The Journal of Wildlife Management; DOI: 10.1002/jwmg.

511

Research Article

Marten Space Use and Habitat Selection in

Managed Coniferous Boreal Forests of Eastern
Canada
MARIANNE CHEVEAU,1,2 Centre for Forest Research and NSERC-UQAT-UQAM Chair in Sustainable Forest Management, Universite´ du
Que´bec en Abitibi-Te´miscamingue, 445 Boulevard de l’Universite´, Rouyn-Noranda, QC, Canada J9X5E4
LOUIS IMBEAU, Centre for Forest Research and NSERC-UQAT-UQAM Chair in Sustainable Forest Management, Universite´ du Que´bec en
Abitibi-Te´miscamingue, 445 Boulevard de l’Universite´, Rouyn-Noranda, QC, Canada J9X5E4
PIERRE DRAPEAU, Centre for Forest Research and NSERC-UQAT-UQAM Chair in Sustainable Forest Management, Universite´ du Que´bec a`
Montre´al, C.P. 8888, Succ. Centre Ville, Montre´al, QC, Canada H3C 3P8
LOUIS BELANGER, Centre for Forest Research, Faculte´ de Foresterie, de Ge´ographie et de Ge´omatique, Universite´ Laval, Pavillon Abitibi-Price,
Que´bec, QC, Canada G1K 7P4

ABSTRACT Effects of habitat loss and fragmentation on the behavior of individual organisms may have
direct consequences on population viability in altered forest ecosystems. The American marten (Martes
americana) is a forest specialist considered as one of the most sensitive species to human-induced dis-
turbances. As some studies have shown that martens cannot tolerate >30–40% clear-cuts within their home
range, we investigated marten space use (home range size and overlap) and habitat selection in landscapes
fragmented by 2 different patterns of timber harvesting in the black spruce boreal forest: dispersed-cut
landscapes (10–80 ha cut-blocks) and clustered-cut landscapes (50–200 ha cut-blocks). We installed radio-
collars on female martens and determined 20 winter home ranges (100% minimum convex polygons and 60–
90% kernels) in dispersed-cut (n ¼ 8) and clustered-cut (n ¼ 12) landscapes. Home range size was not
related to the proportion of clear-cuts (i.e., habitat loss), but rather to the proportion of mixedwood stands
70–120 years old. However, female body condition was correlated to habitat condition inside their home
ranges (i.e., amount of residual forest and recent clear-cuts). At the home range scale, we determined that
mixedwood forests were also among the most used forest stands and the least used were recent clear-cuts and
forested bogs, using resource selection functions. At the landscape scale, home ranges included more
mixedwood forests than random polygons and marten high activity zones were composed of more residual
forest and less human-induced disturbances (clear-cuts, edges, and roads). These results suggest that
mixedwood forests, which occupy approximately 10% of the study area, play a critical role for martens
in this conifer-dominated boreal landscape. We recommend permanent retention or special management
considerations for these isolated stands, as harvesting mixedwood often leads to forest composition conver-
sion that would reduce the availability of this highly used habitat. ß 2013 The Wildlife Society.

KEY WORDS American marten, body condition, boreal forest, Canada, forest management, fragmentation, Martes
americana, mixedwood forests, resource selection, space use.

Habitat loss and fragmentation are widely seen as major space; Ims et al. 1993) may have consequences at broader
threats to wildlife populations (Wiens et al. 1993, Fahrig scales. Previous studies conducted on birds and mammals
and Merriam 1994). Nevertheless, to understand effects of have shown that habitat loss and fragmentation could in-
fragmentation at the population level (Wiens 1995, Fahrig crease individual movement costs (Bélisle et al. 2001, Gobeil
2003), the proximate mechanisms underlying population and Villard 2002) because of constrained movements (Bélisle
responses need to be understood first at the level of individ- and Desrochers 2002, Bowman and Fahrig 2002), increased
uals (Ims 1995, Andreassen et al. 1998). Indeed, fine-scale predation risk (Johnson et al. 2009), and decreased foraging
effects like changes in individual behavior (e.g., in use of efficiency (Turcotte and Desrochers 2003). Hence, these
impacts may reduce breeding success of individuals and, in
Received: 19 July 2011; Accepted: 29 October 2012 turn, influence population viability (Hinsley et al. 1999,
1
Hinsley 2000).
E-mail: [email protected] Ims et al. (1993) predicted 3 potential types of effects on
2
Present address: Direction de la Faune Terrestre et de l’Avifaune,
Direction Ge´ne´rale de l’Expertise sur la Faune et ses Habitats,
individual space use in response to habitat loss and fragmen-
Ministe`re des Ressources Naturelles et de la Faune, 880 Chemin Ste- tation based on modifications of home range size and overlap
Foy, 2e e´tage, Que´bec, QC, Canada G1S 4X4. between individuals. The first type of response may be

Cheveau et al.  Marten Space Use and Habitat Selection 1

expected only for social species, which can tolerate high mammals (Sherburne and Bissonette 1994). Large trees
home range overlap and high population densities. In this and snags also provide resting and denning sites (Wynne
case, a fusion response is possible and implies an increase in and Sherburne 1983, Chapin et al. 1997b) when trees reach
home range overlap, which could be associated with a sufficient size. Finally, martens could be opportunistic in diet
decrease in home range size as habitat becomes more frag- (Martin 1994). Consequently, availability and vulnerability
mented. This response was found, for example, in gray-tailed of prey influence its foraging success, and affect its habitat
vole (Microtus canicaudus) populations (Wolff et al. 1997). selection (Coffin et al. 1997, Andruskiw et al. 2008).
When habitat fragment size approaches the minimum indi- Although key habitat elements can be found in a diversity
vidual requirement, a second type of response known as a of forest stands (Chapin and Harrison 1996), martens select
fission response can be observed, which is characterized by stands with complex vertical and horizontal structure (Payer
less overlap between individual home ranges, and eventually and Harrison 2003). Although such structural complexity
a decrease in home range size. Indeed, individuals in this may be associated with forest age and tree species composi-
case increasingly defend smaller territories to fulfill their tion in some regions, these associations are often site-specific
needs. This type of response has been shown in some bird and cannot be generalized throughout the marten’s range
and mammal species, such as the northern saw-whet owl (Chapin and Harrison 1996). Different forest types fulfill
(Aegolius acadicus; Hinam and St Clair 2008) and European marten needs from the northern to southern limits of its
red squirrel (Sciurus vulgaris; Wauters et al. 1994). Such a geographic range in eastern North America. In the conifer-
response may also be observed if an edge effect creates better ous boreal forest, martens usually select old black-spruce
foraging opportunities in fragmented habitat, which seems to (Picea mariana) stands in Ontario (J. M. Fryxell,
be the case for the European pine marten (Martes martes; University of Guelph, unpublished report), and dense conifer
Mergey et al. 2011, Caryl et al. 2012). If fragment size and mixedwood stands in Newfoundland (Bateman 1986)
becomes less than individual minimum habitat requirements, but show no selection for any specific composition in
an expansion response could be induced. Individuals will Labrador (Smith and Schaefer 2002). In the mixedwood
expand their home range to include more suitable habitats boreal forest of Quebec, martens sometimes select conifer
and eventually several fragments. This scenario implies an forests (Godbout and Ouellet 2008), but they may also select
increase in movements between fragments, depending on mixedwood or deciduous stands (Potvin et al. 2000). In the
their connectivity (distance between patches, matrix hostili- mixedwood forests of Maine, martens select mature stands
ty, and presence of corridors). This response, which seems (>9 m high; Fuller and Harrison 2000), stands that have
the most common, has been found for different species of incurred substantial spruce budworm (Choristoneura fumifer-
birds (Gjerde and Wegge 1989, Carey et al. 1992, Redpath ana) mortality (Chapin et al. 1997a), or mixedwood stands
1995, Siffczyk et al. 2003) and mammals (Selonen et al. (Katnik 1992). However, martens also use partial cuts in
2001, Riley et al. 2003, Irwin 2008). Maine (Fuller and Harrison 2005), second-growth forests in
In some regions, the American marten (Martes americana) Ontario (35–50 yr, Thompson et al. 2008; 50–90 yr,
is considered to be one of the most sensitive species to Bowman and Robitaille 1997), and old plantations in
disturbances induced by timber harvesting (Thompson New Brunswick (20–40 yr, Pelletier 2005), at least during
1991). It could thus be a good candidate for examining the summer. Such patterns of habitat selection that vary
the effects of habitat loss and fragmentation on individual across regions clearly show that habitat loss caused by timber
space use. The marten is not only a predator but also a prey harvesting may lead to different space use strategies follow-
species, which must deal with predation risk from both ing forest management, perhaps depending on which stand
terrestrial and avian predators (Hodgman et al. 1997). types are harvested.
This species is clearly a forest specialist (Buskirk and Although selected habitats may differ among regions,
Powell 1994) that avoids open areas (Koehler and martens clearly avoid open areas with low canopy cover across
Hornocker 1977, Hargis and McCullough 1984, Drew its geographic range. These habitats include clear-cuts
1995). However, across its geographic range, the (Potvin et al. 2000), meadows and burns (Koehler and
American marten is now recognized to use a wide variety Hornocker 1977), as well as low productivity stands
of forest habitats, contrary to previous views that the marten (Smith and Schaefer 2002) or bogs (Bateman 1986). At
was a strict specialist of old coniferous forests (Strickland and the level of individuals, previous studies have concluded
Douglas 1987). Factors that guide marten habitat choice can that martens cannot tolerate >30–40% open areas inside
be summarized in 3 categories (Thompson and Harestad their home range (Chapin et al. 1998, Fuller and
1994): predator avoidance, special habitat features, and prey Harrison 2000, Payer and Harrison 2000, Potvin et al.
availability. Martens prefer to live under a dense canopy 2000). We evaluated marten response to habitat loss and
cover, which provides protection against avian predators fragmentation generated by extensive forest harvesting at the
(Hargis and McCullough 1984, Drew 1995), and avoid individual level (inside their home range) during winter. The
stands with less than 30–40% overhead tree cover forest was harvested with clear-cuts under a spatial clustered
(Koehler and Hornocker 1977, Spencer et al. 1983, Smith pattern and, following an agreement between the Quebec
and Schaefer 2002). Coarse woody debris provides access provincial government and the Cree First Nation govern-
under winter snow cover to resting sites and reproductive ment (Government of Quebec and Grand Council of the
dens (Chapin et al. 1997b), as well as for hunting small Crees [Eeyou Istchee] 2002), which aimed to mitigate

