Ecogfaphy 35: 0 0 1 -0 1 2, 2012
doi: 1 0 .1 1 1 1 /j.l600-0587.2012.07733.x
© 2012 The A uthors. Ecography © 2012 N ordic Society Oikos
Subject E ditor: Eric Post. Accepted 11 July 2012
Resource separation analysis with m oose indicates threats to
caribou in human altered landscapes
Wibke Peters, Mark Hebblewhite, Nicholas DeCesare, Francesca Cagnacci and Marco Musiani
W Peters ([email protected]), M . Hebblewhite a n d N . DeCesare, Wildlife Biology Program, D ept o f Eeosystem Seienees and
Conservation, College o f Forestry a n d Conservation, Univ. o f Montana, Missoula, M T 59812, USA. — P. Cagnaeei, D ept o f Biodiversity a n d
Moleeular Eeology, Researeh a n d Innovation Centre, Eondazione E d m u n d Maeh, via E. M aeh 1, IT -3 8 0 1 0 San Miehele alTAdige, TN, Italy.
— M . M usiani, Faculties o f Environm ental Design a n d Veterinary Medicine, EVD S, Univ. o f Calgary, 2 5 0 0 University D rive NW , Calgary,
A B T 2 N 1 N 4 , Canada.
Species recovery is often im peded by inadequate knowledge o n m echanism s o f c om m unity interactions th a t cause
and exacerbate species endangerm ent. C aribou and w ild reindeer Rangifer tarandus are declining in m any regions o f their
circum polar range likely because o f h um an-induced landscape changes. In general, th eir niche specialization enables
Rangifer to survive in nutrien t-p o o r habitats spatially separated from o ther ungulates a n d their shared predators. Research
has indicated th at shifts in prim ary prey d istribution follow ing hu m an landscape alteration m ay result in spatial overlap
w ith Rangifer. W e studied overlap relationships o f w oodland caribou R. t. caribou a n d moose Alees alees, quantified by
their differential use o f environm ental resources, a n d evaluated the role o f h u m an landscape alteration in spatial sepa­
ration in south-w estern C anada. A nthropogenic conversion o f old-grow th forests to early serai stands is hypothesized
to decrease the spatial separation betw een caribou and moose, the d o m in a n t prey for wolves Canis lupus, contributing
to increased caribou m ortality. R edundancy analysis (RDA) was first used to exam ine coarse scale resource separation
across o u r study area. Second, at a finer spatial scale, we used logistic regression to com pare resource- and spatial separa­
tio n o f sym patric pairs o f 17 moose and 17 caribou. Finally, we tested if the frequency o f predator-caused caribou m or­
talities was higher in regions w ith higher moose resource use. A lthough environm ental resource separation was strong
at the coarser scale, we observed substantial spatial overlap ( > 50% ) at the finer scale. In sum m er we reported a signifi­
cant positive relationship betw een spatial overlap o f moose and caribou and the degree o f h u m an landscape alteration.
M ost im portantly, locations o f caribou m ortalities corresponded w ith areas o f high resource use by moose in sum m er.
Thus, consistent w ith the spatial separation hypothesis, o u r research suggests th a t early successional forest stages may
decrease spatial separation betw een caribou a n d moose, resulting in increased m ortality risk for threatened caribou.
O ver the last century hum ans have significantly im pacted
the global environm ent, leading to dram atic changes in
species distributions and increased extinction rates well
above natural background levels (C hapin et al. 2011).
Ecosystem functions and processes are com m only influ­
enced by interactions am ong species and hum an-induced
changes o f systems such as com petition and trophic inter­
actions can have w ide-ranging ecosystem effects. Direct
(e.g. habitat loss or over-exploitation) and indirect (e.g.
changes in com m unity interactions) mechanisms often
act concurrently and their com bination can drive vulner­
able populations towards extinction (Brook et al. 2008).
Therefore, conservation biologists need to understand the
mechanisms leading to population declines, and complex
interactions am ong those mechanisms to manage and con­
serve species. Unfortunately, how hum an landscape altera­
tion can affect spatial overlap o f species and their use o f
resources, and thus com petitive interactions, often remains
unstudied in m any systems. For example, interactions o f
species in the same trophic level can lead to declines when
indirectly m ediated by a shared predator (DeCesare et al.
2010). A pparent com petition is the process by w hich
two prey species can affect each other’s growth rates
through their contribution to the num erical response
o f a shared predator (H olt and Lawton 1994). Examples
for declining species due to hum an-induced apparent
com petition include Vancouver Island m arm ots M armota
vancouverensis (Bryant and Page 2005), Sierra Nevada
bighorn sheep Ovis canadensis californiana (Johnson
et al. 2012) or the extinct M acquarie Island parakeet
Cyanoramphus novaezelandiae erythrotis (Taylor 1979).
O ne o f the m ost im p o rtan t relationships th at perm it
species coexistence is the differential use o f physical and
biological com ponents (i.e. resources), w hich places sym­
patric species into specific habitats (Rosenzweig 1981).
Following H utchinson (1957), in the absence o f inter­
specific interactions (e.g. predation and com petition) a
species can occupy a geographical region where the abiotic
Early View (EV); 1-EV
conditions (e.g. climate) are suitable for its survival and
reproduction, called the fundam ental niche. W henever
species coexist in geographical space, a species occupies
only a proportion o f its fundam ental niche due to inter­
specific interactions, w hich is called the realized niche.
The behavioral process o f habitat selection (Johnson 1980)
often leads to spatio-tem poral separation am ong spe­
cies (Chesson and Kuang 2008) due to differing resource
preferences (e.g. diet), facilitating their sympatric coexis­
tence through the differentiation o f their realized niches.
Therefore, com m unity studies often focus on assessing
environm ental gradients o f resources, linking resource
variables at used (e.g. global positioning (GPS) collar
data) or occupied locations in geographic space (i.e. habi­
tat) to realized niches in environm ental space (Hirzel and
Le Lay 2008). The arrangem ent o f resources in geographic
space m ay determ ine species distribution and the degree
sym patric species’ realized niches overlap or separate. In
sufficiently heterogeneous habitats, apparently com peting
prey species may be able to spatially separate and thereby
avoid shared predation by isolation in spatial refuges due
this differential resource selection (DeCesare et al. 2010).
Lastly, it is com m only assumed th at a species’ resource
selection and therefore, its geographical distribution, is
in close relationship w ith its ecological requirem ents that
p erm it positive population grow th (Hirzel and Le Lay
2008). However, especially in the face o f hum an land­
scape alteration, animals may occupy suboptim al habitat
where death rates may exceed birth rates (Pulliam 2000).
A n example for the im portance o f hum an-altered
resource distribution driving species declines are caribou
and m any wild reindeer Rangifer tarandus populations,
w hich are decreasing throughout their circum polar range
coincident w ith landscape alteration (e.g. im plem entation
o f infrastructure, m ineral extraction or forest harvesting;
Vors and Boyce 2009). A growing body o f literature
describes how hum an landscape changes have led to shifts
in the distribution o f reindeer and caribou (Vistnes et al.
2001) or altered predator-prey relationships (Seip 1992,
Kojola et al. 2004) and thereby have directly and indi­
rectly decreased population size o f these habitat special­
ists (Nellem ann and Cam eron 1998). As in other parts
o f Rangifer range, conservation o f w oodland caribou is
am ongst the m ost pressing conservation challenges in
Canada. In 2000, boreal and southern m ountain w ood­
land caribou were federally listed as threatened under the
Canadian Species at Risk A ct (SARA). A lthough hum an
landscape change has been identified as the ultim ate cause
for w oodland caribou declines (M cLoughlin et al. 2003),
the proxim al mechanisms are hypothesized to be m edi­
ated by changes in predator—prey dynamics in the large
m am m al com m unity in w hich caribou occur (James et al.
2004). This necessitates a spatial com m unity approach to
caribou conservation.
