Close encounters of the burned kind: Spatiotemporal effects of fire on habitat selection
strategies of woodland caribou (Rangifer tarandus caribou) during winter
by
Kelsey L.M. Russell
B.Sc., University of British Columbia, 2011

THESIS SUBMITTED IN PARTIAL FULFILLMENT OF
THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE
IN
NATURAL RESOURCES AND ENVIRONMENTAL STUDIES
(BIOLOGY)

UNIVERSITY OF NORTHERN BRITISH COLUMBIA
January 2018

© Kelsey L.M. Russell, 2018

ABSTRACT
Forest fire is the primary natural disturbance process influencing the distribution and
abundance of terrestrial lichens across ranges of woodland caribou (Rangifer tarandus
caribou), including the Klaza Caribou Herd in west-central Yukon. I used stand and
understory data to understand variation in the abundance of lichens in burns of various
ages. Focusing on the distribution of individual caribou, I used a dataset of GPS collar
locations to examine resource selection on the winter range and within burns. Results
suggested that burns provided suboptimal habitat for the KCH until 50 years post-fire;
however, analyses focused on the use of burned habitat indicated that they regularly
encountered burns and opportunistically used remnant lichen within the burn perimeter.
The relationship between caribou and burned landscapes is complex and non-linear
indicating that wildlife managers should look beyond burn age to account for the effects
of fire on the availability and quality of caribou habitat.

i

TABLE OF CONTENTS
Abstract ............................................................................................................................ i
Table of Contents .............................................................................................................ii
List of Figures ..............................................................................................................iv
List of Tables............................................................................................................... vii
Acknowledgements ......................................................................................................... x
Chapter One: Thesis Introduction ................................................................................... 1
Research objectives ..................................................................................................... 4
Study area.................................................................................................................... 5
Klaza caribou ............................................................................................................ 6
Forest fire history ...................................................................................................... 9
Chapter Two: Post-fire Dynamics of Forage Lichens on the Klaza Caribou Herd
Range............................................................................................................................ 12
Introduction ................................................................................................................ 12
Methods ..................................................................................................................... 15
Site selection and plot layout .................................................................................. 15
Sampling design ..................................................................................................... 16
Data analyses ......................................................................................................... 19
Results ....................................................................................................................... 22
General plot characteristics .................................................................................... 22
Patterns in the abundance of forage lichens post-fire ............................................. 25
Statistical models .................................................................................................... 28
Discussion.................................................................................................................. 33
Lichen succession and abundance post-fire ........................................................... 34
Influence of stand characteristics on lichen volume ................................................ 36
Using biomass thresholds for changes in caribou selection strategies ................... 38
Conclusions ............................................................................................................ 41
Chapter Three: Variation in Habitat Selection Strategies of Woodland Caribou in
Burns ............................................................................................................................. 42

ii

Introduction ................................................................................................................ 42
Methods ..................................................................................................................... 45
Caribou locations .................................................................................................... 46
RSF and burn-focused model inputs ...................................................................... 47
Winter habitat selection: Resource selection function models ................................ 50
Burn use: Logistic regression models ..................................................................... 53
Results ....................................................................................................................... 56
Resource selection function models ....................................................................... 59
Burn-focused logistic models .................................................................................. 67
Discussion.................................................................................................................. 70
Variation of habitat selection strategies in burns .................................................... 71
Close encounters: How Klaza caribou use burns during winter .............................. 74
Conclusions ............................................................................................................... 78
Chapter Four: Thesis Conclusions ................................................................................ 80
Summary.................................................................................................................... 80
Management considerations and future research ...................................................... 83
Literature Cited .............................................................................................................. 86
Appendix A

Field data collection details .................................................................... 99

Appendix B

Lichen model outputs and prediction .................................................... 103

Appendix C

Lichen biomass estimates on winter ranges of caribou ........................ 109

Appendix D

RSF model outputs............................................................................... 112

Appendix E

Burn-focused logistic model outputs..................................................... 117

iii

List of Figures
Figure 1. The study area, located within the annual range of the Klaza caribou
herd, west-central Yukon................................................................................. 8
Figure 2. Percent of area burned by burn age in the annual range of the Klaza
caribou herd, as of 2015. The numbers above the bars represent the
number of burns within each burn age category (Yukon Fire History
database 2016). ............................................................................................ 10
Figure 3. Recent burns near the study area, located within the annual and late
winter range of the Klaza caribou herd, west-central Yukon (range
polygons developed by Environment Yukon, 2016). ..................................... 11
Figure 4. Schematic diagram of the plot layout for vegetation sampling, with the
quad arrow representing plot centre, trees representing the location of
canopy openness measurements, and the central circle representing the
area measured for tall shrub cover. ............................................................... 16
Figure 5. Burns and vegetation plots sampled in summer 2014 within the Klaza
caribou range, west-central Yukon. ............................................................... 24
Figure 6. Sequence of post-fire lichen succession observed in the Klaza caribou
range, west-central Yukon, summer 2014. .................................................... 25
Figure 7. Coefficients for the count portion of the top-ranked zero-inflated Poisson
model (∆I AICc ≤2) explaining lichen volume (counts) in the Klaza
caribou range, west-central Yukon. All variables defined in Table 2. ............ 30
Figure 8. Coefficients for the binary portion of the top-ranked zero-inflated Poisson
model (∆I AICc ≤2) representing influences on the probability of lichen
absence in the Klaza caribou range, west-central Yukon. All variables
defined in Table 2. ......................................................................................... 31
Figure 9. Biomass of forage lichens within different-aged stands throughout the
Klaza caribou range, west-central Yukon. The two threshold values
(selection = 1,250 kg/ha; avoidance = 400 kg/ha) represent biomass
levels at sites avoided and selected by caribou in the neighbouring
Nelchina herd (Collins et al. 2011). ............................................................... 40
Figure 10.
Example of the sampling design used for the burn-focused models.
Habitat selection units (HSUs) were a grid overlaid on recent (≤50 years
old), occupied burns (plus a 500-m buffer) in the Klaza caribou range,
west-central Yukon. ....................................................................................... 54
Figure 11.
Weighted count of each candidate RSF model, as defined in Table
10, selected as a top model (∆AIC ≤2) for the early winter season
(2012–2016). The number of collared individuals in the Klaza caribou
herd in each season-year is indicated above the bars. ................................. 60
Figure 12.
Weighted count of each candidate RSF model, as defined in Table
10, selected as a top model (∆AIC ≤2) for the late winter season (2013–

iv

2016). The number of collared individuals in the Klaza caribou herd in
each season-year is indicated above the bars. ............................................. 60
Figure 13.
Individual and pooled (across all animals) seasonal selection ratios
(used versus available locations) of burn attributes (burn age, burn
perimeter, and burn size) for Klaza caribou during early (a) and late (b)
winter, 2012–2016. Selection ratios greater than one indicate selection,
whereas values less than one indicate avoidance. ....................................... 66
Figure 14.
Relative mean proportion and duration of burn use by collared
caribou during individual early (EW) and late (LW) winter seasons by
Klaza caribou, west-central Yukon, 2012–2016. All values were scaled
between 0 and 1 using a linear stretch. ......................................................... 67
Figure 15.
GPS collar locations of Klaza caribou during winter (2012–2016) in
relation to recent burns (≤50 years old) and lichen density, west-central
Yukon. ........................................................................................................... 77
Figure 16.
Difference in the observed and predicted volume of forage lichens
generated using the top-ranked zero-inflated Poisson model for
vegetation plots in the Klaza caribou herd range, west-central Yukon. A
value of zero suggests perfect prediction, whereas negative values
indicate model under-prediction and positive values indicate overprediction..................................................................................................... 107
Figure 17.
Difference in the observed and predicted cover of forage lichens
generated using the top-ranked fractional logit models for vegetation
plots in the Klaza caribou herd range, west-central Yukon. A value of
zero suggests perfect prediction, whereas negative values indicate
model under-prediction and positive values indicate over-prediction.
Parameters are defined in Table 2; model statistics are given in Table 6. .. 108
Figure 18.
Averaged, unweighted selection coefficients with 95% confidence
intervals (representing variability of averaged coefficients) for early
winter resource selection function models of individual Klaza caribou.
For very small values, positive and negative (+/-) signs above coefficient
values are provided for clarity. .................................................................... 113
Figure 19.
Averaged unweighted selection coefficients with 95% confidence
intervals (representing variability of averaged coefficients) for late winter
resource selection function models of individual Klaza caribou. For very
small values, positive and negative (+/-) signs above coefficient values
are provided for clarity. ................................................................................ 114
Figure 20.
Number of years during the study period (2012–2016) that individual
Klaza caribou with three years of data demonstrated positive or negative
responses to burn characteristics during the early winter season. .............. 115

v

Figure 21.
Number of years during the study period (2012–2016) that individual
Klaza caribou with three years of data demonstrated positive or negative
responses to burn characteristics during the late winter season. ................ 116

vi

List of Tables
Table 1. Total area of recent burns (≤50 years) in the Klaza caribou herd’s annual
and late winter range (Environment Yukon 2016), during the study
period, 2011–2015. ....................................................................................... 10
Table 2. Variables used in count models of volume of forage lichens in burns
surveyed on the Klaza caribou range, west-central Yukon, 2014.................. 22
Table 3. Vegetation sampling in burns by age class (based on both tree core
results and existing burn mapping), on the Klaza caribou range, westcentral Yukon, 2014. ..................................................................................... 23
Table 4. Mean (standard error, SE) cover, volume, and biomass of forage lichens
in vegetation plots in burns by 15-year intervals, Klaza caribou range,
west-central Yukon, summer 2014. Plots aged 100–149 and ≥150 years
were grouped together rather than by 15-year intervals due to low
sample sizes. ................................................................................................ 26
Table 5. Non-parametric pairwise comparison (using Tukey’s method of
adjustment for multiple comparisons) of lichen volume in different-aged
vegetation plots in burns in the Klaza caribou range, west-central Yukon.
P < 0.05 indicates a significant difference (in bold) between age groups. ..... 27
Table 6. Summary of model selection statistics for the most parsimonious models
(∆i AICc ≤2) of volume and cover of forage lichens in the Klaza caribou
range, west-central Yukon. Parameters are defined in Table 2. .................... 29
Table 7. R2 values (observed versus predicted values) and test statistics from
Wilcoxon sign-rank equality tests for matched pairs for the analysis of
differences between observed and predicted lichen volume and cover
values for the top-ranked (∆i AICc differed by ≤2 points) count and
fractional models. The test revealed no statistically significant
differences in the distribution of observed and predicted values when P
> 0.05. ........................................................................................................... 33
Table 8. Environmental variables used in the candidate RSF and logistic models
for early and late winter distribution and use of burns by the Klaza
caribou herd, west-central Yukon. A single asterisk (*) indicates
variables used only in the RSF models, and ** indicates variables used
only in logistic models. .................................................................................. 50
Table 9. Summary of GPS-collar locations within 500-m of burns in early and late
winter for the Klaza caribou herd, west-central Yukon, 2012–2016. ............. 58
Table 10. Candidate resource selection function models for the Klaza caribou herd
of west-central Yukon, during the early and late winter, 2012–2016. GM
= general model, BM = burn model. .............................................................. 59
Table 11. Weighted counts for each coefficient by early winter season (EW) from
top-ranked (∆AIC ≤2) resource selection models for individual caribou in

vii

the Klaza herd of west-central Yukon. For each year, the number under
the – indicates the weighted count of individual top models that showed
significant avoidance for that parameter; the number under the +
indicates the weighted count of individual top models that showed
significant selection for that parameter. The count of significant β
coefficients (P<0.05) in each year is shown in bold italics for each – or +
coefficient. ..................................................................................................... 63
Table 12. Weighted counts for each coefficient by late winter season (LW) from
top-ranked (∆AIC ≤2) resource selection models for individual caribou in
the Klaza herd of west-central Yukon. For each year, the number under
the – indicates the weighted count of individual top models that showed
significant avoidance for that parameter; the number under the +
indicates the weighted count of individual top models that showed
significant selection for that parameter. The count of significant β
coefficients (P<0.05) in each year is shown in bold italics for each – or +
coefficient. ..................................................................................................... 64
Table 13. Mean Receiver Operating Characteristic (ROC) scores (area under the
curve) for top-ranked (highest AICw) resource selection function (RSF)
models for individual Klaza caribou, 2012–2016. .......................................... 65
Table 14. Candidate logistic regression models used to differentiate Habitat
Selection Units (HSU) with relatively high and low use by monitored
caribou of the Klaza herd during early and late winter, west-central
Yukon, 2012–2016. ....................................................................................... 68
Table 15. Summary of statistics used to select the most parsimonious logistic
regression models for understanding caribou use of Habitat Selection
Units in the Klaza caribou herd range, west-central Yukon. .......................... 69
Table 16. Area under the curve for Receiver Operating Characteristic (ROC)
scores calculated for the top models representing caribou use of Habitat
Selection Units (HSUs) in the Klaza caribou herd range. The threshold
percentiles that differentiated high and low use HSUs are indicated at
the bottom along with the number of HSUs used in each season-year.
LW = late winter, EW = early winter. ............................................................. 70
Table 17. Fire history, terrain, and vegetation variables recorded during field
sampling in 2014 within the Klaza caribou herd annual range, westcentral Yukon. ............................................................................................. 100
Table 18. Protocol for determining the age class of vegetation plots as determined
using mapped burn age and/or tree core ages from trees sampled in
field plots within the Klaza caribou herd annual range, west-central
Yukon. ......................................................................................................... 101

viii

Table 19. Average values (with standard error) of the percent cover of non-lichen
ground cover in vegetation plots in burns by 15-year intervals, Klaza
caribou range, west-central Yukon, summer 2014. Plots aged 100–149
and ≥150 years were grouped together rather than by 15-year intervals
due to low sample sizes. ............................................................................. 102
Table 20. Summary of model selection statistics for candidate a priori models used
to select the most parsimonious count model for understanding the
volume of forage lichens in the Klaza caribou herd range, west-central
Yukon. Parameters are defined in Table 2. ................................................. 104
Table 21. Summary of model selection statistics for candidate a priori models used
to select the most parsimonious fractional logit model (binomial family,
logit link function) for understanding the cover of forage lichens in the
Klaza caribou herd range, west-central Yukon. Parameters are defined
in Table 2. ................................................................................................... 104
Table 22. Coefficients and statistical parameters generated from the top-ranked (∆i
AICc differed by ≤2 points) zero-inflated Poisson (ZIP) regression models
for Cladina spp. and Cetraria spp. lichen volume (dm3/m2) in the Klaza
caribou herd annual range, west-central Yukon. Coefficients and
statistical parameters for the count and binary parts of the ZIP models
are distinguished. ........................................................................................ 105
Table 23. Coefficients and statistical parameters generated from the top ranked
((∆i AICc differed by ≤2 points) fractional logit models for the prediction of
Cladina spp. and Cetraria spp. lichen cover (%, expressed as a
proportion) in the Klaza caribou herd annual range, west-central Yukon. ... 106
Table 24. Lichen biomass of various winter range types, both used and unused
(random) by Rangifer (caribou or reindeer; adapted from Barrier 2011). .... 110
Table 25. Area under the Receiver Operating Characteristic (ROC) curve scores
calculated for each percentile differentiating Habitat Selection Units with
low versus high use by Klaza caribou, west-central Yukon. Values in
bold indicate the highest scores and the corresponding percentile was
used to set the threshold for that season-year. ........................................... 118
Table 26. Coefficients and statistical parameters generated from the top-ranked (∆i
AICc differed by ≤2 points) logistic models representing the relative use
of Habitat Selection Units by caribou during early winter in the Klaza
herd annual range, west-central Yukon (2012–2016). ................................ 119
Table 27. Coefficients and statistical parameters generated from the top-ranked (∆i
AICc differed by ≤2 points) logistic models representing the relative use
of Habitat Selection Units by caribou during late winter in the Klaza herd
annual range, west-central Yukon (2013–2016).......................................... 121

ix

ACKNOWLEDGEMENTS
A number of organizations and individuals have supported this research, and I would
like to take the time to thank them: EDI Environmental Dynamics Inc., Environment
Yukon, Casino Mining Corporation, especially Jesse Duke, Mary Mioska, and Scott
Casselman, the Yukon Fish and Wildlife Enhancement Trust, University of Northern
British Columbia (UNBC), Natural Sciences and Engineering Research Council
(NSERC), Association of Canadian Universities for Northern Studies (ACUNS), W.
Garfield Weston Foundation, International Association of Impact Assessment, Yukon
Research Centre, and the Yukon Foundation. I have a lot of love for the EDI
Environmental Dynamics Inc. Whitehorse office, where I have been continually
supported and encouraged. Special thanks to Mike Setterington for letting me run free
and listening to me drone on about modeling.
I must express the utmost gratitude to my supervisor, Dr. Chris Johnson, for his time,
thoughtfulness, and understanding. Chris always had time for me and his patience and
compassion never seemed to run out. His research and knowledge drew me to Prince
George and UNBC and his passion, professionalism, and openness made my
experience there a pleasure.
Thanks to Dr. Troy Hegel and Dr. Katherine Parker of my supervisory committee for
their contributions to this research. I really appreciated Troy’s advice and time while
working on this research up in Whitehorse, and his expertise was essential throughout
this process. I am also grateful for the wealth of knowledge Kathy possesses and
thoughtfully shared.

x

Special thanks to my over-qualified field assistants, Graeme Pelchat, Don Russell, and
Kyle Russell, my patient helicopter pilots, Melvin Lagersson (Capital Helicopters) and
Scott Smith (Trans North Helicopters).
I must recognize the L-Roc to my K-fold, my sounding board, top supporter, lab mate,
and friend, Laura Grant. I have never met anyone with such incredible determination
and follow-through; you are truly an inspiration. To my parents: thank you for raising me
in a place where caribou still roamed in our backyard, and for always piquing my
curiosity. None of this would have been possible without my partner, Calef Staples.
Your love and encouragement throughout the years has been unrelenting and your
work ethic is inimitable. I will forever be in awe of your passion for the people and things
you love.

xi

CHAPTER ONE: THESIS INTRODUCTION
Populations of woodland caribou (Rangifer tarandus caribou) are of increasing
conservation concern. Habitat loss and alteration, natural and anthropogenic
disturbance, and predation are believed to have contributed to population declines
across much of their distribution (Dyer et al. 2002; Vors et al. 2007; Festa-Bianchet et
al. 2011; Environment Canada 2012). In the Yukon, the Northern Mountain (NM)
population (i.e., Designatable Unit; DU) of woodland caribou is currently assessed as
Special Concern by the Committee on the Status of Endangered Wildlife in Canada
(COSEWIC 2014). Within this population, each herd is subject to a unique suite of
threats that act cumulatively to influence population dynamics.
Caribou in the NM population in the Yukon are distributed throughout mountainous and
valley habitats, generally south of 65 degrees latitude (COSEWIC 2014). These herds
are believed to have one of two wintering strategies: alpine or forest-dwelling,
depending on snow conditions within their range (Kuzyk et al. 1999). During the winter,
these caribou forage primarily on terrestrial lichens from the Cladina, Cetraria, and
Cladonia genera (henceforth referred to as forage lichens), which are found in sprucepine forests or wind-swept alpine or subalpine slopes (Heard and Vagt 1998; Johnson
et al. 2004a; Jones et al. 2007). Forest fires in wintering areas can reduce the
availability of slow-growing lichens, resulting in changes to the distribution of caribou
(Scotter 1970; Thomas et al. 1996; Environment Canada 2012; Anderson and Johnson
2014).

1

The effects of forest fire on caribou winter range have been studied extensively.
Previous research suggested that, in the short-term, forest fire decreases the overall
availability of slow-growing forage lichens (largely from the Cladina genus). Additionally,
the energy costs of movement and foraging increase due to higher amounts of downed
woody debris and greater snow depths resulting from reduced canopy interception
(Schaefer and Pruitt 1991). In the long-term, however, fire may play an important role in
maintaining the diversity and abundance of lichen forage (Klein 1982; Schaefer and
Pruitt 1991).
Klein (1982) stressed the importance of differentiating between short- and long-term
effects of forest fires on caribou habitat, suggesting that fire may reduce an ecosystem’s
potential to support caribou for approximately 50 years, but over the long-term, it likely
benefits caribou habitat by resetting forest succession (Coxson and Marsh 2001).
Similarly, Skoog (1968) suggested that although forest fires can destroy large tracts of
winter habitat, theoretically reducing the potential carrying capacity of the overall range,
caribou populations occur at much lower densities than the maximum determined by the
abundance of forage lichens. Thus, the abundance of forage is not always a primary
cause of population declines. However, more recent research has suggested that with
the onset of global climate change, forest fire intervals may become shorter, further
decreasing the area of mature spruce–lichen forest and the availability of terrestrial
lichens (Stocks et al. 1998; Rupp et al. 2006).
The question of when a burn is no longer a ‘burn’ to caribou is especially important
when assessing the availability of habitat across a herd’s range (Francis et al. 2013).
The consensus is that 50 years is an appropriate benchmark for this relationship (Joly et
2

al. 2003; Dalerum et al. 2007; Sorensen et al. 2008). In 2008, Environment Canada
published its Scientific Review for the Identification of Critical Habitat for Woodland
Caribou (Rangifer tarandus caribou), Boreal Population, in Canada. In that report, a
herd’s total range disturbance was calculated as the area recently burned (in the last 50
years) plus the area of anthropogenic (human-caused) disturbance (Environment
Canada 2008). In 2011, the area considered to be ‘recently burned’ was changed to any
burn that was 40 years or younger; however, this was not a result of new information
regarding habitat suitability, rather it was a constraint imposed by mapping limitations
across jurisdictions (Environment Canada 2011).
Environment Canada (2011) developed a quantitative relationship between total range
disturbance and the probability of persistence of a population of boreal caribou (e.g.,
35% total range disturbance = 60% probability of persistence). Although their report was
focused on boreal caribou, the method of linking a herd’s persistence to range
disturbance has been applied to other populations (e.g., Reid et al. 2013; Environment
Canada 2014), and as suggested by Francis et al. (2013), it is important to better
understand these dynamics as they relate to specific herds and other ecotypes of
caribou.
When considering long-term range conservation and management for caribou, it is
important to quantify and qualify the cumulative change in habitat resulting from natural
disturbance, such as forest fires, and human activities. When considering forest fires,
habitat loss and recovery can be quantified by determining the time and conditions
required for forage lichens to recover post-fire. The magnitude of the effect of forest
fires can be qualified by exploring the suite of behavioural responses caribou display
3

when they encounter burns within their range. This information can then be related to
the predicted effects of environmental change, proposed resource development, or
other anthropogenic disturbances.

Research objectives
The Klaza Caribou Herd (KCH) in west-central Yukon is an example of a NM herd that
has increasingly been subjected to anthropogenic and natural disturbance. My thesis is
focused on this herd, and is part of a larger initiative by Environment Yukon to conduct a
range-wide assessment of the KCH “…to assess risk to population viability, define
management objectives, and identify actions to meet the objectives for focal wildlife
species” (Francis et al. 2013, p. vii). Francis et al. (2013) identified forest fire as a
natural factor that can cumulatively influence caribou habitat, and recommended the
development of a quantitative threshold for ‘recently burned habitat’ specific to the
winter range of the KCH.
The overall goal of this research is to understand the contribution of forest fires to
cumulative habitat change across the winter range of woodland caribou. Although it is
relatively simple to quantify total area burned, there is considerable uncertainty in our
understanding of the ecological conditions that allow caribou to use or re-occupy
habitats that have been changed by fire. I integrated field vegetation data collected in
the summer of 2014 with global positioning system (GPS) collar data for caribou to
investigate the relationship between habitat selection and burns during the early and
late winter seasons. Specifically, I addressed the following research objectives:
1. Examine the environmental characteristics that influence the variability of
lichen abundance within burns.
4

2. Examine range-wide habitat selection strategies during the early and late
winter as well as those strategies employed by caribou that occupied recent
burns.