2 The Journal of Wildlife Management

impacts of forest harvesting on indigenous activities (includ- habitats. Hence, we hypothesized that late-seral and mixed-
ing wildlife harvesting), a dispersed strategy of clear-cutting wood forests, which often have complex structure, will rep-
was also used. We thus investigated the effect of these 2 resent suitable habitat for martens, which will reduce home
clear-cut dispersion patterns (i.e., habitat loss configuration) range size proportionally to the increase of these forest types.
on marten space use (i.e., home range size and overlap).
Because martens do not tolerate intrasexual overlap in their STUDY AREA
territories (Katnik et al. 1994), we predicted that martens Our study area was in northwestern Quebec, eastern Canada
should not respond to fragmentation induced by forest har- (498450 N, 768000 W), within the Waswanipi Cree Model
vesting with a fusion response, as suggested by the results of Forest (approx. 35,000 km2; Fig. 1). The forest was com-
Payer et al. (2004). Similarly, because martens have large posed mainly of black spruce stands with an understory of
habitat requirements, presenting a home range size larger feather moss, sphagnum, and ericaceous shrubs. The
than predicted for its body size (Lindstedt et al. 1986, Waswanipi Cree Model Forest is part of the Canadian
Buskirk and McDonald 1989), we predicted that martens Model Forest Network, which has aimed to develop, test,
would not adopt a fission response. Consequently, we pre- and share solutions in sustainable forest management. It is
dicted that martens would respond to habitat loss and frag- the only Aboriginal-led Model Forest, and was specifically
mentation by expanding their home ranges. In other words, established to minimize conflicts between Cree land users
we predicted that martens will have larger home ranges in the and forest managers through community involvement, par-
clustered-cut landscape than in the dispersed-cut landscape. ticipation, and knowledge sharing. The area was located in
Finally, as responses to forest harvesting may be mediated by the James Bay lowlands, which are characterized by mostly
the habitat quality of the remaining forest cover, we also flat topography (with abundant poorly drained soils), and
investigated marten habitat selection at the landscape and the sparse hills. Despite the fact that forest fires were the major
home range scales. We hypothesized that late-seral and natural regime in this landscape (Bergeron et al. 2001, Le
mixedwood forests would be preferred, whereas human dis- Goff et al. 2007), forest harvesting had been the main source
turbances (clear-cuts, edges and roads) would be avoided, at of disturbances for the past decade. Clear-cuts were mainly
the 2 scales. Because home range size is mainly driven by conducted since 1988 to harvest spruce stands in clustered
habitat productivity at population and individual scales or dispersed patterns. In the clustered-cut landscapes
(McLoughlin and Ferguson 2000), we used the relationship (50–200 ha cut-blocks), the residual forest was composed
between home range size and the proportion of different of narrow corridors 20–100 m wide along streams or be-
habitat types as a predictor to detect suitable and unsuitable tween cut-blocks (Fig. 2). In the dispersed-cut landscapes

Figure 1. Location of the 2 study areas within the Waswanipi Cree Model Forest (WCMF, 498450 N, 768000 W), Quebec, Canada. Study areas were defined as
the minimum concave polygon around capture stations and marten telemetry locations (dots) in (a) the clustered-cut and (b) dispersed-cut landscapes. Marten
home ranges (MCP 100% (dotted), kernels 90% (black), and 60% (gray)) were shown for early winter 2006.

Cheveau et al.  Marten Space Use and Habitat Selection 3

Figure 2. Close up view of the configuration pattern of clear-cuts in (a) the clustered-cut and b) dispersed-cut landscapes within the Waswanipi Cree Model
Forest, Quebec, Canada. We present a subset of marten home ranges as 100% minimum convex polygons (dotted), 90% kernels (black), and 60% kernels (gray)
for early winter 2006. Gray shading indicates lakes and rivers and white areas are clear-cuts.

(10–80 ha cut-blocks), the residual forest was composed of age (cementum analyses; Matson’s Laboratory LLC,
uncut forest blocks equivalent in size to adjacent cut-blocks Milltown, MT, USA; adult 1 yr old, juvenile <1 yr old)
(Fig. 2). Except for regenerating stands from a forest fire and body index (mass/body length, as an index of general
(1986) and timber harvesting (<20 yr old), the landscape was health) for each individual. We installed radio-collars
mainly composed of stands >70 years old. Indeed, the region (Holohil MI-2, Carp, Ontario, Canada; 24 g with mortality
was characterized by a gap in successional stages, as 30- and signal) on female martens (>500 g) during this capture
50-year-old age classes were nearly absent. session. Although we attached 27 radio-collars, we were
For the present study, we focused on 2 landscapes (delin- only able to follow 22 martens for more than 10 locations.
eation defined below): 1 with clustered clear-cuts (518 km2, We performed the telemetry survey using a Cessna-337
498220 N, 76806W) and 1 with dispersed clear-cuts (Cessna, Wichita, KS, USA) airplane and concentrated
(341 km2, 498540 N, 75855W; see Fig. 1). Forest composition our survey during the early winter (20 Nov–19 Dec 2006;
varied slightly between the 2 studied landscapes (Table 1). only 1 location per day). Telemetry surveys were uniformly
The major difference between the 2 was related to the distributed between the morning (0930–1130), midday
proportion of mixedwood and coniferous stands. The (1130–1330), and the afternoon (1330–1600), as martens
clustered-cut landscape included more mixedwood stands are mostly diurnal during winter (Thompson and Colgan
(i.e., 25% of all residual mature forests compared to 8% in 1994). Moreover, we alternately started by each study area.
the dispersed-cut landscape). In contrast, the dispersed-cut We plotted locations on 1:20,000 maps, and then transferred
landscape included more coniferous stands, 92% of all rem- coordinates to ArcGIS 9.3 (Environmental Systems
nant mature forests compared to 75% in the clustered-cut Research Institute, Inc., Redlands, CA, USA). We found
landscape. The overall proportion of clear-cuts was 29% and this method (hand plotting) more precise than using a Global
20% in the clustered- and the dispersed-cut landscapes, Positioning System (GPS), because our landscapes were
respectively. composed of an abundance of easily distinguishable elements
(roads, lakes and rivers, clear-cuts, bogs) making the locali-
METHODS zation in a map reliable. Moreover, we estimated our preci-
sion in locating each marten from the airplane to be <50 m
Marten Telemetry Data (mean ¼ 33 m, median ¼ 12 m), using known locations
We captured martens during the fall season (26 Sep–6 Nov (n ¼ 14), which is much better than previous reported pre-
2006; see Cheveau 2010 for methodology). We estimated cision from aircraft (100 m; Potvin 1998, Payer 1999). One

Table 1. Composition of the 2 studied landscapes (% of the total area of each landscape) in the Waswanipi Cree Model Forest, Quebec, Canada in 2006; C is the
clustered-cut landscape and D is the dispersed-cut landscape.
Residual mature forest Regenerating stands from
Overall Coniferous Coniferous Mixedwood Forested bogs Clear-cuts Clear-cuts Water Non-forested
(%) >120 yr (%) 70–90 yr (%) 70–120 yr (%) (%) <11 yr (%) 11 yr (%) Fire (%) (%) (%)
C 47 9 26 12 10 23 6 10 5
D 52 23 25 4 15 12 8 4 7 3