W hile moose Alees alees and w oodland caribou
R. t. caribou (hereafter caribou) are sym patric throughout
the boreal forest, they are hypothesized to coexist through
their differential use o f resources (Boer 2007). For exam­
ple, caribou diet is com prised o f terrestrial and arboreal
lichens, especially during w inter (Thomas et al. 1996),
and caribou select large contiguous habitat patches o f low
2-EV
productivity, older serai conifer stands where lichen bio­
mass is highest. In m ountain regions, such forests are
generally at higher elevations and caribou often exhibit
seasonal altitudinal m igration (Seip 1992). M oose, in
contrast, are generalist browsers o f early succession shrubs
th at prosper following fire or forestry (Peek 2007). D espite
this broad forage separation, their diet often overlaps in
sum m er w hen b o th species consum e forbs and decidu­
ous plants (Boer 2007). The spatial separation hypothesis
suggests th at the niche specialization by caribou enables
them to survive in n u trien t-p o o r habitats at low densi­
ties, spatially separated from other ungulates and their
predators, w hich reduces the negative effects o f apparent
com petition and increases survival (Bergerud and Page
1987, Seip 1992). However, the conversion o f old-grow th
forests to early serai stage forests is hypothesized to increase
the abundance o f moose, the d om inant prey for wolves
Canis lupus, and thereby increase w olf densities (Kojola
et al. 2004, Serrouya et al. 2011). Predation by wolves is
a leading cause o f caribou m ortality (W ittm er et al. 2005),
and wolves (as well as other predators) select for landcover types w ith high ungulate forage biomass (e.g. shrub
com m unities, burns, logged areas; M osnier et al. 2008,
G urarie et al. 2011) to increase encounter rates w ith prey
(Hebblewhite et al. 2005). Thus, understanding how the
realized niches o f moose and caribou differ in the context
o f spatial separation is key to evaluating the mechanisms of
apparent com petition.
We aimed to determ ine the relationship between the
separation o f realized niches o f moose and caribou and
hum an landscape alteration by relating data o f used animal
locations to environm ental covariates (Chesson and Kuang
2008). As noted above, the geographic distribution o f a
species may n o t always reflect its ecological requirem ents
(i.e. population growth m ay be negative; Hirzel and Le Lay
2008). W hile we did n o t assess growth rates o f moose and
caribou populations, we will use the term ‘resource sepa­
ration’ as a proxy for separation o f realized niches herein.
Further, because ungulates respond to lim iting factors in
their environm ent in a hierarchical fashion across spatial
scales (Senft et al. 1987), hierarchical approaches should
also be considered w hen com paring realized niches o f
sympatric species (Ihl and Klein 2001). First, we tested
for resource separation by assessing the structure o f the
caribou and moose realized niches explained by environ­
m ental variables at a coarser scale across caribou herds
w ithin our study area using redundancy analysis (RDA; ter
Braak 1995). We expected substantial resource separation at
this coarser scale, because research has shown th at caribou
exhibit stronger sensitivity to h um an landscape alteration
at coarser spatial scales (e.g. landscape scale versus w ithinhom e range scale according to Johnson 1980, Polfus et al.
2011, DeCesare et al. 2012b), while moose are expected
to select browse-rich habitats often associated w ith hum an
resource extraction activities (Boer 2007). Second, we
tested w hether b oth resource and spatial separation
between sym patric caribou and moose was lower at finer
spatial scales in h um an altered landscapes by com paring
resource use o f sympatric individual caribou and moose
w ith logistic regression (Latham et al. 2011a) and overlap
indices (Schoener 1974). In contrast to the coarse scale.
we expected greater resource overlap o f paired (i.e. sym­
patric) caribou and moose in regions w ith high hum an
landscape alteration, due to the lim ited capability of
caribou to space away and avoid resource overlap w ith
moose at these finer scales (Bergerud and Page 1987).
Moreover, we expected resource overlap to be higher
during sum m er due to increased forage overlap between
moose and caribou (Boer 2007). Finally, the spatial sepa­
ration hypothesis predicts increased risk o f caribou m or­
tality in areas o f higher probability o f moose occurrence
(M cLoughlin et al. 2005). Therefore, we also examined
the relationship between predation-caused m ortalities o f
caribou and spatial predictions o f the probability o f resource
use by moose and caribou in geographical space.
edrom-Praine Ctsek
■ .4
*
I*
ttle Smoky
nde Cache
cBride
M ethods
Study area
We assessed spatial relationships of caribou and moose in
an approxim ately 3 6 0 0 0 km^ study area in the foothills
and m ountains o f w est-central AB and east-central BC
w ithin the ranges of six declining spatially distinct w ood­
land caribou herds: A La Peche (ALP), Red Rock Prairie
Creek (RPC), Little Smoky (LSM), the Narraway (NAR),
Redwillow (RW) and theT onquin (T O N ) in Jasper N ational
Park (Fig. 1). H um an landscape change varied throughout
the study area w ith low hum an landscape alteration and
a high proportion o f protected areas in the west. The
eastern part o f the study area was characterized by pro­
vincial lands m anaged prim arily for resource extraction,
w ith correspondingly higher hum an landscape alteration in
the form o f oil and gas extraction and forestry exploitation
and therefore high densities o f forest harvesting (mainly
clear-cut block harvesting) and linear developments (e.g.
roads, pipelines, seismic lines). A nthropogenic impacts were
greatest in the LSM caribou range. Elevations followed an
increasing gradient from east to west from about 500 to
> 3000 m. Lower elevations were characterized by mixedw ood forests, com prised mainly of trem bling aspen Populus
tremuloides, lodgepole pine Pinus contorta, w hite spruce
Picea glauca, and black spruce Picea mariana\ while the
western forests in the m ountain region, were dom inated
by lodgepole pine and Engelm ann spruce Picea engelmanii.
Moose and w hite-tailed deer Odocoileus virginianus com ­
prised the m ajority o f the ungulate population, whereas
elk Cervus canadensis, m ule deer Odocoileus hemionus
and w oodland caribou were less com m on. Further, big­
horn sheep Ovis canadensis and m ountain goats Oreamnos
americanus inhabited the m ountain region.
Animal capture
We captured and radio-collared moose via net-gunning
(Barrett et al. 1982) in winters of 2007/2008 and 2008/
2009. We used data from Global Positioning System (GPS)
collars (ATS G 2000 GPS collars; Advanced Telemetry
Systems, Isanti, M N , USA) deployed on ten female and
seven male moose w ithin and adjacent to caribou population
hom e ranges (Fig. 1). We radio-collared female and male
100 km
Caribou GPS Locations
.
+
Protected Areas
Moose GPS Locations
Caribou Mortality Sites
BC
AB
BC-AB Border
•
Towns
Caribou Herd 95% Kernels —
Roads
^
Figure 1. Study area in west central A lberta (AB) and east central
British C olum bia (BC). In this study we used global positioning
(GPS) collar data from 17 moose and 17 caribou. M oose GPS
collars were deployed between winters 2007/2008 and 2009/
2010 w ithin or adjacent to caribou herd hom e ranges. GPS
locations from caribou were collected between winters 2 006/2007
and 2009/2010.
moose to evaluate population-level habitat use and moose
population overlap w ith female caribou. For threatened
caribou populations, female caribou are m ost relevant
for m onitoring population growth rates (DeCesare et al.
2012a). Therefore, we used GPS collar (GPS 3300, 4400,
LO TEK Engineering, Newmarket, O N , Canada) data
from 17 female caribou, captured using the same m ethods
as described above for moose. N et-gunning protocols were
approved by the Univ. o f M ontana Animal Care and Use
Protocol 056-56M H EG S-010207 and 059-09M H W B 122109, Alberta Sustainable Resource Developm ent licenses
no. 21803, 27086, 27088, 27090 and Parks Canada
perm it JN P-2007-952. Both moose and caribou GPS
collars collected locations every two to four hours, which
we re-sampled to a consistent four-hour relocation sched­
ule for one year. Fix rate success o f < 90% can cause
habitat-induced bias in resource selection studies (Frair
et al. 2004). In our study, fix-rates for moose and caribou
were 92.4 and 90.3% respectively. As a result, we did not
correct for habitat-induced fix-rate bias. We conducted
all analysis for two seasons, summ er and winter. W oodland
caribou w ithin our study area are partially m igratory
(M cDevitt et al. 2009), and we defined sum m er (16 M ay 16 October) and w inter (17 O c to b e r-15 May) seasons
according to nonlinear regression analysis o f m ean migra­
tion dates (DeGesare et al. 2012b).
3-EV
Coarse-scale caribou-moose resource separation
We first assessed resource separation o f caribou and moose
populations by m easuring the realized niche position as
a function o f a suite o f environm ental covariates for
caribou and moose used locations across our study area
(i.e. similar to the second order scale described by Johnson
(1980)) using the ordination m ethod o f stepwise redun­
dancy analysis (RDA; ter Braak 1995, Bowman et al. 2010).