Study area
The study area is focused on the annual range of the KCH (Environment Yukon 2014)
in west-central Yukon. The range encompasses approximately 11,095 km2 and is
roughly bounded by the Yukon River to the north and east, the White River to the west,
and the Nisling River to the south (Figure 1). This area lies primarily within the Klondike
Plateau Ecoregion, and to a lesser extent, the Yukon Plateau-Central Ecoregion (Yukon
Ecoregions Working Group [YEWG] 2004; Soil Landscapes of Canada Working Group
2013). The area is characterised by smooth, rolling hills often with ridges and tors
protruding from ridge tops, and narrow, V-shaped valleys (YEWG 2004). Due to the
region’s fluvial origin and absence of glacial scouring, there are very few lakes (YEWG
2004). The Klondike Plateau is also within a zone of extensive discontinuous
permafrost, with 50 to 90% of the soils containing permafrost (McKillop 2013),
especially in valley bottom sediment deposits and upland slopes (YEWG 2013).
Elevations range from 300 m at the Yukon River to approximately 2,000 m at Apex
Mountain, which is located roughly in the centre of the KCH annual range.
The climate within the KCH range is largely continental, with very cold winters and warm
summers (YEWG 2013). Forests of the Klondike Plateau are dominated by white (Picea
glauca) and black spruce (P. mariana) stands, either unmixed or mixed with balsam
poplar (Populus balsamifera; typical of floodplains), paper birch (Betula papyrifera), or
trembling aspen (Populus tremuloides; YEWG 2013). Unlike habitats typical of

5

woodland caribou herds in the southern Yukon, lodgepole pine (Pinus contorta) are
largely absent from the area, except near the Yukon River.
The study area also hosts a number of existing and potential human activities, including
placer mining, mineral exploration, a proposed hard rock mine using existing and new
access roads, existing resource access roads in the north (via the Yukon River) and
southeast parts of the range, hunting, trapping, and year-round motorized recreation.
Klaza caribou
The KCH are part of the NM DU of woodland caribou (COSEWIC 2011), which were
assessed as Special Concern by COSEWIC in 2002 and again in 2014, and federally
listed on Schedule I of the Species at Risk Act in 2005. NM caribou are generally
distinguished from other DUs by seasonal altitudinal migrations and a diet composed
primarily of terrestrial lichens (Heard and Vagt 1998; Environment Canada 2012).
During winter, NM caribou access terrestrial lichens by cratering through snow with their
hooves. NM caribou generally spend the winter season at low elevations in mature
lodgepole pine or black spruce stands or on windswept slopes, and move to higher
elevations for spring and summer (COSEWIC 2011). Klaza caribou winter primarily in
the alpine and subalpine, with the herd’s winter range characterized as a relatively lowsnow area due to its location in the snow-shadow of the St. Elias mountain range
(Farnell et al. 1991; Kuzyk et al. 1999).
Study of the KCH, formerly known as the Klotassin herd (Jingfors 1989), began in the
mid-1980s when the Casino Trail was being constructed to access the Casino Mine
(Farnell et al. 1991). Between 1987 and 1990, 17 very high frequency (VHF) collars

6

were deployed on female caribou in the herd (Farnell et al. 1991), with additional collars
deployed and active until 2000 as part of a study on the neighbouring Aishihik caribou
herd. The Aishihik herd’s range overlaps the southern portion of the KCH range, near
the Nisling River (Hayes et al. 2003). With a restored interest in the development of the
Casino Mine project, Environment Yukon deployed 45 GPS radio collars on Klaza
caribou from 2012 to 2015 (Hegel 2013).
Population estimates were conducted for the KCH from 1989 to 2012, although survey
methods and study areas changed over the years making population estimates difficult
to compare (Hegel 2013). During a 2012 mark-resight survey, the herd was estimated at
1,179 caribou (95% CI = 952–1,461) (Hegel 2013). Although this is higher than previous
estimates, it is not possible to determine a population trend as previous survey areas
and methods are incomparable (Hegel 2013).

7

Figure 1. The study area, located within the annual range of the Klaza caribou herd, west-central Yukon.

8

Forest fire history
Forest fires are considered frequent in the Klondike Plateau, known as the ‘fire belt’ of
the Yukon, with fires regularly occurring every 50 to 200 years (Yukon Government
2010); however, fire cycles in the broader boreal forest region are often estimated to be
between 50 to 100 years (Rowe 1983; Payette 1992). As of 2015, approximately 20% of
the KCH’s annual range was burned in the last 50 years (2,213 km² of 11,095 km²;
Figure 2), with the majority of mapped burns (40 out of 59 burns) having occurred in the
last 20 years (Figure 2). During the study period (2011–2015), recent burns comprised
an average of 19.3 and 23.3% of the annual and late winter ranges, respectively (Table
1). Most burns are located closer to the edge of the herd’s range, with a large area of
higher elevation habitat in the centre of the range remaining unburned in recent years
(Figure 3).

9

35%

Percent of burned area in annual range

16
30%
25%
20%

11
13

15%

8

10%

3
5

5%

1
0

2

0

0%

Burn age (years)

Figure 2. Percent of area burned by burn age in the annual range of the Klaza caribou herd,
as of 2015. The numbers above the bars represent the number of burns within each burn
age category (Yukon Fire History database 2016).
Table 1. Total area of recent burns (≤50 years) in the Klaza caribou herd’s annual and late
winter range (Environment Yukon 2016), during the study period, 2011–2015.
Annual range
Year

Late winter range

Area burned (ha)

Total area (ha) of
2
recent burns (% of
range)

Area burned (ha)

Total area (ha) of
2
recent burns (% of
range)

2011

1,802

206,809 (19.1)

1,733

100,059 (23.2)

2012

156

206,964 (19.1)

0

100,059 (23.2)

2013

924

207,888 (19.2)

914

100,973 (23.4)

2014

0

207,888 (19.2)

0

100,973 (23.4)

2015
9,960
217,847 (20.1)
410
Area burned refers to new burns that had occurred by the end of each calendar year.
2
Total area of recent burns refers to burns that occurred in the last 50 years.

101,383 (23.5)

1

1

1

10

11

Figure 3. Recent burns near the study area, located within the annual and late winter range of the Klaza caribou herd, west-central
Yukon (range polygons developed by Environment Yukon, 2016).

CHAPTER TWO: POST-FIRE DYNAMICS OF FORAGE LICHENS ON
THE KLAZA CARIBOU HERD RANGE
Introduction
Forest fire is the primary natural disturbance process that influences the distribution and
abundance of terrestrial lichens across most boreal and sub-boreal ecosystems
(Coxson and Marsh 2001; Skatter et al. 2014). Lichens are the main food source for
caribou, particularly during winter (Klein 1982; Thomas et al. 1996; Johnson et al. 2001;
Bergerud et al. 2008), thus, a better understanding of the spatiotemporal succession of
lichens can provide guidance for the conservation and management of caribou habitat.
The length of time required for lichens to return to fire-affected areas in sufficient
abundance to support the forage requirements of caribou is thought to be 50–60 years
(Thomas et al. 1996; Joly et al. 2003, 2007; Dalerum et al. 2007; Sorensen et al. 2008;
Collins et al. 2011). However, the characteristics of the fire history (fire intervals,
intensity, severity, etc.) and the biophysical attributes of the ecosystem (landforms and
vegetation) can influence the time and conditions needed for the reestablishment of
forage lichens. At the landscape scale, the time since burn may provide a sufficient
gauge of relative abundance (cover, volume, and/or biomass) of forage, but at the scale
of an individual burn, this relationship can be complicated by site variability.
In forests, lichen productivity is typically highest on sites with coarse, well-drained,
nutrient-poor soils, and a canopy dominated by coniferous trees (Carroll and Bliss 1982;
Coxson and Marsh 2001; Coxson 2015; Haughian and Burton 2015). The sequence of
lichen succession typical of severe burns in the boreal forest generally includes a period
of early succession (10–30 years post-fire) dominated by Cladonia spp., followed by a
12

mid-succession stage (30–80 years post-fire) dominated by ‘reindeer lichens’ (Cladina
mitis/arbuscula, C. rangiferina, and Cladonia uncialis), and finally, a late succession
stage (>80 years post-fire) that is dominated by Cladina stellaris (Rowe and Scotter
1973; Thomas and Kiliaan 1998; Coxson and Marsh 2001; Cichowski and Haeussler
2013; Haughian and Burton 2015). The richness of lichen species may peak around 30
years post-fire and then decrease, as some lichen species, predominantly from the
Cladina genus, exclude others (Holt et al. 2008).
The dynamics of lichen succession are likely variable throughout the range of woodland
caribou, with factors such as latitude, soil texture, moisture regime, and local climate
affecting reestablishment and growth rates (Coxson and Marsh 2001; Coxson 2015;
Haughian and Burton 2015; Girard et al. 2017). At a finer, stand-level scale, lichen
distribution and abundance is patchy and largely influenced by changes in canopy cover
(Sulyma and Coxson 2001). Haughian and Burton (2015) reported that this patchiness
is influenced more by niche processes, such as environmental micro-habitats, than
neutral processes (e.g., random post-fire colonization), suggesting that management
efforts should focus on understanding the stand characteristics that support these
productive lichen patches. The amount of light reaching the forest floor is considered a
significant factor in lichen growth, with fires promoting a more open canopy (Nelson et
al. 2013), resulting in periods of desiccation that prevent competitors, such as mosses,
from establishing in a stand (Cichowski and Haeussler 2013).
Even if fire-lichen dynamics could be accurately predicted to determine quality and
quantity of forage lichens, other factors influence lichen availability, and ultimately, the
selection strategies of caribou. Access to terrestrial lichens during winter can be limited
13

by snow depth, hardness, and density (Collins and Smith 1991; Johnson et al. 2001,
2002). Not only does deep, hard snow restrict movement, but it can also hinder the
accessibility and detection of lichen (Collins and Smith 1991; Johnson et al. 2000,
2001). To reduce predation risk, caribou avoid wolves and their primary prey, moose
(Alces alces; Bergerud et al. 1984; Seip 1992; Johnson et al. 2001; James et al. 2004).
Fire is believed to affect this spatial separation by increasing the overlap between
caribou, moose, and wolves, potentially resulting in higher predation rates (Wittmer et
al. 2005; Robinson et al. 2012).
Despite a number of studies that have explored fire-lichen dynamics, there is still
uncertainty in our understanding of the successional patterns of forage lichens, primarily
as they relate to the quantity and quality of caribou habitat. Many studies exploring
caribou-fire relationships have focused on the boreal or barren-ground ecotypes of
caribou (e.g., Schaefer and Pruitt 1991; Anderson and Johnson 2014). Thus, it is
convenient to extrapolate those results and associated conservation guidance to all
ecotypes of Rangifer. Environment Canada’s (2011) guidance on habitat disturbance,
considers burns ≤50 years old as ‘disturbed’ habitat. However, forest communities
across the range of woodland caribou may have different fire-lichen dynamics that
influence the recovery or succession of habitat. For example, lodgepole pine (Pinus
contorta) forests typical of some NM caribou winter ranges may not reach peak lichen
productivity until 80–120 years post-fire (Coxson and Marsh 2001).
Accurately predicting the burn-specific succession sequence of forage lichens is difficult
due to local variation in environmental conditions, including fire history (Skoog 1968;
Rowe 1983; Payette 1992). However, this is an important consideration when assessing
14

the ability of winter range to support populations of woodland caribou. Thus, I used plotlevel data to identify environmental factors that influenced the variability in lichen
abundance within burns. I focused on the winter range of a population of NM caribou –
the KCH – found across a northern semi-arid ecosystem in west-central Yukon with a
frequent fire-return interval. I quantified the biomass and cover of forage lichens in
burns of a given age. I then investigated the reputed influence of the time since burn on
lichen abundance, in addition to other ecologically plausible factors that may influence
the establishment and succession of terrestrial lichens that serve as forage for
woodland caribou.

Methods
Site selection and plot layout
Field sites were selected from existing burns located within the KCH annual range. The
annual range was considered to be representative of the maximum boundary of current
and future winter range. Forest fires in proximity to the KCH annual range were mapped
as polygons in the 2013 Yukon Fire History database and were available from Wildland
Fire Management (database updated 18 February 2014). I randomly selected 51 burn
polygons, ranging in age from 1 to 63 years old (burned between 1951 and 2013), and
in size from 1 to 97,665 ha. Polygons intersected or were within the buffered (10 km)
annual range of the KCH estimated for 2014. Burn polygons were then separated into
four pre-determined burn age classes (1–10, 11–30, 31–50, and 51–70 years post-fire).
Due to the challenging access (all sites helicopter accessible only), unburned sites were
typically located adjacent to burn plots; however, I ensured that unburned plots were

15

located well outside of burns. Some unburned plots were randomly selected where
landing sites were accessible.
Sample plots were selected according to a random distance and bearing from the edge
of the burn. To avoid edge effects, all plots were located at least 100 m into the burn
interior which was based on an in-field assessment of stand characteristics (e.g., no
apparent transition from burned to unburned forest). Plots were also located at least 100
m from each other. To sample across potential patches of lichen (Haughian and Burton
2015) and avoid sampling bias involved with randomly encountered linear terrain
features, each plot was composed of five parallel 20-m transects, spaced 5 m apart
(Figure 4), and oriented in a random direction.

5m

20 m

T1

T2

T3

T4

T5

Figure 4. Schematic diagram of the plot layout for vegetation sampling, with the quad arrow
representing plot centre, trees representing the location of canopy openness measurements,
and the central circle representing the area measured for tall shrub cover.

Sampling design
Field sampling was conducted in July and August of 2014. At each sample plot I
described the site, including dominant species of tree, tall shrub, and understory; soil
moisture regime; burn description (presence of burned snags or soil); and indications of
16

the presence of foraging by caribou (e.g., foraged/disturbed lichen or pellet groups). Soil
moisture regime was a subjective classification based on general environmental factors,
soil properties, and indicator plants (B.C. Ministry of Forests and Range and B.C.
Ministry of Environment 2010). Additionally, I recorded topographic characteristics for
the plot, including elevation (m), aspect (degrees), and slope (percent) (Appendix A:
Table 17).
I used a marked boot/step-point intercept method to measure percent ground cover of
plant and lichen groups (National Applied Resource Sciences Center 1999). I recorded
ground cover at 0.5-m intervals/steps per 20-m transect (200 total points/plot), with the
dominant species recorded per step. Species percent cover estimates were derived by
taking the average from these 200 points. Lichens were identified to species when
possible; otherwise they were grouped by genus (e.g., Cladonia spp., Peltigera spp.). I
recorded heights, measured as the portion of intact, living thallus (Moen et al. 2007), of
all fruticose lichens (i.e., Cladina spp., Cladonia spp., and Stereocaulon spp.) to the
nearest 0.5 cm each time a dominant lichen was intercepted along a transect. I
considered Cladina arbuscula/mitis (referred to as C. mitis henceforth), C.
rangiferina/stygia (referred to as C. rangiferina henceforth), Cladina stellaris, Cetraria
cucullata, Cet. ericetorum/islandica (grouped as Cet. ericetorum/islandica), and Cet.
nivalis to be primary forage species (henceforth referred to as ‘forage lichens’).
Cladonia uncialis was not differentiated from Cladonia spp. at all sites, thus it was
excluded from the forage lichen group for analyses. Non-lichen ground cover was
classified into discrete groups: graminoids, forbs, bryophytes, dwarf shrubs (<50 cm
tall), mushrooms, litter, soil, and rock (Appendix A: Table 19; Barrier 2011). At the plot

17

centre I conducted a visual estimate of tall shrub cover (>50 cm tall) in a 5.64 m2 (0.01
ha) fixed-area plot and identified the top three most frequent shrub species.
I used a rod relascope to determine stand basal area (m2/ha) in each plot. Canopy
openness was recorded with a levelled fisheye lens camera where I took two
hemispherical photos of the canopy. I used Gap Light Analyzer (V. 2.0) software (GLA
2.0; Frazer et al. 1999) to calculate average percent canopy openness. The time since
burn was determined based on the mapped burn age and/or by using an increment
borer to core five of the most representative conifer trees (if present) within the plot. The
cores were dated at the University of Northern British Columbia’s Enhanced Forestry
Lab. The height of each cored tree was measured using a Vertex hypsometer (Haglöf,
Sweden).
Results from the tree core samples suggested that the stand age of some of the plots
(primarily older burns) did not reflect the stand age determined using burn mapping (i.e.,
the date at which the most recent burn occurred). Of the plots that were inconsistent
with burn mapping, the aged tree cores generally were older than the mapped burn. Of
the 78 plots for which trees were cored, 28 had tree age differences that were equal to
or greater than 50 years apart, and of the plots (n = 48) that were within a mapped burn
boundary, 21 had a minimum tree age that was older than the mapped burn age. To
maintain consistency for determining the plot age class, I developed a protocol to
determine time since burn for vegetation plot age classes (Appendix A: Table 18). Plots
were considered equal to the mapped burn age, minimum core age, median core age,
or maximum core age depending on their location inside or outside of a mapped burn
and the difference between mapped burn age and tree core results.
18

Although I attempted to measure key mechanistic parameters that may influence lichen
recovery and growth following fire, field logistics (e.g. helicopter time and site
accessibility) prevented the testing of all factors known to influence the abundance of
terrestrial lichens. In an effort to assess a large number of sites, at the expense of an
intensive assessment of a few sites, I did not measure variables that required
substantial time and/or complex field equipment (e.g., soil classification, permafrost).
Data analyses
Examining lichen abundance by burn age
I categorized stand age into 15-year intervals, allowing for statistical comparison of age
classes representing the time since last major fire (Skatter et al. 2014). Lichen volume
and biomass were estimated using methods outlined by Fleischman (1990). Volume
was calculated by multiplying percent cover (% = dm2/m2) by height (dm). Lichen
volume is highly correlated with biomass (Fleischman 1990; Kumpula et al. 2000; Moen
et al. 2007), thus biomass estimates were derived for each species by multiplying
Fleischman’s (1990) ratio functions (lichen species weight to volume) by volume
(dm3/m2). The ratios of weight to volume ranged from 9.6 to 20.3, depending on the
species. These values for biomass must be considered cautiously, as no lichen samples
were taken to develop site-specific correction factors for this area. The volume
estimates of Cladina spp. and Cetraria spp. were summed into a final volume estimate
for each plot. Stand age (i.e., time since burn) was considered to be the time at which
the ground burned, and did not account for the lag time between fire and seedling
emergence. I used a non-parametric pairwise comparison test, with Tukey’s method of
adjustment for multiple comparisons, to test for differences in lichen volume among

19

stand age groups (Tukey 1953). All statistical analyses were completed using Stata/IC
(Version 12.1; StataCorp 2011) statistical software.
Statistical model selection and assessment
I used zero-inflated count models (Martin et al. 2005) to investigate ecological factors
hypothesised to explain measured variation in lichen volume. Although the lichen data
collected in this study were not discrete integers, as one observes with counts, these
data were non-negative, overdispersed (i.e., the sample variance is greater than the
mean), and had an excess of zeros (i.e., no lichens present) (Martin et al. 2005;
Richards 2008; Zuur et al. 2007, 2009, 2010). To conform to count distributions, I
rounded all volume estimates to the nearest whole number.
Count models were constructed using either Poisson or negative binomial distributions.
Considering the prevalence of zeros in the data (20%), I used a Vuong test to determine
if a zero-inflated model was appropriate (Vuong 1989; Long and Freese 2001). In
preparation for data analysis, lichen data were summarized for each plot into a single
value representing the total forage lichen group (all Cladina spp. and Cetraria spp.). I
used the terrain and environmental variables collected during field studies to construct
candidate models to explain observed variation in lichen volume (Appendix A: Table 17;
Table 2), with a focus on more mechanistic predictors that allowed a test of ecological
hypotheses (Wiersma et al. 2011). I used quadratic terms to allow for a non-linear
representation of the independent variables where necessary (e.g., tree height; Long
and Freese 2001).

20

Zero-inflated models may be more informative if the same predictors are believed to
influence both the false (undetected because of sampling error) and true zeros (low
frequency of occurrence due to demographics, competition, or habitat quality), or if
there is no clear evidence that distinguishes between the mechanisms affecting species
distribution (e.g., presence/absence) and abundance (Martin et al. 2005). Thus, I used a
consistent set of variables to represent the hypothesized ecological processes
influencing lichen distribution (presence/absence) and abundance (volume counts).
I used fractional logit models (generalized linear models with binomial family and logit
link; Papke and Wooldridge 1996) to investigate ecological factors hypothesised to
explain variation in the percent cover of lichen. This allowed me to determine if a
consistent set of predictor variables influenced the volume and percentage cover of
forage lichens. Percent cover is a common and more rapid measure of lichen
abundance.
To control for a lack of sample independence across plots, all models were clustered by
site ID (i.e., the same mapped burn) allowing for the estimate of robust standard errors
(Long and Freese 2001). I used tolerance scores and a threshold of ≤2 to test predictor
variables for collinearity (Menard 2002). Using deviation coding, I generated a design
matrix of dummy variables to represent contrasts between classes of categorical
predictor variables (Hendrickx 1999).

21

Table 2. Variables used in count models of volume of forage lichens in burns surveyed on the
Klaza caribou range, west-central Yukon, 2014.
Parameter

Abbreviation

Description

Aspect

North+East

Northness (cosine of aspect) and eastness (sine of aspect)

Elevation

Elev

Elevation above sea level (m)

Canopy openness

CO

Percent canopy openness (expressed as a proportion)

Cover type

CT

Dominant cover type in plot (e.g., coniferous, deciduous, open)

Slope angle

S

Percent slope (expressed as a proportion)

Soil moisture regime

SMR

Categorical soil moisture regime (dry, moist, wet)

Stand basal area

SBA

Cross-sectional area of trees in the stand (m )

Tall shrub cover

TSC

Percent of tall shrub (>0.50 m) cover (expressed as a
proportion)

Time since burn

TSB

Number of years since the site burned, estimated using burn
mapping and/or tree core aging

Tree height

TH

Average tree height (m)

2

I used Akaike’s Information Criterion (Akaike 1973) adjusted for small sample sizes
(AICc) to select the most parsimonious model(s) (i.e., the fewest predictor variables to
explain the greatest variation in the data). I used cross-validation to test the predictive
ability of the most parsimonious models. Using a jackknifing procedure, I sequentially
withheld the data from each plot during the model-fitting process; the resulting model
(N-1, where N = number of sample plots) and the withheld record were then used to
generate an independent prediction (Bridger et al. 2016). I regressed the predicted
versus observed values and used the coefficient of determination (R 2) to measure the
variation explained by each model. Also, I used Wilcoxon sign-rank equality tests for
matched pairs to statistically compare the differences in the predicted and observed
lichen values.

Results
General plot characteristics
A total of 100 plots were sampled in 18 different burns and 13 different unburned sites
(Table 3; Figure 5), with plots ranging in age from 1 to 271 years post-fire. Over one
22

third of all mapped burns in the KCH were sampled; 29% of which were recent burns
(<50 years old). Age class 1 (1–10 year old burns) sites were sampled with the lowest
intensity, while sites in age classes 3 to 5 were sampled more frequently to focus efforts
on age classes that were more likely to support forage lichens (Rowe and Scotter 1973;
Thomas and Kiliaan 1998; Coxson and Marsh 2001; Cichowski and Haeussler 2013).
Despite being a high priority for sampling, burns in age class 4 (51–70 years old) were
uncommon within the KCH’s annual range, and sampling efforts were expanded to a
burn on the northeast side of the Yukon River (Figure 5).
Table 3. Vegetation sampling in burns by age class (based on both tree core results and
existing burn mapping), on the Klaza caribou range, west-central Yukon, 2014.
Age class (years)

Number of unique burns/sites

Number of sample plots

1 (1–10)

2

3

2 (11–30)

7

16

3 (31–50)

6

21

4 (51–70)

3

18

5 (>70/unburned)

13

42

TOTAL

31

100

23

24

Figure 5. Burns and vegetation plots sampled in summer 2014 within the Klaza caribou range, west-central Yukon.

Patterns in the abundance of forage lichens post-fire
The succession sequence of lichens began with Cladonia spp. dominating post-fire,
followed by Cladina mitis and C. rangiferina dominating both mid and late successional
stages (Figure 6). The percent cover of forage lichens (Cladina spp. and Cetraria spp.)
increased 40 years post-fire, peaked at 85–99 years, and then decreased in plots that
were >99 years old (Table 4). Similar patterns were observed for both lichen volume
and biomass (Table 4). Lichen volume in 25–39 year old burns was significantly less
than all older age groups (P < 0.05), with the exception of the 55–69 and 100–149 year
age groups (P = 0.98 and 0.32, respectively; Table 5). Conversely, the volume of lichen
in 40–54 year old burns was not significantly different than all older age groups (P >
0.05; Table 5).
18%
Cladina mitis

16%

Cladina rangiferina
Cladina stellaris

14%
Average percent cover

Cetraria spp.
12%

Cladonia spp.

10%
8%
6%
4%
2%
0%
Early succession (10-30 years)

Mid-succession (31-80 years)

Late succession (>80 years)

Figure 6. Sequence of post-fire lichen succession observed in the Klaza caribou range, westcentral Yukon, summer 2014.

25

Table 4. Mean (standard error, SE) cover, volume, and biomass of forage lichens in
vegetation plots in burns by 15-year intervals, Klaza caribou range, west-central Yukon,
summer 2014. Plots aged 100–149 and ≥150 years were grouped together rather than by
15-year intervals due to low sample sizes.
Age group
(years)

n

<25
25–39
40–54
55–69
70–84
85–99
100–149
≥150

18
8
14
18
8
5
16
13

Lichen cover
Mean (%)

SE

2.6
4.2
24.5
9.5
39.4
41.2
22.2
29.3

1.0
2.1
4.4
3.0
7.7
7.9
3.8
6.9

Lichen volume
Mean
3
2
(dm /m )
0.8
2.2
16.1
6.2
22.8
25.1
12.3
17.9

SE
2.4
3.6
2.7
2.4
3.6
4.6
2.5
2.8

Lichen biomass
Mean
(kg/ha)
113
236
1,786
895
2,670
3,061
1,588
2,061

1

SE
52
110
371
283
537
709
306
474

1

Ratios of volume to biomass for each lichen species were derived from Fleischman (1990).

