758–786). The Johns Hopkins
University Press, Baltimore, Maryland, USA.
Anderson, A.K., J.S. Waller, and D.H. Thornton. 2023. Canada lynx occupancy and density in
Glacier National Park. The Journal of Wildlife Management 87:e22383.
Apps, C.D. 2000. Space–use, diet, demographics, and topographic associations of lynx in the
southern Canadian Rocky Mountains: a study. In L. F. Ruggiero, K. B. Aubry, S. W.,
Buskirk, G. M. Koehler, C. J. Krebs, K. S. McKelvey, J. R. Squires (Eds.), Ecology and
Conservation of Lynx in the United States (pp. 351–371). University Press of Colorado,
Boulder, Colorado, USA.
Aubry, K. B., G. M. Koehler, and J. R. Squires. 2000. Ecology of Canada lynx in southern boreal
forests. In L. F. Ruggiero, K. B. Aubry, S. W., Buskirk, G. M. Koehler, C. J. Krebs, K. S.

116

McKelvey, J. R. Squires (Eds.), Ecology and Conservation of Lynx in the United States
(pp. 373–396). University Press of Colorado, Boulder, Colorado, USA.
Bacon, D.W., and D.G. Watts. 1971. Estimating the transition between two intersecting straight
lines. Biometrika 58:525–534.
Banci, V. A. 1987. Ecology and behavior of wolverine in Yukon. MSc. Thesis, University of
British Columbia, Vancouver, British Columbia, Canada.
Barker, R.J., M.R. Schofield, W.A. Link, and J.R. Sauer. 2018. On the reliability of N–mixture
models for count data. Biometrics 74:369–377.
Batary, P., and A. Baldi. 2004. Evidence of an edge effect on avian nest success. Conservation
Biology 18:389–400.
BC FLNRO. 2012. History of the mountain pine beetle infestation in B.C. Ministry of Forests,
Lands and Natural Resource Operations. Province of British Columbia, Canada.
https://www2.gov.bc.ca/assets/gov/farming–natural–resources–and
industry/forestry/forest–health/bark–
beetles/history_of_the_mountain_pine_beetle_infestation.pdf
Blanchet, F.G., K. Cazelles, and D. Gravel. 2020. Co–occurrence is not evidence of ecological
interactions. Ecology Letters 23:1050–1063.
Bowman, J., D. Donovan, and R.C. Rosatte. 2006. Numerical response of fishers to synchronous
prey dynamics. Journal of Mammalogy 87:480–484.
Boyce M.S., P.R. Vernier, S.E. Nielsen, and F.K.A. Schmiegelow. 2002. Evaluating resource
selection functions. Ecological Modelling 157:281–300.
117

Brand, C.J., L.B. Keith, and C.A. Fischer. 1976. Lynx responses to changing snowshoe hare
densities in Alberta. The Journal of Wildlife Management 40:416–428.
Bridger, M.C. 2015. Assessing cumulative impacts of forest development on the abundance and
distribution of furbearers. MSc. Thesis, University of Northern British Columbia, Prince
George, British Columbia, Canada.
Bridger, M.C., C.J. Johnson, and M.P. Gillingham. 2016. Assessing cumulative impacts of forest
development on the distribution of furbearers using expert–based habitat modeling.
Ecological Applications 26:499–514.
Brocke, R.H. 1975. Preliminary guidelines for managing snowshoe hare habitat in the
Adirondacks. Transactions of the Northeast Fish and Wildlife Conference 32:46–66.
Bro–Jorgensen, J., D.W. Franks, and K. Meise. 2019. Linking behaviour to dynamics of
populations and communities: application of novel approaches in behavioural ecology to
conservation. Philosophical Transactions Royal Society Bulletin 374:20190008.
Bryant,J.P., K. Joly, F.S. Chapin III, D.L. Angelis, and K. Kielland. 2014. Can antibrowsing
defense regulate the spread of woody vegetation in arctic tundra? Ecography 37:204–
211..
Bunnell, F.L., and L.A. Dupuis. 1993. Riparian habitats in British Columbia: their nature and
role. In K.H. Morgan, and M.A. Lashmar (Eds.), Riparian Habitat Management and
Research. Proceedings of a workshop by Environment Canada and the British Columbia
Forestry Continuing Studies Network (pp. 7–21). Kamloops, B.C. Canada.

118

Burgar, J.M., F.E.C. Stewart, J.P. Volpe, J.T. Fisher, and A.C. Burton. 2018. Estimating density
for species conservation: Comparing camera trap spatial count models to genetic spatial
capture–recapture models. Global Ecology and Conservation 15:e00411.
Burgos, T., E. Virgos, E.S. Valero, R. Arenas–Rojas, J. Rodriguez–Siles, and M.R. Recio. 2018.
Prey density determines the faecal–marking behaviour of a solitary predator, the Iberian
lynx (Lynx pardinus). Ethology, Ecology, and Evolution 31:219–230.
Burnham, K.P., and D.R. Anderson. 2004. Multimodel inference: Understanding AIC and BIC in
model selection. Sociological Methods and Research 33:261–304.
Burton, A.C., E. Neilson, D. Moreira, R. Ladle, J.T. Fisher, E. Bayne, and S. Boutin. 2015.
Wildlife camera trapping: a review and recommendations for linking surveys to
ecological processes. Journal of Applied Ecology 52:675–685.
Bush, E. and D.S. Lemmen, (Eds.). 2019. Canada’s Changing Climate Report. Government of
Canada, Ottawa, ON. 444 p. www.ChangingClimate.ca/CCCR2019
Buskirk, S.W., L.F. Ruggiero, K.B. Aubry, D.E. Pearson, J.R. Squires, and K.S. McKelvey.
2000. Comparative ecology of lynx in North America. In L. F. Ruggiero, K. B. Aubry, S.
W. Buskirk, G. M. Koehler, C. J. Krebs, K. S. McKelvey, J. R. Squires, (Eds.). Ecology
and Conservation of Lynx in the United States (pp. 397–418). University Press of
Colorado, Boulder, Colorado, USA.
Carignan, V., and M. Villard. 2002. Selecting indicator species to monitor ecological integrity: A
review. Environmental Monitoring and Assessment 78:45–61.

119

Caro, T. 2010. Conservation by Proxy: Indicator, Umbrella, Keystone, Flagship, and Other
Surrogate Species. Island Press, Washington, D.C., U.S.A.
Chandler, R.B., D.A. Crawford, E.P. Garrison, K.V. Miller, and M.J. Cherry. 2022. Modeling
abundance, distribution, movement, and space use with camera and telemetry data.
Ecology 103:e3583.
Chapin, T.G., D.J. Harrison, and D.D. Katnik. 1998. Influence of landscape pattern on habitat
use by American marten in an industrial forest. Conservation Biology 12: 327–1337.
Chisholm, J.D. 2023. Snowshoe hare (Lepus americanus) habitat use in the sub–boreal forests of
North–central British Columbia. MSc. Thesis. Thompson Rivers University, Kamloops,
BC, Canada.
Crooks, K.R. 2002. Relative sensitivities of mammalian carnivores to habitat fragmentation.
Conservation Biology 16:488–502.
Crowley, S., C.J. Johnson, and D. Hodder. 2012. The role of demographic and environmental
variables on the presence of snow tracks by river otters Lontra canadensis. Wildlife
Biology 18:105–112.
Crowley, S.M., D.P. Hodder, and K.W. Larsen. 2013. Canada Lynx (Lynx canadensis) detection
and behaviour using remote cameras during the breeding season. Canadian Field–
naturalist 127: 310–318.
Crowley S.M., and D.P. Hodder. 2017. An assessment of the efficacy of rub stations for
detection and abundance surveys of Canada lynx (Lynx canadensis). Canadian Journal of
Zoology 95:885–890.