In RDA, the ordination axes for the species m atrix are
constrained to be linear com binations o f the columns of
the environm ental m atrix to obtain the best linear com bi­
nations o f environm ental variables that maximise resource
separation between species. Thus, the distribution of
the two species along these environm ental gradients can
be considered as the realized niche w ithin our study area
(i.e. coarse scale). We overlaid a 500 X 500 m sampling
grid (largest resolution o f GIS data sets; Supplem entary
m aterial A ppendix 1, Table A l) onto our study area and
assessed the presence o f GPS locations from each spe­
cies in each grid cell. Because RD A excludes null values
(i.e. grid cells w ithout observations), the sample size for
RD A was constituted by all grid cells used by at least one
species (n „ ^ ^ „ = 6175, n^mter = 7309). G ontinuous envi­
ronm ental variables (Supplem entary m aterial A ppendix 1,
Table A l) were averaged w ithin each grid cell. For cat­
egorical variables (i.e. landcover types) we estim ated pro­
portions w ithin each grid cell in ArcGIS Desktop 9.3.1
software (ESRI, Redlands, GA) and arcsine square root
transform ed them . M onte Garlo perm utation tests were
used to assess the significance o f constraints (999 perm uta­
tions, a = 0.05; ter Braak 1992). Then, we produced sea­
sonal ordination biplots to represent the moose and caribou
assemblage and environm ental resource covariates in realized
niche space (ter Braak 1995). We assessed the fraction that
constrained variance represented o f all covariances between
species and environm ent (ter Braak 1995) and reported
canonical coefficients (GG) to address the influence of
environm ental variables in structuring the ordinations.
Based on the spatial separation hypothesis, caribou and
moose should separate niche space across our study area
and thus, should each be associated w ith a unique set of
environm ental resource covariates. First, we predicted that
the species scores, i.e. the coordinates along the ordina­
tion axes specifying the position o f the species in realized
niche space, w ould be strongly contrasting. We also
expected the axis separating caribou and moose in niche
space to be dom inant over the axis associating the two spe­
cies. W ith respect to specific covariates, we predicted that
caribou w ould be associated w ith higher elevations and
older forest structures, while moose w ould group w ith
variables representing early serai forest stands and hum an
landscape alteration (i.e. clear-cuts, burns and N D V I).
These analyses were perform ed by the R 2.13.1 software w ith
the package Vegan’ 2.00-0 (Oksanen et al. 2011).
Fine-scale caribou-moose resource and spatial
separation
To evaluate separation o f moose and caribou resource
use at a finer scale, we paired each caribou w ith one
4-EV
sympatric moose in or near its respective caribou herd
hom e-range (95% fixed kernel) to m aintain equal avail­
ability o f resources to each pair (i.e. similar to third order
scale described by Johnson (1980)). We used logistic
regression (H osm er and Lemeshow 2000) to m odel seasonal
(i.e. sum m er and winter) differences in the resource use
o f moose and caribou, where caribou used locations were
coded as 1 and moose used locations as 0 (Latham et al.
2011a). This analysis determ ined w hich covariates pre­
dicted similarities and differences in resource use between
paired caribou and moose at smaller spatial scales, measured
by the estim ated (J coefficients from logistic regression.
Negative coefficients indicate less use by caribou com pared
to moose and positive coefficients indicate m ore resource
use by caribou com pared to moose. We again, predicted
th at caribou w ould use higher elevations and lower hum an
landscape alteration than moose. A random intercept
(PO + yOj) for each caribou-m oose pair was used to account
for differences in sample sizes o f GPS locations o f individ­
ual animals using generalized linear mixed-effects models
(GLM M ; Gillies et al. 2006).
We employed a m anual stepwise m odel selection
process described by H osm er and Lemeshow (2000) and
considered candidate covariates (Supplementary material
A ppendix 1, Table A l) previously reported to influence
caribou and moose resource use. All covariates were screened
for collinearity using the Pearson’s correlation coefficient
threshold o f |r| > 0 . 6 , retaining the collinear covariate
w ith the higher log-likelihood, highest coefficient o f deter­
m ination (pseudo R2) and lowest p-values. Further, n o n ­
linear covariates were transform ed upon visual inspection
(Hosm er and Lemeshow 2000). We first conducted
univariate logistic regression analysis, using a p < 0.25 as a
cut-off for the inclusion in model building. Retained cova­
riates entered the multivariate logistic regression modeling
process to build a small subset o f biologically sensible can­
didate models (Hosm er and Lemeshow 2000). We selected
the top m odel using Akaike’s inform ation criterion (AAIG;
Burnham and A nderson 2002). We reported standard­
ized W ald statistics
where we divided the W ald
statistic for each variable by the average o f the absolute
values o f all W ald statistics estimated for all predictor
variables included in the top model (DeGesare et al. 2012b).
These standardized W ald statistics allowed us to compare
the direction and strength o f resource variable use by the
two ungulate species relative to each other between seasonal
models and variables. Like P-coefficients, positive or negative
values o f
indicate resource use for increasing or decreas­
ing values in the predictor variable by caribou relative to
moose, respectively. Statistical analyses were carried out in
STATA 11.0 (StataGorp 2007). To assess the predictive capa­
bilities o f caribou-m oose resource selection models, we con­
ducted k-fold ( k = 5) cross validation (Boyce et al. 2002).
M odels estim ating resource separation by species in
environm ental space, are com m only used to predict the
spatial separation in geographical space (i.e. habitat; Hirzel
and Le Lay 2008, Latham et al. 2011a). We measured
spatial separation o f resource use by moose and caribou
based on fine-scale logistic regression (described above) by
translating environm ental patterns o f resource use (realized
niche) into spatial predicted values (geographical space;
Hirzel and Le Lay 2008). To do this, we developed spa­
tial raster maps at a 30 m resolution based on top logistic
regression models that predicted the probability th at a
pixel was used by caribou or moose using ArcGIS 9.3.1.
Values closer to 0 indicated the highest relative probabil­
ity o f use by moose and conversely, values closer to 1 indi­
cated the highest relative probability o f use by caribou. We
then classified the predicted probabilities o f m oose/caribou
use across the study area into 1 0 equal-sized categories
(Boyce et al. 2002). Finally, we counted the frequency
o f GPS locations by each species in each resource use
category. As an index o f resource separation between
m oose and caribou we calculated Schoener’s overlap index
in these 1 0 ranked resource use categories for animal
locations o f individual pairs moose and caribou. In this
way, Schoener’s overlap index provided a measure o f the
spatial distribution o f each caribou relative to its paired
m oose (Abrams 1980). We predicted th at hum an land­
scape alteration is positively correlated w ith resource over­
lap between caribou and moose. We tested our prediction
by regressing the Schoener’s overlap index for each
caribou-m oose pair versus the intensity o f hum an land­
scape alteration measured as the proportionate area den­
sity o f clear-cut forest per u n it area (at a radius o f 3 km as
% area/ 1 0 0 ; see landscape covariates section; hereafter
referred to as ‘clear-cut density’) at caribou GPS locations.
Because the diet o f caribou and moose can overlap in sum ­
m er w hen bo th species consum e forbs and deciduous veg­
etation (Boer 2007), we predicted that spatial separation
w ould be lower in summer.
Mortality consequences of spatial caribou-moose
overlap
(categories 1—3), overlap (categories 4—7) and caribou
(categories 8—10) habitat. The expected frequency distri­
butions for the chi-squared tests were estim ated based on
the relative proportions o f used caribou GPS locations in
each category.
Landscape covariates
We estimated caribou and moose resource use w ith spatial
covariates, including elevation, slope, aspect, percent snow
cover, normalized difference vegetation index (NDVI)
and landcover types (see DeGesare et al. 2012b and
Supplem entary material A ppendix 1 for details). To address
impacts o f hum an landscape alteration on caribou-m oose
resource separation, we used vector geodatabases to measure
linear features (km km^^; roads, seismic exploration lines,
rail roads, etc.) w ithin 1 km circular neighborhoods
surrounding each raster pixel (Supplementary material
A ppendix 1). We characterized impacts from forest
harvesting by clear-cuts as a landcover class, as well as a
relative index o f clear-cut density. To identify the relevant
spatial scale at w hich clear-cut density had the strongest
effect on spatial caribou-m oose relationships, we conducted
circular neighborhood analysis for clear-cut (proportion­
ate area) density by measuring density surrounding each
raster pixel at concentric radii from 75 to 10000 m. Then,
we fit univariate logistic regression models using density
estimates measured at the varying radii and identified the
m ost predictive radius o f caribou resource use relative to
moose using AIG (Burnham and A nderson 2002).