26

Table 5. Non-parametric pairwise comparison (using Tukey’s method of adjustment for multiple comparisons) of lichen volume in
different-aged vegetation plots in burns in the Klaza caribou range, west-central Yukon. P < 0.05 indicates a significant difference
(in bold) between age groups.
Age group

n

Mean
volume
3
2
(dm /m )

±SE

<25 years

18

0.77

2.40

25–39 years

8

2.22

3.60

40–54 years

14

16.10

2.72

55–69 years

18

6.15

2.40

70–84 years

8

22.75

3.60

85–99 years

5

25.14

4.55

100–149 years

16

12.26

2.54

≥150 years

13

17.87

2.82

Tukey test
statistics
Contrast
t-score
P
Contrast
t-score
P
Contrast
t-score
P
Contrast
t-score
P
Contrast
t-score
P
Contrast
t-score
P
Contrast
t-score
P
Contrast
t-score
P

Age group (years)
25–39

40–54

55–69

70–84

85–99

100–149

≥150

1.45
0.34
1.00

15.33
4.23
<0.001
13.88
3.08
0.05

5.38
1.59
0.76
3.93
0.91
0.98
-9.94
-2.74
0.12

21.98
5.08
<0.001
20.53
4.04
<0.001
6.66
1.48
0.82
16.60
3.84
0.01

24.37
4.74
<0.001
22.92
3.95
<0.001
9.04
1.71
0.68
18.99
3.69
0.01
2.39
0.41
1.00

11.48
3.28
0.03
10.04
2.28
0.32
-3.84
-1.03
0.97
6.10
1.75
0.66
-10.50
-2.38
0.26
-12.88
-2.47
0.22

17.10
4.62
<0.001
15.65
3.42
0.02
1.77
0.45
1.00
11.71
3.16
0.04
-4.89
-1.07
0.96
-7.27
-1.36
0.87
5.61
1.48
0.82
-

27

Statistical models
Model selection
I used data from 100 field sites within the KCH annual range to generate statistical
count models. The alpha parameter was significantly different from zero in all model
sets and Vuong z-tests for each model set were significant (P < 0.05) indicating that a
zero-inflated negative binomial distribution was most supported; however, during the
cross-validation process, the zero-inflated negative binomial (ZINB) models would not
converge, and the models were run using a zero-inflated Poisson (ZIP) model. Although
the coefficient estimates differed, the same top-ranked candidate models were identified
during the model selection process using both distributions.
The most parsimonious model for lichen volume included covariates for time since burn,
canopy openness, stand basal area, tree height, and cover type (Table 6). The topranked model had a large AICc weight (AICcwi = 0.95; Table 6). For lichen cover, there
were seven top-ranked models (∆i AICc ≤2.00) accounting for 79% of the AICc weight
(Table 6).

28

Table 6. Summary of model selection statistics for the most parsimonious models (∆i AICc ≤2)
of volume and cover of forage lichens in the Klaza caribou range, west-central Yukon.
Parameters are defined in Table 2.
Rank

Model parameters

1

Volume models
Time since burn+Canopy openness+Stand basal area+Tree
2
height+Tree height +Cover type

1
2
3
4
5
6
7

Cover models
Canopy openness +Stand basal area+Tree height+Tree
2
height
2
Canopy openness +Tree height+Tree height
2
Tree height+Tree height +Cover type
2
Time since burn+Canopy openness +Tree height+Tree height
Time since burn+Canopy openness +Stand basal area+Tree
2
height+Tree height
2
Time since burn+Tree height+Tree height
2
Canopy openness +Tree height+Tree height +Tall shrub cover

AICci

AICc wi

∆i AICc

782.58

0.95

0.00

71.37

0.17

0.00

71.43
71.79
72.07
72.96

0.16
0.13
0.12
0.08

0.06
0.42
0.70
1.59

73.02
73.36

0.07
0.06

1.65
1.99

Forage lichen volume
Stand basal area, tree height, and cover type were significant predictors of the volume
of forage lichens in the top-ranked ZIP model (Figure 7, Appendix B: Table 20). Lichen
volume increased with stand basal area and the negative quadratic term for tree height
(tree height2) indicated a positive non-linear influence on volume (Appendix B: Table
22). Deciduous and open cover types had a negative influence on lichen volume when
compared to coniferous cover types (Appendix B: Table 22). The covariate for time
since burn had only a small effect and was not statistically significant.
For the binary part of the ZIP model, canopy openness, stand basal area, and cover
type were all significant predictors of lichen presence/absence (Figure 8, Appendix B:
Table 22). For every unit increase in canopy openness (%) and stand basal area
(m2/ha), the probability of lichen absence decreased (Appendix B: Table 22). The
negative quadratic term for tree height (tree height 2) indicated a positive non-linear
29

influence on the probability of lichen absence (Appendix B: Table 22). Relative to
coniferous cover types, deciduous cover types increased the probability of lichen
absence, while open cover types decreased the probability of lichen absence (Appendix
B: Table 22).
2.5
2.0

TSB+CO+SBA+TH+TH²+CT

Count coefficient (95% CI)

1.5
1.0
0.5
0.0
-0.5
-1.0
-1.5
-2.0
-2.5

Figure 7. Coefficients for the count portion of the top-ranked zero-inflated Poisson model
(∆I AICc ≤2) explaining lichen volume (counts) in the Klaza caribou range, west-central
Yukon. All variables defined in Table 2.

30

7.0
TSB+CO+SBA+TH+TH²+CT
Binary coefficient (95% CI)

5.0
3.0
1.0
-1.0
-3.0
-5.0
-7.0
-9.0

Figure 8. Coefficients for the binary portion of the top-ranked zero-inflated Poisson model
(∆I AICc ≤2) representing influences on the probability of lichen absence in the Klaza caribou
range, west-central Yukon. All variables defined in Table 2.

Forage lichen cover
The covariates included in the top-ranked models for percent cover of forage lichens
were the same as those in the top volume model; however, there were fewer predictors
in each candidate model (Appendix B: Table 21). Time since burn was included as a
significant covariate in two of the seven top-ranked models (Appendix B: Table 23).
Canopy openness was included in five of the top-ranked models (Appendix B: Table
23). Similarly, stand basal area was included in two top models and was a positive
influence on cover in both (Appendix B: Table 23). Tree height was included in all top
models (Appendix B: Table 23). The negative quadratic term for tree height (tree
height2) indicated a positive non-linear influence on lichen cover.

31

Model fit
The top-ranked ZIP model explained 54% of the observed variability in lichen volume
(Table 7). For lichen cover, the top-ranked model explained 42% of the observed
variability (Table 7). None of the top-ranked models for either lichen volume or cover
showed a significant difference between observed and predicted values (Table 7).
When plotted against the observed count values, the unstandardized residuals
(predicted-observed lichen volume counts) revealed that the top ZIP model predicted
well in the middle of the value range (10–25 dm3/m2), but over-predicted at lower
volumes, and under-predicted at higher volumes (Appendix B: Figure 16). The same
pattern was observed for most of the top-ranked cover models, with relatively good
prediction between 10 and 40% cover; however the residuals of three of the top-ranked
fractional logit models (CO+TH+TH2, TH+TH2+CT, and CO+TH+TH2+TSC; defined in
Table 2) did not show the same negative trend and were much closer to the observed
values (Appendix B: Figure 17).

32

Table 7. R2 values (observed versus predicted values) and test statistics from Wilcoxon signrank equality tests for matched pairs for the analysis of differences between observed and
predicted lichen volume and cover values for the top-ranked (∆i AICc differed by ≤2 points)
count and fractional models. The test revealed no statistically significant differences in the
distribution of observed and predicted values when P > 0.05.
AICc
rank
1

1
2
3
4
5
6
7

Model
Volume models
Time since burn+Canopy openness+Stand basal area+Tree
2
height+Tree height +Cover type
Cover models
2
Canopy openness +Stand basal area+Tree height+Tree height
2
Canopy openness +Tree height+Tree height
2
Tree height+Tree height +Cover type
2
Time since burn+Canopy openness +Tree height+Tree height
Time since burn+Canopy openness +Stand basal area+Tree
2
height+Tree height
2
Time since burn+Tree height+Tree height
2
Canopy openness +Tree height+Tree height +Tall shrub cover

R

2

Wilcoxon sign-rank
equality tests
z-score

P-value

0.54

-0.85

0.40

0.42
0.33
0.41
0.38
0.44

-0.89
-0.65
-0.29
-0.53
-0.84

0.37
0.52
0.77
0.60
0.34

0.45
0.33

-0.29
-0.58

0.78
0.56

Discussion
Results from this study support previous research concluding that forage lichens require
approximately 40–60 years to recover to levels that support caribou foraging (Schaefer
and Pruitt 1991; Thomas et al. 1996; Joly et al. 2003, 2007, 2010; Collins et al. 2011).
However, I observed that the average volume of forage lichens in burns that are 40–54
years post-fire is no different than in older burns (Table 5). These results also revealed
that a suite of environmental factors may contribute to lichen recovery. Canopy
openness, tree height, cover type, and stand basal area were consistently the best
indicators of lichen volume and cover, but the time since burn did not have a significant
influence on either metric (Table 6). Nonetheless, the time since burn was included in
the top-ranked volume model, suggesting it is an important covariate relative to the
other parameters in the model.

33

Lichen succession and abundance post-fire
The sequence of lichen succession was consistent with stand-replacing burns in the
boreal forest (Rowe and Scotter 1973; Morneau and Payette 1989; Thomas and Kiliaan
1998; Coxson and Marsh 2001; Cichowski and Haeussler 2013; Skatter etal. 2014).
Unlike other findings, however, Cladina stellaris did not dominate the late succession
phase of the post-fire community; instead, C. mitis and C. rangiferina dominated the
older age classes (Figure 6). C. stellaris may be rare in the KCH range because the fire
frequency is too short to allow for the establishment of climax stands. Further, the
majority of stands greater than 80 years old may be transitional, mid-seral stands and/or
the majority of soils in the region may promote an earlier dominance of feather mosses,
thus preventing C. stellaris from dominating.
My results suggest there was considerable variability in the abundance of lichen within a
burn. For example, in plots that were 45 years post-fire, the average biomass was 1,786
kg/ha, but it ranged between 0 and 4,920 kg/ha. One possible explanation for the spatial
variability in the biomass of forage lichens is the accuracy of burn mapping, especially
older burns that were irregularly and/or poorly mapped, and which may not recognise
large unburned patches within a burn perimeter (Yukon Government 2014). Additionally,
although it is assumed most fires in the region are stand-replacing, some areas of a
burn may have been of lower intensity and severity due to a lack of fuel (e.g., sparse
spruce slopes, deciduous stands; Johnstone and Stuart-Chapin 2006). During site
investigations it was apparent that there was considerable variability in burn intensity
and typically the abundance of lichen was greater in low-intensity burns. For example,
evidence of fire (e.g., burned soil or trees) often became less apparent on north or

34

northeast-facing slopes with sparse tree cover, suggesting the lack of fuel (e.g., trees)
resulted in less intense and severe burning. These slopes are also likely underlain by
discontinuous permafrost, enabling lichen species to outcompete vascular plants that
have difficulty establishing and growing in permafrost soils. In combination, these two
factors may have resulted in relatively high levels of lichen abundance in some younger
burns.
Skatter et al. (2014) observed a bi-modal pattern to lichen recovery in northern
Saskatchewan, with higher lichen cover in stands ≤40 years old and >100 years old.
Although my results do not show the same bi-modal pattern, they are similar in that
lichen cover was substantially greater in 41–50 year old stands than 61–70 year old
stands (25% versus 10%, respectively), increasing in stands from 70–110 years-postfire (Table 4). This may indicate that the 41–50 year old plots in the KCH range are
transitional, mid-seral stands (Haughian and Burton 2015). Holt et al. (2008) also
reported that lichen cover, height, and species richness peaked at 30 years post-fire in
northwestern Alaska. Similarly, Joly et al. (2010) highlighted the presence of relatively
high lichen biomass in younger (<51 years old) plots in their Alaskan study area;
however, they considered these to be ‘anomalous’ and not representative of the overall
relationship of stand age and lichen abundance. Nelson et al. (2013) detected high
levels of lichen cover in young burns (<25 years old), but they were exclusively located
on alluvial terraces. These studies (Holt et al. 2008; Joly et al. 2010; Nelson et al. 2013;
Skatter et al. 2014) support my results and are consistent with Haughian and Burton’s
(2015) conclusion that the distribution of lichens is predominantly influenced by
microsite characteristics that may change throughout the successional process.

35

Influence of stand characteristics on lichen volume
Past research suggests that lichens are most productive in open canopy forests with
well-drained and nutrient poor soils (Coxson and Marsh 2001; Sulyma and Coxson
2001; Coxson 2015; Haughian and Burton 2015). Even though soil moisture regime
(Appendix A: Table 17) was documented in my study, forest stand characteristics
proved to be the most influential predictors of lichen volume (Table 6 and Figure 7). In
addition, the time since burn, although not a significant predictor, did improve the
performance of the top-ranked volume model (Table 6). The coefficient estimates from
the ZIP models suggest that the same environmental parameters influence both the
presence and amount (count) of lichen and that the mechanisms influencing the two
processes do not differ (Appendix B: Table 22). The exception to this was canopy
openness, which had a significant influence on the probability of lichen absence, but not
on the volume counts. Results from the models for lichen cover were similar (Table 6),
although there was relatively more model uncertainty in the cover models (i.e., 13
models made up 100% of the weight of evidence for cover versus two models for
volume) and those models were less predictive. While the measured parameters were
able to explain a considerable amount of the observed variability in lichen abundance, it
is important to note that additional, unmeasured parameters, such as soil type and
permafrost, might influence lichen recovery and growth following fire.
Cover type was a strong predictor of lichen volume (Figure 7), with coniferous types
positively influencing abundance. According to Roturier et al. (2017), burned sites with
remnant trees provide the best conditions for lichen establishment and growth. Open
cover types were typically associated with younger burns (<45 years post-fire), whereas

36

coniferous types were associated with older stands (minimum 34 years post-fire). Tree
height was also an influential factor, which may be a result of its ability to broadly
represent two mechanistic processes in a stand: site productivity and light interception.
In combination with stand age, tree height has long been used to calculate site index –
an estimate of site productivity for tree growth (Yukon Government 2006; Skovsgaard
and Vanclay 2008). Additionally, tree height also may represent the amount of light
reaching the forest floor, with taller trees decreasing canopy openness. Considering that
lichens are generally poor competitors, but can thrive in nutrient-poor and dry sites
(Haughian and Burton 2015), the negative coefficient for tree height 2 appears
appropriate, and confirms findings by Lesmerises et al. (2011), who noted that a
combination of landscape variables and stand age, density, and height best estimated
both lichen occurrence and biomass.
A number of stand characteristics were important predictors of lichen abundance in the
KCH range, yet model fit improved with the inclusion a measure of time since burn. This
is consistent with the findings of other studies that reported stand age as influential, but
not necessarily a significant predictor of lichen abundance (Arseneault et al 1997; Joly
et al. 2010; Lesmerises et al. 2011). Some of the lack of explanatory power for stand
age in the count models could be a product of relatively imprecise methods for
identifying the boundaries of older burns (e.g., ~40–60 years old), and aging the stand,
despite using two methods to determine stand age. Visually, the relationship between
stand age and lichen abundance appears to be nonlinear and asymptotic; however, the
quadratic term for stand age did not improve model performance. With the vast majority
of sample plots being less than 150 years old, the scope of my field data may not have

37

captured the complete successional pathway of terrestrial lichens across the range of
the KCH.
Using biomass thresholds for changes in caribou selection strategies
Although lichen cover may be an adequate indication of the succession pattern of
forage lichens following fire, biomass has more influence on caribou distribution and
foraging behaviour (Klein 1982; Thomas et al. 1996; Joly et al. 2003, 2010; Collins et al.
2011). Some studies have estimated lichen cover at the landscape scale (Nelson et al.
2013; Caslys Consulting 2014), but fewer have provided estimates of biomass in burns
of a range of ages (Joly et al. 2010; Collins et al. 2011; Lesmerises et al. 2011). Lichen
biomass studies often are limited to landscapes with low canopy cover, unlike the
majority of areas occupied by woodland caribou (Lesmerises et al. 2011).
Estimates from herds across the circumarctic suggest that the biomass of forage lichens
typically found on the winter range of caribou is highly variable, but tends to be at least
500 kg/ha (Barrier 2011; Appendix C: Table 24). Thomas et al. (1996) observed a
marked increase in the selection by caribou of older forest stands (41–60 versus 61–80
years post fire) and suggested that caribou may respond to a threshold in lichen
biomass. At these sites, the average biomass of Cetraria nivalis increased from 30 to 72
kg/ha and Cladina mitis increased from 547 to 1,029 kg/ha (Thomas et al. 1996). This is
similar to observed values of lichen biomass at sites avoided and selected by caribou in
the Nelchina herd of Alaska, bordering the KCH range (Collins et al. 2011). During
winter, Nelchina caribou selected stands that were more than 80 years post-fire, with
>20% lichen cover, and more than 1,250 kg/ha of forage lichens. Nelchina caribou

38

abandoned previously used winter range after grazing reduced the lichen biomass to
approximately 400 kg/ha (Collins et al. 2011).
My estimates of lichen biomass (using ratios from Fleischman (1990) suggest that in the
KCH range, some plots in 45-year old burns may have enough biomass to support
foraging by caribou (Figure 9). While up to 23% of plots ≤50 years old could
hypothetically be selected by Klaza caribou, no plots in burns less than 45 years old had
lichen biomass levels above the Nelchina herd’s observed selection threshold (Figure
9). The period of the greatest lichen productivity appears to occur between 80 and 110
years post-fire (Figure 9), which is consistent with findings from other studies (Coxson
and Marsh 2001). Further, the amount of variability in lichen biomass observed beyond
45 years is notable (Figure 9), and highlights the importance of considering the
influence of terrain and environmental variables on lichen biomass. Currently 20% of the
annual and 24% of the late winter KCH range is comprised of burns that are less than
50 years old, regarded as disturbed habitat that results in a decline in boreal caribou
(Environment Canada 2008; 2011), thus better understanding these relationships could
have substantial implications for range assessment processes.

39

0

0.2

0.4

0.6

Biomass of forage lichens (kg/ha)

6,000

0.8

1

Selection threshold
Avoidance threshold

5,000
4,000
3,000
2,000
1,000
0
Estimated time since burn (years)

)LJXUH %LRPDVVRIIRUDJHOLFKHQVZLWKLQGLIIHUHQWDJHGVWDQGVWKURXJKRXWWKH.OD]DFDULERX
UDQJHZHVWFHQWUDO<XNRQ7KHWZRWKUHVKROGYDOXHV VHOHFWLRQ NJKDDYRLGDQFH 
NJKD  UHSUHVHQW ELRPDVV OHYHOV DW VLWHV DYRLGHG DQG VHOHFWHG E\ FDULERX LQ WKH
QHLJKERXULQJ1HOFKLQDKHUG &ROOLQVHWDO 

7KHPHDVXUHVRIELRPDVVIRFXVRQTXDQWLW\RQO\QRWDYDLODELOLW\RIIRUDJHOLFKHQV7KLV
GLVWLQFWLRQ PXVW EH PDGH FOHDU HVSHFLDOO\ LQ WKH FRQWH[W RI EXUQV 2WKHU IDFWRUV HJ
VQRZ GHSWK DQG FUXVW SDWFK VL]H DFFHVVLELOLW\ HWF  LQIOXHQFH ERWK WKH DYDLODELOLW\ RI
IRUDJHDQGWKHVWUDWHJLHVWKDWFDULERXHPSOR\ZKHQVHOHFWLQJKDELWDW &ROOLQVDQG6PLWK
 -RKQVRQ HW DO   $FFHVVLQJ OLFKHQULFK SDWFKHV ZLWKLQ EXUQV PD\ SURYH
GLIILFXOW IRU FDULERX GHSHQGLQJ RQ WKH VL]H RI WKH EXUQ SUHVHQFH RI GRZQHG ZRRG\
GHEULVKLJKHUVQRZFRYHUGXHWRDODFNRIWUHHLQWHUFHSWLRQDQGRUWKHDELOLW\RIFDULERX
WRGHWHFWOLFKHQSDWFKHV -RKQVRQHWDO $GGLWLRQDOO\FDULERXDOVRVHOHFWKDELWDW
EDVHGRQDQXPEHURIQRQIRUDJHUHODWHGIHDWXUHVVXFKDVVORSHHOHYDWLRQDQGDVSHFW
-RKQVRQ HW DO   7KLV KLJKOLJKWV WKH QHHG WR IXUWKHU HYDOXDWH KRZ VHOHFWLRQ
VWUDWHJLHVIRUWKH.&+LQEXUQVGLIIHUIURPWKHUHVWRIWKHLUZLQWHUUDQJH


Conclusions
Data from the KCH range revealed that lichen abundance can be relatively high in
stands as young as 45 years post-fire, with biomass levels consistent with stands used
by caribou from other herds. At the landscape level, post-fire stand age provides a
rudimentary guide for the availability of lichen, but my results suggest that alone it does
not explain a large amount of the variability observed at finer scales. Within a burn,
there is considerable variability that is likely driven by the presence of microsites
favourable to lichen recovery. Identifying these lichen-rich patches can be done on a
burn-specific basis using remotely sensed data (Lesmerises et al. 2011; Nelson et al.
2013; Caslys Consulting 2014). However, lichen-rich areas within a burn may be in a
transitional, mid-seral phase of succession, and may have reduced lichen abundance
over time as climax plant communities begin to dominate.
My results add to the growing body of literature that suggests lichen succession cannot
be predicted using only estimates of time since fire. Age-based thresholds, although
easy to use and understand, do not capture the variability of lichen abundance within
burns. An age-based threshold for burns (e.g., 50 years) provides a conservative
evaluation of lichen availability at the landscape level; however, a combination of
environmental factors can explain a larger portion of the total variance in lichen
abundance within and among burns. The challenge for managers is to have a
consistent and robust approach to lichen assessment that can be continually applied
over time and space as natural and anthropogenic changes occur within caribou
ranges.

41

CHAPTER THREE: VARIATION IN HABITAT SELECTION STRATEGIES
OF WOODLAND CARIBOU IN BURNS
Introduction
The spatial and temporal distribution of caribou manifests as a hierarchy of complex
decisions related to the distribution and availability of resources and predation risk
(Johnson 1980; Rettie and Messier 2000; Johnson et al. 2001, 2002; Boyce 2006;
DeCesare et al. 2012). Disturbance from fire can have resounding effects on forest
structure and composition, landscape patterns (Payette 1992), and ultimately influence
key components of wildlife habitat (Schaefer and Pruitt 1991; Joly et al. 2003). Caribou
may avoid winter habitat for approximately 60 years after a fire due to the time required
for terrestrial lichens to recover (Thomas et al. 1996; Joly et al. 2003; Joly et al. 2007;
Collins et al. 2011). Nonetheless, locations of satellite-collared caribou reveal that
recently burned areas are used by some individuals, even during winter when slowgrowing terrestrial lichens are the primary food source (Thomas and Kiliaan 1998;
Anderson and Johnson 2014). The nature of this use is not well-understood and is often
obscured when studies of habitat selection are conducted at the scale of the population.
In such cases, data from multiple individuals are pooled and the collective pattern
suggests avoidance of recent burns, whereas some individuals may use those areas
(e.g., Hegel and O’Donoghue 2015).
Caribou will use habitat within and adjacent to burns during winter (Thomas et al. 1998;
Joly et al. 2003; Anderson and Johnson 2014); however, relative to total range use this
habitat is rarely selected by caribou. The use of recently burned stands has been
attributed to a number of factors, including the availability of vascular forage in younger
42

regenerating burns (11–20 years old; Joly et al. 2003), reduced predation risk as a
result of improved visibility, efficient movement between high-quality habitat patches,
and snow conditions (i.e., depth and hardness) that permit relatively easy access to
terrestrial vegetation (Miller 1976; Thomas et al. 1998). Conversely, others have
suggested that caribou avoid burns because of the presence of predators and other
cervids (e.g., wolves and moose), the lack of forage species used by caribou, and
unfavourable snow conditions (James et al. 2004; Gustine et al. 2006; Gustine and
Parker 2008; Collins et al. 2011).
Resource selection functions (RSFs) have been used to model habitat selection
strategies across a variety of spatial scales (Boyce et al. 2002; Manly et al. 2002), often
providing an indication of patterns of selection at the population level, but seldom at the
individual level (Thomas and Taylor 1990; Gustine et al. 2006; Gillingham and Parker
2008). Examining use and availability of resources by individuals may help to better
explain variation in selection strategies (Thomas and Taylor 1990). Such variation is
believed to be an important survival strategy (Gustine et al. 2006) that allows individuals
to use a range of habitats as they become available. The resource selection strategies
of caribou can vary among individuals (Gustine and Parker 2008), but is largely driven
by decisions related to forage availability and quality as well as predation risk (Apps et
al. 2001; Johnson et al. 2001; Briand et al. 2009). These decisions are complicated by
individual responses to the environment at a range of scales (Rettie and Messier 2000;
Gustine et al. 2006), where the behavioural and physiological plasticity of an individual
is likely more advantageous than a general optimal foraging strategy (Johnson et al.
2001).