120

Davison, S., D.N.M. Donoghue, and N. Galiatsatos. 2020. The effect of leaf–on and leaf–off
forest canopy conditions on LiDAR derived estimations of forest structural diversity.
International Journal of Applied Earth Observations and Geoinformation 92:102160.
DeMars, C.A., and S. Boutin. 2017. Nowhere to hide: Effects of linear features on predator–prey
dynamics in a large mammal system. Journal of Animal Ecology 97:274–284.
De Cáceres, M., P. Legendre, and M. Moretti. 2010. Improving indicator species analysis by
combining groups of sites. Oikos 119:1674–1684.
Delong, C., D. Tanner, and M.J. Jull. 1993. A field guide for site identification and interpretation
for the southwest portion of the Prince George Forest Region. Crown Publications Inc.,
Victoria, B.C.
Desmarchelier, M., M. Cheveau, L. Imbeau, and L. Stephane. 2007. Field use of isoflurane as an
inhalant anesthetic in the American marten (Martes americana). Journal of Wildlife
Diseases 43:719–725.
Doran‐Myers, D., A.J. Kenney, C.J. Krebs, C.T. Lamb, A. Menzies, D. Murray, E. Studd, J.
Whittington, and S. Boutin. 2021. Density estimates for Canada lynx vary among
estimation methods. Ecosphere 12:3774.
Efford M.G., D.K. Dawson, and D.L. Borchers. 2009. Population density estimated from
locations of individuals on a passive detector array. Ecology 90:2676–2682.
Efford M.G., and C.M. Hunter. 2017. Spatial capture–mark–resight estimation of animal
population density. Biometrics 74:411–420.

121

Einoder L., D. Southwell, G. Gillespie, A. Fisher, J. Lahoz–Monfort, and B. Wintle. 2018.
Optimising broad–scale monitoring for trend detection: review and re–design of a long–
term program in northern Australia. In S.M. Legge, et al. (Eds.), Monitoring Threatened
Species and Ecological Communities. CSIRO Publishing, Melbourne, Australia.
Elton, C., and M. Nicholson. 1942. The ten-year cycle in numbers of lynx in Canada. Journal of
Animal Ecology 11:215–244.
Emlen, J.M. 1966. The role of time and energy in food preference. The American Naturalist
100:611–617.
Estevo, C.A., M.B. Nagy–Reis, and J.D. Nichols. 2017. When habitat matters: Habitat
preferences can modulate co–occurrence patterns of similar sympatric species. PLoS
ONE 12(7):e0179489.
Feldhamer, G.A., B.C. Thompson, and J.A. Chapman, (Eds.). 2003. Wild Mammals of North
America: Biology, Management, and Conservation, 2nd ed. The Johns Hopkins
University Press, Baltimore, Maryland. 1216 pp.
Ficetola, G.F., B. Barzaghi, A. Melotto, M. Muraro, E. Lunghi, C. Canedoli, E. Lo Parrino, V.
Nanni, I. Silva–Rocha, A. Urso, M.A. Carretero, D. Salvi, S. Scali, G. Scarì, R. Pennati,
F. Andreone, and R. Manenti. 2018. N–mixture models reliably estimate the abundance
of small vertebrates. Scientific Reports 8:10357.
Fischer, J., and D. Lindenmayer. 2007. Landscape modification and habitat fragmentation: a
synthesis. Global Ecology and Biogeography 16:265–280.

122

Fisher, J.T., B. Anholt, S. Bradbury, M. Wheatley, and J.P. Volpe. 2013. Spatial segregation of
sympatric marten and fishers: the influence of landscapes and species–scapes. Ecography
35:240–248.
Fiske, I., and R. Chandler. 2011. unmarked: an R package for fitting hierarchical models of
wildlife occurrence and abundance. Journal of Statistical Software 43:1–23
FPB (Forest Practices Board). 2009. Biodiversity conservation during salvage logging in the
central interior of BC. Special Report FPB/SR/35.
Fryxell, J.M., A.R.E. Sinclair, and G. Caughley. 2014. Wildlife Ecology, Conservation, and
Management. 3rd edition. John Wily and Sons Ltd, Oxford, U.K.
Fuller, A.K., D.J. Harrison, and J.H. Vashon. 2007. Winter habitat selection by Canada lynx in
Maine: Prey abundance or accessibility? The Journal of Wildlife Management 71:1980–
1986.
Green, D.S., M.T. Farr, K.E. Holekamp, E.D. Strauss, and E.F. Zipkin. 2019. Can hyena
behaviour provide information on population trends of sympatric carnivores.
Philosophical Transactions Royal Society Bulletin 374:20180052.
Hamilton, B.T., B.L. Roeder, K.A. Hatch, K, D.L. Eggett, and D. Tingey. 2015. Why is small
mammal diversity higher in riparian areas than in uplands? Journal of Arid Environments
119:41–50.
Hardin, G. 1960. The competitive exclusion principle. Science 131:1292–1297.
Hargis, C.D., J.A. Bissonette, and D.L. Turner. 1999. The influence of forest fragmentation and
landscape pattern on American martens. Journal of Applied Ecology 36:157–172.
123

Hatler, D.F., and A.M.M. Beal. 2003. Furbearer management guidelines: Lynx (Lynx
canadensis). BC FLNRO. 11 pages.
Hepinstall, J.A., and D.J. Harrison. 2001. Marten as a tool for landscape–scale habitat planning
in Northern Maine. Pages 45–48 In Maine Cooperative Forestry Research Unit 2001
Annual Report. Maine Agricultural and Forest Experiment Station Miscellaneous Report
428. University of Maine, Orono, ME, USA.
Higashide, D., T. Kuriyama, S. Takagi, Y. Nakashima, K. Fukasawa, G. Yajima, M. Kasada, and
M. Yokoyama. 2021. Effectiveness of signs of activity as relative abundance indices for
wild boar. Wildlife Biology wlb.00869.
Hill, J.O., E.J. Pavlik, G.L. Smith, G.M. Burghardt, and P.B. Coulson. 1976. Species–
characteristic responses to catnip by undomesticated felids. Journal of Chemical Ecology
2:239–253.
Hodges, K.E. 2000a. Ecology of snowshoe hares in northern boreal forests. In L. F. Ruggiero, K.
B. Aubry, S. W. Buskirk, G. M. Koehler, C. J. Krebs, K. S. McKelvey, J. R. Squires,
(Eds.). Ecology and Conservation of Lynx in the United States (pp. 117–162). University
Press of Colorado, Boulder, Colorado, USA.
Hodges, K.E. 2000b. Ecology of snowshoe hares in southern boreal and montane forests. In L. F.
Ruggiero, K. B. Aubry, S. W. Buskirk, G. M. Koehler, C. J. Krebs, K. S. McKelvey, J. R.
Squires, (Eds.). Ecology and Conservation of Lynx in the United States (pp. 163–206).
University Press of Colorado, Boulder, Colorado, USA.
Hody, J.W., and R. Kays. 2018. Mapping the expansion of coyotes (Canis latrans) across North
and Central America. ZooKeys 759:81–97.
124