Results
Coarse-scale caribou-moose resource separation
We tested w hether caribou m ortalities occurred w ith
higher frequency in areas o f high overlap between moose
and caribou or in higher quality moose habitat as expected
under the spatial separation hypothesis (M cLoughlin
et al. 2005) using a one-way chi-squared test. In our spa­
tially predictive maps developed above, we classified areas as
caribou habitat where caribou had a high relative probabil­
ity o f use (categories 8 —1 0 ); moose habitat was represented
by categories 1—3 and habitat where bo th species were
predicted to have interm ediate relative probability o f use
by categories 4—7. Long-term m ortality data (1999—2009)
were com piled by A lberta Fish and W ildlife Division and
Parks G anada based on radio-collared (very high frequency
and GPS) caribou. M ortalities used in this analysis were
identified as w olf (n = 32), grizzly/black bear (n = 5),
unknow n confirm ed predator (n = 9) and unknow n
m ortality (n = 72), totaling 59 per season. G onfirm ed
non-predatory m ortalities (e.g. road kill or avalanches)
were excluded. The large num ber o f unknow n m ortality
causes likely resulted from delays in m ortality site investi­
gation due to the remoteness o f the study area. W hile we
can assume that the m ajority o f caribou m ortalities were
predator-caused (W ittm er et al. 2005), we also tested
m ortality using only confirm ed predator-killed cari­
bou (n = 44). We tested the null-hypothesis that caribou
m ortalities were n o t significantly different in moose
We chose a linear response m odel for RDA, because the
D GA first gradient lengths were all < 3 (ter Braak 1995),
and perform ed analyses for both seasons. The m atrix for
the occurrence o f the two ungulate species was related to
the set o f predictor variables (summer: F 2 2 5 1 7 5 = 76.55,
p = 0.005; winter: F 2 2 7 3 0 7 = 5 9 .8 2 , p = 0.005). For both
seasons, the biplots showed a separation o f the realized
niche positions o f caribou and moose and especially in
winter, the am plitude o f environm ental variables describing
caribou occurrence clearly indicated the niche specialisa­
tion by caribou (Fig. 2). D uring summer, the constrained
variance o f the species—environm ent relationship (approx.
2 1 %) was largely explained by the first axis, w hich showed
opposite relative species scores (caribou: 2.43; moose:
—2.45; Table 1) and thus reflected resource separation
between the two species. Environm ental variables that
were negatively correlated w ith axis 1 were associated
w ith moose and vice versa; environm ental variables posi­
tively correlated w ith axis 1 were associated w ith caribou.
Elevation (GG = 0.78), snow (GG = 0.77) and alpine shrub
(GG = 0.54) were strongly positively correlated w ith ordina­
tion axis 1, explaining caribou presence. In contrast, N D V I
( G G = —0.49), closed conifer ( G G = —0.33) and mixed
forests (GG = —0.32), b u t also hum an landscape alteration
(clear-cuts, clear-cut density, and density o f roads and linear
5-EV
1.0
(a)
■
Sum m er
W in te r
SI
Mi
r
DEM
...........................
<
o
ce
M o ose
1.0
C aribo u
C Q ta r '/
N D V r^ /
Lin
-0 .5 -
-
Mu
0.5 -
0 .5 -
Q9
CN
<
o
ce
M o ose
- 0 .5 -
-
-
-
1.0
-0 .5
0.0
C aribo u
0.5
1.0
-
1.0
-
1.0
-0 .5
0.0
1.0
0.5
RDA1
RDM
Figure 2. O rdination biplots for environm ental matrices and the occurrence o f moose and caribou (based on global positioning system
(GPS) collar locations) in 500 m? grid cells for sum m er (16 May—16 O ctober) and w inter (17 O ctober—15 May). Acronym s are explained
in Table 1. GPS collars were deployed on 17 moose and 17 caribou between w inters 2007/2008 and 2009/2010 in w est-central A lberta and
east-central British C olum bia, Canada.
Ta ble 1. S u m m a r y o f r e d u n d a n c y a n a l y s is (RDA) b e t w e e n th e p r e s e n c e o f m o o s e a n d c a r i b o u (i.e. g lo b a l p o s i t i o n i n g s y s te m (GPS) c o l l a r
lo c a ti o n s) a n d a s e t of e n v i r o n m e n t a l v a r i a b l e s in w e s t - c e n t r a l A lb e rta a n d e a s t- c e n tr a l British C o l u m b i a fo r s u m m e r (1 6 M a y - 1 6 O c t o b e r )
a n d w i n t e r (1 7 O c t o b e r - 1 5 M ay ). E ig en v a lu e s, p e r c e n t a g e a n d p r o p o r t i o n o f c o n s t r a i n e d v a r i a n c e a n d c a n o n i c a l c o e f f i c i e n t s o f c o n s t r a i n ­
ing v a r i a b l e s a re listed fo r t h e first t w o a x e s . G P S c o lla rs w e r e d e p l o y e d o n 1 7 m o o s e a n d 1 7 c a r i b o u b e t w e e n w i n t e r s 2 0 0 7 / 2 0 0 8 a n d
2 0 0 9 / 2 0 1 0 in w e s t - c e n t r a l A lb e rta a n d e a s t- c e n tr a l British C o l u m b i a , C a n a d a .
Sum m er
W inter
A xis 1
Axis 2
Axis 1
Axis 2
a s s o c ia t io n
separation
association
separation
E ig en v a lu e s
0.11
2.25E-04
0.08
7 .8 8 E -0 4
P e r c e n t a g e o f c o n s t r a i n e d v a r i a n c e o v e r to ta l v a r i a n c e
0.21
4.50E-04
0 .1 5
1 . 5 8E -0 3
P r o p o r tio n o f c o n s t r a i n e d v a r i a n c e e x p l a i n e d b y o r d i n a t i o n a x e s
0.98
0 .0 2
0.99
0.01
S p e c i e s s c o r e s (c a rib o u )
2 .4 3
-0 .1 1
2.21
0.21
-2 .4 5
-0 .1 1
-2 .0 6
0.23
S p e c i e s s c o r e s (m o o s e )
C o n s t r a i n i n g v a r ia b le s
C a n o n i c a l c o e f fic ie n ts
Elevation (m; DEM)
0.78
0.10
0.38
-0 .3 4
S l o p e (Si)
0.40
0.68
-0 .2 1
-0 .6 0
S n o w (Sn)
0.77
0.19
0 .7 5
-0 .1 0
Eine ar f e a t u r e d e n s i t y (km 1 0 k m - ^ ; Ein)
-0 .2 5
-0 .3 5
0.04
0 .7 5
R o a d d e n s i t y (km 1 0 k m - ^ ; Rd)
-0 .3 0
-0 .2 0
-0 .0 9
0.12
C le a r - c u ts (% are a/1 0 0 w i t h i n r = 3 km; CD)
-0 .2 3
0.1 7
-0 .2 2
0.12
B ar re n (B)
-0 .1 6
0.14
-0 .2 1
-0 .0 5
B ar re n a l p i n e (BA)
0.42
0.14
0.16
-0 .3 0
Burn (Bu)
-0 .2 9
-0 .0 9
-0 .2 7
-0 .0 3
C l o s e d c o n i f e r (CCon)
-0 .3 3
-0 .0 8
-0 .0 3
0.06
C u t (C)
-0 .1 5
0 .2 3
-0 .1 6
-0 .1 1
D e c i d u o u s (D)
-0 .1 9
0.16
-0 .2 7
0.06
0.16
0.07
0.08
-0 .0 8
-0 .1 1
0 .1 2
-0 .1 3
0.06
0. 33
0 .1 3
0.23
-0 .3 2
-0 .3 2
0.2 7
-0 .3 9
0.13
0.09
-0 .4 4
0.17
0.47
0.19
G l a c i e r (G)
H e r b a c e o u s (H)
H e r b a c e o u s a l p i n e (HA)
M i x e d (Mi)
M u s k e g (M)
O p e n c o n i f e r (OC )
S h r u b (S)
S h r u b a l p i n e (SA)
W a t e r (W)
NDVI
6-EV
0.11
-0 .4 1
0.27
-0 .2 3
0.04
-0 .1 6
0.02
0.54
0.20
0.30
-0 .3 4
0 .0 5
-0 .1 0
0.03
-0 .0 7
-0 .4 9
-0 .1 4
-0 .2 3
0.16
(a)
w
c
Sum m er
40 j
■ Caribou
35-
■ Caribou
□ Moose
_o
ro
o
o
W in te r
40
□ Moose
25 —
20
15 - -
0
Q_
1
2
3
4
5
6
7
8
9
1
10
2
Resource use category
3
4
5
6
7
8
9
10
Resource use category
Figure 3. Frequency o f occurrence o f w oodland caribou ( n = 17) and moose (n = 17) global positioning collar (GPS) location data in
10 equal sized bins predicted from the m ost parsim onious generalized linear mixed models w ith a random intercept during sum m er
(16 M ay -1 6 O ctober; (a)) and w inter (17 O c to b er-1 5 May; (b)) in west-central A lberta and east-central British C olum bia, Canada.