43

The late winter season provides unique challenges to the habitat selection strategies of
caribou (Schaefer and Pruitt 1991; Johnson et al. 2001). Cold temperatures and deep,
hard snow reduce accessibility to forage and result in high energetic costs of travel on
and cratering through snow (Fancy and White 1985; Collins and Smith 1991; Schaefer
and Pruitt 1991; Schaefer 1996; Johnson et al. 2001). These energetic demands are
exacerbated by the increasing gestational demands experienced by pregnant caribou
(Parker et al. 2005; Joly et al. 2015), as the majority of fetal mass is accumulated in the
final trimester (Robbins and Robbins 1979), roughly aligning with the late winter season
(Hegel and O’Donoghue 2015). The quality of late winter habitat may influence calf
condition, recruitment, and the survival of juveniles (White 1983; Gaillard et al. 1998;
Post and Klein 1999; Adams 2003).
Habitat selection of caribou during winter in the Yukon has been modeled using fire
history, vegetation cover, and terrain (Florkiewicz et al. 2004, 2007; Barker and Hegel
2012). A recent analysis of winter habitat selection revealed that caribou from the KCH
avoided burns at the landscape scale. However, these models were dependent on a
relatively simple measure of habitat (burned versus unburned) and did not quantify
variation in habitat selection among individuals. The relative importance of some habitat
variables, such as burns, may vary across scales, and those that are most limiting to
individuals should be considered at the broadest scale (Rettie and Messier 2000;
Gustine et al. 2006). More complex measures of burns, including multiple age classes,
location in burns (core versus edge), and burn size may help to reveal the importance of
different burn characteristics at different scales.

44

Snow conditions have an important influence on the seasonal and finer-scale habitat
selection of caribou (Collins and Smith 1991; Russell et al. 1993; Johnson et al. 2001;
Tyler 2010). Not only does deep snow restrict movement and increase energetic costs,
but it can also hinder the accessibility and detection of lichens (Collins and Smith 1991;
Johnson et al. 2000, 2001; Gustine et al 2006). Contrasting habitat selection strategies
between the early and late winter seasons can provide a rudimentary proxy for a snow
depth metric, as snow conditions are believed to be generally more favourable earlier in
the winter (e.g., shallower, less crusty).
In this study, I quantified the habitat selection strategies of the KCH during the winter.
These caribou are found across a landscape with a relatively high frequency of forest
fire. Thus, I examined range-wide habitat selection strategies during the early and late
winter, as well as those strategies employed by caribou that occupied recent burns.
Recognising the potential for variation in strategy, I developed resource selection
functions for individual caribou that were monitored with GPS collars. I hypothesized
that in general, caribou would avoid burns during both early and late winter; however,
some individuals would use burns primarily during the early winter season when snow
conditions were less limiting. Further, I predicted that this use would be concentrated
near high-density patches of lichens at the edge of burns. By focussing on fine-scale
selection by individuals and on locations solely within burns, my results provide a
unique analysis of burn use by caribou during winter.

Methods
I modelled the habitat selection strategies of the KCH at two ecological scales. First, I
used RSFs to model habitat selection of individual caribou across the winter range.
45

Second, I developed logistic regression models that focused exclusively on the use of
recent burns by caribou. These burn-focused models used only early and late winter
locations located within recent burns (≤50 years), plus a 500-m buffer outside burns.
The 500-m buffer captured imprecision in the burn boundaries and accounted for the
influence of edge on habitat selection (Joly et al. 2007).
Caribou locations
From 2012 to 2015, a total of 45 Iridium global positioning satellite (GPS) radio-collars
(ATS; Model G211OE) were deployed on Klaza caribou (all females), with fix rates
programmed at 5-, 8-, or 13-hour intervals. The fix success (i.e., a successful relocation
attempt) was high (>95%; Hegel and O’Donoghue 2015), suggesting that the inference
to caribou behaviour, primarily habitat selection, was not impeded by cover or
topographic bias (Frair et al. 2010).
The seasons of interest, early and late winter, were defined as 1 November to 31
January and 1 February to 30 April, respectively. This is consistent with Hegel and
O’Donoghue’s (2015) work on the KCH and is supported by differences in snow depth
and density (using NASA’s Modern Era Retrospective Analysis for Research and
Applications [MERRA] dataset; Russell et al. 2013) between the two seasons during the
study period (2012–2016). Both snow depth and density on the annual KCH range were
higher during late winter compared to early winter (34% and 23% higher, respectively;
Russell et al. 2013). I assumed that the existing collar data were representative of the
KCH; however, those data represented only a small proportion of individuals in the herd,
over a relatively short time period, and were focused on only one sex and age group
(i.e., adult females).
46

RSF and burn-focused model inputs
I used knowledge of the ecology of woodland caribou, with a focus on factors that may
explain the observed distribution of the KCH during the early and late winter seasons, to
select candidate environmental variables. These variables represent broad mechanistic
processes for habitat selection: vegetation, disturbance as a function of forest fire, and
terrain.
As part of an initiative to better understand habitat use by caribou, a model representing
the spatial extent of seven lichen cover classes, ranging from 0 to >50% cover, was
mapped within the KCH range (Caslys Consulting 2014). The imagery was acquired
before the collaring program began in 2012, thus any areas that burned after 22 June
2009 were not represented. Updated Landsat 7 ETM scenes (4 August 2010, 25 June
2013, and 17 July 2015) were used to examine the post-2009 burns and determine if
there were any discrepancies (e.g., large, unburned patches) within the recently
mapped burn polygons. Where unburned patches were identified within burn polygons, I
used ArcMap (version 10.3) to digitize those patches and merge them with the 2009
lichen cover raster (original, unburned lichen cover pixel values were re-assigned). For
the purposes of this analysis, I assumed no lichen cover in the remaining burn
polygons. I used a moving-window operation to calculate the density of lichen cover in
areas defined by the mean 95th centile movement distance of 8-hr fixes across all years
during both early and late winter seasons (distance = 4,432 m, SD = 765 m). Thus, my
measure of lichen density was considered as the mean percent cover of lichen within a
4,432-m window and served as the principal land cover for the analysis.

47

Caribou are known to select habitat in proximity to watercourses during winter, as the
hard, unobstructed surface may allow for more efficient travel (Schaefer and Pruitt
1991; Thomas et al. 1998; Hegel and O’Donoghue 2015). Thus, I used watercourses
from Geomatics Yukon at the 1:1,000,000 scale to determine distance to major
watercourses (rivers and named creeks) from telemetered caribou locations.
I generated a number of variables that characterized fire history. Based on the existing
literature and findings from the second chapter of this thesis, recently burned areas
were considered as any forest stands that were ≤50 years post-fire (Environment
Canada 2008). For the RSF models, habitat was represented as burned or unburned
(binary term), density of burn patches within a seasonal home range, and distance to
burns (Anderson and Johnson 2014). For the burn-focused logistic models, burn
variables included time since burn and a categorical term for burn perimeter (core, inner
500 m, and outer 500 m). To represent the burn perimeter, I generated two 500-m
buffers inside and outside mapped burn polygons (Joly et al. 2003). The area beyond
the inner 500-m buffer was considered the core of the burn, whereas the area beyond
the outer 500-m buffer was considered unburned, and was assumed to be beyond the
influence of the burn ecotone. The ‘unburned’ perimeter class was not used for the
burn-focused model.
Thomas (1991) reported that during winter, caribou moved through individual burns up
to 25 km wide, but avoided large regions with a high density of recent burns. Similarly,
Dalerum et al. (2007) suggested that caribou avoided burns at the scale of the home
range. Thus, I used a moving window to calculate the density of burns for each seasonyear. I used the 95% kernel estimate (ArcView 3.2a Home Range Extension) to
48

calculate the seasonal range for each individual. Seasonal range areas for each
individual were derived by summing the total area of all estimated kernels (one to
seven). These values were then averaged across individuals for each season-year and
served to define the extent of the moving window.
Aspect and slope were derived from a 30-m resolution digital elevation model (DEM;
Yukon Department of Environment, Information Management & Technology Branch
2015). I used those data to calculate the Topographic Position Index (TPI) representing
discrete terrain classes: valley, gentle slope, steep slope, and ridgeline (Jenness 2006;
Dickson and Beier 2007; Johnson et al. 2015) in ArcView (V. 3.2a, ESRI 2005). I used
my knowledge of the study area in combination with visual aids (hillshade overlaid with
TPI model outputs, Google Earth) to select representative values of ridgelines, valleys,
and the slope gradient. This resulted in a valley threshold of -20 TPI units, a ridgeline
threshold of 100 TPI units, and a slope value of 15 degrees (circular neighbourhood
with 1500-m radius).
I used tolerance scores and a threshold of ≤2 to test predictor variables (Table 8) for
excessive multicollinearity (Menard 2002). Using deviation coding, I generated a design
matrix of dummy variables to represent contrasts between classes of categorical
predictor variables (Hendrickx 1999).

49

Table 8. Environmental variables used in the candidate RSF and logistic models for early and
late winter distribution and use of burns by the Klaza caribou herd, west-central Yukon. A
single asterisk (*) indicates variables used only in the RSF models, and ** indicates
variables used only in logistic models.
Ecological variable

Unit

Description

Slope

degrees

Derived from DEM

Aspect (eastness)

n/a

Derived from DEM, sine-transformed from aspect
values (radians)

Aspect (northness)

n/a

Derived from DEM, cosine-transformed from aspect
values (radians)

Topographic position index

n/a

Derived from DEM to create categorical classes: valley,
ridgeline, steep slope, gentle slope

Lichen density

percent

2009 Landsat imagery representing 0–>50% lichen
cover & reclassified areas post-2009

Distance to water

M

Distance to major water courses

Burn

Binary

Mapped burn polygons (1–50 years post-fire) and
‘unburned’ areas (>50 years)

Burn age**

Years

Mapped burn polygons (1–50 years post-fire) and
‘unburned’ areas (>50 years) within 500-m of burn
polygons

Burn density*

ha/km

Distance to burns*

M

Distance to the nearest mapped burn polygon

Core**

n/a

Burned area >500-m inside the burn boundary

Burn edge**

n/a

Burned area ≤500-m inside the burn boundary

Outer buffer**

n/a

Unburned area ≤500-m outside the burn boundary

Burn variables

2

Density of recently burned area (≤50 years post-fire)
within individual’s home range

Location of burn boundary**

Winter habitat selection: Resource selection function models
Determining use and availability
I determined resource availability (random locations) by calculating the 95th percentile
movement rate of an individual within a 5-, 8-, or 13-hour interval (depending on the
individual’s GPS collar fix rate) for each season-year. This provided the radius for a
circle, centred on each caribou location, representing an individual’s potential for
movement within an ecologically relevant time frame (Arthur et al. 1996; Johnson et al.

50

2005; Gustine et al. 2006; Gillingham and Parker 2008). To reduce bias in the estimated
parameters (Northrup et al. 2013), I paired each caribou location with five locations
selected randomly from within the circle (Johnson et al. 2005). This resulted in sample
sizes of 48,430 and 242,150 used and available (random) locations, respectively.
Statistical model selection and assessment
I constructed a series of individual and season-year specific RSF models that
represented plausible hypotheses explaining the distribution of collared caribou during
early and late winter. I used paired (one use location to five random locations)
conditional logistic regression to parameterise the RSF models (Manly et al. 2002;
Johnson et al. 2005).
The most parsimonious model(s) were identified using Akaike’s Information Criterion
difference corrected for small sample sizes (∆AICc; Akaike 1973; Burnham and
Anderson 1998) and Akaike weights (AICcwi). Models with a ∆AICc of ≤2 were
considered plausible (Richards 2005; Symonds and Moussalli 2011). Then, for all
caribou in each season-year, I tabulated the number of times each of the five candidate
models was determined to be plausible (∆AICc ≤2.00). I used weighted counts (each
individual divided by its total number of top models) to account for bias resulting from a
different number of top models per individual (one to five top models possible). The
unweighted coefficients for each individual’s single top model were averaged over each
season-year to provide an indication of overall consistency in herd selection strategies
(Anderson and Johnson 2014). Error was represented using 95% confidence intervals
derived from the variance of averaged coefficients. Variance was estimated using the
following equation (Marzluff et al. 2004):
51

̂) =
̅̅̅̅
Var (𝛽
∑
j
(𝑛−1)
1

𝑛
𝑖=1

̂
̅ 2
(𝛽̂
𝑖𝑗 − 𝛽j )

where β represents individual coefficient values, and n represents the number of
animal-model combinations. Evaluating predictive capability is essential, as model
selection is a relative measure of the most parsimonious model within the candidate set,
not a measure of model fit. I used the Receiver Operating Characteristic (ROC) with
independent cross validation to assess the predictive accuracy of the RSF models.
Using a jackknifing procedure, I sequentially withheld the data for one cluster group
(one used and five paired availability locations) during the model-fitting process; the
resulting model (N-1) and the withheld record were then used to generate an
independent prediction. I then calculated the area under the curve (AUC) generated
from the nonparametric ROC. An AUC of 0.50 suggests a model has no predictive
power, while a score of 1.00 suggests perfect prediction (Boyce et al. 2002). The ROC
was a conservative measure of predictive accuracy as it assumed a case-control
sample design (Boyce et al. 2002). For each individual caribou and season-year, model
fit was assessed for the top-selected model only (i.e., the highest AICcwi value).
Selection ratios for burn variables
For the RSF models, categorical burn age classes, burn perimeter use, and burn size
classes (small, medium, and large) were initial model inputs; however, several of these
classes had to be dropped due to near-complete separation (i.e., complete avoidance of
an available resource; Gillingham and Parker 2008). To incorporate these resources of
interest, I determined the ratio of used (GPS collar locations) to available (random)
locations across years for each season; with available locations divided by five (the

52

number of available locations generated per used location; Gillingham and Parker
2008). The selection ratio for each individual caribou was treated as a replicate after
which I averaged ratios to represent selection for the sampled population.
Burn use: Logistic regression models
Approximately 12% of winter locations (n = 5,663) were located within 500 m of the
boundaries of a mapped burn polygon. Thus, I used a grid of habitat selection units
(HSUs) overlaid on burns to define individual sampling units for modeling the use of
burns by caribou relative to ecological variables (Figure 10). HSUs were contained
within a defined area of use (AOU) where one or more caribou were located within a
500-m buffer of recent burns within the KCH annual range. With all burns within the
AOU identified, I used two techniques to define the size of the HSU: 1) the 2.5th centile
area (representing the lower end of values in the 95% central range) of burns in the
AOU, and 2) the smallest individual burn with >10 caribou locations, >1 individual, and
>1 season of use. Both techniques resulted in a HSU size of 47 ha (a length of 686 m
per side).

53

54

Figure 10. Example of the sampling design used for the burn-focused models. Habitat selection units (HSUs) were a grid overlaid on
recent (≤50 years old), occupied burns (plus a 500-m buffer) in the Klaza caribou range, west-central Yukon.

I quantified use as the product of the proportion of collared caribou and the median
duration of use of individual caribou within each HSU:

𝑯𝑺𝑼𝒊 =

𝑵(𝐥𝐨𝐜𝐚𝐭𝐢𝐨𝐧𝐬)𝒊𝒌
𝑵(𝐜𝐚𝐫𝐢𝐛𝐨𝐮)𝒊
̃(
∙ 𝑿
)
∑𝒋𝒌 𝑵(𝐥𝐨𝐜𝐚𝐭𝐢𝐨𝐧𝐬)
∑𝒋 𝑵(𝐜𝐚𝐫𝐢𝐛𝐨𝐮)

The proportion of individual caribou in each HSU was defined as the total number of
collared caribou in that HSU (N(caribou)i) divided by the total number of collared caribou
occupying any HSU during that season-year (∑jN(caribou)). The relative duration of use
reflected the number of collar locations in that HSU for each individual (N(locations)ik)
divided by the total number of locations in any HSU (∑ikN(locations)) for that individual
̃ ) for all collared
in that season-year. I calculated the median relative duration of use (𝑿

caribou that occupied the HSU. For example, if Caribou #1 had three locations in HSU
#A1 and a total of 30 locations in all HSUs during late winter 2014, ‘use’ would equal
0.1. Given the relatively small number of samples in each HSU, the median value was
then derived from all individuals in a given HSU. To prevent one metric, proportion of
caribou or median proportion of locations, having a greater influence on the final score
for the HSU, prior to calculating the product of these two, I performed a linear stretch to
scale the values of each metric between 0 and 1 (Johnson et al. 2004b):

𝒘
̂=

𝒘(𝒙) − 𝒘𝒎𝒊𝒏
𝒘𝒎𝒂𝒙 − 𝒘𝒎𝒊𝒏

Where w(x) is the proportion of caribou or median proportion of locations and wmin and
wmax are the smallest and largest use values, respectively.

55

After multiplying the two stretched proportions, the values characterizing use were
relatively small, thus I transformed each continuous HSU score into a binary variable.
For this purpose, I used the ROC to identify the classification breakpoint between HSUs
with high (1) versus low (0) use. I fit the most parameterized model (greatest number of
predictor variables) with the dependent variable defined by percentile values of the HSU
scores (e.g., 10th, 25th...90th). The percentile with the greatest area under the curve
defined the breakpoint for the continuous HSU values for that season-year. To avoid
creating an unbalanced data set (e.g., considerably more high (1) versus low (0) use
values as the threshold approaches the 5th percentile), I selected threshold values
identified between the 10th and 90th percentiles.
Statistical model selection and assessment
I used logistic regression (binomial family, logit link function) to identify environmental
and terrain factors that differentiated HSUs with high (1) versus low (0) use (Table 8).
As with the RSF analysis, I used AICc to select the most parsimonious model(s) and the
jackknifing cross-validation procedure to test the predictive ability of those models. All
statistical analyses were completed using Stata/IC (Version 12.1; StataCorp 2011)
statistical software with the desmat add-in (Hendrickx 1999) for deviation coding of
categorical variables.

Results
Forty-two GPS-collared female caribou provided 48,430 locations across nine seasonyears from 2012–2016 (Table 9). Only two individuals were collared during late winter
2012, providing an insufficient sample for modelling habitat selection during that
season; those data were excluded from seasonal comparisons across years. Most
56

collared individuals had more than one season-year of location data, and some had up
to four season-years. Of these locations, 5,663 (12%) were located within 500 m of
recent burns (≤50 years old), and included locations from 64–100% of collared
individuals, depending on year and season (early and late winter; Table 9).

57

Table 9. Summary of GPS-collar locations within 500-m of burns in early and late winter for the Klaza caribou herd, west-central
Yukon, 2012–2016.
Season

Early
winter

Late
winter

Year

Total collared
individuals (within
500 m of burns)

Locations within
500-m of burns

Total locations

% of locations
within 500 m of
burns

2012/2013

28(18)

572

7,159

8

2013/2014

24(19)

466

6,965

7

2014/2015

28(19)

295

8,234

4

2015/2016

10(9)

323

2,618

12

All years

-

1,656

24,976

7

2012

2(2)

158

257

61

2013

27(23)

1,019

6,775

15

2014

23(20)

1,235

6,882

18

2015

27(24)

1,199

7,136

17

2016

10(10)

396

2,404

16

All years

-

4,007

23,454

17

Number of individuals
with x years of data
1

2

3

4

12

12

18

0

11

12

15

2

58

Resource selection function models
Five candidate models were developed to explore individual seasonal habitat selection
strategies for the KCH (Table 10), totaling 885 individual by season-year model
combinations. Due to the high number of individual-season-year combinations, the
candidate model set was limited to five, relatively complex models. These models
represented ecologically plausible hypotheses explaining the distribution of collared
caribou without comprising the inherent complexity of resource selection (Boyce and
McDonald 1999).
Table 10. Candidate resource selection function models for the Klaza caribou herd of westcentral Yukon, during the early and late winter, 2012–2016. GM = general model, BM = burn
model.
Model name

Model parameters
2

2

2

2

2

2

2

2

2

2

GM1

Eastness+Northness+Slope+Slope +Elevation+Elevation +TPI+Lichen density+Distance
2
to water+Distance to water

BM1

Eastness+Northness+Slope+Slope +Elevation+Elevation +TPI+Lichen density+Distance
2
2
to water+Distance to water +Burn density+Distance to burns+Distance to burns +Burned

BM2

Eastness+Northness+Slope+Slope +Elevation+Elevation +TPI+Lichen density+Distance
2
to water+Distance to water +Burned

BM3

Eastness+Northness+Slope+Slope +Elevation+Elevation +TPI+Lichen density+Distance
2
2
to water+Distance to water + Distance to burns+Distance to burns +Burned

BM4

Eastness+Northness+Slope+Slope +Elevation+Elevation +TPI+Lichen density+Distance
2
to water+Distance to water +Burn density

Model selection
For the early winter season, there was considerable model selection uncertainty, with
no single model consistently selected more than others (Figure 11). Model BM1 (the
most parameterized model) was selected as a top model in three of four years. For the
late winter season, all models that included burn variables (BM1–4) were selected as
top models more often than the general model (GM1). Model BM1 was selected as a
top model (∆AIC ≤2) for all years, with BM3, differing only by the exclusion of burn
density, as the second most selected model (Table 10, Figure 12).
59

Weighted count of top models (∆AIC ≤2)

9.00
8.00

n=28

n=28

n=24

7.00
6.00
BM1

5.00

n=10

4.00

BM2
BM3

3.00

BM4

2.00

GM1

1.00
0.00

Season-year

Weighted count of top models (∆AIC ≤2)

Figure 11. Weighted count of each candidate RSF model, as defined in Table 10, selected as a
top model (∆AIC ≤2) for the early winter season (2012–2016). The number of collared
individuals in the Klaza caribou herd in each season-year is indicated above the bars.

18.00

n=27

16.00
14.00

n=23

12.00

BM1

n=27

BM2

10.00

BM3

8.00

n=10

6.00

BM4
GM1

4.00
2.00
0.00
LW 2013

LW 2014

LW 2015

LW 2016

Season-year

Figure 12. Weighted count of each candidate RSF model, as defined in Table 10, selected as a
top model (∆AIC ≤2) for the late winter season (2013–2016). The number of collared
individuals in the Klaza caribou herd in each season-year is indicated above the bars.

60

Winter habitat selection coefficients
During early winter, individual caribou consistently demonstrated nonlinear selection for
lichen density, elevation, and distance to water (Table 11; Appendix D: Figure 18).
Caribou consistently avoided valleys, demonstrated a nonlinear avoidance of increasing
slopes, and variation among individuals in selection or avoidance of aspect, ridgelines,
and gentle and steep slopes varied across years. Individuals generally avoided recent
burns and selected for areas further from burns; however, individual strategies for burn
density varied across the years.
During late winter, RSF models indicated that the KCH collectively demonstrated
nonlinear selection for increasing lichen density, elevation, slope, and distance to water,
and southern aspects and ridgelines (Table 12; Appendix D: Figure 19). Valleys and
north-facing aspects were consistently avoided, while selection strategies for east and
west-facing slopes and gentle and steep slopes were inconsistent across all years. The
selection of recent burns was variable across year and among caribou. Individuals
generally avoided recent burns in 2014 and 2015, and selected for recent burns in
2016, whereas the overall selection strategy of individuals monitored during 2013 was
unclear. Overall, individuals selected for increasing distance to burns in all years except
in 2013, lower burn density in 2014 and 2016, and higher burn density in 2013 and
2015.
For the burn-related variables (burned/unburned, burn density, and distance to burns),
individuals with three season-years of data were examined for consistency in selection
strategies across years (Appendix D: Figure 20 and Figure 21). During early winter, one
individual consistently selected for increasing distance to burns, while another
61

consistently selected for decreasing distance to burns (N = 5; Appendix D: Figure 20d),
and one individual consistently avoided burned areas (N = 13; Appendix D: Figure
20a,b). One individual consistently selected for decreasing burn density (N = 13;
Appendix D: Figure 20c). During late winter, five of nine individuals consistently selected
for increasing distance to burns (Appendix D: Figure 21d), two and three individuals
consistently selected for burned and unburned areas, respectively, and one individual
consistently selected for increasing burn density (N = 12; Appendix D: Figure 21a,b).