4 The Journal of Wildlife Management

marten traveled without establishing a territory, and a second landscape (third- and second-order selection, respectively,
died after being tracked to only 11 locations before we could Johnson 1980; design III and II, respectively, Thomas and
estimate its home range size. We excluded these 2 individuals Taylor 1990). Using digital forest cover maps, known to
from home range analysis. We finally obtained 20 home accurately represent dominant tree species as well as age
ranges (more than 18 locations/marten), including 8 in classes in boreal regions (Potvin et al. 1999), we defined 9
the dispersed-cut landscape and 12 in the clustered-cut land cover types (fine-grained habitats) that were known to
landscape. The Institutional Committee of the Canadian influence marten habitat use at this scale on the basis of
Council for Animal Care of Université du Québec en available literature: 1) recent clear-cuts (<11 yr old), 2) old
Abitibi-Témiscamingue approved the trapping protocol clear-cuts (11 yr old), 3) regenerating stands after fire, 4)
and all manipulations of live martens. lakes and large rivers, 5) forested bogs, 6) mature and late-
We used kernels as home ranges, which we computed with seral mixedwood forests >70 years old (either coniferous or
RANGES 6 software (Kenward et al. 2003), for habitat deciduous dominated, mainly 70–90 and 120-yr age classes),
selection analyses (Fig. 1). We defined 90% kernels as rep- 7) late-seral conifer forests (>120-yr-old), 8) conifer forests
resenting winter home ranges (hereafter home ranges) and 70–90 years old (mainly black spruce forests), and 9) non-
60% kernels as delineating high activity zones (hereafter core forested areas (see Table 1). Clear-cuts represent net habitat
areas). We used fixed kernels with a least-squares cross- loss for martens, given that they avoid open and low regen-
validation (LSCV) estimator for the smoothing parameter eration areas (Buskirk and Powell 1994). We decided to split
(Seaman and Powell 1996), which seems to produce the least recent (<10 yr, which roughly correspond to <1-m high
bias and the smallest surface fit error (Seaman et al. 1999). regeneration in the region) and old clear-cuts (>10 yr),
Because the reference smoothing parameter tends to overes- considering the snow accumulation of 1 m in the study
timate range areas (Worton 1989), we multiplied it by a area. An old fire (1986) occurred in the northern part of
fractional value. We first estimated this fractional multiplier 1 study area; therefore, we considered that 20-year-old post-
by LSCV, which reaches an inflection point for each home fire stands (approx. 3 m high) represented habitat more
range, and then used the median of these multipliers (0.69) as permeable to marten movements than clear-cut regeneration
a fixed multiplier for all the home ranges, as suggested by blocks <1 m in height. Poorly drained forested bogs repre-
Kenward et al. (2003). Despite the fact that LSCV tends not sented natural open areas that could be avoided by martens in
to perform well with fewer than 30 locations (Seaman et al. the same way as human-induced open areas. We retained
1999), all our home ranges, except 1, reached an inflection, late-seral coniferous forests in this analysis as martens prefer
indicating that a local minimum had been found. We forests with complex vertical and horizontal structure
also delineated 100% minimum convex polygons (MCP) (Chapin et al. 1997a), attributes that are frequently found
to compare home range sizes with previous published in such habitat type, and to a lesser extent, in mature (70- to
studies. 90-yr-old) coniferous forests. We classified late-seral mixed-
We analyzed the relationship between marten home range wood forests as such, because we assumed that the mixed-
and core area size and landscape type (dispersed vs. clustered wood component could be selected by martens since it was
clear-cuts), age, body mass, and body index of martens using preferred in the southern portion of the boreal forest of
generalized linear models (GLM, 10 models, see Appendix). eastern Canada (Potvin et al. 2000). Such stands potentially
We determined model selection using Akaike’s Information present greater prey richness, as shown for other vertebrates
Criterion (AIC) and multimodel inference (Burnham and (Hobson and Bayne 2000). Because martens are opportunis-
Anderson 2002). We considered models within 2 AIC of tic in their food habits, we expected that a locally more
the top model to be supported then we assessed the magni- diverse prey community could influence marten habitat
tude of the effect using 95% unconditional confidence use. We derived all age classes from forest cover maps
intervals around model-averaged estimates (Burnham and (MRNF 2009). We also grouped land cover types in
Anderson 2002). We verified goodness-of-fit for the global coarse-grained habitats. We merged recent and old clear-
models (models including all non-correlated variables) using cuts (clear-cuts); late-seral conifer, 70- to 90-year-old coni-
residuals, as suggested by Burnham and Anderson (2002), fer, and mixedwood forests (remnant forests); and recent
and respected all assumptions for GLM. We also tested clear-cuts, old clear-cuts, fire regenerated stands, and bogs
the relationship between home range size and its composi- (open areas). We extracted habitat maps from digitized forest
tion (% of each land cover type, described later, 14 models, maps (1:20,000), rasterized and imported in ASCII format
see Appendix). We conducted the same analysis for marten to R 2.9.2 (R Development Core Team), where we per-
body index and related this metric to home range composi- formed habitat selection analyses. We used 10-m  10-m
tion (same as for the previous analysis), landscape type, and pixels for the home range scale analysis and 100-m  100-m
age (16 models). pixels for the landscape scale analysis.

Scales and Habitat Types Habitat Selection

We analyzed habitat selection at 2 scales: 1) at the home We analyzed habitat selection at the home range and core
range scale, we compared habitat use and availability within area scale using resource selection functions (RSFs) for a
each home range; and 2) at the landscape scale, we compared separate sampling of available and used units (SP-A protocol;
home range composition and habitat availability within each Manly et al. 2002, Zuur et al. 2009). We based RSFs on the

Cheveau et al.  Marten Space Use and Habitat Selection 5

comparison between habitat composition (fine-grained hab- landscape using R. We calculated edge and road densities
itats only) at marten telemetry locations (habitat used) and at in ArcGIS 9.3.
1,000 random points (habitat available) that we identified Finally, we compared landscape connectivity between the 2
within each individual home range. We realized all analyses landscape types (clustered and dispersed-cut) using the prob-
in R 2.9.2 (packages adehabitat, maptools, and sp). We ability of connectivity index (PC) developed by Saura and
defined available units as all pixels (10 m  10 m) inside Pascual-Hortal (2007), using Conefor Sensinode (CS) 2.2
a 50-m buffer around the random points, and used units as all (Pascual-Hortal and Saura 2006, Saura and Pascual-Hortal
pixels inside a 50-m buffer around the marten locations. We 2007). This index considers connectivity from a functional
tolerated no overlap between units that were used and those perspective (With et al. 1997) by accounting for the focal
that were available, as suggested by Thomas and Taylor species’ movement abilities. It combines habitat loss and
(2006). We discarded habitat types that represented spatial arrangement of the residual habitat, within a graph
<1.5% of the home range, as suggested by Erickson et al. theory framework (Saura and Pascual-Hortal 2007). We
(1998). For each landscape type, we performed a logistic delineated the area of each landscape by the minimum
regression using a generalized linear mixed model (GLMM; concave polygons described previously. Here, we defined
Gillies et al. 2006; package lme4) with a random intercept for potential habitat as the intact residual forest (7 m). To
each individual. We predicted RSF values using the partic- compute PC of a landscape from a marten’s perspective, CS
ular exponential form, which was suggested by Manly et al. 2.2 requires 3 input values. First, we provided species dis-
(2002). Then, we estimated relative probabilities of use for persal ability, which is the maximum distance traveled during
each habitat type derived from RSF values and scaled these to dispersion. In our study area, we found that some unsettled
between 0 and 1. martens dispersed over a distance of >100 km. Dispersal
At the landscape scale, we assessed habitat selection by abilities were thus not a restricting factor at the scale of our
comparing the composition of marten home ranges (habitat analysis. Second, the probability of connectivity requires the
used) from random polygons (habitat available) within each maximum distance that the species can cross in non-habitat
landscape. The delineation of the landscape is critical at this conditions, which was set at 600 m for marten, as deter-
level (Johnson 1980). We defined each of the 2 landscapes as mined from the available literature (see Snyder and
the minimum concave polygon around all capture points and Bissonette 1987). Third, we provided an associated proba-
marten telemetry locations to which we added a 1-km influ- bility of crossing non-habitat with respect to distance be-
ence zone (Fig. 1), which represented the sampled study area. tween suitable patches. We extrapolated these values from
We generated random polygons with the same size distri- Snyder and Bissonette (1987), who showed that 87% of
bution as marten home ranges (Gaussian distribution defined crossings were over distances <250 m. As an example based
by mean and standard deviation, bounded by minimum and on these reference values, the probability of crossing a 150-m
maximum) within each landscape, using SELES (Spatially gap is estimated to be only 30%, using a negative exponential
Explicit Landscape Event Simulator; Fall and Fall 1996). distribution. Last, Chapin et al. (1998) found that marten
We performed 1,000 iterations of 8 (for the dispersed-cut did not use habitat patches less than 2.7 ha in area.
landscape) or 12 (for the clustered-cut landscape) random Consequently, we did not consider these small patch sizes
polygons for the size distributions of home ranges and core in this analysis.
areas in each landscape. Available units corresponded to all
pixels (100 m  100 m) within random polygons, whereas RESULTS
used units were all pixels within marten home ranges or core
areas. Home Range Size
We characterized available habitat using both composition Mean (SE) marten home range size (MCP) was
(fine- and coarse-grained habitats) and configuration varia- 5.69  1.17 km2 (2.12–11.39 km2) in the dispersed-cut
bles: edge density (m/ha), road density (m/ha), and the landscape and 4.23  0.89 km2 (0.85–9.82 km2) in the clus-
proportion of residual forest core area (%), which excluded tered-cut landscape. Mean (SE) 90% kernel size was
a 50-m interior buffer. Edge density (edges between clear- 3.50  0.97 km2 (0.70–7.78 km2) in the dispersed-cut
cut and forest) is an index of the shape and size of clear-cuts. landscape and 2.49  0.57 km2 (0.36–6.66 km2) in the clus-
Martens tend to minimize edges inside their home range in tered-cut landscape. Mean (SE) 60% kernel size (core area)
the boreal forest (Potvin et al. 2000). Road density was was 1.72  0.48 km2 (0.30–3.91 km2) in the dispersed-cut
included as such open areas could represent barriers to mar- landscape and 1.15  0.26 km2 (0.17–3.00 km2) in the clus-
ten movements. We retained residual forest core area fol- tered-cut landscape.
lowing Potvin et al. (2000), who found that marten home Home range and core area size were not related to land-
ranges had larger proportions of core area of residual mature scape type (clustered- vs. dispersed-cut), female age, body
forest than that which was typically available in the land- mass, or body index (see Table 2). Home range and core area
scape. We calculated the proportion of each habitat type size also were not related to the proportion of different clear-
(mean percentage for 8 or 12 polygons) within the random cuts (recent, old, and combined clear-cuts) within the home
polygons. We then compared the mean percentage of each range or core area, nor were they related to the proportion of
habitat type within marten home ranges with the distribution late-seral forests or 70- to 90-year-old conifer forests
of mean percentage from the 1,000 iterations, for each (Table 2). Home range and core area size were best predicted