Holbrook, J.D., J.R. Squires, B. Bollenbacher, R. Graham, L.E. Olson, G. Harvey, S. Jackson, R.
L. Lawrence, and S.L. Savage. 2019. Management of forests and forest carnivores:
Relating landscape mosaics to habitat quality of Canada lynx at their range periphery.
Forest Ecology and Management 437:411–425.
IUCN 2020. The IUCN Red List of Threatened Species. Version 2020–1.
https://www.iucnredlist.org. Downloaded on 19 March 2020.
Johnson, C.J., K.L. Parker, D.C. Heard, and M.P. Gillingham. 2002a. Movement parameters of
ungulates and scale–specific responses to the environment. Journal of Animal Ecology
71:225–235.
Johnson, C.J., K.L. Parker, D.C. Heard, and M.P. Gillingham. 2002b. A multi–scale behavioral
approach to understanding the movements of woodland caribou. Ecological Applications
2:1840–1860.
Jung, T.S. 2022. Raven (Corvus corax) as a novel food item for lynx (Lynx canadensis). The
Canadian Field-Naturalist 136:17–19.
Jung, T.S., M.J.L. Peers, R. Drummond, and S.D. Taylor. 2023. Dining with a glutton: an
intraguild interaction between scavenging wolverine (Gulo gulo) and lynx (Lynx
canadensis). Ecosphere 14:e4491.
Jung, T.S., and B.G. Slough. 2011. The status of fisher (Martes pennanti) at the Northwestern
edge of their range: Are they increasing and expanding in the Yukon. Northwestern
Naturalist 92:57–64.

125

Keith, L.B. 1963. Wildlife’s Ten–year Cycle. University of Wisconsin Press, Madison,
Wisconsin, USA.
Keith, L.B., A.W. Todd, C.J. Brand, R.S. Adamick, and D.H. Rusch. 1977. An analysis of
predation during a cyclic fluctuation of snowshoe hares. International Congress of Game
Biologists 13:151–175.
Kendall, K.C., and K.S. McKelvey. 2008. Chapter 6: Hair collection. In R.A. Long, P. MacKay,
J.C. Ray, and W.J. Zielinski, (Eds.). Noninvasive Survey Methods for North American
Carnivores (pp. 135–176). Island Press, Washington, D. C., USA.
Kenney, A.J., S. Boutin, T.S. Jung, D.L. Murray, N. Johnson, and C.J. Krebs. 2024. Motion‐
sensitive cameras track population abundance changes in a boreal mammal community in
southwestern Yukon, Canada. The Journal of Wildlife Management 88:e22564.
Kidwai, Z., J. Jimenez, C.J. Louw, H.P. Nel, and J.P. Marshal. 2019. Using N–mixture models to
estimate abundance and temporal trends of black rhinoceros (Diceros bicornis L.)
populations from aerial counts. Global Ecology and Conservation 19:e00687.
King, T.W., C. Vynne, D. Miller, S. Fisher, S. Fitkin, J. Rohrer, J.I. Ransom, and D. Thornton.
2020. Will lynx lose their edge? Canada lynx occupancy in Washington. The Journal of
Wildlife Management 84:705–725.
Koehler, G.M. 1990. Population and habitat characteristics of lynx and snowshoe hares in north
central Washington. Canadian Journal of Zoology 68:845–851.

126

Koetke, L.J., D.P. Hodder, and C.J. Johnson. 2024. Using camera traps and N-mixture models to
estimate population abundance: Model selection really matters. Methods in Ecology and
Evolution 15:900–915.
Krebs, C.J., B.S. Gilbert, S. Boutin, A.R.E. Sinclair, and J.N.M. Smith. 1986. Population biology
of snowshoe hares. Demography of food supplemented populations in the southern
Yukon, 1976–84. Journal of Animal Ecology 55:963–982.
Krebs, C.J., S. Boutin, and R. Boonstra. 2001. Ecosystem Dynamics of the Boreal Forest: the
Kluane Project: Oxford University Press, 544 p.
Krebs C.J., S. Boutin, R. Boonstra, D.L. Murray, T.S. Jung, M. O’Donoghue, B.S. Gilbert, P.M.
Kukka, S.D. Taylor T. Morgan, R. Drummond, A.R.E. Sinclair, and A.J. Kenney. 2023.
Long-term monitoring in the boreal forest reveals high spatio-temporal variability among
primary ecosystem constituents. Frontiers in Ecology and Evolution 11:1187222.
Krebs, C.J., R. Boonstra, and S. Boutin. 2017. Using experimentation to understand the 10–year
snowshoe cycle in the boreal forest of North America. Journal of Animal Ecology 87:87–
100.
Krebs, C.J., J. Bryant, M. O’Donoghue, K. Kielland, F. Doyle, C. McIntyre, … and H. Golden.
2014. What factors determine cyclic amplitude in the snowshoe hare cycle? Canadian
Journal of Zoology 92:1039–1048.
Kremsater, L., and F.L. Bunnell. 1999. Edge effects: Theory, evidence and implications to
management of western North American forests. In J.A. Rochelle, L.A. Lehmann, and J.
Wisniewski, (Eds.). Forest Fragmentation: Wildlife and Management Implications (pp.
117–153). Brill, Leiden, Netherlands.
127

Krohn, W.B., K.D. Elowe, and R.B. Boone. 1995. Relations among fishers, snow, and martens:
Development and evaluation of two hypotheses. The Forestry Chronicle 71:97–105.
Kupferman, C.A., A.P. Crupi, L.P. Waits, and S.L. Gilbert. 2021. Spatial and temporal
partitioning of mustelids in Southeast Alaska. Ecosphere 12:e03827.
Lambeck, R.J. 1997. Focal species: A multi–species umbrella for nature conservation.
Conservation Biology 11:849–856.
Larsen–Gray, A.L., and C. Loehle. 2022. Relationship between riparian buffers and terrestrial
wildlife in the eastern United States. Journal of Forestry 120:336–357.
Li, H., C.J. Johnson, J.F. Rex, and M. Todd. 2024. Long–term riparian forest loss around
streams, lakes, and wetlands in ecologically diverse managed and unmanaged landscapes.
Forest Ecology and Management 562:121931.
Lidicker, W.Z. Jr. 1999. Responses of mammals to habitat edges: an overview. Landscape
Ecology 14:333–343.
Lindenmayer, D.B., T. Lavery, and B.C. Scheele. 2022. Why we need to invest in large–scale,
long–term monitoring programs in landscape ecology and conservation biology. Current
Landscape Ecology Reports 7:137–146.
Link, W.A., M.R. Schofield, R.J. Barker, and J.R. Sauer. 2018. On the robustness of N–mixture
models. Ecology 99:1547–1551.
Lino, A., C. Fonseca, D. Rojas, E. Fischer, and M.J.R. Pereira. 2019. A meta–analysis of the
effects of habitat loss and fragmentation on genetic diversity in mammals. Mammalian
Biology 94:69–76.
128