Values closer to 1 indicate high relative probability o f use by moose and conversely, values closer to 10 indicate high relative probability of
use by caribou. Areas w ith high shared use o f b oth species indicate low resource separation. GPS collars were deployed between winters
2007/2008 and 2009/2010.
features) were correlated w ith moose presence (Table 1 ,
Fig. 2 ). The second axis (and therefore resource over­
lap between the species) explained a very low proportion
o f variance ( < 0 . 1 %). Thus, indicating that our set o f
variables better explained the separation between species
than their overlap. Similarly, during winter, the con­
strained variance o f the species-environm ent relation­
ship (approx. 15%) was largely explained by the first axis
(> 9 9 % ), which also measured resource separation o f the
two species indicated by the opposite scores o f 2 . 2 1 for
caribou and —2.06 for moose (Table 1, Fig. 2). Variables
that were strongly related to the presence o f caribou
were snow (CC = 0.75), elevation (C C = 0.38) and alpine
shrubs (C C = 0.30), while moose were associated w ith
mixed forests (CC = -Q .3 9 ), N D V I (CC = - 0 .2 3 ) , burns
and deciduous forests (C C = —0.27 for both variables).
w inter from 17 individuals o f each species. The average
num ber o f locations per caribou and per moose in sum m er
was 848 (SE = 37.58) and 801 (S E = 14.09), and in w in­
ter 1165 (SE = 47.37) and 1 2 0 2 (SE = 38.44) respectively.
In general, resource use by caribou and moose and the
degree o f resource separation differed only slightly between
seasons (Table 2 , Fig. 3). Caribou used higher elevations than
moose (also, the highest standardized z-values) during both
season, although resource separation due to elevation was
weaker during winter. Moose also tended to use areas w ith
increased hum an landscape alteration and green, broad­
leaved forage (NDVI; z^^j
= —1.00 during sum m er
and Zjjj
= —1.26 during winter), whereas caribou
avoided these features relative to moose (Table 2 ). For
example, the relationship between clear-cut density and
caribou resource use was strongly negative during winter
(Z s td _ C le a rC u tD e n s = -l-2 6 )
Fine-scale caribou-moose resource and spatial
separation
At a finer scale, we evaluated resource separation w ith
14420 caribou and 13615 moose GPS locations in sum ­
mer and 19809 caribou and 2 0 4 3 7 moose locations in
Sum m er
(a)
and
SUmmer
(z,,d_ClearCutDens =
—0.90). In general, caribou and moose differed in responses
to hum an landscape alteration less in sum m er than winter.
D uring winter, caribou used areas w ith higher probabilities
o f being covered by snow (z^^j snow ~^-4l) and occurred
more often in open conifer (z^^j openCon ~ 1-48) and mixedforsts (Zjjj Mixed ~ 1-14) than moose. The m ost parsim oni­
ous generalized linear mixed models cross validated very
W in te r
(b)
1
0.8
0.6
« 0. 4
0. 4
0.2
w
r2
0 0.2
= 0.3521
0
0
0.02
0.04
0.06
0.08
Average CC density at caribou
GPS locations
0.1
0
0.02
0.04
0.06
0.08
0.1
0.12
Average CC density at caribou
GPS locations
Figure 4. Schoener’s overlap index values, where values closer to 0 indicate low overlap between caribou and moose resource use and values
closer to 1 indicate high overlap, versus proportionate area density o f clear-cuts (GG) measured w ithin 3 km radii (% area/100) for sum m er
(a) and w inter (b).
7-EV
well, confirming their predictive capacity w ith average
Spearmans rho o f 0.86 (p = 0.002) during sum m er and
0.98 ( p < 0.0001) during winter.
Schoeners overlap index was high in all 10 resource
categories in both seasons (Fig. 3; C’summer^
Q-inter ^ 0.672), indicating high resource overlap. The pro­
portion o f caribou locations falling into moose resource
categories (categories 1—3) was higher during sum m er
(25%) than during w inter (2 0 %), b u t also the propor­
tion o f caribou locations falling into caribou resource use
categories (categories 8—10) was higher (summer = 5 4 %
and w inter = 42% ; Fig. 3). Consequently, the proportion
o f caribou locations falling into interm ediate resource
categories (i.e. categories 4—7) was lower during sum m er
(21%) than during w inter (38%). As predicted, spatial
separation between moose and caribou was negatively cor­
related w ith clear-cut density (Fig. 4). This relationship
was significant during sum m er (R^ = 0.35, F^j J5 ^ = 8.15,
p = 0.012), b u t n o t during w inter (R^ = 0.17, F^j
= 3.12,
p = 0.097).
Mortality consequences of increased caribou-moose
overlap
We observed differences am ong the proportions o f caribou
mortalities falling into the caribou, moose and interm edi­
ate resource use categories during sum m er ( % ^ ( 2 5 9 ) ^ 23.70,
p < 0.0001). The greatest proportion o f caribou killed by
predators occurred in categories o f highest moose resource
use (53%; Fig. 5). In winter, the relationship between
resource use categories and the frequency o f caribou
mortalities was n o t statistically significant (%2^^ 5 9 ) = 9.36,
p = 0.406). The analysis exclusively w ith predator-caused
mortalities broadly confirmed this pattern w ith 65, 1 2
and 23% o f the mortalities occurring in moose, interm e­
diate and caribou resource categories, respectively in sum ­
m er and 11, 44 and 44% in winter. Again, the relationship
was only significant in the sum m er (summer:
=
22.69, p < 0.0001; winter: (2 . 18) ■ 0.42, p = 0.807).
<n
I
is 30
ra
I
I
O bserved
Expected
25
g 20
JD
■§ 15
o
•S 10
o
O
5
0
p
m
Cat. 1-3 I Cat. 4 -7 |cat. 8-10 Cat. 1-3 |I Cat. 4 -7 |cat.
|Cat. 8-10
Sum m er
W in te r
Figure 5. O bserved w oodland caribou m ortalities in categories
(cat.) o f the relative probability o f resource use by caribou relative
to moose, where cat. 1—3 have the lowest probability o f use by
caribou, b u t the highest by moose, cat. 4—7 have interm ediate
probabilities o f use by both species and cat. 8—10 have highest
probabilities o f use by caribou, while lowest for moose. For further
explanation, please see the m ethods section.
8-EV
D iscussion
Conservation biologists often need to understand com ­
plex ecological interactions to identify threats for declining
species and prescribe m anagem ent actions to reverse
declines. In this study, we present an example for compara­
tive analyses o f resource use and spatial overlap between a
threatened species (caribou) and a generalist species (moose),
thriving in hum an altered landscapes. Surprisingly, despite
the often suggested im portance o f understanding the
mechanisms o f apparent com petition for caribou declines,
direct comparisons o f moose and caribou resource use are
rare. In this example, we showed that hum an landscape
alteration was positively correlated w ith resource overlap
o f coexisting species, w hich may be especially im portant in
systems o f shared predation. W oodland caribou isolate
themselves from other more abundant prim ary prey species
and their shared predators to reduce the negative effect of
predation (Bergerud and Page 1987, Seip 1992). O u r results
confirmed that caribou and moose generally use different
resources as expected under the spatial separation hypoth­
esis, b u t we also showed that the strength o f resource
separation varies across different spatial scales and seasons,
and the geographical realization o f resource separation could
be influenced by the intensity o f hum an landscape change.
Resource selection varies w ith scale and consequently,
niche relationships between sym patric species may also
change at different scales (Ihl and Klein 2001). Scaledependent resource selection has been dem onstrated for
caribou in previous studies. For example, several research­
ers concluded th at caribou broadly select resources to
m inim ize predation risk at coarse spatial scales and maxi­
mize forage benefits at smaller spatial scales (Rettie and
Messier 2000, Apps et al. 2001). O u r RD A suggested that
caribou and moose separated through their asymmetric
occurrence in the m ultidim ensional space o f environm ental
variables at coarse spatial scales in b o th seasons, and thus,
occupied contrasting realized niches (Hirzel and Le Lay
2008). This niche separation was indicated by the oppo­
site species scores for the first axis and th at alm ost all o f
the total variance was explained by the Eigenvalue o f the
first axis ( > 9 9 % ; Table 1). A t this coarse scale, resource
overlap, indicated by the similar species scores for cari­
bou and moose along the second axis, was weak (variance
explained by the Eigenvalue < 0 .1 % ) during b oth seasons
(Table 1). Nonetheless, at a finer scale, our analysis o f
sympatric caribou and moose pairs indicated substantial
spatial overlap due to similar use o f environm ental resource
variables in geographical space in all 1 0 resource use
categories (Fig. 3). Further, we observed a positive relation­
ship between resource use overlap and intensity o f hum an
landscape alteration, especially during sum m er (Fig. 4).
Finally, our results suggested that caribou also experience
increased m ortality risk w hen their resource use over­
laps w ith moose at finer spatial scales in summer, coinci­
dent w ith similar forage requirem ents during this season.
Thus, caribou m ay avoid the m ost lim iting factor to fitness,
i.e. overlap w ith moose at coarser scales (i.e. num eric
response by predators), b u t maximize forage at finer scales
w hich could result in spatial overlap w ith moose especially
in sum m er (Rettie and Messier 2000, Apps et al. 2001).