6 The Journal of Wildlife Management

Table 2. Model-averaged estimates (beta), unconditional standard errors (SE), and 95% unconditional confidence intervals (CI) associated with variables that
explained female marten home range (60% and 90% kernels) size, in the Waswanipi Cree Model Forest, Quebec, Canada in 2006.
60% kernels 90% kernels
Beta SE 95% CI Beta SE 95% CI
Landscape 0.59 0.53 (0.45, 1.63) 1.13 1.12 (1.08, 3.33)
Body index 0.04 0.19 (0.42, 0.33) 0.14 0.39 (0.91, 0.63)
Body mass 0.001 0.01 (0.01, 0.01) 0.002 0.01 (0.02, 0.02)
Age 0.06 0.64 (1.19, 1.32) 0.53 1.34 (3.16, 2.09)
MIX70120a 0.03 0.01 (0.06, 0.01)b 0.08 0.03 (0.13, 0.02)b
CONIF120a 0.01 0.02 (0.03, 0.06) 0.07 0.05 (0.03, 0.17)
CONIF7090a 0.02 0.02 (0.01, 0.05) 0.03 0.04 (0.04, 0.10)
Recent clear-cuts 0.01 0.02 (0.05, 0.03) 0.001 0.04 (0.08, 0.08)
Old clear-cuts 0.003 0.02 (0.03, 0.04) 0.02 0.04 (0.06, 0.10)
Forested bogs 0.05 0.03 (0.001, 0.10) 0.13 0.07 (0.004, 0.26)
Regenerating stands from fire 0.02 0.04 (0.10, 0.06) 0.07 0.08 (0.23, 0.10)
Non-forested 0.003 0.05 (0.09, 0.0–9) 0.02 0.13 (0.23, 0.27)
Water 0.01 0.03 (0.06, 0.07) 0.004 0.06 (0.13, 0.12)
Residual forest 0.01 0.02 (0.05, 0.02) 0.02 0.04 (0.09, 0.05)
Clear-cuts 0.01 0.02 (0.05, 0.03) 0.01 0.04 (0.07, 0.09)
Open areas 0.01 0.02 (0.03, 0.05) 0.04 0.04 (0.04, 0.12)

a
MIX70120: mixedwood stands 70–120 yr, CONIF120: coniferous stands >120 yr, CONIF7090: coniferous stands 70–90 yr.
b
Confidence interval excludes 0, meaning that the variable had a substantial effect in explaining home range size.

by the proportion of mixedwood forests within the home increase body index from 14.4 to 17. On the contrary, an
range or core area (Appendix). Indeed, this variable had a increase from 0% to 50% of recent cuts in the home range is
substantial negative effect in explaining home range and core predicted to reduce body index from 16.5 to 13.7.
area size (see Table 2). An increase from 0% to 35% of
mixedwood in the home range is predicted to reduce
home range size from 5 km2 to 1 km2 in clustered-cut Habitat Selection
landscape and from nearly 4 km2 to 2 km2 in dispersed- Within home ranges, all habitat types (except forested bogs
cut landscape (Fig. 3). in the dispersed-cut landscape) were more frequently selected
We found no overlap in female home ranges (see Fig. 1). Of by marten than were recent clear-cuts (Table 3). The most
the 20 home ranges, only 3 were composed of >40% clear- utilized habitat in the dispersed-cut landscape was late-seral
cuts (2–53%), including 1 with 40% or more recent clear-cuts conifer forests followed by mixedwood forest and conifer
(0–53%). Female body index was positively related to the forests 70–90 years old (Table 4). In the clustered-cut land-
overall proportion of remnant forest within the home range scape, old clear-cuts were highly used (Table 4), but this
(best model, Akaike weight [wi] ¼ 0.33; b ¼ 0.05  0.02, effect was almost entirely driven by 1 individual (#462),
95% CI ¼ 0.01, 0.09). Moreover, female body index was which had established its home range in a 30-year-old
negatively related to the proportion of recent cuts within the area regenerating after clear-cutting. When we excluded
home range (second best model, wi ¼ 0.23; b ¼ 0.06  this individual, old clear-cuts were clearly less frequently
0.02, 95% CI ¼ 0.10, 0.02). An increase from 30% to used by other martens (relative probability of use
90% of remnant forest in the home range is predicted to [RPU] ¼ 0.284) and the most utilized habitat was mixed-
wood forest (RPU ¼ 1), closely followed by late-seral coni-
fer forests (RPU ¼ 0.825) and conifer forests 70–90 years
8 clustered clearcuts
dispersed clearcuts
Table 3. Mixed-effects logistic regression model of habitat selection by
Home range size (km²)

6 martens with their selection coefficients and standard errors, in the 2 land-
scapes of the Waswanipi Cree Model Forest, Quebec, Canada in 2006.

4 Dispersed-cut Clustered-cut
Habitat typea landscape landscape
Old clear-cuts 0.473  0.084 1.956  0.053b
2 Forested bogs 0.698  0.090 0.464  0.060
MIX70120 1.413  0.069 1.937  0.037
CONIF120 1.541  0.064 1.880  0.045
0 CONIF7090 1.336  0.063 1.823  0.035
0 10 20 30 40 50 60 Regenerating stands from fire 1.103  0.080
% Mixedwood Water 0.661  0.074

a
Figure 3. Relationship between home range size of female martens (90% Reference category was recent clear-cuts. MIX70120: mixedwood stands
kernels) and percentage of mixedwood forests in the clustered-cut (solid line) 70–120 yr, CONIF120: coniferous stands >120 yr, CONIF7090: co-
and dispersed-cut (dashed line) landscapes, in the Waswanipi Cree Model niferous stands 70–90 yr.
b
Forest, Quebec, Canada in 2006. 1.048  0.086 when marten #462 is ignored.

Cheveau et al.  Marten Space Use and Habitat Selection 7

Table 4. Predicted resource selection functions (RSFs) and relative prob- random polygons (Table 5). Finally, we found that marten
abilities of use (RPU) of the different habitat types, for martens in the 2
landscapes of the Waswanipi Cree Model Forest, Quebec, Canada in 2006.
home ranges were always more connected than the surround-
ing landscape in the dispersed-cut landscape (n ¼ 8/8, mean
Dispersed-cut Clustered-cut home range PC ¼ 0.31 vs. landscape PC ¼ 0.10) and
landscape landscape
almost always in the clustered-cut landscape (n ¼ 8/10,
Habitat typea RSF RPU RSF RPU
mean home range PC ¼ 0.16 vs. landscape PC ¼ 0.03).
Recent clear-cuts 0.021 0.120 0.022 0.000
Old clear-cuts 0.034 0.265 0.153 1.000 DISCUSSION
Forested bogs 0.010 0.000 0.034 0.097
MIX70120 0.087 0.865 0.150 0.979 Our first prediction concerning home range expansion as a
CONIF120 0.098 1.000 0.142 0.915 response to habitat loss and fragmentation (i.e., increasing
CONIF7090 0.080 0.793 0.134 0.855 proportion of clear-cuts) was not supported. Despite con-
Regenerating stands from fire 0.064 0.603
trasting proportions of clear-cuts within marten home ranges
a
MIX70120: mixedwood stands 70–120 yr, CONIF120: coniferous in the 2 landscape types (dispersed: mean ¼ 12%, range: 2–
stands >120 yr, CONIF7090: coniferous stands 70–90 yr. 44%; clustered: mean ¼ 28%, range: 3–53%), we found no
relationship between home range size and the proportion of
clear-cuts, nor did we find a relationship between home
old (RPU ¼ 0.780). In contrast, martens avoided forested range size and landscape type. Although marten home ranges
bogs and recent clear-cuts in the 2 landscape types (Table 4). were located in areas showing better connectivity and greater
In the dispersed-cut landscape, marten winter home ranges amounts of habitat core areas in both landscape types, we
and core areas included significantly more mixedwood for- suggest that, at this level of remnant forest habitat (dispersed:
ests, 70–90-year-old conifer forests, overall remnant forests, mean ¼ 67%, range: 40–91%; clustered: mean ¼ 53%,
and core areas of remnant forests than did the random range: 32–81%), configuration effects were possibly still
polygons. Likewise, winter home ranges included marginally unpronounced. Considering that we found no overlap be-
less overall clear-cuts and lesser edge and road densities tween female home ranges, our results also show limited
(Table 5). Core areas also had significantly less recent support for a fusion response following habitat loss.
clear-cuts, overall clear-cuts, and lesser edge and road densi- Similarly, the marginally lower edge density within home
ties, as well as marginally less open areas (Table 5). In the ranges does not support a fission response that would be
clustered-cut landscape, winter home ranges and core areas consistent with greater foraging opportunities along edges, as
were composed of more mixedwood forests than were the observed in the case of the European pine marten (Mergey