62

Table 11. Weighted counts for each coefficient by early winter season (EW) from top-ranked (∆AIC ≤2) resource selection models
for individual caribou in the Klaza herd of west-central Yukon. For each year, the number under the – indicates the weighted
count of individual top models that showed significant avoidance for that parameter; the number under the + indicates the
weighted count of individual top models that showed significant selection for that parameter. The count of significant β
coefficients (P<0.05) in each year is shown in bold italics for each – or + coefficient.
Parameter

EW 2012/2013

EW 2013/2014

EW 2014/2015

EW 2015/2016

Burned

12.13

P<0.05
9.98

+
6.49

P<0.05
4.00

7.66

P<0.05
5.66

+
11.80

P<0.05
7.64

14.55

P<0.05
8.15

+
4.73

P<0.05
1.33

2.25

P<0.05
0.50

+
1.83

P<0.05
1.00

Unburned

6.49

4.00

12.13

9.98

11.80

7.64

7.66

5.66

4.73

1.33

14.05

8.15

1.83

1.00

2.25

0.50

Burn density

8.89

4.83

3.82

2.83

7.73

2.66

2.33

3.49

1.83

10.80

1.33

0.25

0.25

4.83

1.50

Distance to burns

4.08

3.00

8.07

5.24

3.58

1.50

12.48

10.15

6.41

4.66

9.14

2.16

0.75

0.75

Distance to burns²

8.90

6.91

3.25

2.00

13.56

10.15

2.50

0.50

9.81

3.66

5.74

1.00

1.00

0.50

Lichen density

4.31

1.00

23.58

11.96

2.98

20.96

9.65

0.00

27.91

22.43

1.00

Lichen density²

21.26

10.96

6.63

1.99

20.21

5.99

3.73

Elevation

4.98

22.91

14.31

2.00

Elevation²

20.42

13.65

7.47

0.66

21.94

17.95

2.00

Slope

21.07

6.98

6.82

2.00

18.96

7.16

4.98

Slope²

16.93

4.00

10.96

3.00

12.97

0.99

Northness

21.91

10.47

5.98

1.00

12.98

Eastness

13.98

4.74

13.91

2.31

17.13

Ridgeline

10.95

1.99

16.94

1.00

Steep slope

10.95

16.94

Gentle slope

10.48

17.41

Valley

5.72

22.17

2.00

3.00

Distance to water

8.97

2.82

18.92

9.96

7.47

Distance to water²

20.92

11.63

6.97

3.99

15.96

21.94

27.25
17.29

16.44

0.00

0.66
27.91

4.00

7.99
22.60

7.99

5.00

1.00

0.00

8.99

27.91

20.95

0.00

8.99

8.99

0.00

0.99

25.92

10.95

1.99

8.99

5.33

0.00

10.97

3.00

9.98

5.00

10.96

1.00

16.95

8.98

6.81

2.00

11.60

7.82

1.00

16.12

4.16

6.00

1.00

17.94

3.00

11.97

1.99

11.97

8.99

17.93

1.00

2.00

4.99

10.96

2.99

4.00

1.33

16.31

5.00

3.99

9.32

1.00

18.59

3.98

7.33

7.97

1.00

19.94

4.99

3.00

5.99

2.00

13.94

1.99

13.97

3.00

1.00

7.99

1.00

21.26

5.00

2.00

6.99

2.66

3.99

0.99

5.00

1.00

20.94

5.33

6.65

1.00

16.47

6.99

13.97

4.99

13.94

5.96

5.00

6.99

7.98

1.33

10.95

5.30

16.96

5.32

3.99

6.99
1.00

4.99
5.00

1.00

1.00

2.00
1.00

1.66

63

Table 12. Weighted counts for each coefficient by late winter season (LW) from top-ranked (∆AIC ≤2) resource selection models for
individual caribou in the Klaza herd of west-central Yukon. For each year, the number under the – indicates the weighted count of
individual top models that showed significant avoidance for that parameter; the number under the + indicates the weighted count
of individual top models that showed significant selection for that parameter. The count of significant β coefficients (P<0.05) in
each year is shown in bold italics for each – or + coefficient.
Parameter

LW 2013

LW 2014

LW 2015

LW 2016

Burned

11.00

P<0.05
5.50

+
11.99

P<0.05
6.33

8.99

P<0.05
1.99

+
8.16

P<0.05
6.33

11.09

P<0.05
5.09

+
10.31

P<0.05
3.99

1.00

P<0.05
1.00

+
8.15

P<0.05
2.00

Unburned

11.99

6.33

11.00

5.50

8.16

6.33

8.99

1.99

10.31

3.99

11.09

5.09

8.15

2.00

1.00

1.00

Burn density

2.50

1.00

14.33

13.33

7.49

4.99

8.00

6.00

7.69

4.53

5.69

3.00

1.66

0.66

2.83

1.33

Distance to burns

6.16

2.33

17.50

12.50

4.66

3.33

13.16

9.83

7.50

5.50

8.72

3.86

2.99

2.33

5.00

3.00

Distance to burns²

19.50

12.50

4.16

0.66

13.49

10.50

4.33

2.00

10.72

4.66

5.50

3.00

5.00

3.00

2.99

2.33

Lichen density

3.00

23.99

14.00

4.00

18.98

14.48

5.16

21.78

11.47

0.83

9.15

5.99

Lichen density²

23.50

6.00

3.00

21.45

5.49

0.33

8.49

20.99

17.99

25.95

23.96

0.00

15.50

3.49
26.99

16.98
24.99

7.66

Elevation

0.00

Elevation²

26.99

24.99

0.00

Slope

9.00

1.00

17.99

7.49

3.00

Slope²

22.99

12.99

4.00

1.00

20.98

10.99

Northness

25.99

23.99

1.00

20.98

16.98

Eastness

18.00

9.00

8.99

1.99

10.00

Ridgeline

4.00

1.00

22.99

10.00

1.00

Steep slope

14.99

1.00

12.00

Gentle slope

13.00

Valley

20.99

3.00

6.00

Distance to water

12.00

5.00

14.99

Distance to water²

17.49

8.99

9.50

13.99

1.99
21.98

19.99

1.00
19.98

6.99

2.00

11.98

0.99
25.95

23.96

0.99

7.98

1.00

18.96

19.96

9.98

6.98

6.98

5.99

1.49
9.98

9.98

9.98

9.98

0.00

4.00

1.00

5.98

1.00

7.98

2.33

2.00

1.00

5.98

2.32

4.00

24.94

23.94

2.00

12.98

4.00

12.96

1.00

13.98

4.99

3.98

6.00

2.00

1.00

21.98

12.99

4.00

1.00

22.94

5.98

5.00

4.98

1.99

12.48

2.99

10.50

1.00

17.44

1.99

9.50

1.00

4.98

5.00

15.99

3.00

6.99

12.63

3.49

14.31

1.50

3.98

19.98

3.99

3.00

20.28

5.66

6.66

5.99

3.00

1.00

19.98

13.99

5.16

1.00

21.78

13.96

2.00

7.98

4.00

17.32

12.99

5.66

1.00

21.94

8.97

5.00

1.00

7.99

1.99

3.99

2.00

6.48

6.00
1.00

3.50
2.00

64

Model fit
ROC scores were generally higher for the top-ranked (highest overall AICw) late winter
RSF models when compared to early winter models. Not all seasonal RSFs had good
predictive power, but the average scores across the late winter seasons consistently
exceeded the 0.7 threshold for good model performance (Table 13).
Table 13. Mean Receiver Operating Characteristic (ROC) scores (area under the curve) for
top-ranked (highest AICw) resource selection function (RSF) models for individual Klaza
caribou, 2012–2016.
Season-year

Individuals

ROC (mean)

ROC (min)

ROC (max)

All years (early winter)

89

0.70

0.53

0.87

EW 2012/2013

28

0.68

0.58

0.87

EW 2013/2014

24

0.67

0.53

0.79

EW 2014/2015

28

0.72

0.63

0.84

EW 2015/2016

9

1

0.76

0.70

0.82

All years (late winter)

87

0.75

0.61

0.91

LW 2013

27

0.77

0.67

0.87

LW 2014

23

0.76

0.61

0.88

LW 2015

27

0.74

0.66

0.89

LW 2016

10

0.76

0.67

0.91

1

The jackknifing procedure for one individual would not converge, thus it was excluded.

Seasonal selection ratios
Seasonal selection ratios were derived for measures of burn age, burn perimeter, and
burn size (Figure 13). During early winter, many individual caribou selected the 21–30
and 31–40 year burn age classes. During late winter, there was relatively little variability
around the pooled ratio estimates for all burn age classes, with the exception of the 1–
10 year age class, which showed more selection than other age classes. For both early
and late winter, there was considerable individual variation in the selection or avoidance
of the outer 500-m perimeter of burns, although individuals selected this class more
than others (Figure 13). Similarly, individuals had a variable response to small burns,
with greater selection of this size class compared to others.
65

D  (DUO\ZLQWHU
6

 ±\HDUV
 ±\HDUV
 ±\HDUV
 ±\HDUV
 ±\HDUV
 !\HDUV8QEXUQHG

Selection Ratio

5
4
3

E  /DWHZLQWHU
10
9
8
7
6
5
4
3
2
1
0

Individuals
Pooled

2
1
0
1

2

3
4
Burn age class

Selection Ratio

4
 &RUH
 ,QQHUP
 2XWHUP
 8QEXUQHG

3

5

6

Pooled

1

1

0

0
4



6

4

Pooled

3
2
1
0
1

2
Burn size class

5

6



1

2
3
Burn perimeter class

4

1

2
Burn size class

3



9
8
7
6
5
4
3
2
1
0

Individuals

 /DUJH
 0HGLXP
 6PDOO

5

3
4
Burn age class

3
2

2
3
Burn perimeter class

2

4

Individuals

2

1

Selection Ratio

1



3





)LJXUH ,QGLYLGXDO DQG SRROHG DFURVV DOO DQLPDOV  VHDVRQDO VHOHFWLRQ UDWLRV XVHG YHUVXV DYDLODEOH
ORFDWLRQV RIEXUQDWWULEXWHV EXUQDJHEXUQSHULPHWHUDQGEXUQVL]H IRU.OD]DFDULERXGXULQJHDUO\ D DQG
ODWH E ZLQWHU±6HOHFWLRQUDWLRVJUHDWHUWKDQRQHLQGLFDWHVHOHFWLRQZKHUHDVYDOXHVOHVVWKDQ
RQHLQGLFDWHDYRLGDQFH




Burn-focused logistic models
During early winter, the proportion of collared caribou in a HSU was relatively consistent
across years, with the highest use by collared individuals in 2015/2016. The mean
duration of use was lowest in 2015/2016 (Figure 14). During late winter, the mean
proportion of collared caribou in a burn (i.e., HSU) varied throughout the study period,
but was greatest in 2016, and the mean duration of use was greatest in 2013 (Figure
14). Overall mean combined use (proportion and use) of burns by the KCH was highest
during late winter 2013, although the use of burns during late winter decreased

Relative use of burns (stretched)

considerably from 2014 to 2016 (Figure 14).

0.40

Proportion of use

0.35

Duration of use

0.30
0.25
0.20
0.15
0.10
0.05
0.00

Figure 14. Relative mean proportion and duration of burn use by collared caribou during
individual early (EW) and late (LW) winter seasons by Klaza caribou, west-central Yukon,
2012–2016. All values were scaled between 0 and 1 using a linear stretch.

Model selection
A set of eight candidate models (M1–8) were developed to explain the relative use of
HSUs by caribou, with all but one (M1) containing variables that represented burn
characteristics (Table 14). For each season-year, one to three models were identified as

67

a top model (∆i AICc ≤2.00). Model M2 (Table 14) was selected the most often across all
season-years (seven times), whereas models M4 and M7 (Table 14), which contained
the most parameters, were never selected as a top model (Table 15). Model M2, which
explained use with burn age, lichen density, and burn perimeter, had relatively few
parameters compared to the other models, but it was selected more often than a
simpler model (M3) that did not include burn age.
Table 14. Candidate logistic regression models used to differentiate Habitat Selection Units
(HSU) with relatively high and low use by monitored caribou of the Klaza herd during early
and late winter, west-central Yukon, 2012–2016.
Model name
M1
M2
M3
M4
M5
M6
M7
M8

Model Parameters
2
Northness+Eastness+Slope+Elevation+Elevation +TPI+Lichen density+Lichen
2
density +Distance to water
2
Burn age+Lichen density+Lichen density +Burn perimeter
2

Lichen density+Lichen density +Burn perimeter
2
Northness+Eastness+Slope+Elevation+Elevation +TPI+Lichen density+Lichen
2
density +Distance to water+Burn age+Burn perimeter
2
2
Slope+Elevation+Elevation +Lichen density+Lichen density +Burn age+Burn perimeter
2

2

Elevation+Elevation +TPI+Lichen density+Lichen density +Burn age+Burn perimeter
2
Northness+Eastness+Slope+Elevation+Elevation +TPI+Lichen density+Lichen
2
density +Distance to water+Burn perimeter
2
2
Elevation+Elevation +Lichen density+Lichen density +Burn age+Burn perimeter

TPI = Topographic Position Index

68

Table 15. Summary of statistics used to select the most parsimonious logistic regression
models for understanding caribou use of Habitat Selection Units in the Klaza caribou herd
range, west-central Yukon.
Season-year
Early winter 2012/2013
Early winter 2013/2014

Early winter 2014/2015

Early winter 2015/2016

Late winter 2013
Late winter 2014
Late winter 2015
Late winter 2016

Rank

Model name

AICci

AICc wi

∆i AICc

1

M2

197.57

0.80

0.00

1

M2

113.91

0.33

0.00

2

M8

113.96

0.32

0.05

3

M6

115.57

0.14

1.66

1

M6

59.38

0.42

0.00

2

M2

59.85

0.33

0.48

1

M6

134.48

0.39

0.00

2

M1

136.00

0.18

1.51

3

M3

136.01

0.18

1.53

1

M3

255.04

0.64

0.00

2

M2

256.88

0.26

1.84

1

M8

179.86

0.84

0.00

1

M5

224.80

0.65

0.00

2

M8

226.64

0.26

1.84

1

M2

230.28

0.65

0.00

Caribou use of burns
In early winter, burn age, burn perimeter use, lichen density, and TPI classes (ridgeline,
steep slope, gentle slope, and valley) were significant predictors of the use of HSUs by
caribou (Appendix E: Table 26). Use increased with burn age, lichen density, and within
the inner and outer 500 m of burns, and decreased within the core of burns. Compared
to other TPI classes, ridges and gentle slopes had a positive influence on use, whereas
steep slopes did not. The influence of valleys on use was not consistently positive or
negative (Appendix E: Table 26). In late winter, burn age, burn perimeter use (core,
inner 500 m, and outer 500 m), slope, and elevation were significant predictors of the
use of HSUs by caribou (Appendix E: Table 27). Use increased with burn age,
elevation, and within the outer 500 m of burns, whereas the core of the burn and
decreasing slope had a negative influence on use (Appendix E: Table 27).

69

Model fit
The ability of logistic regression models to explain caribou use of burns, ranged from
relatively poor (AUC = 0.55, SE = 0.05) to very good (AUC = 0.89, SE = 0.04); however
most were considered fair (i.e., ~0.70; Table 16). On average, models for early winter
had more power to explain the distribution of caribou in burns than the models for late
winter (𝑥̅ = 0.73, SD = 0.08; 𝑥̅ = 0.66, SD = 0.07, respectively).
Table 16. Area under the curve for Receiver Operating Characteristic (ROC) scores calculated
for the top models representing caribou use of Habitat Selection Units (HSUs) in the Klaza
caribou herd range. The threshold percentiles that differentiated high and low use HSUs are
indicated at the bottom along with the number of HSUs used in each season-year. LW = late
winter, EW = early winter.
Model name

EW
2012/2013

EW
2013/2014

EW
2014/2015

EW
2015/2016
0.75

LW
2013

LW
2014

LW
2015

-

-

-

LW
2016

M1

-

-

-

M2

0.71

0.60

0.83

M3

-

-

-

0.73

0.61

-

-

-

M4

-

-

-

-

-

-

-

-

M5

-

-

-

-

-

-

0.73

-

M6

-

0.68

0.89

0.74

-

-

-

-

-

-

-

-

-

-

M7

0.61

0.55

M8

-

0.67

-

-

-

0.76

0.69

-

Threshold percentile

50

10

90

75

75

10

90

25

n (HSUs used)

172

200

125

139

234

397

388

210

Discussion
These findings substantiate conclusions from other studies suggesting that caribou
(both woodland and migratory) collectively avoid recent burns during winter (Schaefer
and Pruitt 1991; Thomas et al. 1996; Joly et al. 2003, 2007, 2009; Collins et al. 2011;
Hegel and O’Donoghue 2015). While much of the previous research has focused on
population-level selection patterns, I observed considerable variation in habitat selection

70

strategies among individuals, years, and winter seasons. That variation was most
apparent when considering burn-related resources.
A large number of past works have documented the winter ecology of caribou on
burned landscapes. However, this is one of the first studies to quantify the resource
selection strategies of caribou that have chosen to occupy burns. For the KCH, 12% of
locations were located within 500 m of recent burns and 20% of the annual range was
burned in the last 50 years. Active use of burns is not a trivial component of the winter
distribution and behaviour of these caribou and likely other herds that are found in fireprone ecosystems. Use of burns by the KCH varied by season and year; however,
burns were used more consistently during early winter. Recent burns (≤50 years old)
are likely sub-optimal habitat for the KCH (see Chapter 2), but observed selection for
areas with a high density of lichen suggests that caribou seek out remnant and reestablishing terrestrial lichen across burned areas.
Variation of habitat selection strategies in burns
As expected, there were differences in the selection and avoidance of habitats among
individual caribou (Gillingham and Parker 2008; Lesmerises and St-Laurent 2017);
however, selection coefficients revealed some consistent strategies. During late winter,
caribou consistently selected for areas with relatively greater lichen density and higher
elevations, and avoided northern aspects (Table 12; Appendix D: Figure 19). Although
somewhat less consistent, caribou typically selected for steeper slopes, ridgelines, and
areas further from major watercourses (Table 12; Appendix D: Figure 19). During early
winter, the selection strategies of caribou were more variable, with only increasing

71

elevation and lichen density being consistently important (Table 11; Appendix D: Figure
18).
The collective response of the KCH to burns is similar to other studies that used pooled
models to examine the selection strategies of caribou during the winter (Robinson et al.
2012; Hegel and O’Donoghue 2015). Averaged model coefficients indicated that the
KCH avoided burns in all years except early and late winter 2015/2016 (Appendix D:
Figure 18 and Figure 19). This could be a result of the reduced number of collared
individuals in 2015/2016 (n = 10) compared to other years (n = 23–28). With a small
sample of individuals, the average response to burns could be skewed by outliers
(Gillingham and Parker 2008). Also, as the collective response of the KCH is merely an
averaged value, there is likely bias towards individuals that had more model selection
uncertainty (i.e., more top models selected). Contrary to the averaged selection
strategies of caribou, counts of significant burn coefficients (when adjusted for the
number of top models per individual) demonstrated that more individuals were selecting
for burns in three out of four late winter seasons (2013, 2014, and 2016; Table 12),
whereas more individuals were selecting for burns in only one out of four early winter
seasons (2014; Table 11). The discrepancy between averaged coefficients and
coefficient counts suggests that negative responses to burns, as indexed by the
magnitude of the coefficients, may be greater than positive responses. The scope of this
analysis was also limited to averaging coefficients from the top selected model only to
provide a simple indication of overall consistency in herd selection strategies. Using
AICc weights to average all plausible models for each individual may have reduced the
discrepancy between averaged coefficients and coefficient counts.

72

Selection for the factors that define habitat can vary among scales (Rettie and Messier
2000; Johnson et al. 2001), thus inferences made at one scale do not necessarily hold
true at another (Apps et al. 2001; Johnson et al. 2001; Shepard et al. 2007). While other
studies have focused on selection strategies at the scale of the seasonal, home, or
annual range (e.g., Dalerum et al. 2007; Anderson and Johnson 2014; Hegel and
O’Donoghue 2015), I examined habitat selection at the individual level. This allowed me
to better understand the fine-scale foraging behaviour of individual caribou in relation to
burns. Population-level analyses represent the average response of the sample, yet no
one individual may be represented by the ‘average’ animal. Although more difficult to
interpret, the generation of RSF models for each individual allowed me to document the
full range of selection strategies demonstrated by collared caribou.
Individuals were not uniform in their selection strategies across seasons or years, with
few exhibiting consistent selection or avoidance of burns (Appendix D: Figure 20 and
Figure 21). This variability is likely due to a number of interacting factors, including
nutritional and maternal condition, predation risk, landscape diversity, and differences in
winter or range conditions among years. Lone caribou behave differently than adult
females with calves (Bergerud et al. 1984; Rettie and Messier 2000; Lesmerises et al.
2016), with resource use most variable when females and calves are more vulnerable to
predation or when food resources are limited (Gustine and Parker 2008). This variability
creates a difficult challenge for wildlife managers: meaningfully incorporating individual
variability in selection strategies into population-level management objectives.
Habitat selection strategies were more variable and the use of burns was greater during
early winter than during late winter. That variability in habitat selection also resulted in
73

reduced predictive accuracy of individual models compared to late winter models (Table
13). During early winter, snow depths are lower and energetic reserves are higher,
allowing for greater ease of movement on the landscape. By late winter, caribou have
reduced forage intake and instead focus on minimizing energetic costs, especially those
related to movement (Gustine et al. 2006).
Snow conditions vary annually and seasonally, influencing the distribution of caribou.
Previous late winter RSF models completed for the KCH were limited by the lack of a
snow depth metric (EDI 2013; Hegel and O’Donoghue 2015). I used the contrast
between early and late winter as a coarse representation of snow conditions throughout
a typical winter. Hegel and O’Donoghue (2015) noted that typical snow conditions on
the KCH range were reversed in the winter of 2013, with low snow depths at high
elevations and deeper snow at low elevations. This year-to-year variation in snow
conditions affects seasonal distribution, potentially resulting in different parts of the
range being used, including burns. Although I could not measure range-wide snow
conditions, such data could considerably improve our understanding of selection
strategies among years.
Close encounters: How Klaza caribou use burns during winter
Analyses focused on the use of burned habitat suggested that the KCH did not actively
select recent burns. Instead, caribou appear to regularly encounter burns, primarily
small ones (Figure 13), and opportunistically use remnant lichens within the burn
perimeter. During early winter, burn use models with higher threshold values (e.g., 90th
percentile; 2014/2015, 2015/2016) performed better than lower threshold models (e.g.,
10th percentile); however, there was no apparent pattern for late winter models
74

(Appendix E: Table 25). Lower threshold values of high versus low use HSUs were
representative of more general use, whereas higher threshold values represented more
selective use of HSUs. This indicates that caribou in burns during early winter are using
them more selectively, and caribou in burns during late winter are encountering and
using them randomly, potentially for traveling between high density lichen patches. Joly
et al. (2003) found a similar pattern, and suggested it was a result of early successional
vascular forage being more accessible in burns during the early winter season. The use
of early seral habitats, including burns, was also observed by Anderson and Johnson
(2014), indicating that early successional forage in burns, in addition to lichen, may
supplement winter diets. This is paralleled by the fact that burn age was generally a
poor predictor of burn use by caribou (Appendix E: Table 26 and Table 27), suggesting
that use is temporally non-linear and reflects the successional patterns of both lichen
and non-lichen forage species (e.g., herbs, horsetails, shrubs, etc.).
My data suggest that caribou focus their use of burns at the edges (both the inner and
outer 500 m); typically in proximity to areas of high lichen density (Figure 15), which is
consistent with findings from other herds (Joly et al. 2003, 2007; Anderson and Johnson
2014). The core of burns is often where the most intense and severe effects to
vegetation and overall forest structure are observed (Joly et al. 2007). Additionally, the
core of burns may be more exposed to sun and wind, resulting in crusty, high-density
snow that restricts movement (Thomas et al. 1998). Use of burn edges and strong
avoidance of the core indicates that edges not only provide nutritional advantages, but
may also facilitate caribou movement within the winter range (Thomas et al. 1998).

75

The analysis of burn use is limited to revealing ecological factors that explain the
intensity of occurrence of caribou in burns; we cannot infer the behaviour of caribou that
occupy burns. Movement analyses may provide some insight on the activities of caribou
as they interact with burned habitats (e.g., foraging, quickly transiting burns, bedding
down). Although beyond the scope of this research, this highlights the need to better
understand caribou movement patterns as they relate to burned and unburned habitats.

76

77

Figure 15. GPS collar locations of Klaza caribou during winter (2012–2016) in relation to recent burns (≤50 years old) and lichen
density, west-central Yukon.

Conclusions
On average, burns were avoided during winter; however, individual Klaza caribou may
benefit from the use of burns during early winter and to a lesser degree during late
winter when access to remnant patches of lichen is limited by snow conditions. I
observed that individual Klaza caribou exhibited considerable variability in how they
selected habitat during winter, especially during early winter, when snow conditions
were presumably more favourable to movement and accessing terrestrial lichens or
vascular plants (Pruitt 1959; Fleishman 1990; Joly et al. 2003). Recent burns (≤50
years) were used by Klaza caribou, but similar to the conclusions from Chapter Two of
this thesis, the individual RSF and burn-focused models demonstrated that burn age
alone did not provide a consistent indication of habitat selection or burn use. In addition
to age, managers should consider the season of use, the effect of burn size, and the
burn landscape — burn edges near lichen patches are used more than the core of
burns, but areas generally further from burns are selected.
This study demonstrates that collective behaviour does not necessarily mirror that of
individuals. Although the variability of habitat selection strategies among individuals may
hinder the applicability of these models to management (Gustine and Parker 2008), it is
important to understand that variation (Thomas and Taylor 1990) and assess
differences between pooled and individual models. As noted by others, (e.g., Gustine et
al. 2006; Gustine and Parker 2008), wildlife managers should aim to identify this
variability within a population by using a multi-scale approach to define availability,
quantifying individual variation, and then pooling similar strategies. This will result in

78

better predictions of habitat selection that account for the diversity of strategies within
the population.

