You are on page 1of 11

Journal of Mammalogy, 93(2):464474, 2012

Martens in the matrix: the importance of nonforested habitats for

forest carnivores in fragmented landscapes
Australian Research Centre for Urban Ecology, School of Botany, University of Melbourne, Parkville, Victoria 3010,
Australia (FMC)
Biological and Environmental Sciences, University of Stirling, Stirling, FK9 4LA, Scotland, United Kingdom
Forest Research, Centre for Human and Ecological Sciences, Northern Research Station, Roslin, Midlothian, EH25 9SY,
Scotland, United Kingdom (CPQ)
* Correspondent:

The intervening landscape between patches of forest (i.e., matrix) has enormous potential to mitigate the
negative effects of forest fragmentation. However, to release this potential requires understanding of how
individual species perceive matrix. Here we investigated use of matrix by pine martens (Martes martes) in a
region with low forest cover (Scotland). We radiotracked 11 martens to determine their habitat preferences, then
combined our data with those published from 5 additional Scottish landscapes to examine how home-range size
(i.e., population density) and diet of martens varied with forest edge density (i.e., fragmentation). Our tracking
showed that although mature forest was the most preferred habitat, certain matrix habitats (scrub and tussock
grassland) also were consistently selected. These 2 habitats provided martens with fundamental resources that
are of limited availability within intensively managed plantation forests: den sites and primary prey (Microtus
agrestis). Our synthesis of data across landscapes indicated martens benefit from supplemental resources in
matrix habitats; consumption of small mammals increased with fragmentation and coincided with an initial
increase in marten population densities. However, population densities of martens decreased once
fragmentation passed a threshold level. Our results demonstrate that habitat complementation at the
landscape-scale is essential for some forest-associated species. Resource supplementation from matrix habitats
may be particularly important in regions with a long history of low-forest cover or where forest cover is now
dominated by afforested plantations, which may lack essential resources.

Key words: afforestation, foraging ecology, functional connectivity, habitat complementation, habitat selection, home
range, Martes martes, plantation, resource subsidy, Scotland

E 2012 American Society of Mammalogists

DOI: 10.1644/11-MAMM-A-149.1

Conceptual models of forest fragmentation that contrast (Dunford and Freemark 2004). Functional connectivity may
habitat patches with a uniformly unsuitable matrix (i.e., the be maintained even within highly fragmented landscapes
nonhabitat surrounding habitat patches) ignore the importance depending on how an organism perceives and responds to the
of heterogeneity within the matrix (Kupfer et al. 2006). intervening matrix between remnants (Belisle 2005; Ricketts
Although binary fragmentation models may be appropriate for 2001). In some cases, the matrix may provide alternative or
some forest specialists, for many species the matrix forms a supplementary resources (e.g., food or nest sites) that support
continuum of habitats of varying permeability (Fischer et al. greater population densities than would be expected within
2004; Haila 2002). Matrix permeability is determined by the remnants alone (i.e., habitat complementation [e.g., Andren
structural similarity between matrix habitats and remnant 1994]). Ignoring the ecological qualities of the matrix may
habitats, the perception of which is both species- and context- therefore result in a gross underestimation of its importance to
dependent (Kupfer et al. 2006). Because ecological processes
within remnants (such as dispersal and foraging) vary as a
function of their surroundings, the matrix has great potential to
mitigate the negative effects of habitat loss and fragmentation