Litvaitis, J.A., J.A. Sherburne, and J.A. Bisonette. 1985. Influence of understory characteristics
on snowshoe hare habitat use and density. The Journal of Wildlife Management 49:866–
873.
Long, R.A., P. MacKay, J. Ray, and W. Zielinski. 2008. Non–invasive Survey Methods for
Carnivores. Island Press, Washington D.C., USA.
MacArthur, R.H., and E.R. Pianka. 1966. On optimal use of a patchy environment. American
Naturalist 100:603–609.
MacKenzie, D.I., L.L. Bailey, and JD. Nichols. 2004. Investigating species co–occurrence
patterns when species are detected imperfectly. Journal of Animal Ecology 73:546–555.
Mackenzie D.I., and J.A. Royle. 2005. Designing occupancy studies: general advice and
allocating survey effort. Journal of Applied Ecology 42:1105–1114.
McDaniel, G.W., K.S. McKelvey, J.R. Squires, and L.F. Ruggiero. 2000. Efficacy of lures and
hair snares to detect lynx. Wildlife Society Bulletin 28:119–123.
McKelvey, K.S., S.W. Buskirk, and C.J. Krebs. 2000. Theoretical insights into the population
viability of lynx. In L.F. Ruggiero, K.B. Aubry, S.W. Buskirk, G.M. Koehler, C.J. Krebs,
K.S. McKelvey, J.R. Squires, (Eds.). Ecology and Conservation of Lynx in the United
States (pp. 397–418). University Press of Colorado, Boulder, Colorado, USA.
McLaren, M.A., I.D. Thompson, and J.A. Baker. 1998. Selection of vertebrate wildlife indicators
for monitoring sustainable forest management in Ontario. The Forestry Chronicle
74:241–248.

129

McLellan, S.R., J.H. Vashon, E.L. Johnson, S.M. Crowley, and A.D. Vashon. 2018. Fisher
predation on Canada lynx in the Northeastern United States. The Journal of Wildlife
Management 82:1775–1783.
Menard, S. 2001. Applied Logistic Regression Analysis. Sage University Paper Series on
Quantitative Applications in the Social Sciences, 07–106. Thousand Oaks, California:
Sage.
Mills, L.S., K. Pilgrim, M.K. Schwartz, and K. McKelvey. 2000. Identifying lynx and other
North American felids based on MtDNA analysis. Conservation Genetics 1:285–289.
Monterroso, P., L.N. Rich, A. Serronha, P. Ferreras, and P.C. Alves. 2014. Efficiency of hair
snares and camera traps to survey mesopredator populations. European Journal of
Wildlife Research 60:279–289.
Moore, R.D., and J.S. Richardson. 2003. Progress towards understanding the structure, function,
and ecological significance of small stream channels and their riparian zones. Canadian
Journal of Forest Research 33:1349–1351.
Morgan, K.H., and M.A. Lashmar. 1993. Riparian habitat management and research.
Environment Canada, Fraser River Action Plan. Delta. British Columbia.
Moriarty, K.M., W.J. Zielinski, and E.D. Forsman. 2011. Decline in American marten occupancy
rates at Sagehen Experimental Forest, California. The Journal of Wildlife Management
75:1774–1787.

130

Moriarty, K.M., C.W. Epps, M.G. Betts, D.J. Hance, J.D. Bailey, and W.J., Zielinski. 2015.
Experimental evidence that simplified forest structure interacts with snow cover to
influence functional connectivity for Pacific martens. Landscape Ecology 30:1865–1877.
Mortelliti, A., A.M. Brehm, and B.E. Evans. 2022. Umbrella effect of monitoring protocols for
mammals in the Northeast US. Scientific Reports 12:1893.
Mowat, G., K.G. Poole, and M. O’Donoghue. 2000. Ecology of lynx in northern Canada and
Alaska. In L.F. Ruggiero, K.B. Aubry, S.W. Buskirk, G.M. Koehler, C.J. Krebs, K.S.
McKelvey, and J.R. Squires, (Eds.). Ecology and Conservation of Lynx in the United
States (pp. 265–306). University Press of Colorado, Boulder, Colorado, USA.
Murray, D.L. 2003. Snowshoe hare and other hares. In G. Feldhamer, B. Thompson, and J.
Chapman, (Eds.). Wild Mammals of North America: Biology, Management, and
Conservation, 2nd ed (pp. 147–175). The Johns Hopkins University Press, Baltimore,
Maryland, USA.
Murray D.L., and S. Boutin. 1991. The influence of snow on lynx and coyote movements: does
morphology affect behavior? Oecologia 88:463-469
Murray, D.L., S. Boutin, and M. O’Donoghue. 1994. Winer habitat selection by lynx and coyotes
in relation to snowshoe hare abundance. Canadian Journal of Zoology 72:1444–1451.
Murray, D.L, T.D. Steury, and 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.

131

Murray, D.L., J. Gobin, R. Scully, and D.H. Thorton. 2023. Conventional niche overlap
measurements are not effective for assessing interspecific competition. Frontiers in
Ecology and Evolution 11:1281108.
Naiman, R.J., H. Decamps, and M. Pollock. 1993. The role of riparian corridors in maintaining
regional biodiversity. Ecological Applications 3:209–212.
Nams, V.O., and M. Bourgeois. 2004. Fractal analysis measures habitat use at different spatial
scales: and example with American marten. Canadian Journal of Zoology 82:1738–1747.
Neilson, E.W., T. Avgar, A.C. Burton, K. Broadley, and S. Boutin. 2018. Animal movement
affects interpretation of occupancy models from camera trap surveys of unmarked
animals. Ecosphere 9:e02092.
Nellis, C.H. and L.B. Keith. 1968. Hunting activities and success of lynx in Alberta. Journal of
Wildlife Management. 36:718-722.
Nellis, C.H., S.P. Wetmore, and L.B. Keith. 1972. Lynx–prey interactions in central Alberta. The
Journal of Wildlife Management 36:320–328.
Noss, R.F. 1999. Assessing and monitoring forest biodiversity: A suggested framework and
indicators. Forest Ecology and Management 115:135–146.
O’Donoghue, M., E. Hofer, and F.I. Doyle. 1995. Predator versus predator. Natural History
104:6–9.
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.

132

O’Donoghue, M., S. Boutin, C.J. Krebs, D.L. Murray, and E.J. Hofer. 1998a. Behavioural
responses of coyotes and lynx to the snowshoe hare cycle. Oikos 82:169–183.
O’Donoghue, M., S. Boutin, C.J. Krebs, G. Zuleta, D.L. Murray, and E.J. Hofer. 1998b.
Functional responses of coyotes and lynx to the snowshoe hare cycle. Ecology 79:1193–
1208.
O’Donoghue, M., S. Boutin, D.L. Murray, C.J. Krebs, E.J. Hofer, U. Breitenmoser, C.
Breitenmoser–Wuersten, G. Zuleta, C. Doyle, and V.O. Nams. 2001. Coyotes and lynx.
In C. J. Krebs, S. Boutin, and R. Boonstra, (Eds.). Ecosystem Dynamics of the Boreal
Forest: The Kluane Project (pp. 276–323). Oxford University Press, New York, New
York, USA.
Organ, J.F., J.H. Vashon, J.E. McDonald Jr., A.D. Vashon, S.M. Crowley, W.J. Jakubas, G.J.
Matula, Jr., and A.L. Meehan. 2008. Within–stand selection of Canada lynx natal dens in
northwest, Maine, USA. The Journal of Wildlife Management 72:1514–1517.
Parker, G.R., J.W. Maxwell, and L.D. Morton. 1983. The ecology of lynx (Lynx Canadensis) on
Cape Breton Island. Canadian Journal of Zoology 61:770–786.
Parminter, J.V. 2000. From timber to biodiversity: the evolution of British Columbia's forest
inventory program. Forest History Today, Fall 2000: pp. 12–19.
Parsons, M.A., J.C. Lewis, B. Gardner, T. Chestnut, J.I. Ransom, D.O. Werntz, and L.R. Prugh.
2019. Habitat selection and spatiotemporal interactions of a reintroduced mesopredator.
The Journal of Wildlife Management 83:1172–1184.