Ta ble 2. M o d e l c o e f fic ie n ts (P), s t a n d a r d e rr ors (SE) a n d s t a n d a r d i z e d W a l d st at is ti cs (Zj,j) fr o m t h e m o s t p a r s i m o n i o u s g e n e r a l i z e d li n e a r
m i x e d m o d e l s w i t h a r a n d o m in t e r c e p t d e s c r i b i n g d if f e r e n c e s in h a b ita t u s e by w o o d l a n d c a r i b o u ( d e p e n d e n t v a r i a b l e = 1) a n d m o o s e
( d e p e n d e n t v a r i a b l e = 0) in w e s t - c e n t r a l A lb e rta a n d e a s t- c e n tr a l British C o l u m b i a , C a n a d a . H a b i t a t use w a s c o m p a r e d d u r i n g s u m m e r
(16 M a y - 1 6 O c t o b e r ) a n d w i n t e r (17 O c t o b e r - 1 5 May) a n d fr o m 2 0 0 7 to 2 0 0 9 . C l o s e d c o n i f e r w a s th e r e f e r e n c e c a t e g o r y for l a n d ­
c o v e r t y p e s.
Sum m er
Ele va tio n (1 GO m)
A spect w est-east
A s p e c t n o rt h - s o u t h
A spect north-south sq u ared
S n o w (winter)
W inter
P
SE
0.63
0.011
4.19
-0 .2 4
-
0.024
-
-0 .7 3
-
2std
P
SE
0.38
-
0.010
-
2.67
0.1 7
0.019
0.61
-0 .2 8
0.037
-0 .5 2
9.92
0.2 77
2.41
2std
-
Eine ar fe a tu r e d e n s i t y (km km^^)
R o a d d e n s i t y (km km^^j
-0 .2 2
0.021
-0 .7 8
-
-
-
-0 .6 1
0.069
-0 .6 5
-0 .1 0
0.048
-0 .1 4
C le a r - c u ts (% a r e a w i t h i n 3 km radius )
-2 .9 5
0.242
-0 .9 0
-3 .5 5
0.189
-1 .2 6
-
-
-
-1 .2 3
0.153
-0 .5 4
-1 .0 5
-1 .9 4
0.162
-0 .8 0
-1 .4 0
0.114
-0 .8 3
B arren
Burn
Cut
-
0.110
-
-0 .7 1
-
D eciduous
-
-
-
-1 .1 9
0.128
-0 .6 2
-0 .7 3
0.136
-0 .3 9
-1 .0 4
0.124
0.092
0.18
-
-0 .5 6
0.22
-1 .1 8
0.089
-0 .9 7
-1 .2 9
0.076
-1 .1 4
M uskeg
0.78
0.069
0.83
O p e n c o n if e r
0.43
0.59
H erbaceous
H e rb a c e o u s alpine
M ix e d
-
-
-
Shrub
-
0.054
NDVI
-2 .3 2
0.1 71
-1 .0 0
-3 .8 4
0.205
-1 .2 6
M odel intercept
-7 .9 2
0.543
-1 .0 8
-1 1 .6 3
0.469
-1 .6 6
-
In our study, caribou associated w ith variables repre­
senting alpine and coniferous habitats and moose w ith
variables representing deciduous foraging habitats at b oth
scales, b u t the m agnitude o f the predictor variables was
scale- and season dependent. These differences may arise
because o f differences in m igratory strategies between
sym patric ungulates. For example, as predicted by the spa­
tial separation hypothesis (Seip 1992), elevation was one
o f the m ain factors separating caribou and moose niches,
b u t its effect strongly varied w ith season. In our study area,
five out o f six caribou populations are partially m igratory
(i.e. only p art o f the population migrates; M cD evitt et al.
2009) and individuals may leave high elevation m ountain
sum m er ranges to exploit m ature and old conifer forests
in the foothills during winter, reflected by the weaker
coefficients for elevation during w inter com pared to sum ­
m er at both scales (Table 1, 2). As a result, m igration may
constrain caribou from spatially separating at coarser spa­
tial scales during winter, b u t interestingly, caribou seem to
separate from moose at finer spatial scales during this
season, possibly because o f large forage differences between
moose and caribou during this season (Boer 2007). Thus,
caribou may have adopted seasonal separation strate­
gies w hich could differ w ith spatial scale (Ihl and Klein
2001). O u r results indicate such seasonal scale-dependent
avoidance strategies by caribou by a m uch stronger avoid­
ance o f N D V I and clear-cut densities by caribou compared
to moose during w inter at the finer scale, b u t n o t the coarser
scale. Further, our results suggest that snow may be a strong
driver in the fine scale resource separation during w inter that
allows caribou to spatially separate from sympatric moose.
C aribou are well adapted to harsh w inter conditions and
their large, crescent-shaped hooves and long legs allow them
to dig through snow to access lichen, and ease locom otion
-
0.93
0.042
1.48
-0 .5 4
0.0 41
-0 .9 0
over snow (Klein et al. 1987). In contrast, moose movements
have been found to be im peded at snow depths exceeding
60 cm, because o f the associated high energy cost for this
larger bodied ungulate (Renecker and Schwartz 2007).
There is a general consensus th at caribou avoid
hum an landscape alteration in the literature and our selec­
tion coefficients from logistic regression and canonical
coefficients from RD A were consistent w ith previous stud­
ies. Nevertheless, studies assessing how landscape altera­
tion may affect resource use overlap between sympatric
moose and caribou are rare (Bowman et al. 2010). For
example, several studies have suggested th at caribou avoid
roads and seismic lines (Dyer et al. 2001, Vistnes and
N ellem ann 2001). Also, clear-cuts in O ntario have signifi­
cantly displaced caribou from harvested areas (Vors et al.
2007, Bowman et al. 2010). H abitat alteration in caribou
range leads to the direct loss and fragm entation o f cari­
bou habitat w ith th at preferred by other ungulate species
(W ittm er et al. 2005). In agreem ent w ith the hypothesized
effect o f hum an landscape change on resource separation,
the two ungulate species in our study had higher resource
use and spatial overlap (Schoeners C) in regions w ith
increased clear-cuts per u n it area (Fig. 4).
Overlap o f sym patric prey species can result in concur­
rent occurrence o f exploitative (shared resources consum p­
tion) and apparent com petition (shared predators; H olt
and Lawton 1994). However, the degree o f overlap between
species does n o t necessarily equal the am ount to w hich
they com pete w ith each other, b u t rather the degree co­
existing species are similar in their environm ental resource
use (Sale 1974). In general, direct habitat loss is unlikely
to lim it forage for w oodland caribou because m ost popu­
lations are hypothesized to be well below the forage car­
rying capacity (M cLoughlin et al. 2003, W ittm er et al.
9-EV
2005). A lthough we did n o t specifically assess diet com ­
position and foraging by moose and caribou, habitat use
o f bo th species seemed to indicate that exploitative com ­
petition between the two species is unlikely, especially
during winter. In a stable isotope diet study conducted by
Ben-David et al. (2001) in Alaska, moose and caribou
stable isotope ratios were significantly different from each
other in late sum m er—autum n and winter. However, in
early and m id sum m er caribou may also feed on simi­
lar forbs and deciduous vegetation (Apps et al. 2001)
as moose. Consequently, high fine-scale resource overlap
during sum m er we observed can be explained by potentially
overlapping forage and elevation preferences o f the two
species during sum m er (Boer 2007). In contrast, we m ust
assume that forage overlap did n o t result in similar resource
overlap in w inter at finer scales, b u t rather the lim ited
availability o f undisturbed caribou habitat as previously
suggested (W ittm er et al. 2005). For example, between
23 and 38% o f the w inter or perm anent ranges o f five cari­
b ou herds we studied were altered by forestry based on sat­
ellite imagery (Alberta Sustainable Resource D evelopm ent
and A lberta Conservation Association [ASRD and ACA]
2 0 1 0 ).
In summary, while interpretation o f resource
(niche) overlap indices as measures o f com petition between
species has been subject to debate (Abrams 1980), we feel
that our use o f overlap indices rather as a measure for the
distribution o f moose and caribou w ith respect to resources
is a good approach to measure spatial overlap.
C aribou survival and population growth may be signifi­
cantly reduced in regions w ith increased landscape altera­
tion and thus, increased spatial overlap between caribou,
prim ary prey and wolves (M cLoughlin et al. 2005, Vors and
Boyce 2009). For example, the abundance o f wolves is pre­
dom inantly determ ined by the biomass o f their ungulate
prey and therefore, should be more abundant in landscape
im pacted by hum an alteration (Hebblewhite et al. 2007).
O u r results suggested higher m ortality frequencies for
caribou in moose resource use categories during summer.