organisms within remnants and its potential to serve as Habitat loss and fragmentation are landscape-scale process-
functional habitat. To better understand how wildlife popula- es, yet few studies of their effects on populations of European
tions persist within fragmented landscapes it is necessary to mammals have been conducted at an appropriate scale (i.e.,
determine if, how, and why animals utilize matrix habitats. where individual landscapes are the study unitsMortelliti
The European pine marten (Martes martes) is often et al. 2010). To better understand the landscape-level effects
described as a forest-dependent specialist of late-successional of habitat loss and fragmentation on marten populations, we
forests, an association that may be partly due to a large compiled data from previous studies to investigate how the
number of habitat studies for martens occurring within the diet and population density of martens varied among
forest-dominated landscapes of Fennoscandia (Brainerd 1990; landscapes with differing amounts of forest cover and forest
Kurki et al. 1998; Lindstrom 1989; Storch et al. 1990), and to edge densities (the latter as a measure of fragmentation).
implied similarities with congeneric American martens (M. Marten populations are vulnerable to fragmentation (Bright
americana) within the equally forested regions of North 2000), yet increased foraging opportunities in forest gaps may
America and Canada (e.g., Buskirk and Powell 1994). By allow population densities to increase with low levels of
contrast, the landscapes of western Europe are much less fragmentation, provided forest habitats remain sufficiently
forested (Food and Agriculture Organization of the United connected, although a sharp population decline is expected
Nations 2003). Yet martens are present in western Europe, once fragmentation reaches a threshold level (Brainerd 1990;
occurring in landscapes with as little as 4% forest cover Thompson and Harestad 1994).
(Balharry 1993). Although their habitat preferences are much
less studied, research indicates martens in this region are less
dependent on forests than elsewhere in their range (e.g., MATERIALS AND METHODS
Pereboom et al. 2008). In Scotland, for instance, where forest Study area.The 6,000-ha study area was located around
covers just 17% of land area (Malcolm et al. 2001), the Morangie Forest, a managed coniferous plantation in north-
primary prey of martens is Microtus agrestis (Balharry 1993; eastern Scotland, United Kingdom (57.8uN, 4.1uW). The
Caryl 2008; Lockie 1961), a species of vole that is surrounding landscape has a low human population density (5
characteristic of open tussock grassland and is absent from people/km2) and comprises farmland, woodland, and open
forest interiors (Hansson 1978). This contrasts with dietary heath uplands, with elevations ranging from 0 to 400 m above
preferences of martens within the more forested regions of sea level. The region has an oceanic climate with cool, wet
temperate and boreal Europe, where the most important prey summers (July X 5 13uC) and mild winters (February X 5
are forest-dwelling Myodes (Zalewski 2004). Understanding 0uCMetOffice 2008). The forest, which covers around
how martens utilize different matrix habitats may facilitate 3,000 ha, is dominated by 1st-rotation stands planted on
conservation management of this species in regions where formerly bare ground (61%) with fewer stands on historically
forest cover is low. Yet despite this dietary indication that wooded land (34%, of which just 4% has been continuously
martens are utilizing the nonforested matrix, previous research forested since about 1900). Lowland areas are dominated by
on habitat use by Scottish martens has focused on their use of thinned stands of native Pinus sylvestris that contain dense
forest habitats while treating the matrix as uniformly ground cover by Calluna vulgarisVaccinium myrtillus heath.
unsuitable (Balharry 1993; Halliwell 1997). Exposed uplands are dominated by unthinned stands of exotic
In this study we investigated how both forest and matrix conifers (e.g., Pinus contorta and Picea sitchensis) managed
habitats are utilized by martens by measuring the structural on a large-scale clear-cut and replant system. Deciduous trees
and trophic components of habitats occupied by martens. (e.g., Betula pendula and Sorbus aucuparia) are patchily
Predation and winter severity are major limiting factors to distributed throughout the forest. Shrubby species (e.g., Ulex
European marten populations that directly influence habitat europaeus, Rhododendron ponticum, and Salix) frequently
selection (Helldin 1998; Zalewski and Jedrzejewski 2006). We encroach into open areas. Unplanted areas are dominated by
therefore predicted martens would select structurally complex heaths of blanket bogs (e.g., C. vulgaris and Eriophorum
habitats that provide protective cover from predators and vaginatum) and moorland grasses (e.g. Nardus stricta and
temperature extremes (Buskirk and Powell 1994), irrespective Molinea caerulae) at higher elevations, and by damp
of whether those habitats occur within forest or the matrix. We grasslands (e.g., Holcus lanatus and Deschampsia cespitosa
also predicted martens would select habitats in which their [all communities according to Rodwell 1998]) at lower
preferred prey was found (i.e., tussock grasslandCaryl elevations.
2008). Matrix permeability is often related to body size; Marten capture and telemetry.Between January 2006 and
smaller-bodied species tend to be more prone to predation in July 2007 we set cage traps (Tomahawk Live Trap Co.,
the matrix, whereas large-bodied species view the matrix as Tomahawk, Wisconsin) at prebaited locations along forest
less of a barrier to movement (Gehring and Swihart 2003). roads throughout the study area. Captured martens were
Because there is considerable sexual size dimorphism within immobilized with an intramuscular injection of Ketalar and
Martes species, we predicted smaller-bodied females would be Domitor, antagonized by Antisedan (Pfizer Inc., New York,
more risk-averse than males and would avoid those habitats New York), using dosage rates recommended for similarly
that offered the least structural protection from predators. sized mustelids (Fournier-Chambrillon et al. 2003). Captured

TABLE 1.Habitat definition criteria based on structural and trophic resources. Forest habitats were classed as those with .30% canopy.