133

Pauli, J.N., P.J. Manlick, J.M. Tucker, G. Bradley Smith, P.G. Jensen, and J.T. Fisher. 2022.
Competitive overlap between martens Martes americana and Martes caurina and fishers
Pekania pennanti: a range wide perspective and synthesis. Mammal review 52:392–409.
Pearce J., and S. Ferrier. 2000. Evaluating the predictive performance of habitat models
developed using logistic regression. Ecological Modeling 133:225–245.
Peers, M.J.L., Y.N. Majchrzak, A.K. Menzies, E.K. Studd, G. Bastille-Rousseau, R. Boonstra,
M. Humphries, T.S. Jung, A.J. Kenney, C.J. Krebs, D.L. Murray, and S. Boutin. 2020.
Climate change increases predation risk for a keystone species of the boreal forest.
Nature Climate Change s41558-020-00908-4.
Peers, M.J.L., D.H. Thornton, and D.L. Murray. 2012. Reconsidering the specialist–generalist
paradigm in niche breadth dynamics: Resource gradient selection by Canada lynx and
bobcat. PLoS ONE 7:e51488.
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 Bulletin 280:20132495.
Perault, D.R., and M.V. Lomolino. 2000. Corridors and mammal community structure across a
fragmented, old–growth forest landscape. Ecological Monographs 70:401–422.
Poole, K.G. 1994. Characteristics of an unharvested lynx population during a snowshoe hare
decline. The Journal of Wildlife Management 58:608–618.
Poole, K.G., L.A. Wakelyn, and P.N. Nicklen. 1996. Habitat selection by lynx in the Northwest
Territories. Canadian Journal of Zoology 74:845–850.

134

Poole, K.G. 1997. Dispersal patterns of lynx in the Northwest Territories. The Journal of
Wildlife Management 61:497–505.
Poole, K.G. 2003. A review of the Canada Lynx, Lynx canadensis, in Canada. Canadian Field–
Naturalist 117:360–376.
Popescu, V.D., P. de Valpine, and R.A. Sweitzer. 2014. Testing the consistency of wildlife data
types before combining them: the case of camera traps and telemetry. Ecology and
Evolution 4:933–943.
Powell, R.A., S.W. Buskirk, and W.J. Zielinski. 2003. Fisher (Martes pennanti) and Marten
(Martes americana). In G. Feldhamer, B. Thompson, and J. Chapman, (Eds.). Wild
mammals of North America: Biology, Management, and Conservation, 2nd ed. (pp. 635–
649). The Johns Hopkins University Press, Baltimore, Maryland, USA.
Pozzanghera, C.B., K.J. Sivy, M.S. Lindberg, and L.R. Pugh. 2016. Variable effects of snow
conditions across boreal mesopredator species. Canadian Journal of Zoology 94:697–705.
Priol, P.M., M.J. Mazerolle, L. Imbeau, P. Drapeau, C. Trudeau, and J. Ramiere. 2014. Using
dynamic N–mixture models to test cavity limitation on northern flying squirrel
demographic parameters using experimental nest box supplementation. Ecology and
Evolution 44:2165–2177.
Rich L.N., M.J. Kelly, R. Sollmann, A.J. Noss, L. Maffei, R.L. Arispe, A. Paviolo, C.D. De
Angelo, Y.E. Di Blanco, and M.S. Di Bitetti. 2014. Comparing capture–recapture, mark–
resight, and spatial mark–resight models for estimating puma densities via camera traps.
Journal of Mammalogy 95:382–391.

135

Roberts, N.M., and S.M. Crimmins. 2010. Bobcat population status and management in North
America: evidence of large–scale population increase. The Journal of Fish and Wildlife
Management 1:169–174.
Rodgers, A.R., and A.R.E. Sinclair. 1997. Diet choice and nutrition of captive snowshoe hares
(Lepus americanus): Interactions of energy, protein, and plant secondary compounds.
Ecoscience 4:163–169.
Rohner C., F.L. Doyle, and J.N.M Smith. 2001. Great Horned Owls. In C.J. Krebs, S. Boutin,
and R. Boonstra, (Eds.). Ecosystem Dynamics of the Boreal Forest: The Kluane Project,
Chapter 15 (pp. 339–376). Oxford University Press, Oxford, UK.
Rosenzweig, M.L. 1981. A theory of habitat selection. Ecology 2:327–335.
Roth, J.D., J.D. Marshall, D.L. Murray, D.M. Nickerson, and T.P. Steury. 2007. Geographical
gradients in diet affect population dynamics of Canada lynx. Ecology 88:2736–2743.
Royle, J.A. 2004. N–mixture models for estimating population size from spatially replicated
counts. Biometrics 60:108e115.
Royle, J.A., and R.M. Dorazio. 2006. Hierarchical models of animal abundance and occurrence.
Journal of Agricultural, Biological, and Environmental Statistics 11:249–263.
Saunders, J.K. 1963. Food habits of the lynx in Newfoundland. The Journal of Wildlife
Management 27:384–390.
Schmidt, K., and R. Kowalczyk. 2006. Using scent–marking stations to collect hair samples to
monitor Eurasian lynx populations. Wildlife Society Bulletin 34:462–466.

136

Scully, A.E., S. Fisher, D.A.W. Miller, and D.H. Thornton. 2018. Influence of biotic interactions
on the distribution of Canada lynx (Lynx canadensis) at the southern edge of their range.
Journal of Mammalogy 99:760–772.
Sheriff, M.J., C.J. Krebs, and R. Boonstra. 2009. The sensitive hare: Sublethal effects of predator
stress on reproduction in snowshoe hares. Journal of Animal Ecology 78:1249–1258.
Sheriff, M.J., C.J. Krebs, and R. Boonstra. 2010. The ghosts of predators past: Population cycles
and the role of maternal programming under fluctuating predation risk. Ecology 91:2983–
2994.
Sheriff, M.J., C.J. Krebs, and R. Boonstra. 2011. From pattern to process: How fluctuating
predation risk impacts the stress axis of snowshoe hares during the 10–year cycle.
Oecologia 166:593–605.
Sibly, R.M., H.M.R. Nott, and D.J. Fletcher. 1990. Splitting behaviour into bouts. Animal
Behaviour 39:63–69.
Sinclair, A.R.E., C.J. Krebs, J.N.M. Smith, and S. Boutin. 1988. Population biology of snowshoe
hares. III. Nutrition, plant secondary compounds and food limitation. Journal of Animal
Ecology 57:787–806.
Slough, B.G., and G. Mowat. 1996. Lynx population dynamics in an untrapped refugium. The
Journal of Wildlife Management 60:946–961.
Squires, J.R., N.J. DeCesare, J.A. Kolbe, and L.F. Ruggiero. 2008. Hierarchical den selection of
Canada lynx in Western Montana. 72:1497–1506.