Table 5. Mean  standard error of the proportion of the different habitat variables for random polygons (n ¼ 1,000 simulations of 8 or 12 polygons) compared
to the mean proportion of habitat variables within marten home ranges (90% and 60% kernels) in the 2 landscapes of the Waswanipi Cree Model Forest, Quebec,
Canada in 2006. Values with an asterisk showed significant (or nearly significant) differences between random polygon and home range composition (P ¼ 0.05).
Dispersed-cut landscape Clustered-cut landscape

90% kernels 60% kernels 90% kernels 60% kernels

Home Home Home Home
Random range Random range Random range Random range
mean  SEa mean Pb mean  SEa mean Pb mean  SEa mean Pb mean  SEa mean Pb
Fine-grained habitats
Recent clear-cuts 12.3  3.8 7.1 0.081 12.4  4.7 4.5 0.032 22.8  8.3 21.3 0.444 23.5  9.3 19.1 0.324
Old clear-cuts 8.2  4.9 5.2 0.284 8.3  5.7 6.6 0.445 6.0  5.0 6.1 0.574 6.5  5.8 4.5 0.438
Forested bogs 14.8  4.3 11.1 0.203 14.5  5.2 10.9 0.257 9.9  3.9 5.4 0.118 10.2  4.9 4.9 0.129
MIX70120c 4.4  2.6 13.5 0.005 4.3  2.9 13.2 0.004 11.6  4.4 21.5 0.023 11.4  5.5 23.6 0.020
CONIF120c 22.4  5.1 17.1 0.848 22.5  6.0 17.4 0.790 8.8  3.4 4.8 0.902 8.4  4.0 5.0 0.794
CONIF7090c 25.1  6.4 36.3 0.049 25.1  7.1 38.1 0.040 26.1  5.7 27.3 0.427 25.9  7.0 27.8 0.366
Regenerating stands from fire 3.4  4.4 5.4 0.716 3.7  4.9 5.0 0.682
Non-forested 2.7  1.4 3.0 0.665 2.5  1.7 3.5 0.774 5.0  2.2 3.9 0.352 4.9  2.6 5.1 0.585
Water 6.7  5.4 1.3 0.104 6.8  6.3 0.8 0.139 9.8  6.7 9.7 0.559 9.4  7.3 10.1 0.579
Coarse-grained habitats
Clear-cuts 20.5  5.2 12.3 0.052 20.7  6.0 11.1 0.044 28.6  6.9 27.3 0.437 29.9  9.7 23.6 0.284
Residual forests 52.0  6.2 67.0 0.002 52.3  6.6 68.6 0.003 46.6  6.1 53.7 0.118 45.5  9.0 56.4 0.117
Open areas 38.6  5.7 28.8 0.063 38.8  7.3 27.1 0.051 38.5  7.0 32.7 0.198 39.9  10.0 28.5 0.139
Habitat configuration
Edge density 23.2  4.3 16.2 0.053 25.1  5.4 13.5 0.013 30.0  6.6 27.4 0.351 33.3  8.3 26.2 0.195
Road density 10.2  2.0 7.4 0.074 11.2  2.6 4.8 0.005 12.0  3.4 12.0 0.559 14.0  4.2 12.3 0.392
Core area of residual forests 30.7  5.3 44.2 0.006 29.4  6.5 44.1 0.014 28.3  5.9 31.7 0.273 27.6  8.9 31.8 0.292

a
Mean % for all habitat variables except edge and road densities (m/ha).
b
Probabilities associated with rejection of the hypothesis. We hypothesized that forested areas (mixedwood stands 70–120 yr, coniferous stands 70–90 yr and
>120 yr, remnant forests and residual forest core area) were more abundant in marten home ranges than in random polygons and the opposite for all other
variables.
c
MIX70120: mixedwood stands 70–120 yr, CONIF120: coniferous stands >120 yr, CONIF7090: coniferous stands 70–90 yr.

8 The Journal of Wildlife Management

et al. 2011, Caryl et al. 2012). However, the use of mixed- Previous studies have shown that American martens cannot
wood stands in the study area was consistent in both land- tolerate more than 30–40% of their home range in open and
scape types. Although such stands are a rare habitat type regenerating areas (Chapin et al. 1998, Payer and Harrison
(only 4% and 12%) in our dominant coniferous landscapes, 1999, Fuller and Harrison 2000, Potvin et al. 2000,
marten selected mixedwood stands both at the home range Dumyahn et al. 2007). Marten populations in our part of
and the landscape scales of investigation. With regards to the boreal forest evolved in the context where unburned
habitat factors influencing marten habitat use (Thompson forests in post-fire landscapes generally reach low propor-
and Harestad 1994), we suggest that mixedwood stands are tions (7–37% of the area; Perron et al. 2009), although they
key habitats, especially in our boreal landscapes, presenting a are surrounded by an unburned forest matrix, which is
complex forest structure that reduces the risk of predation dominated by late-seral cover (Cyr et al. 2010). Even though
and provides increased prey availability and denning sites. our 2 studied landscapes did not cross such habitat loss
According to Hély et al. (2000), coarse woody debris accu- thresholds, our measured female body index was positively
mulation in boreal mixedwood stands is influenced by time related with the amount of residual forests, and negatively
since fire and canopy composition. Mixedwood stands dom- related to the amount of recent clear-cuts within individual
inated by trembling aspen (Populus tremuloides) produce a home ranges. A similar relationship between habitat quality
large amount of coarse woody debris, mainly when the stands and marten body condition was found for dispersing juve-
are between 80 and 130 years old (Hély et al. 2000), as niles, who experienced poorer body condition in regenerating
dominance of trembling aspen in the canopy declines around than in uncut forest landscapes (Johnson et al. 2009). Such
80 years after fire in our study region (Bergeron et al. 1998). results suggest that in landscapes where the availability of
Consequently, our 70–120-year-old mixedwood forests are mixedwood forests is the same, dispersed clear-cut patterns
at their peak of structural complexity. In parallel, snowshoe would provide a better configuration of residual habitats than
hares (Lepus americanus), which browse almost exclusively on in aggregated clear-cut patterns.
deciduous species in our study region (Jacqmain 2003), are
known to be at greater density in mixedwood stands in the MANAGEMENT IMPLICATIONS
same study area (Jacqmain et al. 2007). In winter, snowshoe Mixedwood forests that are embedded within coniferous
hares contribute significantly to marten diet in eastern North landscapes are selected by martens, even if they are rare at
America (Thompson and Colgan 1990, Cumberland et al. the landscape scale. Cree land users have already recognized
2001). We suggest that the apparent absence of a range the importance of mixedwood forests for wildlife and have
expansion response to habitat loss and fragmentation found asked for specific management guidelines for these stands
in this study can thus be explained by the greater proportion (Government of Quebec and Grand Council of the Crees
of residual mixedwood stands in clustered-cut landscape [Eeyou Istchee] 2002). Furthermore, harvesting mixedwood
(3-fold greater in clustered-cut than in dispersed-cut stands often leads to forest composition conversion (mixed-
landscape). The overwhelming selection of this habitat by wood to conifer or deciduous stand) and results in permanent
martens is likely to dampen other effects that are related to loss of this specific habitat, which was recognized as one of
habitat loss. the major ecological issue related to forest management in
As expected, martens preferred forest types older than Canada (Jetté et al. 2008). Our results therefore provide
70 years and avoided open areas like recent clear-cuts and compelling evidence that mature and late-seral mixedwood
forested bogs in our study area, corroborating patterns that stands require special management considerations, including
have already been shown in other regions (open bogs, Raine the conservation of permanent retention blocks as they rep-
1983, Gosse et al. 2005; recent clear-cuts, Buskirk and resent critical habitat for many wildlife species in the boreal
Powell 1994, Thompson and Harestad 1994, Poole et al. coniferous forest of eastern Canada.
2004). Thompson and Colgan (1994) found that martens
avoid clear-cuts for at least 40 years to limit predation risk ACKNOWLEDGMENTS
(Thompson 1994, Drew 1995) and because their preferred This study could not have been conducted without the
prey were absent or at low density in these habitats collaboration of the Waswanipi Cree Model Forest and
(Thompson and Colgan 1994). At the landscape scale, we Cree trappers from Waswanipi, who allowed us access to
also found that martens avoided human disturbances such as their traditional hunting territories. We especially thank the
clear-cuts, especially recent ones, as well as edges and roads, Waswanipi Cree Model Forest staff and the 2 Cree co-
within their high activity zones (core areas) in the dispersed- researchers, N. Ottereyes and J. Awashish, for translations
cut landscape. Such avoidance was not detected by our and their help throughout the project. This project was
analyses in the clustered-cut landscape likely because delin- funded by Quebec Ministry of Natural Resources and
eated kernels included clear-cuts around remnant forest Wildlife—Volet 1, by FQRNT Fonds forestier grants to
strips, even though martens did not use clear-cuts. In addi- P.D. (and colleagues) and L.I. (and colleagues), the
tion, we used coarse-scale maps (100-m  100-m pixels) for Fondation de la Faune du Québec, a NSERC Discovery
habitat selection analysis at the landscape scale, mainly grant to P.D., and the NSERC-UQAT-UQAM Industrial
because of computational limitations. The resolution of Chair in Sustainable Forest Management. M.C. benefited
pixels could locally omit narrow linear structures like forest from a NSERC graduate scholarship. We thank D.
strips, which were usually 20–100 m wide. Kneeshaw, J.-F. Robitaille, and I. Thompson for insightful