79

CHAPTER FOUR: THESIS CONCLUSIONS
Summary
Lichens are essential winter forage for woodland caribou, with almost 80% of the KCH
winter diet composed of terrestrial lichens (Farnell et al. 1991). Terrestrial lichens
consumed by caribou are susceptible to forest fires and are known to grow slowly when
compared to other early successional vegetation (Bliss and Wien 1972). Due to the slow
growth and recovery of terrestrial lichens, forest fires are the most dominant natural
disturbance process influencing the habitat use and distribution of caribou in winter
(Scotter 1970; Thomas et al. 1996; Environment Canada 2012; Anderson and Johnson
2014). While forest fires reduce the abundance of forage lichens in the short term, in the
long term fire can enhance overall lichen diversity and productivity (Kershaw 1978; Klein
1982).
Caribou may avoid recently burned habitat until forage lichens recover. Recent burns
also influence snow conditions and may inhibit movement due to increased deadfall
(Schaefer and Pruitt 1991; Thomas et al. 1998). Approximately 50 years is required for
burned areas to regenerate enough lichen to be used consistently by caribou (Thomas
et al. 1996; Joly et al. 2003, 2007; Collins et al. 2011; Environment Canada 2011).
Nonetheless, 12% of collar locations from the KCH were located within 500 m of recent
burns (≤50 years old), suggesting that some caribou have chosen to occupy burns
despite the apparent disadvantages to foraging and movement. The use of recent burns
has been attributed to a number of factors, including the availability of vascular forage in
younger regenerating burns (11–20 years old; Joly et al. 2003), reduced predation risk
as a result of improved visibility, efficient movement between high-quality habitat
80

patches, and selection of remnant patches of forage within the burn boundary (Miller
1976; Thomas et al. 1998).
Cumulative disturbance to boreal caribou ranges has been linked to a herd’s probability
of persistence (Environment Canada 2011). Although a different ecotype, NM caribou
may also be limited by disturbance within their ranges. For the KCH, where
approximately 24% of the current late winter range is disturbed by recent burns (Table
1), knowledge of caribou-fire dynamics is essential to better manage cumulative
landscape change. The diverse landscapes within the KCH’s annual range result in
burns of different intensities, severities, and ultimately, value as habitat.
I studied the determinants of lichen succession post-fire (Chapter Two) and collective
and individual habitat selection strategies in burns (Chapter Three) to determine when
and how burned landscapes are used by caribou. In Chapter Two I used vegetation field
data collected on the KCH range to evaluate lichen abundance in different age burns. I
then tested the ability of a number of environmental characteristics to explain the
variation in lichen abundance in burns of a similar age.
With respect to lichen succession in burns, I observed that:


The sequence of lichen succession was consistent with stand-replacing burns in
the boreal forest; however, Cladina stellaris did not dominate the late succession
phase of the post-fire community. The relative absence of this late-seral lichen
suggests that the fire interval was too short to allow for the establishment of
climax stands across the range of the KCH.

81



Burns of the same age had considerable variation in lichen abundance. This
variation was best explained by greater canopy openness, tree height, stand
basal area, and coniferous cover types. Although not a significant parameter in
models of volume or percent cover of forage lichen, time since burn improved
model performance; however, models performed better with it included as a
predictor, suggesting it does help to explain some variability in lichen abundance.



50 years provides a conservative estimate of lichen recovery post-fire; however,
other environmental characteristics should be used in combination with burn age
to better assess the abundance of forage lichens on a burn-specific basis.

In Chapter Three I explored the factors explaining the use of recent burns by the KCH. I
applied locations from GPS-collared caribou to resource selection functions and
quantified the habitat selection strategies of caribou during the early and late winter
relative to topography, density of terrestrial lichen, and fire history. I used those RSFs to
understand variation in selection among individual caribou as well as the population, as
represented by the pooled sample of caribou locations. Further, I isolated locations of
caribou that occurred in or adjacent to burns and determined factors that were important
for explaining the relative use of Habitat Selection Units that were representative of
burned habitat.
With respect to habitat selection strategies, I observed that:


Averaged selection coefficients indicated that the KCH avoided burned habitats
during winter in most years. In some season-years, more individual caribou

82

selected for burned habitat; however, this selection was weak compared to
individuals who avoided burns (Appendix D: Figure 18 and Figure 19).


Caribou used burns selectively during early winter when snow conditions were
less limiting, and focused their use at the edges and in proximity to areas of high
lichen density.



During late winter, when snow was deeper, denser, and crustier, caribou
encountered and used burns randomly.



Variability in selection strategies was observed for individual caribou that were
monitored for multiple seasons and among caribou within any one season. Few
caribou exhibited consistent selection or avoidance of burns across seasons and
years, suggesting that, on average, burns provided sub-optimal habitat
(Appendix E: Table 26 and Table 27).

Together, the findings from Chapters Two and Three suggest that the relationship
between forest fire and the KCH is both complex and non-linear. In general, burns in the
KCH annual range are not beneficial to caribou during winter until at least 50 years
post-fire; however, there was considerable variability in lichen abundance observed
within recent burns. When caribou encounter recent burns during winter, they can take
advantage of available forage; however, the ability to do so likely decreases when snow
conditions deteriorate as winter progresses.

Management considerations and future research
In this research, I explored a limited set of methods for understanding the resource
selection strategies and use of burns by woodland caribou during the winter. I
83

recommend exploring other measures of caribou distribution and behaviour. Movement
parameters, an index of behaviour, can improve our understanding of how caribou use
landscapes altered by forest fire. Few studies have explicitly explored the relationship
between movement and fire history (e.g. Thomas et al. 1998). Winter is an energetically
stressful time for caribou (Fancy and White 1985; Schaefer and Pruitt 1991; Schaefer
1996; Johnson et al. 2001) and burn history is important for range planning and
management, thus, it is essential to understand how or if caribou alter movement and
behaviour in response to burns. This would include migratory movements among
seasonal ranges.
Additionally, the availability of some resources influences diet and habitat selection
strategies (Boyce and McDonald 1999). An analysis of caribou faecal pellets collected
during winter could provide further insight into winter forage characteristics and perhaps
habitat selection strategies on a burned landscape. Winter diet should be considered on
a herd-specific basis. The use of early-seral habitats, including burns (Joly et al. 2003;
Anderson and Johnson 2014), suggests that early successional forage may play an
important role in winter diet for some herds. A singular focus on lichen abundance in
burns could result in a more limited understanding of caribou-fire dynamics.
In both Chapters Two and Three, my results provide further evidence that burn age or
time since fire, often used for conservation and management planning purposes, is not
a comprehensive metric for understanding the spatiotemporal effects of fire on the
habitat selection of caribou during winter. It has become increasingly apparent that
wildlife managers need to look beyond simple age thresholds to better account for the
effects of fire on the availability and quality of habitat for caribou.
84

This research demonstrates that the relationship between caribou and burns is
complicated and requires careful consideration in the context of cumulative landscape
change. Where snow conditions allow, caribou will use sites within burns that provide
lichen forage or other components of winter diet. In high-priority areas, wildlife
managers should develop a more comprehensive inventory of existing and new burns
that quantify not only burn area, but also lichen cover, forest cover type, canopy
openness, some measure of burn intensity or severity, and through digital terrain
models, characterisation of terrain (i.e. slope, aspect, and elevation). Finally, when
evaluating caribou habitat, managers face the difficult challenge of recognising the
variability in caribou behaviour while developing tractable land-use strategies that
represent the requirements of the population. As a starting point, we can assess
differences between pooled and individual habitat selection models. The challenge of
interanimal variability becomes less when most animals act similarly or respond to only
a few habitat features differently. Where there are considerable differences in resource
selection among animals we can identify and manage for the dominant selection
strategy or strategies. Alternatively or in combination with that approach, we can bound
management prescriptions according to the range of selection strategies demonstrated
by monitored individuals or weight the value we place on particular elements or
components of habitat according to observed selection. In the case of the KCH during
late winter, this would mean a general focus on lichen density, elevation, slope, distance
to water, aspect, and the contrast between valleys and ridgelines, while other elements,
such as burn characteristics, may need to be assessed on an annual basis.

85

LITERATURE CITED
Adams, L.G. 2003. Marrow fat deposition and skeletal growth in caribou calves. Journal
of Wildlife Management 67:20–24.
Akaike H. 1973. Information theory and an extension of the maximum likelihood
principle. In: Petrov, B.N. and F. Caski, eds, Proceeding of the Second
International Symposium on Information Theory, Akaemiai Kiado, Budapest.
267–281 pp.
Anderson, T.A. and C.J. Johnson. 2014. Distribution of barren-ground caribou during
winter in response to fire. Ecosphere 5:140.
Apps, C.D., B.N. McLellan, T.A. Kinley, and J.P. Flaa. 2001. Scale-dependent habitat
selection by mountain caribou, Columbia mountains, British Columbia. Journal of
Wildlife Management 65:65–77.
Arseneault, D., N. Villeneuve, C. Boismenu, Y. Leblanc, and J. Deshaye. 1997.
Estimating lichen biomass and caribou grazing on the wintering grounds of
northern Québec: an application of fire history and Landsat data. Journal of
Applied Ecology 34:65–78.
Arthur, S.M., B.F. Manly, L.L. McDonald, and G.W. Garner. 1996. Assessing habitat
selection availability changes. Ecology 77:215–227.
Barker, O. and T. Hegel. 2012. Habitat selection by Forty Mile caribou in the Dawson
region late winter. Yukon Fish and Wildlife Branch, Environment Yukon,
Whitehorse, Yukon. 25 pp.
Barrier, T.A. 2011. Factors influencing the distribution of Bathurst barren-ground caribou
(Rangifer tarandus groenlandicus) during winter. M.Sc. Thesis, University of
Northern British Columbia.
Bergerud, A.T., S.N. Luttich, and L. Camps. 2008. The Return of Caribou to Ungava.
McGill-Queen’s University Press, Montréal, Québec.
Bergerud A.T., H.E. Butler, and D.R. Miller. 1984. Antipredator tactics of calving
caribou: dispersion in mountains. Canadian Journal of Zoology 62:1566–1575.
Bliss, L.C. and R.W. Wien. 1972. Plant community responses to disturbances in the
western Canadian Arctic. Canadian Journal of Botany 50:1097–1109.
Boyce, M.S. 2006. Scale for resource selection functions. Diversity and Distributions
12:269–276.
86

Boyce, M.S. and L.L. McDonald. 1999. Relating populations to habitats using resource
selection functions. Trends in Ecology & Evolution 14:489–490.
Boyce, M.S., P.R. Vernier, S.E. Nielsen, and F.K. Schmiegelow. 2002. Evaluating
resource selection functions. Ecological Modelling 157:281–300.
Briand, Y., J.-P. Ouellet, C. Dussault, and M.-H. St-Laurent. 2009. Fine-scale habitat
selection by female forest-dwelling caribou in managed boreal forest: empirical
evidence of a seasonal shift between foraging opportunities and antipredator
strategies. Ecoscience 16:330–340.
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.
British Columbia Ministry of Forests and Range and British Columbia Ministry of
Environment. 2010. Field manual for describing terrestrial ecosystems. 2nd ed.
Forest Science Program, Victoria, B.C. Land Management Handbook No. 25.
Burnham, K.P. and D.R. Anderson. 1998. Model selection and inference: A practical
information-theoretic approach. Springer, New York, USA.
Carroll, S.B. and L.C. Bliss. 1982. Jack pine-lichen woodland on sandy soils in northern
Saskatchewan and northeastern Alberta. Canadian Journal of Botany 60:2270–
2282.
Caslys Consulting. 2014. Mapping lichen forage in the Klaza caribou herd range using a
revised land cover: Summary report. Submitted to Department of Fish and
Wildlife, Environment Yukon. Whitehorse, Y.T. 32 pp.
Cichowski, D. and S. Haeussler. 2013. The response of caribou terrestrial forage
lichens to mountain pine beetles and forest harvesting in the east Ootsa and
Entiako areas: Annual Report 2012/13 – Year 11. 49 pp.
COSEWIC. 2014. COSEWIC assessment and status report on the caribou Rangifer
tarandus, Northern Mountain population, Central Mountain population and
Southern Mountain population in Canada. Committee on the Status of
Endangered Wildlife in Canada. Ottawa. xxii + 113 pp.
COSEWIC. 2011. Designatable units for caribou (Rangifer tarandus) in Canada.
Committee on the Status of Endangered Wildlife in Canada. Ottawa. 88 pp.
Collins, W.B. and T.S. Smith. 1991. Effects of wind-hardened snow on foraging by
reindeer (Rangifer tarandus). Arctic 44:217–222.

87

Collins, W.B., B.W. Dale, L.G. Adams, D.E. McElwain, and K. Joly. 2011. Fire, grazing
history, lichen abundance, and winter distribution of caribou in Alaska's taiga.
Journal of Wildlife Management 75:369–377.
Coxson, D.S. 2015. Using partial-cut harvesting to conserve terrestrial lichens in
managed landscapes. Canadian Wildlife Biology and Management 4:150–162.
Coxson, D.S. and J. Marsh. 2001. Lichen chronosequence (postfire and postharvest) in
lodgepole pine (Pinus contorta) forests of northern interior British Columbia.
Canadian Journal of Botany 79:1449–1464.
Crête, M., C. Morneau, and R. Nault. 1990. Biomasse et especes de lichens terrestres
pour le caribou dans le nord du Québec. Canadian Journal of Botany 68:2047–
2053.
Dalerum, F., S. Boutin, and J.S. Dunford. 2007. Wildfire effects on home range size and
fidelity of boreal caribou in Alberta, Canada. Canadian Journal of Zoology 85:26–
32.
DeCesare, N.J., M. Hebblewhite, F. Schmiegelow, D. Hervieux, G.J. McDermid, L.
Neufeld, M. Bradley, J. Whittington, K.G. Smith, L.E. Morgantini, M. Wheatley,
and M. Musiani. 2012. Transcending scale dependence in identifying habitat with
resource selection functions. Ecological Applications 22:1068–1083.
Dickson, B.G. and P. Beier.2007. Quantifying the influence of topographic position on
cougar (Puma concolor) movement in southern California, USA. Journal of
Zoology 271:270–277.
Dyer, S.J., J.P. O’Neill, S.M. Wasel, and S. Boutin. 2002. Quantifying barrier effects of
roads and seismic lines on movements of female woodland caribou in
northeastern Alberta. Canadian Journal of Zoology 80:839–845.
EDI Environmental Dynamics Inc. 2013. Casino Project: Wildlife Baseline Report.
Prepared for Casino Mining Corporation, Vancouver, BC by EDI, Whitehorse, YT.
18 October 2013.
Edmonds, E.J. and M. Bloomfield. 1984. A study of woodland caribou (Rangifer
tarandus caribou) in West Central Alberta, 1979–1983. Alberta Energy and
Natural Resources, Fish and Wildlife Division, Edmonton, Alberta.
Environment Canada. 2014. Recovery Strategy for the Woodland Caribou, Southern
Mountain population (Rangifer tarandus caribou) in Canada [Proposed]. Species
at Risk Act Recovery Strategy Series. Environment Canada, Ottawa. viii + 68 pp.

88

Environment Canada. 2012. Management Plan for the Northern Mountain Population of
Woodland Caribou (Rangifer tarandus caribou) in Canada. Species at Risk Act
Management Plan Series. Environment Canada, Ottawa. vii + 79 pp.
Environment Canada. 2011. Scientific assessment to inform the identification of critical
habitat for woodland caribou (Rangifer tarandus caribou), boreal population, in
Canada: 2011 update. Ottawa, Ontario, Canada. 102 pp. plus appendices.
Environment Canada. 2008. Scientific review for the identification of critical habitat for
woodland caribou (Rangifer tarandus caribou), boreal population, in Canada.
August 2008. Ottawa: Environment Canada. 72 pp. plus 180 pp appendices.
Environmental Systems Research Institute (ESRI). 2005. ArcView Release 3.2a.
Redlands, CA.
Farnell, R., R. Sumanik, J. McDonald, B. Gilroy. 1991. The distribution, movements,
demography, habitat characteristics of the Klaza caribou herd in relation to the
Casino Trail development, Yukon Territory. Technical Report TR-91-3,
Whitehorse, YT.
Fancy, S.G. and R.G. White. 1985. Energy expenditures by caribou while cratering in
snow. Journal of Wildlife Management 49:987–993.
Festa-Bianchet, M., J.C. Ray, S. Boutin, and A. Gunn. 2011. Conservation of caribou
(Rangifer tarandus) in Canada : an uncertain future. Canadian Journal of Zoology
89:419–434.
Fleischman, S.J. 1990. Lichen availability on the range of an expanding caribou
(Rangifer tarandus) population in Alaska. MSc. thesis. University of Alaska,
Fairbanks.
Florkiewicz, R., R. Maraj, T. Hegel, and M. Waterreus. 2007. The effects of human land
use on the winter habitat of the recovering Carcross woodland caribou herd in
suburban Yukon Territory, Canada. Rangifer 17:181–197.
Florkiewicz, R.F., N. Flynn, N. MacLean, S.R. Francis, J.Z. Adamczewski, and V.
Loewen. 2004. Little Rancheria caribou in the Yukon: Evaluation of winter habitat
quality and habitat use. Department of Environment, Government of Yukon.
Whitehorse, Yukon. 61 pp.
Frair, J.L., J. Fieberg, M. Hebblewhite, F. Cagnacci, N.J. DeCesare, and L. Pedrotti.
2010. Resolving issues of imprecise and habitat-biased locations in ecological
analyses using GPS telemetry data. Philosophical Transactions of the Royal
Society B 365:2187–2200.
89

Francis, S., T. Antoniuk, J. Nishi, and S. Kennett. 2013. Range assessment as a
cumulative effects management tool: A recommended approach for Environment
Yukon. Prepared for Environment Yukon. Yukon Fish and Wildlife Branch Report
MRC-13-01, Whitehorse, Yukon, Canada.
Frazer, G.W., C.D. Canham, and K.P. Lertzman. 1999. Gap Light Analyzer (GLA),
Version 2.0: Imaging software to extract canopy structure and gap light
transmission indices from true-colour fisheye photographs, user’s manual and
program documentation. Copyright © 1999: Simon Fraser University, Burnaby,
British Columbia, and the Institute of Ecosystem Studies, Millbrook, New York.
Gaare, E. and T. Skogland. 1980. Lichen-reindeer interaction studied in a simple case
model. Proceedings of the International Reindeer/Caribou Symposium 2:47–56.
Gaillard, J., M. Festa-Bianchet, and N.G. Yoccoz. 1998. Population dynamics of large
herbivores: variable recruitment with constant adult survival. Trends in Ecology
and Evolution 13:58–63.
Gillingham, M.P. and K.L. Parker. 2008. The importance of individual variation in
defining habitat selection by moose in northern British Columbia. Alces 44:7–20.
Girard, F., S. Payette, and A. Delwaide. 2017. Patterns of early postfire succession of
alpine, subalpine and lichen-woodland vegetation: 21 years of monitoring from
permanent plots. Forests 8:1–14.
Gustine, D.D. and K.L. Parker. 2008. Variation in the seasonal selection of resources by
woodland caribou in northern British Columbia. Canadian Journal of Zoology
86:812–825.
Gustine, D.D. K.L. Parker, R.J. Lay, M.P. Gillingham, and D.C. Heard. 2006.
Interpreting resource selection at different scales for woodland caribou in winter.
Journal of Wildlife Management 70:1601–1614.
Haughian, S.R. and P.J. Burton. 2015. Microhabitat associations of lichens,
feathermosses, and vascular plants in a caribou winter range, and their
implications for understory development. Botany 93:221–231.
Hayes, R.D., R. Farnell, R.M.P. Ward, J. Carey, M. Dehn, G.W. Kuzyk, A.M. Baer, C.L.
Gardner, and M. O’Donoghue. 2003. Experimental reduction of wolves in the
Yukon: Ungulate responses and management implications. Wildlife Monographs
152:1–35.
Heard, D.C. and K.L. Vagt. 1998. Caribou in British Columbia: a 1996 status report.
Rangifer Special Issue No. 10:159–172.
90

Hegel, T. 2013. Inventory studies of the Klaza Caribou herd – 2012 activities. Yukon
Department of Environment. 19 pp.
Hegel, T. and M. O’Donoghue. 2015. Late-winter habitat selection and distribution of the
Klaza caribou herd. Yukon Department of Environment. 31 pp.
Helle, T. 1981. Studies on the wild forest reindeer and semi-domestic reindeer in
Finland. Acta Universitatis Ouluensis A 12:1–34.
Hendrickx, J. 1999. Using categorical variables in Stata. Stata Technical Bulletin STB52.
Holt, E.A., B. McCune, and P. Neitlich. 2008. Grazing and fire impacts on macrolichen
communities of the Seward Peninsula, Alaska, U.S.A. The Bryologist 111:68–83.
James, A.R.C., S. Boutin, D.M. Hebert, and A.B. Rippin. 2004. Spatial separation of
caribou from moose and its relation to predation by wolves. Journal of Wildlife
Management 68:799–809.
Jenness, J. 2016. Topographic Position Index (tpi_jen.avx) extension for Arcview 3.x,
v.1.3a. Flagstaff, AZ.
Jingfors, K. 1989. Wildlife management plan for the Casino Trail area. Prepared for the
Government of Yukon, Department of Renewable Resources, Fish and Wildlife
Branch, Habitat and Research Section. Whitehorse, YT.17 pp.
Johnson, C.J., L.P.W. Ehlers, and D.R. Seip. 2015. Witnessing extinction – Cumulative
impacts across landscapes and the future loss of an evolutionarily significant unit
of woodland caribou in Canada. Biological Conservation 186:176–186.
Johnson, C.J., M.S. Boyce, R.L. Case, H.D. Cluff, R.J.Gau, A. Gunn, and R. Mulders.
2005. Quantifying the cumulative effects of human developments: a regional
environmental assessment for sensitive Arctic wildlife. Wildlife Monographs
160:1–36.
Johnson, C.J., K.L. Parker, D.C. Heard, and D.R. Seip. 2004a. Movements, foraging
habits, and habitat use strategies of northern woodland caribou during winter:
Implications for forest practices in British Columbia. BC Journal of Ecosystems
and Management 5:2235.
Johnson, C.J., D.R. Seip, and M.S. Boyce. 2004b. A quantitative approach to
conservation planning: using resource selection functions to map the distribution
of mountain caribou at multiple spatial scales. Journal of Applied Ecology
41:238–251.

91

Johnson, C.J., K.L. Parker, D.C. Heard, and M.P. Gillingham. 2002. A multiscale
behavioral approach to understanding the movements of woodland caribou.
Ecological Applications 12:1840–1860.
Johnson, C.J., K.L. Parker, and D.C. Heard. 2001. Foraging across a variable
landscape: behavioral decisions made by woodland caribou at multiple spatial
scales. Oecologia 127:590–602.
Johnson, C.J., K.L. Parker, and D.C. Heard. 2000. Feeding site selection by woodland
caribou in north-central British Columbia. Rangifer 12:159–172.
Johnson, D.H. 1980. The comparison of usage and availability measurements for
evaluating resource preference. Ecology 61:65–71.
Johnstone, J.F. and F. Stuart-Chapin III. 2006. Effects of soil burn severity on post-fire
tree recruitment in boreal forest. Ecosystems 9:14–31.
Joly, K. S.K. Wasser, and R. Booth. 2015. Non-invasive assessment of the
interrelationships of diet, pregnancy rate, group composition, and physiological
and nutritional stress of barren-ground caribou in late winter. PLoS ONE 10:
e0127586.
Joly, K., F.S. Chapin III, and D.R. Klein. 2010. Winter habitat selection by caribou in
relation to lichen abundance, wildfires, grazing, and landscape characteristics in
northwest Alaska. Ecoscience 17:321–333.
Joly, K., P. Bente, and J. Dau. 2007. Response of overwintering caribou to burned
habitat in northwest Alaska. Arctic 60:401–410.
Joly, K., B.W. Dale, W.B. Collins, and L.G. Adams. 2003. Winter habitat use by female
caribou in relation to wildland fires in interior Alaska. Canadian Journal of
Zoology 81:1192–1201.
Jones, E.S., M.P. Gillingham, D.R. Seip, and D.C. Heard. 2007. Comparison of
seasonal habitat selection between threatened woodland caribou ecotypes in
central British Columbia. Rangifer Special Issue No. 17:111–128.
Kershaw, K.A. 1978. The role of lichens in boreal tundra transition areas. The Bryologist
81:294–306.
Klein, D.R. 1982. Fire, lichens, and caribou. Journal of Range Management 35:390–
395.