137

Squires, J.R., and T. Laurion. 2000. Lynx home range and movements in Montana and
Wyoming: Preliminary results. In L.F. Ruggiero, K.B. Aubry, S.W. Buskirk, G.M.
Koehler, C.J. Krebs, K.S. McKelvey, and J.R. Squires, (Eds.). Ecology and Conservation
of Lynx in the United States (pp. 337–350). University Press of Colorado, Boulder,
Colorado, USA.
Squires, J.R., L.E. Olson, D.L. Turner, N.J. DeCesare, and J.A. Kolbe. 2012. Estimating
detection probability for Canada lynx Lynx canadensis using snow-track surveys in the
northern Rocky Mountains, Montana, USA. Wildlife Biology 18:215–224.
Steenweg, R., et al. 2017. Scaling–up camera traps: monitoring the planet’s biodiversity with
networks of remote sensors. Frontiers in Ecology and the Environment 15:26–34.
Stewart, F.E.C, J.T. Fisher, A.C. Burton, and J. Volpe. 2018. Species occurrence data reflect the
magnitude of animal movements better than the proximity of animal space use.
Ecosphere 9:e02112.
Stewart, F.E.C., J.P. Volpe, and J.T. Fisher. 2019. The debate about bait: A red herring in
wildlife research. The Journal of Wildlife Management 83:985–992.
Strohm, S., and R. Tyson. 2009. The effect of habitat fragmentation on cyclic population
dynamics: a numerical study. Bulletin of Mathematical Biology 71:1323–1348.
Suffice, P., M. Cheveau, L. Imbeau, M.J. Mazerolle, H. Asselin, and P. Drapeau. 2020. Habitat,
climate, and fisher and marten distributions. The Journal of Wildlife Management
84:277–292.

138

Sun, C., J.M. Burgar, J.T. Fisher, and A.C. Burton. 2022. A cautionary tale comparing spatial
count and partial identity models for estimating densities of threatened and unmarked
populations. Global Ecology and Evolution 38:e02268.
Thomas, J.P., M.L. Reid, R.M.R. Barclay, and T.S. Jung. 2019. Salvage logging after an insect
outbreak reduces occupancy by snowshoe hares (Lepus americanus) and their primary
predators. Global Ecology and Conservation 17:e00562.
Thompson, I.D., J. Fryxell, and D.J. Harrison. 2012. Improved insights into the use of habitat by
American martens. Pages 209–230 in K.B. Aubry, W.J. Zielinski, M.G. Raphael, G.
Proulx, and S.W. Buskirk. Biology and Conservation of Martens, Sables, and Fishers: A
New Synthesis. 1st ed. Cornell University Press, Ithaca, New York, USA.
Tigner, J., E.M. Bayne, and S. Boutin. 2015. American marten respond to seismic lines in
Northern Canada at two spatial scales. PLoS ONE 10(3):e0118720.
Todd, N.B. 1962. Inheritance of the catnip response in domestic cats. Journal of Heredity 53:54–
56.
Tucker, A.O., and S.S. Tucker. 1988. Catnip and the catnip response. Economic Botany 42:214–
231.
Tyson, R., S. Haines, and K.E. Hodges. 2010. Modelling the Canada lynx and snowshoe hare
population cycle: the role of specialist predators. Theoretical Ecology 3:97–111.
Uenoyama, R., T. Miyazaki, J.L. Hurst, R.J. Beynon, M. Adachi, T. Murooka, I. Onoda, Y.
Miyazawa, R. Katayama, T. Yamashita, S. Kaneko, T. Nishikawa, and M. Miyazaki.

139

2021. The characteristic response of domestic cats to plant iridoids allows them to gain
chemical defense against mosquitoes. Science Advances 7:eabd9135.
Vashon, J.H., A.L. Meehan, W.J. Jakubas, J.F. Organ, A.D. Vashon, C.R. McLaughlin, G.J.
Matula Jr., and S.M. Crowley. 2008a. Spatial ecology of a Canada lynx population in
northern Maine. The Journal of Wildlife Management 72:1479–1478.
Vashon, J.H., A.L. Meehan, J.F. Organ, W.J. Jakubas, C.R. McLaughlin, A.D. Vashon, and S.M.
Crowley. 2008b. Diurnal habitat relationships of Canada lynx in an intensively managed
private forest landscape in northern Maine. The Journal of Wildlife Management
72:1488–1496.
Vitense, K., A.J. Wirsing, R.C. Tyson, and J.J. Anderson. 2016. Theoretical impacts of habitat
loss and generalist predation on predator–prey cycles. Ecological Modeling 327:85–94.
Vogt, D., D. Hofer, A. Ryser, M. Kolliker, and U. Breitenmoser. 2016. Is there a trade–off
between scent marking and hunting behaviour in a stalking predator, the Eurasian lynx,
Lynx lynx? Animal Behaviour 117:59–68.
Ward, R.M.P., and C.J. Krebs. 1985. Behavioural responses of lynx to declining snowshoe hare
abundance. Canadian Journal of Zoology 63:2817–2824.
Wasser, L., R. Day, L. Chasmer, and A. Taylor. 2013. Influence of vegetation structure on lidar–
derived canopy height and fractional cover in forested riparian buffers during leaf–off
and leaf–on conditions. PLoS ONE 8:e54776.
Whittaker, M.E., and V.G. Thomas. 1983. Seasonal levels of fat and protein reserves of
snowshoe hares in Ontario. Canadian Journal of Zoology 61:1339–1345.

140

Wiens, J.A. 1977. On competition and variable environments: Populations may experience
"ecological crunches" in variable climates, nullifying the assumptions of competition
theory and limiting the usefulness of short–term studies of population patterns. American
Scientist 65:590–597.
Wiens, J.A., N.C. Stenseth, B. Van Horne, and R.A. Ims. 1993. Ecological mechanisms and
landscape ecology. Oikos 66:369–380.
Willmer, J.N.G., T. Puttker, and J.A. Prevedello. 2022. Global impacts of edge effects on species
richness. Biological Conservation 272:109654.
Wolfe, M.L., N.V. Debyle, C.S. Winchell, and T.R. McCabe. 1982. Snowshoe hare cover
relationships in northern Utah. The Journal of Wildlife Management 46:662–670.
Wolff, J.O. 1980. The role of habitat patchiness in the population dynamics of snowshoe hares.
Ecological Monographs 50:111–130.
Woollard, T.F., D.J. Harrison, E.M. Simons–Legaard, and K.E. Fagan. 2024. Functional
responses in American marten habitat selection indicate cumulative effects of progressive
habitat change. Ecosphere 15:e4715.
Yahner, R.H. 1988. Changes in wildlife communities near edges. Conservation Biology 2:333–
339.
Yan, C., N.C. Stenseth, C.J. Krebs, and Z. Zhang. 2013. Linking climate change to population
cycles of hares and lynx. Global Change Biology 19:3263–3271.