Cheveau et al.  Marten Space Use and Habitat Selection 9

comments made on a previous version of this manuscript. Coffin, K. W., Q. J. Kujala, R. J. Douglass, and L. R. Irby. 1997. Interactions
among marten prey availability, vulnerability, and habitat structure.
The authors also thank W. F. J. Parsons for revision of the Pages 199–210 in G. Proulx, H. N. Bryant, and P. M. Woodard, editors.
English. Martes: taxonomy, ecology, techniques, and management. The Provincial
Museum of Alberta, Edmonton, Alberta, Canada.
Cumberland, R. E., J. A. Dempsey, and G. J. Forbes. 2001. Should diet be
LITERATURE CITED based on biomass? Importance of larger prey to the American marten.
Andreassen, H. P., K. Hertzberg, and R. A. Ims. 1998. Space-use responses Wildlife Society Bulletin 29:1125–1130.
to habitat fragmentation and connectivity in the root vole Microtus oeco- Cyr, D., S. Gauthier, D. A. Etheridge, G. J. Kayahara, and Y.
nomus. Ecology 79:1223–1235. Bergeron. 2010. A simple Bayesian Belief Network for estimating the
Andruskiw, M., J. M. Fryxell, I. D. Thompson, and J. A. Baker. 2008. proportion of old-forest stands in the Clay Belt of Ontario using the
Habitat-mediated variation in predation risk by the American marten. provincial forest inventory. Canadian Journal of Forest Research 40:573–
Ecology 89:2273–2280. 584.
Bateman, M. C. 1986. Winter habitat use, food habits and home range size Drew, G. S. 1995. Winter habitat selection by American marten (Martes
of the marten, Martes americana, in western Newfoundland. Canadian americana) in Newfoundland: why old growth? Dissertation, Utah State
Field Naturalist 100:58–62. University, Logan, USA.
Bélisle, M., and A. Desrochers. 2002. Gap-crossing decisions by forest birds: Dumyahn, J. B., P. A. Zollner, and J. H. Gilbert. 2007. Winter home-range
an empirical basis for parameterizing spatially-explicit, individual-based characteristics of American marten (Martes americana) in Northern
models. Landscape Ecology 17:219–231. Wisconsin. American Midland Naturalist 158:382–394.
Bélisle, M., A. Desrochers, and M. J. Fortin. 2001. Influence of forest cover Erickson, W. P., T. L. McDonald, and R. Skinner. 1998. Habitat selection
on the movements of forest birds: a homing experiment. Ecology 82:1893– using GIS data: a case study. Journal of Agricultural, Biological and
1904. Environmental Statistics 3:296–310.
Bergeron, Y., S. Gauthier, V. Kafka, P. Lefort, and D. Lesieur. 2001. Fahrig, L. 2003. Effects of habitat fragmentation on biodiversity. Annual
Natural fire frequency for the eastern Canadian boreal forest: consequences Review of Ecology and Systematics 34:487–515.
for sustainable forestry. Canadian Journal of Forest Research 31:384–391. Fahrig, L., and G. Merriam. 1994. Conservation of fragmented populations.
Bergeron, Y., P. J. H. Richard, C. Carcaillet, S. Gauthier, M. Flannigan, and Conservation Biology 8:50–59.
Y. T. Prairie. 1998. Variability in fire frequency and forest composition in Fall, J., and A. Fall. 1996. SELES: a spatially explicit landscape event
Canada’s southeastern boreal forest: a challenge for sustainable forest simulator. Pages 104–112 in Proceedings of the third international con-
management. Conservation Ecology 2:6. ference on integrating geographic information systems and environmental
Bowman, J., and L. Fahrig. 2002. Gap crossing by chipmunks: an experi- modeling, Santa Fe, New Mexico, USA.
mental test of landscape connectivity. Canadian Journal of Zoology Fuller, A. K., and D. J. Harrison. 2000. Influence of partial timber harvesting
80:1556–1561. on American marten and their primary prey in northcentral Maine.
Bowman, J. C., and J. F. Robitaille. 1997. Winter habitat use of American University of Maine, Orono, USA.
martens Martes americana within second-growth forest in Ontario, Fuller, A. K., and D. J. Harrison. 2005. Influence of partial timber harvesting
Canada. Wildlife Biology 3:17–105. on American martens in north-central Maine. Journal of Wildlife
Burnham, K. P., and D. R. Anderson. 2002. Model selection and multi- Management 69:710–722.
model inference: a practical information-theoretic approach. Springer, Gillies, C. S., M. Hebblewhite, S. E. Nielsen, M. A. Krawchuck, C. L.
New York, New York, USA. Aldridge, J. L. Frair, D. J. Saher, C. E. Stevens, and C. L. Jerde. 2006.
Buskirk, S. W., and L. L. McDonald. 1989. Analysis of variability in home- Application of random effects to the study of resource selection by animals.
range size of the American marten. Journal of Wildlife Management Journal of Animal Ecology 75:887–898.
53:997–1004. Gjerde, I., and P. Wegge. 1989. Spacing pattern, habitat use and survival
Buskirk, S. W., and R. A. Powell. 1994. Habitat ecology of fishers and of capercaillie in a fragmented winter habitat. Ornis Scandinavica 20:
American martens. Pages 283–296 in S. W. Buskirk, A. S. Harestad, M. 219–225.
G. Raphael, and R. A. Powell, editors. Martens, sables, and fishers— Gobeil, J. F., and M. A. Villard. 2002. Permeability of three boreal forest
biology and conservation. Cornell University Press, Ithaca, New York, landscape types to bird movements as determined from experimental
USA. translocations. Oikos 98:447–458.
Carey, A. B., S. P. Horton, and B. L. Biswell. 1992. Northern spotted owls: Godbout, G., and J. P. Ouellet. 2008. Habitat selection of American marten
influence of prey base and landscape character. Ecological Monographs in the logged landscape of the southern fringe of the boreal forest.
62:223–250. Écoscience 15:332–342.
Caryl, F. M., C. P. Quine, and K. J. Park. 2012. Martens in the matrix: the Gosse, J. W., R. Cox, and S. W. Avery. 2005. Home-range characteristics
importance of nonforested habitats for forest carnivores in fragmented and habitat use by American martens in eastern Newfoundland. Journal of
landscapes. Journal of Mammalogy 93:464–474. Mammalogy 86:1156–1163.
Chapin, T. G., and D. J. Harrison. 1996. Influence of landscape pattern, Government of Quebec and Grand Council of the Crees [Eeyou Istchee].
forest type, and forest structure on use of habitat by marten in Maine. 2002. Agreement concerning a new relationship between le
National Council of the Paper Industry for Air and Stream Improvement Gouvernement du Québec and the Crees of Québec. Gouvernement
(NCASI), Technical Bulletin No. 728, Research Triangle Park, North du Québec, Québec, Canada.
Carolina, USA. Hargis, C. D., and D. R. McCullough. 1984. Winter diet and habitat
Chapin, T. G., D. J. Harrison, and D. D. Katnik. 1998. Influence of selection of marten in Yosemite National Park. Journal of Wildlife
landscape pattern on habitat use by American marten in an industrial Management 48:140–146.
forest. Conservation Biology 12:1327–1337. Hély, C., Y. Bergeron, and M. D. Flannigan. 2000. Coarse woody debris in
Chapin, T. G., D. J. Harrison, and D. M. Phillips. 1997a. Seasonal habitat the southeastern Canadian boreal forest: composition and load variations
selection by marten in an untrapped forest preserve. Journal of Wildlife in relation to stand replacement. Canadian Journal of Forest Research
Management 61:707–717. 30:674–687.
Chapin, T. G., D. M. Phillips, D. J. Harrison, and E. C. York. 1997b. Hinam, H. L., and C. C. St Clair. 2008. High levels of habitat loss and
Seasonal selection of habitats by resting martens in Maine. Pages 166–181 fragmentation limit reproductive success by reducing home range size and
in G. Proulx, H. N. Bryant, and P. M. Woodard, editors. Martes: provisioning rates of northern saw-whet owls. Biological Conservation
taxonomy, ecology, techniques, and management. The Provincial 141:524–535.
Museum of Alberta, Edmonton, Alberta, Canada. Hinsley, S. A. 2000. The costs of multiple patch use by birds. Landscape
Cheveau, M. 2010. Multiscalar effects of forest fragmentation by timber Ecology 15:765–775.
harvesting on American marten in the boreal forest of eastern Canada. Hinsley, S. A., P. Rothery, and P. E. Bellamy. 1999. Influence of woodland
Dissertation, Université du Québec en Abitibi-Témiscamingue, Rouyn- area on breeding success in great tits Parus major and blue tits Parus
Noranda, Québec, Canada. caeruleus. Journal of Avian Biology 30:271–281.