Structural vegetation layers Small mammala

Habitat Description Ground Understory Canopy (%) Density (no./ha) Species present
Closed Coniferous stands of 1744 years ool. Absent Absent .70 27.7 Mg, A
Predominantly exotic species (e.g., Picea).
Mature Coniferous stands . 45 years old that have Present Absent 3070 18.7 Mg, A, S
received .1 thinning treatment. Predominantly
native Pinus silvestris (63%). Well-established
field vegetation and occasional shrub layer.
Deciduousb Dominated by Betula pendula or Prunus avium, Absent Absent .30
typically little to no understory vegetation.
Scrub Formed by several habitats that shared a similar level Present Present ,30 19.5 Mg, A, S
of structural complexity within field and shrub
layers, including dense shrubs (e.g., Ulex europaeus),
wind-thrown debris, clear-cut debris, and
regenerating stands (,17 years).
Tussock Ungrazed grassland, typical of low elevations. Present Absent Absent 18 Ma, S
Contained high-productivity grasses and rushes
(e.g., HolcusDeschampsiaRodwell 1998).
Heath Common on unplanted uplands; dominated by heaths Present Absent Absent 3.2 Mg
(e.g., Calluna vulgaris) and low-productivity
upland grasses (e.g., NardusCarexRodwell 1998).
Agriculture Includes tilled fields but predominantly comprised of Present Absent Absent 0
grazed pasture.
Small mammal species (Mg 5 Myodes glareolus, A 5 Apodemus, Ma 5 Microtus agrestis, S 5 Sorex) densities taken from Caryl (2008).
Deciduous woodland was not trapped because patches of this habitat were too small and too few within the study area.

animals were weighed and aged (juvenile or adult  1 year) landscape, is an arbitrary decision (Garshelis 2000). We
according to body size and tooth wear. Adult martens were defined the study area as the minimum convex polygon
fitted with very-high-frequency radiocollars (TW-3; Biotrack containing the home ranges of all martens buffered by a
Ltd., Dorset, United Kingdom) and uniquely identified with a distance equal to the average length of home ranges belonging
passive transponder implanted subcutaneously into the neck. to females (1.8 km). We included this additional buffer to
Animal capture and handling were consistent with guidelines make composition of the habitats from which martens could
of the American Society of Mammalogists (Sikes et al. 2011). select their home ranges more representative of those within
All procedures were preapproved by the University of Stirling the greater landscape. Within the 5,902-ha study area, land-
ethical review committee and adhered to strict Government cover was classified into 7 habitat types based upon ground,
compliance under the Animals (Scientific Procedures) Act shrub, and canopy vegetation using forest inventory maps
1986 regarding the welfare of animals in scientific studies (Forestry Commission, Edinburgh, United Kingdom), digital
(British Home Office License PIL60/10174). All legal land classifications (LCM2000; Land Cover Map 2000; Centre
requirements of working with a protected species were met for Ecology and Hydrology, Edinburgh, United Kingdom),
(Scottish Natural Heritage Scientific License 6146). and orthorectified aerial photographs (0.5 m2 resolution taken
One animal was tracked per night, on foot or by car, for up to 2005; Forestry Commission, United Kingdom) before being
16 h. Locations of martens were estimated at intervals  20 min ground-truthed with field visits. We defined 4 matrix habitats
with simultaneous biangulations. Preliminary tests showed the where tree canopy was ,30% (scrub, tussock, heath, and
mean accuracy of this method was 55 m (SE 5 9 m) and mean agriculture), and 3 forest habitats where tree canopy was
bearing error was 12u (SD 5 10uZimmerman and Powell .30% (closed-canopy forest, mature coniferous forest, and
1995). We used Locate III software (Nams 2006) to estimate deciduous woodland; Table 1). The selection of 30% canopy
animal locations and 95% confidence ellipses from biangulations. cover as a threshold distinguishing between matrix and forest
Locations with confidence ellipses . 10 ha were excluded from habitats was arbitrary but based on the canopy-cover threshold
further analysis. Remaining locations had confidence ellipses of at which habitat suitability increases for American martens
2 ha 6 0 SE, 58% of which were 1 ha. We entered marten (Hargis et al. 1999).
locations into ArcView GIS (ESRI, Redlands, California) for Habitat selection of martens.We determined use of habitat
analysis with the Home Range Extension (Rodgers and Carr by martens at 2 spatial scales; 1st comparing the proportion of
1998). We used a 95% minimum convex polygon (Harris et al. habitats found within individual home ranges to their
1990) to estimate the home range of each marten. Martens were availability within the study area based on the number of
tracked until plots of home-range size over time reached an hectares of each habitat, referred to herein as home-rangelevel
asymptote; those that did not were excluded from further analysis. selection (Design IIThomas and Taylor 1990); then compar-
Habitat mapping.Delimiting a study area, hence deter- ing the proportion of locations in each habitat with their
mining the availability of habitats to martens within the availability within home ranges, referred to herein as location-