141

Appendix 1.1. Full model results, AICc scores, and AICc weights (w) for multinomial logistic
regression models representing habitat use of Canada lynx (Lynx canadensis) and American
marten (Martes americana) using GPS collars and camera traps in central British Columbia,
Canada, 2015–2016 and 2020–2021.
Model
LYNX WINTER
Cover Stand Scale
Disturbance + Cover
Topography + Cover
Riparian + Cover
Canopy Cover (50 m)
Disturbance + Topography 1
Topography
Disturbance + Topography 2
Disturbance
Null
LYNX SNOW-FREE
Cover Stand Scale
Topography + Cover
Riparian + Cover
Canopy Cover (50 m)
Disturbance + Cover
Disturbance + Topography 1
Disturbance + Topography 2
Topography
Disturbance
Null
MARTEN WINTER
Topography + Cover
Disturbance + Topography 1
Disturbance + Topography 2
Topography
Riparian + Cover
Cover Stand Scale
Canopy Cover (50 m)
Disturbance + Cover
Disturbance
Null

K

AICc

ΔAICc

w

5
4
7
4
4
5
4
4
3
1

8489.4
8741.5
8756.5
8786.3
8834.5
9311.3
9351.2
9353.6
9488.0
9528.4

0.0
252.1
267.1
296.9
345.1
821.9
861.8
864.2
998.6
1039.0

1.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00

5
7
4
4
4
5
4
4
3
1

8248.4
8265.8
8324.9
8354.4
8356.8
8614.3
8628.3
8680.0
8794.9
8816.0

0.0
17.5
76.5
106.0
108.4
365.9
379.9
431.6
546.5
567.6

1.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00

7
6
4
4
4
5
4
4
3
1

1789.3
1803.7
1807.8
1813.0
1954.5
1958.1
2010.5
2015.0
2062.7
2067.1

0.0
14.4
18.5
23.7
165.2
168.8
221.2
225.7
273.4
277.8

1.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00

142

Appendix 2.1. Full model rankings, AICc scores, and AICc weights (w) for N-mixture models
predicting abundance of Canada lynx (Lynx canadensis) and snowshoe hare (Lepus americanus)
at camera traps in central British Columbia, Canada, 2015–2016 and 2020–2022.

143

Appendix 2.2. AICc scores, and AICc weights (w) for spatial-capture mark-resight (SCMR)
models predicting Canada lynx (Lynx canadensis) density at camera traps in central British
Columbia, Canada, 2015–2016 and 2020–2022.
Model
2020
Detection probability- marking and
resighting
Forest age
Null
2022
Detection probability- marking and
resighting
Null
Forest age

K

AICc

ΔAICc

w

5
7
4

443.6
479.7
481.7

0.0
36.1
38.1

1.000
0.000
0.000

5
4
7

449.6
458.3
466.8

0.0
9.5
18.0

0.992
0.008
0.000

144

Appendix 3.1. Full model rankings, AICc scores, and AICc weights (w) for multinomial logistic
regression models representing habitat use and co-occurrence of Canada lynx (Lynx canadensis),
sympatric mesopredators, and prey at camera traps in central British Columbia, Canada during a
winter of low hare abundance (2020–2021).
HARE LOW WINTER
Model
COYOTE - LYNX
Riparian + Cover
Canopy Cover
Edge + Cover
Canopy Height + Riparian
Canopy Height + Ground Cover
Disturbance + Canopy Height
Prey- 3 Species
Disturbance
Prey- 2 species
null
Disturbance + Edge
WOLVERINE - LYNX
Riparian + Cover
Edge + Cover
Canopy Cover
Canopy Height + Ground Cover
Canopy Height + Riparian
Disturbance + Canopy Height
Prey- 2 species
Prey- 3 Species
null
Disturbance
Disturbance + Edge
FISHER - LYNX
Riparian + Cover
Canopy Height + Ground Cover
Canopy Cover
Canopy Height + Riparian
Edge + Cover
Disturbance + Canopy Height
Prey- 2 species
Prey- 3 Species
null
Disturbance

K

AICc

ΔAICc

w

4
5
5
4
4
3
4
4
3
1
3

111.3
118.7
120.0
123.8
132.5
133.0
143.6
148.1
148.9
152.5
153.9

0.0
7.4
8.7
12.5
21.2
21.7
32.3
36.8
37.6
41.2
42.6

0.961
0.024
0.012
0.002
0.000
0.000
0.000
0.000
0.000
0.000
0.000

4
5
5
4
4
3
3
4
1
4
3

143.8
146.1
146.2
153.9
157.0
160.5
181.2
181.9
182.2
182.8
182.9

0.0
2.3
2.4
10.1
13.2
16.7
37.4
38.1
38.4
39.0
39.1

0.621
0.192
0.182
0.004
0.001
0.000
0.000
0.000
0.000
0.000
0.000

4
4
5
4
5
3
3
4
1
4

148.1
149.0
150.4
150.7
151.0
156.3
176.5
176.5
179.3
177.8

0.0
0.9
2.3
2.6
2.9
8.2
28.4
28.4
31.2
29.7

0.406
0.255
0.128
0.109
0.096
0.007
0.000
0.000
0.000
0.000
145

Appendix 3.1 (continued)
Disturbance + Edge
HARE - LYNX
Riparian + Cover
Canopy Height + Riparian
Canopy Cover
Edge + Cover
Disturbance + Canopy Height
Canopy Height + Ground Cover
Disturbance
null
Disturbance + Edge
RED SQUIRREL - LYNX
Riparian + Cover
Canopy Height + Riparian
Canopy Cover
Edge + Cover
Canopy Height + Ground Cover
Disturbance + Canopy Height
Disturbance
Disturbance + Edge
null

3

178.4

30.3

0.000

4
4
5
5
3
4
4
1
3

139.5
144.9
145.8
146.4
155.7
156.0
178.7
180.6
181.5

0.0
5.4
6.3
6.9
16.2
16.5
39.2
41.1
42.0

0.874
0.060
0.037
0.028
0.000
0.000
0.000
0.000
0.000

4
4
5
5
4
3
4
3
1

140.3
148.0
149.2
150.0
156.7
160.1
176.8
181.7
183.9

0.0
7.7
8.9
9.7
16.4
19.8
36.5
41.4
43.6

0.961
0.020
0.011
0.008
0.000
0.000
0.000
0.000
0.000

146

Appendix 3.2. Full model rankings, AICc scores, and AICc weights (w) for multinomial logistic
regression models representing habitat use and co-occurrence of Canada lynx (Lynx canadensis),
sympatric mesopredators, and prey at camera traps in central British Columbia, Canada during a
winter of high hare abundance (2015–2016).
HARE HIGH WINTER
Model
COYOTE - LYNX
Canopy Cover
Riparian + Cover
Edge + Cover
Canopy Height + Ground Cover
Disturbance + Canopy Height
Canopy Height + Riparian
Disturbance
Disturbance + Edge
Prey- 2 species
Prey- 3 Species
null
WOLVERINE - LYNX
Canopy Cover
Edge + Cover
Riparian + Cover
Canopy Height + Ground Cover
Disturbance + Canopy Height
Canopy Height + Riparian
Disturbance + Edge
Disturbance
Prey- 3 Species
Prey- 2 species
null
HARE - LYNX
Riparian + Cover
Canopy Cover
Edge + Cover
Canopy Height + Riparian
Canopy Height + Ground Cover
Disturbance + Canopy Height
Disturbance
Disturbance + Edge
null
RED SQUIRREL - LYNX