Similarly, overlapping resource use by caribou and moose
m ay also be the reason for increased predator-caused cari­
bou mortalities in caribou hom e ranges that had proportion­
ately less old forests and more mid-aged forest com pared to
surviving caribou, reported in a study by W ittm er et al.
(2007). Further, wolves have been shown to use roads and
other linear features as travel routes that can increase pre­
dation efficiency. In that way, wolves m ight dom inate the
scale at w hich moose and caribou partition resources due to
increased range and speed o f m ovements in hum an altered
landscapes, especially in the presence o f linear features
(Gurarie et al. 2011, W hittington et al. 2011). We would
expect this effect to be strongest during the snow free
period w hen m ovem ent rates o f wolves are highest, which
m ay be the reason why caribou mortalities were significantly
higher in moose resource use categories during sum m er
in com parison to w inter (functional response by wolves).
A lthough we feel confident that spatial separation
between moose and caribou is lower in landscapes w ith
h um an landscape alteration, some characteristics o f our
study design may affect our results. In general, resource
selection studies should be interpreted cautiously because of
the com m on assum ption that resource selection is directly
10-EV
linked to fitness, w hich cannot be generalized (van H orne
1983). Furtherm ore, while high levels o f overlap in resource
use are often used to infer com petition (Sale 1974), it is
essentially the ratio o f the density o f consum er individu­
als (i.e. moose and caribou) relative to the resource base
(i.e. habitat) that determines the strength o f com peti­
tive interactions (Abrams 1980). In our study area,
caribou populations experience negative growth rates in
landscapes altered by hum ans (Vors and Boyce 2009). In
contrast, we observed only one death o f a total sample o f
33 radio-collared m oose that were m onitored for at least one
year each. Further, moose are expected to persist in high densi­
ties in regions w ith im proved forage following clear-cutting
(Lavsund et al. 2003), suggesting high moose population
viability in stark contrast to caribou population declines.
Thus, despite the untested assum ption about moose density
relating to highly selected moose habitats, our results are
indicative for higher moose density in caribou ranges as a
result o f increased h um an landscape alteration (Peek
2007). Also, we did n o t take w ithin-population heterogene­
ity in resource use that may occur w ith partial m igration
into account. But, we can expect that n o t only resource
overlap may be elevated for non-m igratory versus m igratory
animals (Seip 1992), b u t also predation risk, if m igration
allows animals to escape from predation at broad spatial
scales (Hebblewhite and M errill 2009). Therefore, further
investigation o f exclusively sedentary caribou and moose
w ould be necessary to determ ine niche overlap during sum ­
m er in the foothills. Lastly, Latham et al. (201 lb ) described
the potential role o f distribution shifts o f w hite tailed deer
in response to hum an landscape alteration as prim ary prey
species for wolves in north-eastern AB and suggested that
also m anagem ent o f w hite tailed deer populations need to
be considered if the aim is to conserve caribou populations
in AB. In this study, we only present research on the relation­
ship between moose and caribou, b u t o f course com m unity
interactions are complex and other factors such as climate
(e.g. rain on snow events) or other prim ary prey and preda­
tors may play an im portant role in our and other study areas
and should be evaluated (Vors and Boyce 2009).
Especially in hum an altered landscapes, species may
be unable to adapt to novel m ortality risks that were n o t
present in their evolutionary history (Schlaepfer et al. 2002).
W ith the encroachm ent o f hum an landscape alteration,
caribou refugia from moose, and hence predators like
wolves, are com prom ised and their spatial separation
strategy m ay be less effective. This could potentially result
in destabilizing the relationship between predators and
prey as predicted by the spatial separation hypothesis.
However, predator control does n o t appear to be an effective
long-term conservation strategy, if increased predation pres­
sure is only the ‘sym ptom ’ o f hum an landscape alteration
(Vors and Boyce 2009) or possibly wolves are also pro­
tected (Gurarie et al. 2011). Failure to address the habitatbased root-causes o f caribou declines, will likely result in
continuous long-term caribou population decrease. Thus,
if the goal is long-term recovery o f caribou populations
(ASRD and AGA 2010), the integrity o f caribou refugia
(old coniferous forest in our study area and throughout the
boreal forest) and the connectivity between these refuges in
already compromised caribou ranges should be maximized
(W ittm er et al. 2005, Vors et al. 2007). We described spatio'tem poral changes in the distribution o f prey species,
b u t other factors, such as changes in plant phenology or
extreme weather events due to climate change can affect
persistence o f globally declining caribou and reindeer
(Vors and Boyce 2009). Integrating these global changes in
wild Rangifer conservation will be a challenging task
determ ining the fate o f this ecologically and economically
im portant species. As in other biomes, also the absence of
changes in policy and hum an behavior will continue to alter
biodiversity in the boreal forest biome.
Acknowledgements — Financial and logistical support was provided
by the Alberta Conservation Association, A lberta D ept o f
Sustainable Resource D evelopm ent, British C olum bia M inistry
o f the E nvironm ent, C anadian Association o f Petroleum
Producers, NASA grant no. N N X 1 1 A 0 4 7 G , Foothills Research
Inst., Petroleum Technology Alliance o f C anada, Parks Canada,
Royal D utch Shell Canada, Univ. o f Alberta, Univ. o f Calgary,
Univ. o f M ontana, W eyerhaeuser Com pany, W orld W ildlife Fund,
Fulbright Com m ission, G erm an Academ ic Exchange Service and
the Philanthropic E ducational O rganization. W e th an k J. Berger,
P. R. K rausm an, and H . R obinson for helpful discussion and
reviews o f previous versions o f this m anuscript. We also thank
S. Hazenberg, D . Hervieux, D . H obson, L. Neufeld, M . Russell,
D . Seip, S. Slater, K. Sm ith, D . Stepnisky, B. W eckw orth, and
J. W ilm shurst for their assistance w ith data collection and project
m anagem ent. W e th an k the subject editor, Eric Post, for helpful
com m ents for im proving our m anuscript.
R eferences
Abrams, P. 1980. Some com m ents o n m easuring niche overlap.
—Ecology 61: 44—49.
Apps, C. D . et al. 2001. Scale-dependent habitat selection by
m ountain caribou, C olum bia M ountains, British C olum bia.
—J. W ildl. M anage. 65: 65—77.
A SRD and ACA 2010. Status o f the w oodland caribou {Rangifer
tarandus caribou) in Alberta: update 2010. —Alberta Sustain­
able Resource D evelopm ent, W ildlife Status R eport no. 30,
E dm onton, AB.
Barrett, M . W. et al. 1982. Evaluation o f a hand-held net gun to
capture large m am m als. —W ildl. Soc. Bull. 10: 108—114.
Ben-David, M . et al. 2001. U tility o f stable isotope analysis in
studying foraging ecology o f herbivores: examples from moose
and caribou. —Alees 37: 421—434.
Bergerud, A. T. and Page, R. E. 1987. D eplacem ent and dispersion
o f parturient caribou at calving as antipredator tactics. —Can.
J. Zool. 65: 1597-1606.
Boer, A. H . 2007. Interspecific relationships. —In: Fianzm ann, A. W.
and Schwartz, C. C. (eds), Ecology and m anagem ent o f the
N orth Am erian moose. Univ. Press o f Colorado, pp. 337—349.
Bowm an, J. et al. 2010. Roads, logging, and large-mam mal
com m unity o f an eastern Canadian boreal forest. — Can. J.
Zool. 88: 4 5 4 -4 6 7 .
Boyce, M . S. et al. 2002. Evaluating resource selection functions.
—Ecol. M odel. 157: 2 8 1 -3 0 0 .
Brook, B. W. et al. 2008. Synergies am ong extinction drivers under
global change. —Trends Ecol. Evol. 23: 453—460.
Bryant, A. A. and Page, R. E. 2005. T im ing and causes o f m ortal­
ity in the endangered Vancouver Island m arm ot {Marmota
vancouverensis). — Can. J. Zool. 83: 674—682.
B urnham , K. P. and Anderson, D . R. 2002. M odel selection
and m ultim odel inference: a practical inform ation-theoretic
approach. —Springer.
Chapin, F. S. et al. 2011. Consequences o f changing biodiversity.
—N ature 405: 234—242.
Chesson, P. and Kuang, J. J. 2008. The interaction between preda­
tio n and com petition. —N ature 456: 235—238.
DeCesare, N . et al. 2010. Endangered, apparently: the role o f
apparent com petition in endangered species conservation.
—A nim . Conserv. 13: 353—362.
DeCesare, N . et al. 2012a. Estim ating ungulate recruitm ent and
growth rates using age ratios. —J. W ildl. Manage. 76: 144—153.
DeCesare, N . et al. 2012b. Transcending scale-dependence in
identifying habitat and threats for endangered species. —Ecol.
Appl. in press.