10 The Journal of Wildlife Management

Hobson, K. A., and E. Bayne. 2000. Breeding bird communities in boreal Pascual-Hortal, L., and S. Saura. 2006. Comparison and development of
forest of western Canada: consequences of ‘‘unmixing’’ the mixedwoods. new graph-based landscape connectivity indices: towards the prioritization
Condor 102:759–769. of habitat patches and corridors for conservation. Landscape Ecology
Hodgman, T. P., D. J. Harrison, D. M. Phillips, and K. D. Elowe. 1997. 21:959–967.
Survival of American marten in an untrapped forest preserve in Maine. Payer, D. C. 1999. Influences of timber harvesting and trapping on habitat
Pages 86–99 in G. Proulx, H. N. Bryant, and P. M. Woodard, editors. selection and demographic characteristics of American marten.
Martes: taxonomy, ecology, techniques, and management. The Provincial Dissertation, University of Maine, Orono, USA.
Museum of Alberta, Edmonton, Alberta, Canada. Payer, D. C., and D. J. Harrison. 1999. Influence of timber harvesting and
Ims, R. A. 1995. Movement patterns related to spatial structures. Pages 85– trapping on habitat selection and demographic characteristics of marten.
109 in L. Hansson, L. Fahrig, and G. Merriam, editors. Mosaic landscapes University of Maine, Orono, USA.
and ecological processes. Chapman and Hall, London, United Kingdom. Payer, D. C., and D. J. Harrison. 2000. Structural differences between
Ims, R. A., J. Rolstad, and P. Wegge. 1993. Predicting space use responses to forests regenerating following spruce budworm defoliation and clear-
habitat fragmentation: can voles Microtus oeconomus serve as an experi- cut harvesting: implications for marten. Canadian Journal of Zoology
mental model system (EMS) for capercaillie grouse Tetrao urogallus in 30:1965–1972.
boreal forest? Biological Conservation 63:261–268. Payer, D. C., and D. J. Harrison. 2003. Influence of forest structure on
Irwin, M. T. 2008. Diademed sifaka (Propithecus diadema) ranging and habitat use by American marten in an industrial forest. Forest Ecology and
habitat use in continuous and fragmented forest: higher density but lower Management 179:145–156.
viability in fragmented? Biotropica 40:231–240. Payer, D., D. Harrison, and D. Phillips. 2004. Territoriality and home-range
Jacqmain, H. 2003. Rabbit Habitat Project: analyse biologique et autochtone fidelity of American martens in relation to timber harvesting and trapping.
de la restauration de l’habitat du lièvre d’Amérique après coupe sur la Pages 99–114 in D. J. Harrison, A. K. Fuller, and G. Proulx, editors.
Terre des Cris de Waswanipi. Thesis, Université Laval, Québec, Québec, Martens and fishers (Martes) in human-altered environments: an interna-
Canada. [In French.] tional perspective. Springer, New York, New York, USA.
Jacqmain, H., L. Bélanger, S. Hilton, and L. Bouthillier. 2007. Bridging Pelletier, A. M. 2005. Préférences d’habitat estival, structure d’âge et repro-
native and scientific observations of snowshoe hare habitat restoration after duction d’une population de martres d’Amérique (Martes americana) dans
clearcutting to set wildlife habitat management guidelines on Waswanipi un secteur forestier aménagé de manière intensive au Nord-Ouest du
Cree land. Canadian Journal of Forest Research 37:530–539. Nouveau-Brunswick. Thesis, Université de Moncton, Edmunston, New
Jetté, J. P., M. A. Vaillancourt, A. Leduc, and S. Gauthier. 2008. Ecological Brunswick, Canada. [In French.]
issues related to forest management. Pages 1–10 in S. Gauthier, M. A. Perron, N., L. Bélanger, and M. A. Vaillancourt. 2009. Spatial structure of
Vaillancourt, A. Leduc, L. De Grandpré, D. Kneeshaw, H. Morin, P. forest stands and remnants under fire and timber harvesting regimes. Pages
Drapeau, and Y. Bergeron, editors. Ecosystem management in the boreal 129–154 in S. Gauthier, M. A. Vaillancourt, A. Leduc, L. De Grandpré,
forest. Presses de l’Université du Québec, Québec, Canada. D. Kneeshaw, H. Morin, P. Drapeau, and Y. Bergeron, editors. Ecosystem
Johnson, C. A., J. M. Fryxell, I. D. Thompson, and J. A. Baker. 2009. management in the boreal forest. Presses de l’Université du Québec,
Mortality risk increases with natal dispersal distance in American martens. Québec, Canada.
Proceedings of the Royal Society of London, Series B 276:3361–3367. Poole, K. G., A. D. Porter, A. De Vries, C. Maundrell, S. D. Grindal, and C.
Johnson, D. H. 1980. The comparison of usage and availability measure- C. St Clair. 2004. Suitability of a young deciduous-dominated forest for
ments for evaluating resource preference. Ecology 61:65–71. American marten and the effects of forest removal. Canadian Journal of
Katnik, D. D. 1992. Spatial use, territoriality, and summer-autumn selection Zoology 82:423–435.
of habitat in an intensively harvested population of martens on commercial Potvin, F. 1998. La martre d’Amérique (Martes americana) et la coupe à
forestland in Maine. Thesis, University of Maine, Orono, USA. blanc en forêt boréale: une approche télémétrique et géomatique.
Katnik, D. D., D. J. Harrison, and T. P. Hodgman. 1994. Spatial relations in Ministère de l’Environnement et de la Faune, Direction de la faune
a harvested population of marten in Maine. Journal of Wildlife et des habitats, Service de la faune terrestre, Québec, Québec, Canada.
Management 58:600–607. [In French.]
Kenward, R. E., A. B. South, and S. S. Walls. 2003. Ranges6 v1.2: for the Potvin, F., L. Bélanger, and K. Lowell. 1999. Validité de la carte écofor-
analysis of tracking and location data. Anatrack Ltd, Wareham, United estière pour décrire les habitats fauniques à l’échelle locale: une étude de cas
Kingdom. en Abitibi-Témiscamingue. Forestry Chronicle 75:851–859. [In French.]
Koehler, G. M., and M. G. Hornocker. 1977. Fire effects on marten habitat Potvin, F., L. Bélanger, and K. Lowell. 2000. Marten habitat selection in a
in the Selway-Bitterroot Wilderness. Journal of Wildlife Management clearcut boreal landscape. Conservation Biology 14:844–857.
41:500–505. Raine, R. M. 1983. Winter habitat use and responses to snow cover of fisher
Le Goff, H., M. D. Flannigan, Y. Bergeron, and M. P. Girardin. 2007. (Martes pennanti) and marten (Martes americana) in southeastern
Historical fire regimes shift related to climate teleconnections in the Manitoba. Canadian Journal of Zoology 61:25–34.
Waswanipi area, Central Quebec, Canada. Journal of Wildland Fire Redpath, S. M. 1995. Fragmentation and the individual: tawny owls Strix
16:607–618. aluco in woodland patches. Journal of Animal Ecology 64:652–661.
Lindstedt, S. L., B. J. Miller, and S. W. Buskirk. 1986. Home range, time, Riley, S. P., R. M. Sauvajot, T. K. Fuller, E. C. York, D. A. Kamradt, C.
and body size in mammals. Ecology 67:413–418. Bromley, and R. K. Wayne. 2003. Effects of urbanization and habitat
Manly, B. F. J., L. L. McDonald, D. L. Thomas, T. L. McDonald, and W. fragmentation on bobcats and coyotes in southern California.
P. Erickson. 2002. Resource selection by animals: statistical design and Conservation Biology 17:566–576.
analysis for field studies. Kluwer Academic Publishers, Norwell, Saura, S., and L. Pascual-Hortal. 2007. A new habitat availability index to
Massachusetts, USA. integrate connectivity in landscape conservation planning: comparison
Martin, S. K. 1994. Feeding ecology of American martens and fishers. Pages with existing indices and application to a case study. Landscape and
297–315 in S. W. Buskirk, A. S. Harestad, M. G. Raphael, and R. A. Urban Planning 83:91–103.
Powell, editors. Martens, sables, and fishers—biology and conservation. Seaman, D. E., J. J. Millspaugh, B. J. Kernohan, G. C. Brundige, K. J.
Cornell University Press, Ithaca, New York, USA. Raedeke, and R. A. Gitzen. 1999. Effects of sample size on kernel home
McLoughlin, P. D., and S. H. Ferguson. 2000. A hierarchical pattern of range estimates. Journal of Wildlife Management 63:739–747.
limiting factors helps explain variation in home range size. Écoscience Seaman, D. E., and R. A. Powell. 1996. An evaluation of the accuracy of
7:123–130. kernel density estimators for home range analysis. Ecology 77:2075–2085.
Mergey, M., R. Helder, and J. J. Roeder. 2011. Effect of forest fragmenta- Selonen, V., I. K. Hanski, and P. C. Stevens. 2001. Space use of the Siberian
tion on space-use patterns in the European pine marten (Martes martes). flying squirrel Pteromys volans in fragmented forest landscapes. Ecography
Journal of Mammalogy 92:328–335. 24:588–600.
MRNF. 2009. Normes de cartographie écoforestière: troisième inventaire Sherburne, S. S., and J. A. Bissonette. 1994. Marten subnivean access point
écoforestier. Ministère des Ressources naturelles et de la Faune, Forêt use: response to subnivean prey levels. Journal of Wildlife Management
Québec, Direction des inventaires forestiers, Québec, 95p. 58:400–405.