K

AICc

ΔAICc

w

5
4
5
4
3
4
4
3
3
4
1

138.5
145.4
145.5
151.1
155.5
159.2
161.0
163.0
165.4
166.0
168.1

0.0
6.9
6.9
12.5
17.0
20.7
22.4
24.5
26.8
27.4
29.6

0.939
0.030
0.029
0.002
0.000
0.000
0.000
0.000
0.000
0.000
0.000

5
5
4
4
3
4
3
4
4
3
1

112.8
116.4
119.5
128.5
130.0
134.3
134.9
136.1
139.8
141.3
145.0

0.0
3.6
6.7
15.7
17.2
21.5
22.1
23.3
27.0
28.5
32.2

0.833
0.138
0.029
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000

4
5
5
4
4
3
4
3
1

144.3
150.6
156.0
157.7
164.4
166.7
167.3
174.2
177.6

0.0
6.3
11.7
13.4
20.1
22.4
23.0
29.9
33.3

0.956
0.040
0.003
0.001
0.000
0.000
0.000
0.000
0.000

147

Appendix 3.2 (continued)
Canopy Cover
Riparian + Cover
Disturbance
Disturbance + Canopy Height
Disturbance + Edge
Edge + Cover
Canopy Height + Ground Cover
Canopy Height + Riparian
null

5
4
4
3
3
5
4
4
1

151.1
158.4
160.6
170.8
171.0
173.9
177.2
178.3
185.0

0.0
7.3
9.5
19.7
19.9
22.8
26.1
27.2
33.9

0.966
0.025
0.008
0.000
0.000
0.000
0.000
0.000
0.000

148

Appendix 3.3. Full model rankings, AICc scores, and AICc weights (w) for multinomial logistic
regression models representing habitat use and co-occurrence of Canada lynx (Lynx canadensis),
sympatric mesopredators, and prey at camera traps in central British Columbia, Canada during a
snow-free period of low hare abundance (2020–2021).
HARE LOW SNOW-OFF
Model
COYOTE - LYNX
Edge + Cover
Riparian + Cover
Canopy Cover
Canopy Height + Riparian
Canopy Height + Ground Cover
Disturbance + Canopy Height
Prey- 3 Species
Disturbance + Edge
Prey- 2 species
Disturbance
null
WOLVERINE - LYNX
Riparian + Cover
Edge + Cover
Canopy Cover
Canopy Height + Ground Cover
Prey- 3 Species
Canopy Height + Riparian
Prey- 2 species
Disturbance
null
FISHER - LYNX
Edge + Cover
Riparian + Cover
Canopy Cover
Prey- 3 Species
Disturbance + Canopy Height
Prey- 2 species
Canopy Height + Ground Cover
Canopy Height + Riparian
Disturbance + Edge
null
Disturbance
HARE - LYNX

K

AICc

ΔAICc

w

5
4
5
4
4
3
4
3
3
4
1

123.6
131.2
133.6
135.8
137.1
141.0
143.1
145.3
147.6
151.5
152.9

0.0
7.6
10.0
12.2
13.5
17.4
19.5
21.7
24.0
27.9
29.3

0.969
0.021
0.006
0.002
0.001
0.000
0.000
0.000
0.000
0.000
0.000

4
5
5
4
4
4
3
4
1

108.2
111.0
112.8
117.3
117.4
117.6
118.3
132.0
133.0

0.0
2.8
4.6
9.1
9.2
9.4
10.1
23.8
24.8

0.719
0.181
0.074
0.008
0.007
0.007
0.005
0.000
0.000

5
4
5
4
3
3
4
4
3
1
4

145.7
146.5
147.3
151.9
154.4
154.6
154.7
158.2
168.6
169.7
169.7

0.0
0.8
1.6
6.2
8.7
8.9
9.0
12.5
22.9
24.0
24.0

0.451
0.310
0.201
0.021
0.006
0.005
0.005
0.001
0.000
0.000
0.000

149

Appendix 3.3 (continued)
Riparian + Cover
Edge + Cover
Canopy Cover
Disturbance + Canopy Height
Canopy Height + Riparian
Canopy Height + Ground Cover
Disturbance
null
Disturbance + Edge
RED SQUIRREL - LYNX
Edge + Cover
Riparian + Cover
Canopy Cover
Disturbance + Canopy Height
Canopy Height + Ground Cover
Canopy Height + Riparian
Disturbance + Edge
null
Disturbance

4
5
5
3
4
4
4
1
3

148.9
151.3
153.2
158.7
159.8
160.4
172.7
174.2
175.8

0.0
2.4
4.3
9.8
10.9
11.5
23.8
25.3
26.9

0.703
0.205
0.082
0.005
0.003
0.002
0.000
0.000
0.000

5
4
5
3
4
4
3
1
4

163.8
165.6
167.5
171.1
172.3
173.5
183.7
184.5
186.9

0.0
1.8
3.7
7.3
8.5
9.7
19.9
20.7
23.1

0.620
0.253
0.098
0.016
0.009
0.005
0.000
0.000
0.000

150

Appendix 3.4. Full model rankings, AICc scores, and AICc weights (w) for multinomial logistic
regression models representing habitat use and co-occurrence of Canada lynx (Lynx canadensis),
sympatric mesopredators, and prey at camera traps in central British Columbia, Canada during a
snow-free period of high hare abundance (2015–2016).
HARE HIGH SNOW-OFF
Model
COYOTE - LYNX
Disturbance
Riparian + Cover
Prey- 3 Species
Disturbance + Canopy Height
Canopy Cover
Edge + Cover
Canopy Height + Riparian
Canopy Height + Ground Cover
Disturbance + Edge
Prey- 2 species
null
HARE - LYNX
Edge + Cover
Disturbance + Edge
Canopy Cover
Riparian + Cover
Canopy Height + Ground Cover
Disturbance
Disturbance + Canopy Height
Canopy Height + Riparian
null
RED SQUIRREL - LYNX
Disturbance
Riparian + Cover
Canopy Cover
Edge + Cover
Disturbance + Canopy Height
Canopy Height + Ground Cover
Canopy Height + Riparian
Disturbance + Edge
null

K

AICc

ΔAICc

w

4
4
4
3
5
5
4
4
3
3
1

141.8
145.8
149.2
150.1
150.8
151.0
152.5
153.6
154.7
157.4
159.3

0.0
4.0
7.4
8.3
9.0
9.2
10.7
11.8
12.9
15.6
17.5

0.827
0.113
0.021
0.013
0.009
0.008
0.004
0.002
0.001
0.000
0.000

5
3
5
4
4
4
3
4
1

132.4
148.3
149.7
150.1
156.5
157.7
158.2
162.7
167.1

0.0
15.9
17.3
17.7
24.1
25.3
25.8
30.3
34.7

0.999
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000

4
4
5
5
3
4
4
3
1

151.6
159.4
161.5
162.5
165.5
166.5
166.9
168.6
172.1

0.0
7.8
9.9
10.9
13.9
14.9
15.3
17.0
20.5

0.967
0.020
0.007
0.004
0.001
0.001
0.000
0.000
0.000

151