92

Kumpula, J., A. Colpaert, and M. Nieminen. 2000. Condition, recovery rate, and
productivity of lichen (Cladonia spp.) ranges in the Finnish Reindeer
Management Area. Arctic 53:152–160.
Kuzyk, G.W., M.M. Dehn, and R.S. Farnell. 1999. Body-size comparisons of alpine- and
forest-wintering woodland caribou herds in the Yukon. Canadian Journal of
Zoology 77:1017–1024.
Lesmerises, R. and M-H. St-Laurent. 2017. Not accounting for interindividual variability
can mask selection patterns: a case study on black bears. Oecologia:1–11.
Lesmerises, F. C.J. Johnson, and M.-H. St-Laurent. 2016. Refuge or predation risk?
Alternate ways to perceive hiker disturbance based on maternal state of female
caribou. Ecology and Evolution 2016:1–10.
Lesmerises, R. J-P. Ouellet, and M-H. St-Laurent. 2011. Assessing terrestrial lichen
biomass using ecoforest maps: a suitable approach to plan conservation areas
for forest-dwelling caribou. Canadian Journal of Forest Research 41:632–642.
Long, J.S. and J. Freese. 2001. Regression models for categorical dependent variables
using Stata. Stata Press, College Station, Texas, U.S.A. 288 pp.
Manly, B.F., L.L. McDonald, and D.L. Thomas. 2002. Resource Selection by Animals:
Statistical Design and Analysis for Field Studies. Second edition. Chapman-Hall,
London, U.K.
Martin, T.G., B.A. Wintle, J.R. Rhodes, P.M. Kuhnert, S.A. Field, S.J. Low-Choy, A.J.
Tyre, and H.P. Possingham. 2005. Zero tolerance ecology: Improving ecological
inference by modeling the source of zero observations. Ecology Letters 8:1235–
1246.
Marzluff, J.M., J.J. Millspaugh, P. Hurvitz, and M.S. Handcock. 2004. Relating
resources to a probabilistic measure of space use: forest fragments and Steller’s
Jays. Ecology 85:1411–1427.
McKillop, R., D. Turner, K. Johnston, and J. Bond. 2013. Property-scale classification of
surficial geology for soil geochemical sampling in the unglaciated Klondike
Plateau, west-central Yukon. Yukon Geological Survey, Open File 2013-15, 85
pp., including appendices.
Menard, S. 2002. Applied logistic regression analysis. Sage university paper series:
Quantitative applications in the social sciences. Sage Publications, Thousand
Oaks, California, USA. 111 pp.

93

Miller, D.R. 1976. Biology of the Kaminuriak population of barren-ground caribou. Part
3: Taiga winter range relationships and diet. Canadian Wildlife Service Technical
Report Series No. 36, Canadian Wildlife Service, Prairie and Northern Region,
Edmonton, Alberta.
Moen, J., O. Danell, and R. Holt. 2007. Non-destructive estimation of lichen biomass.
Rangifer 27:41–46.
Morneau, C. and S. Payette. 1989. Postfire lichen-spruce woodland recovery at the limit
of the boreal forest in northern Quebec. Canadian Journal of Botany 67:2770–
2782.
National Applied Resource Sciences Center. 1999. Sampling vegetation attributes.
Interagency Technical Reference 1734-4. U.S. Department of the Interior,
Bureau of Land Management. Denver, Colorado. 164 pp.
Nelson, P.R., C. Roland, M.J. Macander, and B. McCune. 2013. Detecting continuous
lichen abundance for mapping winter caribou forage at landscape spatial scales.
Remote Sensing of Environment 137:43–54.
Northrup, J.M., M.B. Hooten, C.R. Anderson Jr., and G. Wittemyer. 2013. Practical
guidance on characterizing availability in resource selection functions under a
use-availability design. Ecology 94:1456–1463.
Papke, L.E. and J.M. Wooldridge. 1996. Econometric methods for fractional response
variables with an application to 401(K) plan participation rates. Journal of Applied
Econometrics 11:619–632.
Parker, K.L., P.S. Barboza, and T.R. Stephenson. 2005. Protein conservation in female
caribou (Rangifer tarandus): Effects of decreasing diet quality during winter.
Journal of Mammalogy 86:610–622.
Payette, S. 1992. Fire as a controlling process in the North American boreal forest. In: A
systems analysis of the global boreal forest. H.H. Shugart, R. Leemans, and G.B.
Bonan, eds. Cambridge University Press, Cambridge. 144–169 pp.
Post, E. and D.R. Klein. 1999. Caribou calf production and seasonal range quality
during a population decline. The Journal of Wildlife Management 63:335–345.
Pruitt, W.O. 1959. Snow as a factor in the winter ecology of the barren ground caribou
(Rangifer arcticus). Arctic 12:158–179.

94

Reid, D.G., S.R. Francis, and T. Antoniuk. 2013. Application of herd viability models for
boreal woodland caribou (Rangifer tarandus caribou) to a northern mountain
caribou herd. Canadian Wildlife Biology and Management 2:67–79.
Rettie, W.J. and F. Messier. 2000. Hierarchical habitat selection by woodland caribou:
its relationship to limiting factors. Ecography 23:466–478.
Richards, S.A. 2008. Dealing with overdispersed count data in applied ecology. Journal
of Applied Ecology 45:218–227.
Rickbeil, G.J.M., T. Hermosilla, N.C. Coops, J.C. White, and M.A. Wulder. 2017.
Barren-ground caribou (Rangifer tarandus groenlandicus) behaviour after recent
fire events; integrating caribou telemetry data with Landsat fire detection
techniques. Global Change Biology 23:1036–1047.
Richards, S.A. 2005. Testing ecological theory using the information-theoretic approach:
Examples and cautionary results. Ecology 86:2805–2814.
Robbins, C.T. and B.L. Robbins. 1979. Fetal and neonatal growth patterns and maternal
reproductive effort for ungulates and subungulates. The American Naturalist
114:101–116.
Robinson, H.S., M. Hebblewhite, N.J. DeCesare, J. Whittington, L. Neufeld, M. Bradley,
and M. Musianai. 2012. The effect of fire on spatial separation between wolves
and caribou. Rangifer 20:277–294.
Roturier, S., S. Ollier, L.-E. Nutti, U. Bergsten, and H. Winsa. 2017. Restoration of
reindeer lichen pastures after forest fire in northern Sweden: Seven years of
results. Ecological Engineering 108:143–151.
Rowe, J.S. 1983. Concepts of fire effects on plant individuals and species. In: The role
of fire in northern circumpolar ecosystems. R.W. Wein and D.A. MacLean, eds.
John Wiley & Sons Ltd. 135-154 pp.
Rowe, J.S. and G.W. Scotter. 1973. Fire in the boreal forest. Quaternary Research
3:444–464.
Rupp, T.S., M. Olson, L.G. Adams, B.W. Dale, K. Joly, J. Henkelman, W.B. Collins, and
A.M. Starfield. 2006. Simulating the influences of various fire regimes on caribou
winter habitat. Ecological Applications 16:1730–1743.
Russell, D.E., P.H. Whitfield, J. Cai, A. Gunn, R.G. White, and K. Poole. 2013.
CARMA’s MERRA-based caribou range climate database. Rangifer 33:145–152.

95

Russell, D.E., A.M. Martell, and W.A.C. Nixon. 1993. Range ecology of the Porcupine
caribou herd in Canada. Rangifer Special Issue 13:1–168.
Schaefer, J.A. 1996. Canopy, snow, and lichens on woodland caribou range in
southeastern Manitoba. Rangifer 9:239–244.
Schaefer, J.A. and W.O. Pruitt Jr. 1991. Fire and woodland caribou in southeastern
Manitoba. Wildlife Monographs 116:1–39.
Scotter, G.W. 1970. Wildfires in relation to the habitat of barren-ground caribou in the
taiga of northern Canada. Proceedings of the annual Tall Timbers Fire Ecology
Conference 10:85–106.
Seip,

D.R. 1992. Factors limiting woodland caribou populations and their
interrelationships with wolves and moose in southeastern British-Columbia.
Canadian Journal of Zoology 70:1494–1503.

Shepard, L. F. Schmiegelow, and E. Macdonald. 2007. Managing fire for woodland
caribou in Jasper and Banff National Parks. Rangifer, Special Issue 17:129–140.
Skatter, H.G., J.L. Kansas, M.L. Charlebois, and B. Balicki. 2014. Recovery of terrestrial
lichens following wildfire in the boreal shield of Saskatchewan: Early seral forage
availability for woodland caribou (Rangifer tarandus caribou). Canadian Wildlife
Biology and Management 3:1–14.
Skoog, R.O. 1968. Ecology of the caribou (Rangifer tarandus granti) in Alaska.
University of California, Berkeley. Berkeley, California. 699 pp.
Skovsgaard, J.P. and J.K. Vanclay. 2008. Forest site productivity: a review of the
evolution of dendrometric concepts for even-aged stands. Forestry 81:13–31.
Sorensen, T., P.D. McLoughlin, D. Hervieux, E. Dzus, J. Nolan, B. Wynes, and S.
Boutin. 2008. Determining sustainable levels of cumulative effects for boreal
caribou. Journal of Wildlife Management 72:900–905.
StataCorp. 2011. Stata Statistical Software: Release 12. College Station, TX: StataCorp
LP.
Stocks B.J., M.A. Fosberg T.J. Lynham, L. Mearns, and B.M. Wotton. 1998. Climate
change and forest fire potential in Russian and Canadian boreal forests. Climatic
Change 38:1–13.
Sulyma, R. and D.S. Coxson. 2001. Microsite displacement of terrestrial lichens by
feather moss mats in late seral pine-lichen woodlands of north-central British
Columbia. The Bryologist 104:505–516.
96

Symonds, M.R.E. and A. Moussalli. 2011. A brief guide to model selection, multimodel
inference and model averaging in behavioural ecology using Akaike’s information
criterion. Behavioral Ecology and Sociobiology 65:13–21.
Thomas, D.C. 1991. Adaptations of barren-ground caribou to snow and burns. In:
Butler, C. and Mahoney, S. P., eds. 4th North American caribou workshop. St.
John's, Newfoundland. Newfoundland and Labrador Wildlife Division,
Department of Environment and Lands. Newfoundland, 529 pp.
Thomas, D.C. and H.P.L. Kiliaan. 1998. Fire-caribou relationships: (IV) Recovery of
habitat after fire on winter range of the Beverly Herd. Technical Report Series.
No. 312. Canadian Wildlife Service, Prairie and Northern Region. Edmonton, AB.
115 pp.
Thomas, D.C., H.P.L. Kiliaan, and T.W.P. Trottier. 1998. Fire-caribou relationships: (III)
Movement patterns of the Beverly herd in relation to burns and snow. Technical
Report Series No. 311. Canadian Wildlife Service, Prairie and Northern Region.
Edmonton, AB. 176 pp.
Thomas, D.C., S.J. Barry, and G. Alaie. 1996. Fire-caribou-winter range relationships in
northern Canada. Rangifer 16:57–67.
Thomas, D.L. and E.J. Taylor. 1990. Study designs and tests for comparing resource
use and availability. The Journal of Wildlife Management 54:322–330.
Tukey, J.W. 1953. The problem of multiple comparisons. Unpublished manuscript,
Princeton University.
Tyler, N.J.C. 2010. Climate, snow, ice, crashes, and declines in populations of reindeer
and caribou (Rangifer tarandus L.). Ecological Monographs 80:197–219.
Vors, L.S., J.A. Schaefer, B.A. Pond, A.R. Rodgers, and B.R. Patterson. 2007.
Woodland caribou extirpation and anthropogenic landscape disturbance in
Ontario. Journal of Wildlife Management 71:1249–1256.
Vuong, Q.H. 1989. Likelihood ratio tests for model selection and non-nested
hypotheses. Econometrica 57:307–333.
White, R.G. 1983. Foraging patterns and their multiplier effects on productivity of
northern ungulates. Oikos 40:377–384.
Wiersma, Y.F., F. Huettmann, and C.A. Drew. 2011. Chapter 1: Introduction. Landscape
modelling of species and their habitat: History, uncertainty, and complexity:
Concepts and applications. In: Predictive species and habitat modelling in

97

landscape ecology, Drew, C.A., Y.F. Wiersma, and F. Huettmann, eds. Springer
Science and Business Media, New York.
Wittmer, H.U., A.R.E. Sinclair, and B.N. McLellan. 2005. The role of predation in the
decline and extirpation of woodland caribou. Oecologia 144:257–267.
Yukon Ecoregions Working Group (YEWG). 2004. Klondike Plateau. In: Ecoregions of
the Yukon Territory: Biophysical properties of Yukon landscapes, C.A.S. Smith,
J.C. Meikle and C.F. Roots (eds.), Agriculture and Agri-Food Canada, PARC
Technical Bulletin No. 04-01, Summerland, BC, 159–168.
Yukon Government. 2014. Yukon fire history metadata. Community Services, Protective
Services Branch, Wildland Fire Management. 5 pp.
Yukon Government. 2010. Driving the fire belt: North Klondike Highway. Public
brochure
available
at:
http://www.env.gov.yk.ca/publicationsmaps/documents/firebrochureforweb2010.pdf. Accessed 20 November 2014.
Yukon Government Department of Community Services and Environment Yukon.
8 pp.
Yukon Government. 2006. Yukon vegetation inventory manual, version 2.1. Forest
Management Branch, Energy, Mines, and Resources.53 pp.
Zuur, A.F., E.N. Ieno, and C.S. Elphick. 2010. A protocol for data exploration to avoid
common statistical problems. Methods in Ecology and Evolution 1:3–14.
Zuur, A.F., E.N. Ieno, N.J. Walker, A.A. Saveliev, and G.M. Smith. 2009. Mixed effects
models and extensions in ecology with R. Springer, New York, NY, USA.
Zuur, A.F., E.N. Ieno, and G.M. Smith. 2007. Analysing ecological data. Springer, New
York, NY, USA.

98

Appendix A Field data collection details

99

Table 17. Fire history, terrain, and vegetation variables recorded during field sampling in 2014
within the Klaza caribou herd annual range, west-central Yukon.
Variable

Fire-related variables
Burn age class
1-10 years
11–20 years
21–30 years
31–50 years
51–70 years
>70 years/unburned
Burn size
Distance to burn edge/closest burn
Terrain variables
Elevation
Aspect
Slope
Slope position
1
Moisture regime
Vegetation variables
Top tree species
Tree height
Stand basal area
Top tall shrub species
Cover of tall shrubs
Canopy openness
Ground cover
Bryophytes
Dwarf shrubs
Forbs
Graminoids
Horsetails
Mushrooms
Litter
Soil or Rock
Lichens

Lichen height
1

Description
Derived from mapped polygons from the 2013 Yukon Fire History database +
tree core samples

Age post-fire derived from tree core samples
Size of burn polygon in hectares, from Yukon Fire History database
Distance in metres from plot centre to nearest edge of burn or closest burn (if
in unburned plot)
Metres above sea level
0–360° (converted to northness and eastness using cosine and sine)
Percent slope angle
Level, depression, toe, lower, mid, upper, crest
0 = very xeric, 1 = xeric, 2 = subxeric, 3 = submesic, 4 = mesic, 5 =
subhygric, 6 = hygric, 7 = subhydric
Most dominant tree species in plot
Average height (m) of the most dominant tree species
Basal area factor; determined using a rod relascope
Most dominant tall shrub species in plot (>0.5 m in height)
Percent cover of tall shrubs (> 0.5 m in height)
Percent canopy openness; determined using Gap Light Analyzer software
Percent cover of non-vascular plants
Percent cover of shrubs <0.5 m tall
Percent cover of herbaceous flowering plants
Percent cover of grasses, sedges, and rushes
Percent cover of plants in the genus Equisetum (spore-producing, vascular)
Percent cover of fungus species (aboveground fruiting bodies)
Percent cover of leaf, needle, or small woody debris
Percent cover of exposed mineral soil or rock (bedrock or large boulders)
Percent cover of Cladina mitis/arbuscula (not distinguished in the field), C.
rangiferina, C. stellaris, Cladonia spp., Cetraria cucullata, C. ericetorum/
islandica (not distinguished in the field), C. nivalis, Stereocaulon spp., and
Peltigera spp.
Height (cm) of lichen thallus measured from beginning of living stem for all
species except Peltigera spp.

During modeling, some categorical predictor variables had too few samples per category, thus requiring fewer categories
per variable. The categories for moisture regime were re-categorized to: Dry (very xeric, xeric, subxeric, submesic), Moist
(mesic), and Wet (subhygric, hygric, subhydric).

100

Table 18. Protocol for determining the age class of vegetation plots as determined using
mapped burn age and/or tree core ages from trees sampled in field plots within the Klaza
caribou herd annual range, west-central Yukon.
Tree core result

Assigned age

Justification/assumptions

Minimum core age within 10
years of mapped burn age

Mapped burn age

Some trees established post-burn, but there
may have been some survivors.

Minimum or maximum core age
>10 years younger than mapped
burn age

Mapped burn age

It may have taken longer for trees to establish
post-burn, and there may have been some
survivors.

Minimum core age >10 years
older than mapped burn age

Minimum core
age

Suggests that no new trees established postburn, and the stand is made up of survivors.

Maximum-minimum core age ≤50
years

Median core age

Suggests that the stand is relatively even-aged,
and the median value is representative of the
site age.

Maximum-minimum core age ≥50
years

Maximum core
age

Suggests the stand is multi-aged and it is
assumed the oldest tree core represents the
age class

Inside mapped burn

Outside mapped burn

101

Table 19. Average values (with standard error) of the percent cover of non-lichen ground cover in vegetation plots in burns by 15-year intervals,
Klaza caribou range, west-central Yukon, summer 2014. Plots aged 100–149 and ≥150 years were grouped together rather than by 15-year
intervals due to low sample sizes.
Age
group
(years)

n

Bryophyte

<25
25–39
40–54
55–69
70–84
85–99
100–149
≥150

18
8
14
18
8
5
16
13

%
11.5
5.6
17.7
16.1
15.4
17.8
23.8
23.1

SE
2.7
1.3
4.2
3.8
3.6
4.2
5.6
5.4

Dwarf
shrub
%
29.7
45.3
33.7
37.1
21.8
21.1
28.6
28.3

SE
7.0
10.7
7.9
8.8
5.1
5.0
6.7
6.7

Forb
%
4.6
3.9
1.9
8.4
2.9
1.5
4.8
3.6

SE
1.1
0.9
0.4
2.0
0.7
0.4
1.1
0.9

Graminoid

Horsetail

Mushroom

%
16.9
10.6
8.3
8.8
11.8
7.7
7.3
8.2

%
1.0
0.0
0.5
1.6
0.1
0.2
0.4
1.5

%
0.2
0.0
0.1
0.1
0.1
0.0
0.0
0.1

SE
4.0
2.5
1.9
2.1
2.8
1.8
1.7
1.9

SE
0.2
0.0
0.1
0.4
0.0
0.0
0.1
0.3

SE
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0

Litter
%
3.6
9.1
6.2
10.0
3.6
0.0
2.9
0.8

SE
0.8
2.2
1.5
2.4
0.8
0.0
0.7
0.2

Soil
%
10.4
1.8
0.2
0.6
0.0
0.0
0.1
0.0

SE
2.4
0.4
0.1
0.1
0.0
0.0
0.0
0.0

Rock
%
5.6
18.8
10.7
0.0
0.0
0.0
15.6
0.0

SE
1.3
4.4
2.5
0.0
0.0
0.0
3.7
0.0

Fern
%
0.0
0.0
0.0
2.8
0.0
0.0
0.0
0.0

SE
0.0
0.0
0.0
0.7
0.0
0.0
0.0
0.0

102

Appendix B Lichen model outputs and prediction

103

Table 20. Summary of model selection statistics for candidate a priori models used to select
the most parsimonious count model for understanding the volume of forage lichens in the
Klaza caribou herd range, west-central Yukon. Parameters are defined in Table 2.
Rank
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

Model parameters
2
TSB+CO+SBA+TH+TH +CT
2
CO+SBA+TH+TH +CT
2
TSB+TH+TH +CT
2
TH+TH +CT
2
TSB+CO+SBA+TH+TH +Elev+North+East
2
TSB+CO+SBA+TH+TH +TSC
2
CO+SBA+TH+TH +Elev+North+East
2
TSB+CO+SBA+TH+TH +SMR
2
TSB+CO+SBA+TH+TH
2
CO+SBA+TH+TH +SMR
2
CO+SBA+TH+TH +TSC
2
TSB+CO+TH+TH +TSC
2
CO+SBA+TH+TH
2
TSB+CO+TH+TH
2
CO+TH+TH +TSC
2
CO+TH+TH
TSB+SMR+S
TSB+Elev+North+East+S
SMR+S
Elev+North+East+S

AICci
782.58
788.57
808.89
829.70
884.66
888.61
893.06
893.38
894.35
902.93
905.96
913.05
914.07
917.37
950.64
959.94
1267.83
1290.62
1331.04
1369.99

AICc wi
0.95
0.05
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00

∆i AICc
0.00
6.00
26.31
47.12
102.08
106.04
110.48
110.81
111.78
120.36
123.38
130.48
131.49
134.80
168.06
177.37
485.26
508.04
548.47
587.41

Table 21. Summary of model selection statistics for candidate a priori models used to select
the most parsimonious fractional logit model (binomial family, logit link function) for
understanding the cover of forage lichens in the Klaza caribou herd range, west-central
Yukon. Parameters are defined in Table 2.
Rank
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

Model parameters
2
CO+SBA+TH+TH
2
CO+TH+TH
2
TH+TH +CT
2
TSB+CO+TH+TH
2
TSB+CO+SBA+TH+TH
2
TSB+TH+TH +CT
2
CO+TH+TH +TSC
2
CO+SBA+TH+TH +TSC
2
CO+SBA+TH+TH +CT
2
TSB+CO+TH+TH +TSC
2
TSB+CO+SBA+TH+TH +TSC
2
TSB+CO+SBA+TH+TH +CT
2
CO+SBA+TH+TH +SMR
TSB+SMR+S
2
TSB+CO+SBA+TH+TH +SMR
SMR+S
2
CO+SBA+TH+TH +Elev+North+East
TSB+Elev+North+East+S
2
TSB+CO+SBA+TH+TH +Elev+North+East
Elev+North+East+S

AICci
71.37
71.43
71.79
72.07
72.96
73.02
73.36
73.45
73.94
74.15
75.15
75.95
77.60
79.08
79.54
79.67
80.05
81.70
81.85
83.67

AICc wi
0.17
0.16
0.13
0.12
0.08
0.07
0.06
0.06
0.05
0.04
0.03
0.02
0.01
0.00
0.00
0.00
0.00
0.00
0.00
0.00

∆i AICc
0.00
0.06
0.42
0.70
1.59
1.65
1.99
2.08
2.57
2.78
3.78
4.58
6.23
7.71
8.16
8.30
8.68
10.33
10.48
12.30

104

Table 22. Coefficients and statistical parameters generated from the top-ranked (∆i AICc
differed by ≤2 points) zero-inflated Poisson (ZIP) regression models for Cladina spp. and
Cetraria spp. lichen volume (dm3/m2) in the Klaza caribou herd annual range, west-central
Yukon. Coefficients and statistical parameters for the count and binary parts of the ZIP
models are distinguished.
Parameter

β

Robust SE

Z

P

95% CI
Lower

Upper

Count
Time since burn

0.002

0.001

1.27

0.21

-0.001

0.005

Canopy openness

0.011

0.009

1.21

0.23

-0.007

0.028

Stand basal area

0.034

0.016

2.12

0.03

0.003

0.065

Tree height

0.271

0.101

2.69

0.01

0.073

0.469

-0.022

0.006

-3.55

0.00

-0.034

-0.010

Deciduous

-0.736

0.342

-2.15

0.03

-1.406

-0.066

Open

-0.697

0.327

-2.13

0.03

-1.338

-0.057

Coniferous

1.433

0.265

5.41

0.00

0.914

1.952

Constant

-0.460

0.900

-0.51

0.61

-2.224

1.304

Time since burn

-0.015

0.016

-0.95

0.34

-0.046

0.016

Canopy openness

-0.076

0.036

-2.13

0.03

-0.146

-0.006

Stand basal area

-0.665

0.140

-4.75

0.00

-0.939

-0.390

Tree height

1.997

0.995

2.01

0.05

0.046

3.948

-0.076

0.035

-2.21

0.03

-0.144

-0.009

Deciduous

3.543

0.481

7.37

0.00

2.601

4.485

Open

-6.514

0.983

-6.62

0.00

-8.441

-4.586

Coniferous

2.971

0.968

3.07

0.00

1.073

4.869

Constant

-7.080

7.808

-0.91

0.37

-22.383

8.223

Tree height
Cover type

2

Binary

Tree height
Cover type

2

105

Table 23. Coefficients and statistical parameters generated from the top ranked ((∆i AICc
differed by ≤2 points) fractional logit models for the prediction of Cladina spp. and Cetraria
spp. lichen cover (%, expressed as a proportion) in the Klaza caribou herd annual range,
west-central Yukon.
AICc rank

1

2

3

4

5

6

7

Parameter
Canopy openness
Stand basal area
Tree height
2
Tree height
Constant
Canopy openness
Tree height
2
Tree height
Constant
Tree height
2
Tree height
Cover type
Deciduous
Open
Coniferous
Constant
Time since burn
Canopy openness
Tree height
2
Tree height
Constant
Time since burn
Canopy openness
Stand basal area
Tree height
2
Tree height
Constant
Time since burn
Tree height
2
Tree height
Cover type
Deciduous
Open
Coniferous
Constant
Canopy openness
Tree height
2
Tree height
Shrub cover
Constant