Cheveau et al.  Marten Space Use and Habitat Selection 11

Siffczyk, C., L. Brotons, K. Kangas, and M. Orell. 2003. Home range Thompson, I. D., and P. W. Colgan. 1994. Marten activity in uncut and
size of willow tits: a response to winter habitat loss. Oecologia 136:635– logged boreal forests in Ontario. Journal of Wildlife Management 58:280–
642. 288.
Smith, A. C., and J. A. Schaefer. 2002. Home-range size and habitat Thompson, I. D., and A. S. Harestad. 1994. Effects of logging on American
selection by American marten (Martes americana) in Labrador. martens, and models for habitat management. Pages 355–367 in S. W.
Canadian Journal of Zoology 80:1602–1609. Buskirk, A. S. Harestad, M. G. Raphael, and R. A. Powell, editors.
Snyder, J. E., and J. A. Bissonette. 1987. Marten use of clear-cuttings and Martens, sables, and fishers—biology and conservation. Cornell
residual forest stands in western Newfoundland. Canadian Journal of University Press, Ithaca, New York, USA.
Zoology 65:169–174. Turcotte, Y., and A. Desrochers. 2003. Landscape-dependent response to
Spencer, W. D., R. H. Barrett, and W. J. Zielinski. 1983. Marten habitat predation risk by forest birds. Oikos 100:614–618.
preferences in the northern Sierra Nevada. Journal of Wildlife Wauters, L., P. Casale, and A. A. Dhondt. 1994. Space use and dispersal of
Management 47:1181–1186. red squirrels in fragmented habitats. Oikos 69:140–146.
Strickland, M. A., and C. W. Douglas. 1987. Marten. Pages 530–546 in Wiens, J. A. 1995. Landscape mosaics and ecological theory. Pages 1–26 in
M. Novak, editor. Wild furbearer management and conservation L. Hansson, L. Fahrig, and G. Merriam, editors. Mosaic landscapes and
in north America. Ministry of Natural Resources, Ottawa, Ontario, ecological processes. Chapman and Hall, London, United States.
Canada. Wiens, J. A., N. C. Stenseth, B. Van Horne, and R. A. Ims. 1993. Ecological
Thomas, D. L., and E. J. Taylor. 1990. Study designs and tests for com- mechanisms and landscape ecology. Oikos 66:369–380.
paring resource use and availability. Journal of Wildlife Management With, K. A., R. H. Gardner, and M. G. Turner. 1997. Landscape connec-
54:322–330. tivity and population distribution in heterogeneous environments. Oikos
Thomas, D. L., and E. J. Taylor. 2006. Study designs and tests for com- 78:151–169.
paring resource use and availability. Journal of Wildlife Management Wolff, J. O., E. M. Schauber, and W. D. Edge. 1997. Effects of habitat loss
70:324–336. and fragmentation on the behavior and demography of gray-tailed voles.
Thompson, I. D. 1991. Could marten become the spotted owl of eastern Conservation Biology 11:945–956.
Canada? Forestry Chronicle 67:136–140. Worton, B. J. 1989. Kernel methods for estimating the utilization distribu-
Thompson, I. D. 1994. Marten populations in uncut and logged boreal tion in home-range studies. Ecology 70:164–168.
forests in Ontario. Journal of Wildlife Management 58:272–280. Wynne, K. M., and J. A. Sherburne. 1983. Summer home range use by adult
Thompson, I. D., J. A. Baker, C. Jastrebski, J. Dacosta, J. Fryxell, and D. marten in northwestern Maine. Canadian Journal of Zoology 62:941–943.
Corbett. 2008. Effects of post-harvest silviculture on use of boreal forest Zuur, A. F., E. N. Ieno, N. J. Walker, A. A. Saveliev, and G. M. Smith.
stands by amphibians and marten in Ontario. Forestry Chronicle 84: 2009. Mixed effects models and extensions in ecology with R. Springer,
741–747. New York, New York, USA.
Thompson, I. D., and P. W. Colgan. 1990. Prey choice by marten during a
decline of prey abundance. Oecologia 83:443–451. Associate Editor: John Squires.

Appendix: Model selection for female marten home range size in the Waswanipi Cree Model Forest, Quebec, Canada in 2006. We modeled landscape
(dispersed-cut vs. clustered-cut) and intrinsic effects separately from habitat effects. We report the number of parameters in the model (K), Akaike’s Information
Criterion adjusted for small sample size (AICc), relative difference in AICc values compared to the top-ranked model (DAICc), and AICc model weight (wi).
60% kernels 90% kernels
Models K AICc DAICc wi Models K AICc DAICc wi
Landscape and intrinsic effects
Landscape 3 66.38 0.00 0.29 Landscape 3 95.79 0.00 0.26
Body mass 3 67.63 1.25 0.16 Body index 3 96.68 0.89 0.17
Age 3 67.65 1.27 0.16 Body mass 3 96.72 0.93 0.16
Body index 3 67.72 1.34 0.15 Age 3 96.73 0.93 0.16
Landscape þ body index 4 69.33 2.96 0.07 Landscape þ age 4 98.28 2.48 0.07
Landscape þ age 4 69.46 3.09 0.06 Landscape þ body index 4 98.61 2.81 0.06
Landscape þ body mass 4 69.54 3.16 0.06 Landscape þ body mass 4 98.96 3.17 0.05
Age þ body index 4 70.78 4.40 0.03 Age þ body index 4 99.81 4.02 0.03
Landscape þ age þ body index 5 72.89 6.51 0.01 Landscape þ age þ body index 5 101.61 5.82 0.01
Landscape þ body mass þ age 5 73.07 6.69 0.01 Landscape þ body mass þ age 5 101.89 6.09 0.01
Habitat effectsa
MIX70120 3 60.63 0.00 0.55 MIX70120 3 89.34 0.00 0.49
MIX70120 þ CONIF120 4 63.67 3.05 0.12 MIX70120 þ CONIF120 4 91.10 1.76 0.20
Forested bogs 3 63.96 3.33 0.10 CONIF120 3 93.09 3.75 0.08
CONIF120 3 66.50 5.87 0.03 Forested bogs 3 93.15 3.81 0.08
CONIF7090 3 66.61 5.98 0.03 CONIF120 þ CONIF7090 4 95.24 5.89 0.03
Residual forest 3 67.18 6.55 0.02 Open areas 3 95.69 6.35 0.02
Open areas 3 67.22 6.60 0.02 Regenerating stands from fire 3 96.11 6.77 0.02
Regenerating stands from fire 3 67.40 6.77 0.02 Residual forest 3 96.15 6.81 0.02
CONIF120 þ CONIF7090 4 67.55 6.93 0.02 CONIF7090 3 96.52 7.18 0.02
Recent clear-cuts 3 67.66 7.04 0.02 Old clear-cuts 3 96.57 7.23 0.01
Old clear-cuts 3 67.71 7.08 0.02 Clear-cuts 3 96.64 7.30 0.01
Water 3 67.72 7.09 0.02 Non-forested 3 96.76 7.42 0.01
Non forested 3 67.74 7.11 0.02 Water 3 96.78 7.44 0.01
Clear-cuts 3 67.74 7.11 0.02 Recent clear-cuts 3 96.78 7.44 0.01
a
MIX70120: mixedwood stands 70–120 yr, CONIF120: coniferous stands >120 yr, CONIF7090: coniferous stands 70–90 yr.

12 The Journal of Wildlife Management