Dyer, S. J. et al. 2001. Avoidance o f industrial developm ent
by w oodland caribou. —J. W ildl. M anage. 65: 531—542.
Frair, J. L. et al. 2004. Removing GPS collar bias in habitat
selection studies. —J. Appl. Ecol. 41: 201—212.
Gillies, C. et al. 2006. A pplication o f random effects to the study
o f resource selection by animals. —J. Anim. Ecol. 75: 887—898.
Gurarie, E. et al. 2011. Sum m er m ovements, predation and
habitat use o f wolves in hu m an m odified boreal forests.
—Oecologia 165: 891—903.
Hebblew hite, M . and Merrill, E. H . 2009. Trade-offs between
predation risk and forage differ betw een m igrant strategies in
a m igratory ungulate. —Ecology 90: 3445—3454.
Hebblew hite, M . et al. 2005. Spatial decom position o f predation
risk using resource selection functions: an example in a wolf—
elk p redator-prey system. —O ikos 111: 101—111.
Hebblew hite, M . et al. 2007. C onditions for caribou persistence
in the wolf—elk—caribou systems o f the C anadian Rockies.
—Rangifer 17: 7 9 -9 1 .
Hirzel, A. H . and Le Lay, G. 2008. H abitat suitability m odelling
and niche theory. —J. Appl. Ecol. 45: 1372—1381.
Holt, R. D . and Lawton, J. H . 1994. The ecological consequences o f
shared natural enemies. —A nnu. Rev. Ecol. Syst. 25: 495—520.
Hosmer, D . W. and Lemeshow, S. (eds) 2000. A pplied logistic
regression. —Wiley.
H utchinson, G. E. 1957. C oncluding remarks. — C old Spring
H arb o u r Symp. Q u an t. Biol. 22: 4 1 5 M 2 7 .
Ihl, C. and Klein, D . R. 2001. H abitat and diet selection by
m uskoxen and reindeer in western Alaska. —J. W ildl. Manage.
65: 9 6 4 -9 7 2 .
James, A. R. C. et al. 2004. Spatial separation o f caribou from
moose and its relation to predation by wolves. — J. W ildl.
M anage. 68: 799—809.
Johnson, D . H . 1980. The com parison o f usage and availability
m easurem ents for evaluating resources preference. — Ecology
61: 6 5 -7 1 .
Johnson, H . E. et al. 2012. A pparent com petition lim its the
recovery o f an endangered ungulate: quantifying the direct and
indirect effects o f predation. —Oecologia in press.
Klein, D . R. et al. 1987. Factors determ ining leg length in Rangifer
tarandus. —J. M am m al. 68: 642—655.
Kojola, I. et al. 2004. Predation o n European w ild forest reindeer
{Rangifer tarandus) by wolves {Canis lupu^ in Finland. —J.
Zool. 263: 2 2 9 -2 3 5 .
Latham , A. D . M . et al. 2011a. H abitat selection and spatial
relationships o f black bears {Ursus americanus) w ith w oodland
caribou {Rangifer tarandus caribou) in northeastern Alberta.
—Can. J. Z ool. 89: 2 6 7 -2 7 7 .
Latham , A. D . M . et al. 201 lb . Invading w hite-tailed deer change
wolf—caribou dynam ics in northeastern Alberta. — J. W ildl.
M anage. 75: 2 0 4 -2 1 2 .
Lavsund, S. et al. 2003. Status o f moose populations and challenges
to moose m anagem ent in Scandinavia. —Alees 39: 109—130.
M cD evitt, A. D . et al. 2009. Survival in the Rockies o f an endan­
gered hybrid swarm from diverged caribou {Rangifer tarandus)
lineages. —M ol. Ecol. 18: 665—679.
M cLoughlin, P. D . et al. 2003. Declines in populations o f w ood­
land caribou. —J. W ildl. M anage. 67: 755—761.
11-EV
M cLoughlin, P. D . et al. 2005. Relating predation m ortality to
broad'Scale habitat selection. —J. A nim . Ecol. 74: 701—707.
Mosnier, A. et al. 2008. Black bear adaptation to low productivity
in the boreal forest. —Ecoscience 15: 4 8 5 ^ 9 7 .
N ellem ann, C. and Cam eron, R. D . 1998. C um ulative impacts o f
an evolving oil-field complex on the distribution o f calving
caribou. —Can. J. Zool. 76: 1425—1430.
O ksanen, J. et al. 2011. Vegan: com m unity ecology package.
—R package.
Peek, J. M . 2007. H abitat relationships. — In: Franzm ann, A. W.
and Schwartz, C. C. (eds). Ecology and m anagem ent o f the
N orth American moose. Univ. Press o f Colorado, pp. 351—375.
Polfus, J. L. et al. 2011. Identifying indirect habitat loss and
avoidance o f hum an infrastructure by northern m ountain
w oodland caribou. —Biol. Conserv. 144: 2637—2646.
Pulliam, H . R. 2000. O n the relationship betw een niche and
distribution. —Ecol. Lett. 3: 349—361.
Renecker, L. A. and Schwartz, C. C. 2007. Food habits and
feeding behavior. —In: Franzm ann, A. W. and Schwartz, C. C.
(eds). Ecology and m anagem ent o f the N o rth A m erican moose.
Univ. Press o f C olorado, pp. 403—439.
Rettie, W. J. and Messier, F. 2000. Hierarchical habitat selection
by w oodland caribou: its relationship to lim iting factors.
—Ecography 23: 466—478.
Rosenzweig, M . L. 1981. A theory o f habitat selection. —Ecology
62: 32 7 -3 3 5 .
Sale, P. F. 1974. Overlap in resource use, and interspecific com pe­
tition. —Oecologia 17: 245—256.
Schlaepfer, M . A. et al. 2002. Ecological and evolutionary traps.
—Trends Ecol. Evol. 17: 474—480.
Schoener, T. W. 1974. Resource partitioning in ecological com ­
m unities. —Science 185: 27—39.
Seip, D . R. 1992. Factors lim iting w oodland caribou populations
and their interrelationships w ith wolves and moose in south­
eastern B ritish-Colum bia. —Can. J. Zool. 70: 1494—1503.
Senft, R. L. et al. 1987. Large herbivore foraging and ecological
hierarchies. —Bioscience 37: 789—795, 798—799.
Supplem entary material (Appendix E7733
oikosoffice.lu.se/appendix>). A ppendix 1.
12-EV
at < w w w .
Serrouya, R. et al. 2011. Developing a population target for an
overabundant ungulate for ecosystem restoration. — J. Appl.
Ecol. 48: 93 5 -9 4 2 .
StataGorp 2007. Stata statistical software: release 10. —StataGorp LP.
Taylor, R. H . 1979. H ow the M acquarie Island parakeet became
extinct. —N . Z. J. Ecol. 2: 42—45.
ter Braak, G. (ed.) 1992. Perm utation versus bootstrap significance
tests in m ultiple regression and AN OVA. - Springer,
ter Braak, G. 1995. C hapter 5: ordination. —In: Jongm an, R. et al.
(eds). D ata analysis in com m unity and landscape ecology.
C am bridge Univ. Press, pp. 91—173.
Thom as, D . L. et al. 1996. The diet o f w oodland caribou
populations in west-central Alberta. — Rangifer Spec. Issue 9:
3 3 7 -3 4 2 .
van H orne, B. 1983. D ensity as a m isleading indicator o f habitat
quality. —J. W ildl. M anage. 47: 893—901.
Vistnes, 1. and N ellem ann, G. 2001. Avoidance o f cabins, roads,
and power lines by reindeer during calving. —J. W ildl. M anage.
65: 9 1 5 -9 2 5 .
Vistnes, 1. et al. 2001. W ild reindeer: impacts o f progressive
infrastructure developm ent o n distribution and range use.
- Polar Biol. 24: 5 3 1-537.
Vors, L. S. and Boyce, M . S. 2009. Global declines o f caribou
and reindeer. —Global C hange Biol. 15: 2626—2633.
Vors, L. S. et al. 2007. W oodland caribou extirpation and
anthropogenic landscape disturbance in O ntario. —J. W ildl.
M anage. 71: 1249-1256.
W h ittington, J. et al. 2011. C aribou encounters w ith wolves
increases near roads and trails: a tim e-to-event approach.
- J. Appl. Ecol. 48: 1535-1542.
W ittm er, H . U. et al. 2005. Population dynam ics o f the
endangered m ountain ecotype o f w oodland caribou (Rangifer
tarandus caribou) in British Colum bia, Canada. —Can. J. Zool.
83: 4 0 7 M 1 8 .
W ittm er, H . U. et al. 2007. Changes in landscape com position
influence the decline o f a threatened w oodland caribou
population. —J. A nim . Ecol. 76: 568—579.