β

Robust SE

Z

P

0.044
0.108
0.527
-0.032
-7.231
0.029
0.578
-0.032
-5.891
0.222
-0.019

0.012
0.038
0.106
0.008
1.132
0.010
0.112
0.008
0.958
0.152
0.009

3.66
2.86
4.99
-4.07
-6.39
2.9
5.17
-3.88
-6.15
1.46
-2.20

-1.168
-0.427
1.595
-2.974
0.006
0.028
0.498
-0.030
-5.915
0.005
0.040
0.088
0.482
-0.031
-7.032
0.005
0.205
-0.020

0.400
0.344
0.331
0.384
0.003
0.009
0.116
0.008
0.847
0.002
0.011
0.038
0.110
0.008
1.083
0.002
0.145
0.008

-1.119
-0.318
1.437
-3.187
0.033
0.587
-0.032
-0.006
-6.097

0.404
0.336
0.334
0.375
0.010
0.115
0.008
0.003
0.952

95% CI

0.000
0.004
0.000
0.000
0.000
0.004
0.000
0.000
0.000
0.145
0.027

Lower
0.020
0.034
0.320
-0.048
-9.449
0.009
0.359
-0.048
-7.768
-0.076
-0.037

Upper
0.067
0.183
0.734
-0.017
-5.014
0.049
0.797
-0.016
-4.014
0.520
-0.002

-2.92
-1.24
4.82
-7.74
2.52
3.18
4.29
-3.67
-6.98
1.92
3.53
2.33
4.39
-3.95
-6.49
2.14
1.41
-2.36

0.004
0.214
0.000
0.000
0.012
0.001
0.000
0.000
0.000
0.055
0.000
0.020
0.000
0.000
0.000
0.032
0.158
0.018

-1.953
-1.100
0.946
-3.727
0.001
0.011
0.270
-0.046
-7.575
0.000
0.018
0.014
0.267
-0.046
-9.155
0.000
-0.080
-0.036

-0.384
0.246
2.244
-2.221
0.011
0.045
0.725
-0.014
-4.254
0.009
0.063
0.163
0.698
-0.016
-4.910
0.010
0.489
-0.003

-2.77
-0.95
4.31
-8.50
3.26
5.1
-3.9
-1.86
-6.41

0.006
0.344
0.000
0.000
0.001
0.000
0.000
0.063
0.000

-1.911
-0.976
0.783
-3.922
0.013
0.361
-0.049
-0.012
-7.963

-0.327
0.340
2.091
-2.452
0.053
0.813
-0.016
0.000
-4.232

106

30
3UHGLFWHGREVHUYHGOLFKHQYROXPH
GPñPð 

TSB+CO+TH+TH²+SBA+CT
20
10
0
-10
-20
-30
-40
0

10

20

30

40

2EVHUYHGOLFKHQYROXPH GPñPð 

50

60



)LJXUH 'LIIHUHQFHLQWKHREVHUYHGDQGSUHGLFWHGYROXPHRIIRUDJHOLFKHQVJHQHUDWHGXVLQJ
WKH WRSUDQNHG ]HURLQIODWHG 3RLVVRQ PRGHO IRU YHJHWDWLRQ SORWV LQ WKH .OD]D FDULERX KHUG
UDQJH ZHVWFHQWUDO <XNRQ $ YDOXH RI ]HUR VXJJHVWV SHUIHFW SUHGLFWLRQ ZKHUHDV QHJDWLYH
YDOXHVLQGLFDWHPRGHOXQGHUSUHGLFWLRQDQGSRVLWLYHYDOXHVLQGLFDWHRYHUSUHGLFWLRQ



CO+TH+TH²

CO+SBA+TH+TH²
0.50

0.50

0.00

0.00

-0.50

3UHGLFWHGREVHUYHGOLFKHQFRYHU  

0.00

0.10

0.20

0.50

0.30

0.40

0.50

0.60

0.70

0.80

-0.50
0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.60

0.70

0.80

0.60

0.70

0.80

TSB+CO+TH+TH²

TH+TH²+CT

0.50

0.00

0.00

-0.50
0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

-0.50
0.00

0.10

0.20

TSB+CO+SBA+TH+TH²
0.50

0.00

0.00

-0.50

-0.50

0.10

0.20

0.30

0.40

0.50

0.40

0.50

TSB+TH+TH²+CT

0.50

0.00

0.30

0.60

0.70

0.80

0.60

0.70

0.80

0.00

0.10

0.20

0.30

0.40

0.50

CO+TH+TH²+TSC

0.50
0.00
-0.50
0.00

0.10

0.20

0.30

0.40

0.50



2EVHUYHGOLFKHQFRYHU  
)LJXUH 'LIIHUHQFHLQWKHREVHUYHGDQGSUHGLFWHGFRYHURIIRUDJHOLFKHQVJHQHUDWHGXVLQJWKHWRSUDQNHGIUDFWLRQDOORJLWPRGHOVIRUYHJHWDWLRQSORWV
LQWKH.OD]DFDULERXKHUGUDQJHZHVWFHQWUDO<XNRQ$YDOXHRI]HURVXJJHVWVSHUIHFWSUHGLFWLRQZKHUHDVQHJDWLYHYDOXHVLQGLFDWHPRGHOXQGHU
SUHGLFWLRQDQGSRVLWLYHYDOXHVLQGLFDWHRYHUSUHGLFWLRQ3DUDPHWHUVDUHGHILQHGLQ7DEOHPRGHOVWDWLVWLFVDUHJLYHQLQ7DEOH



Appendix C Lichen biomass estimates on winter
ranges of caribou

109

Table 24. Lichen biomass of various winter range types, both used and unused (random) by
Rangifer (caribou or reindeer; adapted from Barrier 2011).
Location
Northwest
Territories
Northwest
Territories
Northwest
Territories
Northwest
Territories
N.
Saskatchewan
N.
Saskatchewan
N. Manitoba
Ungava

Herd
Bathurst
Bathurst
Beverly (west block)
Beverly (east block)

Range Type
Used unburned
forest
Random
unburned forest
Used, >60 years
old
Used, >60 years
old

Biomass
(kg/ha)

Reference

No

2,412

Barrier 2011

No

2,516

Barrier 2011

No

2,594

Thomas et al. 1996

No

6,250

Thomas et al. 1996

Thallus

a

Beverly/Qamanirjuaq

Lichen-dominant

No

810

Scotter 1970

Qamanirjuaq

Lichen-dominant

Yes

5,850

Miller 1976

Qamanirjuaq
Leaf River

Lichen-dominant
Lichen-dominant

Yes
Unknown

4,270
1,223

Ungava

George River

All present

Unknown

3,170

N. Quebec

George/Leaf River

Forests <30
years

No

530

N. Quebec

George/Leaf River

All present

No

2,800

N. Quebec

George/Leaf River

Forests >90
years

No

8,010

N. Quebec

George/Leaf River

Lichen-dominant

No

5,440

Northern
Yukon

Miller 1976
Crete et al. 1990
Bergerud et al.
2008
Arseneault et al.
1997
Arseneault et al.
1997
Arseneault et al.
1997
Arseneault et al.
1997

Porcupine

All present

Yes

508

Russell et al. 1993

All present

Unknown

100–850

Fleischman 1990

All present

Unknown

>2,000

Fleischman 1990

Unknown

3,007

Joly et al. 2010

Unknown

1,260

Joly et al. 2010

Unknown

818

Joly et al. 2010

No

520

Unknown

11,000

No

38–
1,272

Helle 1981
Gaare and
Skogland 1980

Central Alaska
Central Alaska

Delta (traditional
areas)
Delta (peripheral
areas)

Used unburned
forest
Random
unburned forest
Random burned
forest
Heath forest
Climax forest
stands
Random burned
forest

NW Alaska

Western Arctic

NW Alaska

Western Arctic

NW Alaska

Western Arctic

N. Finland

Reindeer

Norway

Reindeer

b

Reindeer

Alberta

b

Woodland

Used

Unknown

4,017

Alberta

b

Woodland

Random

Unknown

Random

Wolverine

Used

Unknown

Nelchina

Used
Unused,
abandoned
stands

Unknown

1,730–
3,450
1,250

Unknown

400

Finland

British
b
Columbia
b
Alaska
b

Alaska

Nelchina

Kumpula et al. 2000
Edmonds and
Bloomfield 1984
Edmonds and
Bloomfield 1984
Johnson et al. 2001
Collins et al. 2011
Collins et al. 2011

110

1082–
Briand et al. 2009
3054
a
Indicates whether or not dead bases of lichen thalli were included in the biomass estimates (Barrier
2011).
b
Addition to Barrier’s (2011) original table.
Quebec

b

Forest-dwelling

Used

Unknown

111

Appendix D RSF model outputs

112

RSF selection coefficent (β)
RSF selection coefficent (β)

EW 2012/2013

2.40

1.40
0.40

- + -

+

- + - - - - - - + + - + -

-0.60
-1.60
-2.60

2.00
1.50
1.00
0.50
0.00
-0.50
-1.00
-1.50
-2.00

EW 2014/2015
+ + -

+

+ +
+ - + - - + - +
-

1.00
0.80
0.60
0.40
0.20
0.00
-0.20
-0.40
-0.60
-0.80
-1.00

0.40
0.30
0.20
0.10
0.00
-0.10
-0.20
-0.30
-0.40

EW 2013/2014
-

+ -

+

- + -

-

-

-

-

+ +

EW 2015/2016
+

+ -

-

+

-

-

+

+ +

-

+

-

-

+ -

+
+ -

113

Figure 18. Averaged, unweighted selection coefficients with 95% confidence intervals (representing variability of averaged
coefficients) for early winter resource selection function models of individual Klaza caribou. For very small values, positive and
negative (+/-) signs above coefficient values are provided for clarity.

RSF selection coefficient (β)

0.60
0.40
0.20

+ +
- -

-

0.00

+

-

+

1.50
+

-

+ -

0.50

-

-

-0.60

-1.50

LW 2015

0.00

-0.50
-1.00
-1.50

+ + -

+

- + - + -

-

- + -

+
-

-

+ -

+

- + - + -

+

-

-

-

+ -

-0.50
-1.00

0.50

+

0.00

-0.40

1.00

LW 2014

1.00

+

-

-0.20

1.50
RSF selection coefficient (β)

+

LW 2013

0.50
0.40
0.30
0.20
0.10
0.00
-0.10
-0.20
-0.30
-0.40
-0.50

LW 2016

+
- + -

-

+

+ +

- + -

-

+
-

+ -

-

114

Figure 19. Averaged unweighted selection coefficients with 95% confidence intervals (representing variability of averaged
coefficients) for late winter resource selection function models of individual Klaza caribou. For very small values, positive and
negative (+/-) signs above coefficient values are provided for clarity.

Number of years (+/- β)
Number of years (+/- β)

3
2
1
0

-1

3

a) Burned
EW 2012-13
EW 2013-14
EW 2014-15
EW 2015-16

2
1
0

-1

-2

-2

-3

-3

3
2
1
0
-1
-2
-3

c) Burn density

3

EW 2012-13
EW 2013-14
EW 2014-15
EW 2015-16

1

2

0
-1
-2

b) Unburned
EW 2012-13
EW 2013-14
EW 2014-15
EW 2015-16

d) Distance to burns
EW 2012-13
EW 2013-14
EW 2014-15
EW 2015-16

-3

Figure 20. Number of years during the study period (2012–2016) that individual Klaza caribou with three years of data demonstrated
positive or negative responses to burn characteristics during the early winter season.

115

Number of years (+/- β)

3

a) Burned

2

2
1

1

LW 2013
LW 2014
LW 2015
LW 2016

0
-1
-2

0
-1
-2

b) Unburned
LW 2013
LW 2014
LW 2015
LW 2016

-3

-3

Number of years (+/- β)

3

3

c) Burn density

2
1
0
-1
-2
-3

3

d) Distance to burns

2
LW 2013
LW 2014
LW 2015
LW 2016

1
0
-1

LW 2013
LW 2014
LW 2015

-2
-3

Figure 21. Number of years during the study period (2012–2016) that individual Klaza caribou with three years of data demonstrated
positive or negative responses to burn characteristics during the late winter season.

116

Appendix E Burn-focused logistic model outputs

117

Table 25. Area under the Receiver Operating Characteristic (ROC) curve scores calculated for
each percentile differentiating Habitat Selection Units with low versus high use by Klaza
caribou, west-central Yukon. Values in bold indicate the highest scores and the
corresponding percentile was used to set the threshold for that season-year.

Percentile
of combined
use metric
5
10
25
50
75
90
95
99

Early winter

2012/13
0.45
0.60
0.65
0.69
0.56
0.53
0.66
-

Late winter

2013/14

2014/15

2015/16

2013

2014

2015

2016

0.71
0.65
0.64
0.60
0.58
0.42
-

0.71
0.66
0.80
0.98
0.00
0.78

0.74
0.72
0.75
0.44
0.44
-

0.47
0.47
0.50
0.57
0.59
0.57
0.50
-

0.76
0.76
0.67
0.66
0.62
0.56
0.63
0.46

0.58
0.55
0.66
0.66
0.58
0.71
0.75
0.01

0.42
0.42
0.58
0.50
0.54
0.35
0.21
-

Note: To avoid creating an unbalanced data set, only ROC scores identified within the 10
percentiles were considered.

th

to 90

th

118

Table 26. Coefficients and statistical parameters generated from the top-ranked (∆i AICc
differed by ≤2 points) logistic models representing the relative use of Habitat Selection Units
by caribou during early winter in the Klaza herd annual range, west-central Yukon (2012–
2016).
Year

2012/2013

AICc
rank

1

1

2013/2014
2

Parameter

β

Robust SE

Z

P

Burn age

0.158

0.036

4.380

Lichen density

0.488

0.245

Lichen density²

-0.036

Core

95% CI
Lower

Upper

0.000

0.087

0.228

2.000

0.046

0.009

0.968

0.018

-1.960

0.050

-0.072

0.000

0.468

0.468

1.000

0.317

-0.449

1.384

Inner 500

-0.250

0.372

-0.670

0.503

-0.979

0.480

Outer 500

-0.218

0.336

-0.650

0.517

-0.877

0.441

Constant

-1.935

0.639

-3.030

0.002

-3.188

-0.682

Burn age

0.938

0.956

0.980

0.326

-0.935

2.811

Lichen density

-0.229

0.256

-0.890

0.371

-0.731

0.273

Lichen density²

0.012

0.020

0.570

0.568

-0.028

0.052

Core

-0.632

0.496

-1.280

0.202

-1.604

0.339

Inner 500

0.433

0.428

1.010

0.312

-0.406

1.273

Outer 500

0.199

0.436

0.460

0.647

-0.655

1.053

Constant

-0.224

2.887

-0.080

0.938

-5.883

5.435

Elevation

-0.102

0.072

-1.400

0.160

-0.243

0.040

Elevation²

0.000

0.000

1.540

0.125

0.000

0.000

Valley

1.257

0.572

2.190

0.028

0.135

2.378

Gentle slope

-0.272

0.504

-0.540

0.590

-1.259

0.716

Steep slope

0.224

0.726

0.310

0.758

-1.200

1.648

Ridge

-1.209

1.121

-1.080

0.281

-3.407

0.988

Lichen density

-0.297

0.303

-0.980

0.328

-0.891

0.297

Lichen density²

0.019

0.022

0.870

0.383

-0.024

0.061

Burn age

1.252

0.932

1.340

0.179

-0.575

3.078

Core

-0.294

0.581

-0.510

0.613

-1.433

0.846

Inner 500

0.448

0.497

0.900

0.368

-0.526

1.422

Outer 500

-0.154

0.464

-0.330

0.740

-1.063

0.756

Constant

50.528

39.869

1.270

0.205

-27.615

128.671

Elevation

-0.105

0.059

-1.770

0.076

-0.221

0.011

Elevation²

0.000

0.000

1.840

0.066

0.000

0.000

Lichen density

-0.244

0.262

-0.930

0.353

-0.758

0.271

Lichen density²

0.013

0.021

0.630

0.529

-0.027

0.053

Burn perimeter use

Burn perimeter use

Burn perimeter use

3

119

Year

AICc
rank

Parameter

β

Robust SE

Z

P

Burn age

0.977

0.891

1.100

Core

-0.505

0.511

Inner 500

0.315

Outer 500

95% CI
Lower

Upper

0.273

-0.770

2.724

-0.990

0.323

-1.505

0.496

0.436

0.720

0.470

-0.539

1.168

0.190

0.462

0.410

0.681

-0.716

1.096

Constant

56.786

32.605

1.740

0.082

-7.118

120.690

Burn age

0.251

0.061

4.150

0.000

0.133

0.370

Lichen density

1.262

0.591

2.140

0.033

0.105

2.420

Lichen density²

-0.060

0.035

-1.730

0.084

-0.127

0.008

Core

-9.615

0.699

-13.750

0.000

-10.985

-8.244

Inner 500

4.647

0.560

8.300

0.000

3.550

5.744

Outer 500

4.967

0.594

8.370

0.000

3.804

6.131

Constant

-14.515

2.670

-5.440

0.000

-19.748

-9.282

Elevation

-0.054

0.055

-0.990

0.322

-0.161

0.053

Elevation²

0.000

0.000

1.100

0.271

0.000

0.000

Valley

2.723

1.325

2.050

0.040

0.126

5.320

Gentle slope

5.701

0.851

6.700

0.000

4.033

7.368

Steep slope

-12.436

1.186

-10.490

0.000

-14.760

-10.112

Ridge

4.013

0.771

5.210

0.000

2.502

5.523

Lichen density

2.089

0.747

2.800

0.005

0.625

3.553

Lichen density²

-0.100

0.041

-2.440

0.015

-0.181

-0.020

Burn age

0.377

0.091

4.140

0.000

0.199

0.556

Core

-8.831

0.919

-9.610

0.000

-10.632

-7.030

Inner 500

4.541

0.720

6.310

0.000

3.131

5.951

Outer 500

4.290

0.651

6.590

0.000

3.014

5.565

Constant

7.127

34.934

0.200

0.838

-61.342

75.596

Northness

0.218

0.348

0.620

0.532

-0.465

0.900

Eastness

-0.383

0.340

-1.130

0.259

-1.050

0.283

Slope

0.044

0.068

0.650

0.516

-0.089

0.177

Elevation

0.077

0.045

1.720

0.085

-0.011

0.166

Elevation²

0.000

0.000

-1.790

0.074

0.000

0.000

Valley

-0.746

0.502

-1.490

0.137

-1.729

0.237

Gentle slope

-0.191

0.518

-0.370

0.712

-1.207

0.824

Steep slope

-1.398

0.631

-2.220

0.027

-2.634

-0.162

Ridge

2.336

0.524

4.460

0.000

1.308

3.363

Lichen density

2.277

0.485

4.690

0.000

1.325

3.228

Burn perimeter use

1

2014/2015

2

Burn perimeter use

Burn perimeter use

2015/2016

1

120

Year

AICc
rank

Parameter

β

Robust SE

Z

P

Lichen density²

-0.151

0.036

-4.230

Distance to water

0.000

0.000

Constant

-53.985

Lichen density
Lichen density²

95% CI
Lower

Upper

0.000

-0.222

-0.081

-1.540

0.123

-0.001

0.000

28.401

-1.900

0.057

-109.651

1.680

1.803

0.398

4.530

0.000

1.023

2.582

-0.123

0.028

-4.340

0.000

-0.179

-0.068

Core

0.093

0.429

0.220

0.828

-0.747

0.934

Inner 500

-0.381

0.346

-1.100

0.271

-1.059

0.297

Outer 500

0.288

0.322

0.890

0.372

-0.343

0.918

Constant

-6.643

1.253

-5.300

0.000

-9.099

-4.187

Elevation

0.056

0.042

1.320

0.187

-0.027

0.139

Elevation²

0.000

0.000

-1.370

0.171

0.000

0.000

Valley

-1.137

0.555

-2.050

0.041

-2.225

-0.048

Gentle slope

-0.340

0.396

-0.860

0.391

-1.116

0.437

Steep slope

-1.096

0.533

-2.060

0.040

-2.141

-0.051

Ridge

2.572

0.567

4.540

0.000

1.461

3.684

Lichen density

2.295

0.496

4.630

0.000

1.323

3.267

Lichen density²

-0.153

0.036

-4.190

0.000

-0.224

-0.081

Burn age

-0.026

0.023

-1.130

0.258

-0.071

0.019

Core

0.620

0.562

1.100

0.270

-0.481

1.722

Inner 500

-0.657

0.430

-1.530

0.127

-1.500

0.186

Outer 500

0.037

0.381

0.100

0.923

-0.709

0.783

Constant

-40.593

26.575

-1.530

0.127

-92.679

11.493

Burn perimeter use
2

3

Burn perimeter use

Table 27. Coefficients and statistical parameters generated from the top-ranked (∆i AICc
differed by ≤2 points) logistic models representing the relative use of Habitat Selection Units
by caribou during late winter in the Klaza herd annual range, west-central Yukon (2013–
2016).
Year

2013

AICc
rank

1

Parameter
Lichen density
Lichen density²
Burn perimeter use
Core
Inner 500
Outer 500
Constant

β

Robust SE

Z

P

0.120
0.012

0.202
0.016

0.600
0.720

-0.207
0.168
0.039
-2.058

0.286
0.241
0.230
0.562

-0.730
0.700
0.170
-3.660

95% CI

0.551
0.472

Lower
-0.275
-0.020

Upper
0.515
0.438

0.468
0.486
0.865
0.000

-0.767
-0.305
-0.412
-3.159

0.352
0.641
0.490
-0.958

121

Year

AICc
rank

2

2014

1

1

Parameter

β

Robust SE

Z

P

Lichen density

0.156

0.208

0.750

Lichen density²
Burn age

0.009
-0.007

0.017
0.013

0.540
-0.520

Core

-0.151

0.291

Inner 500

0.151

0.242

Outer 500

0.000

Constant

95% CI
Lower

Upper

0.454

-0.252

0.564

0.591
0.600

-0.024
-0.031

0.042
0.018

-0.520

0.604

-0.722

0.420

0.620

0.533

-0.324

0.626

0.233

0.000

1.000

-0.457

0.457

-1.989

0.570

-3.490

0.000

-3.105

-0.872

Elevation

-0.030

0.024

-1.260

0.208

-0.077

0.017

Elevation²

0.000

0.000

1.120

0.261

0.000

0.000

Lichen density

0.140

0.313

0.450

0.654

-0.473

0.753

Lichen density²

0.010

0.036

0.280

0.782

-0.060

0.080

Burn age

0.083

0.029

2.890

0.004

0.026

0.139

Core

-0.771

0.283

-2.720

0.006

-1.326

-0.216

Inner 500

-0.007

0.289

-0.020

0.981

-0.573

0.560

Outer 500

0.778

0.351

2.210

0.027

0.090

1.466

Constant

17.182

11.927

1.440

0.150

-6.194

40.559

Slope

-0.063

0.031

-2.000

0.045

-0.125

-0.001

Elevation

0.076

0.028

2.770

0.006

0.022

0.130

Elevation²

0.000

0.000

-2.840

0.005

0.000

0.000

Lichen density

0.541

0.336

1.610

0.108

-0.118

1.200

Lichen density²

-0.029

0.027

-1.080

0.281

-0.082

0.024

Burn age

-0.027

0.015

-1.810

0.070

-0.055

0.002

Core

-0.584

0.472

-1.240

0.216

-1.508

0.341

Inner 500

-0.259

0.357

-0.730

0.467

-0.958

0.440

Outer 500

0.843

0.295

2.860

0.004

0.264

1.421

Constant

-40.099

13.973

-2.870

0.004

-67.485

-12.713

Elevation

0.065

0.026

2.480

0.013

0.014

0.116

Elevation²

0.000

0.000

-2.590

0.010

0.000

0.000

Lichen density

0.574

0.326

1.760

0.078

-0.064

1.212

Lichen density²

-0.029

0.026

-1.130

0.257

-0.080

0.021

Burn age

-0.024

0.014

-1.750

0.081

-0.051

0.003

Core

-0.534

0.467

-1.140

0.253

-1.449

0.382

Inner 500

-0.308

0.359

-0.860

0.391

-1.011

0.395

Outer 500

0.842

0.291

2.890

0.004

0.271

1.412

Burn perimeter use

Burn perimeter use

Burn perimeter use

2015

2

Burn perimeter use

122

Year

2016

AICc
rank

1

Parameter

β

Robust SE

Z

P

Constant

-35.082

13.393

-2.620

Burn age

0.034

0.012

Lichen density

-0.417

Lichen density²

95% CI
Lower

Upper

0.009

-61.332

-8.833

2.760

0.006

0.010

0.058

0.260

-1.600

0.109

-0.926

0.093

0.037

0.021

1.790

0.073

-0.003

0.077

Core

-0.207

0.299

-0.690

0.488

-0.794

0.379

Inner 500

-0.047

0.244

-0.190

0.848

-0.525

0.431

Outer 500

0.254

0.240

1.060

0.291

-0.217

0.725

Constant

1.254

0.675

1.860

0.063

-0.069

2.577

Burn perimeter use

123