TABLE 2.Marten home-range sizes, forest cover metrics, and small mammal consumption at 6 sites across Scotland. Home-range sizes are
represented by mean (6 SE) 100% minimum convex polygons with sample size of animals tracked given in parentheses. Forest edge density
was calculated as a measure of the degree of forest fragmentation as meters of forest edge per hectare of forest. Forest cover shows the
proportion of the 300-km2 landscape with forest cover. The occurrence of small mammals in the diet was standardized as the frequency of small
mammal occurrences (% FO) divided by the number of scats sampled within each landscape (also shown).

Location Home range (X 6 SE: km2) Small mammals in diet

Forest edge Forest cover
Landscape Latitude Longitude Males Females density (m/ha) (%) (% FO) No. scats
1. Kinlochewe 57.6 25.3 23.63 6 3.68 (7) 8.83 6 1.13 (5) 147.05 4.1 92.5 723
2. Strathglass 57.3 24.9 6.28 6 1.00 (6) 3.57 6 0.49 (6) 91.34 16.5 44.7 581
3. Novar 57.7 24.7 3.04 6 0.98 (2) 2.03 6 0.27 (2) 91.32 18.3 55.8 1938
4. Morangie 57.8 24.1 5.63 6 1.65 (4) 0.70 6 0.12 (7) 103.13 36.9 49.5 2,249
5. Glen Troolb 55.0 24.7 8.36 6 1.77 (3) 4.52 6 0.79 (2) 73.43 46.4 34.0 100
6. Minnochb 55.1 24.6 32.86 6 12.08 (3) 10.61 6 1.23 (3) 65.84 47.7 34.0 100
Sites 1 and 2 (Balharry 1993), 3 (Halliwell 1997), 4 (present study; Caryl 2008), and 5 and 6 (Bright & Smithson 1997).
Bright and Smithson (1997) did not report separate dietary data for their 2 sites but reported no difference in the frequency of occurrence of prey items between sites.

level selection (Design IIIThomas and Taylor 1990). from scat-content analysis conducted simultaneously with
Locations were differentiated as being active or static tracking within each landscape. In total, we had home-range
depending on the physical state of martens when a location and dietary data for martens in 6 landscapes across Scotland
was recorded and separate analyses were conducted for each. (including our data from Morangie, dietary data for which
To ensure independence of static locations, only 1 static came from Caryl [2008]; Table 2; Fig. 1).
location was recorded per denning event and subsequent We standardized the dietary importance of small mammals
locations were disregarded until the animal moved again. in each landscape as the frequency of small mammal
Home ranges were defined by the 95% minimum convex occurrences in scats divided by the number of scats analyzed
polygon buffered by a distance of 55 m (our mean telemetry within that landscape to account for inconsistencies in
error). Locations were treated as an ellipse rather than a single reporting results. We provide a relative measure of the density
point, with the habitat of greatest proportional cover considered of resident adult martens in each landscape using the mean
used at that location (Nams 1989). Because female martens are sex-specific home-range size (100% minimum convex
morphologically and energetically more constrained than polygon). We take this measure to be the approximate inverse
males, they may be more selective (Buskirk and Powell of breeding marten population density assuming that martens
1994). We therefore examined sex-specific habitat selection display intersexual territoriality, that only adult martens
rather than pooling data across sexes (Garshelis 2000). Habitat defend territories, and that home ranges within each sex abut
preferences and differences in preferences between sexes were each other contiguously (Balharry 1993; Caryl 2008). We
examined with Bonferroni-adjusted 90% confidence limits recognize that high population density does not necessarily
(Cherry 1998). To avoid unequal weighting, we randomly indicate high habitat quality (Van Horne 1983), but by
selected an equal number of active locations (males: n 5 22; focusing on adult resident martens we effectively ignore the
females: n 5 28) and static locations (n 5 8 for both sexes) juvenile, dispersing, or transient individuals that often
from each individual (Thomas and Taylor 2006). Selection of artificially inflate population densities in poor-quality habitat.
critical resources is likely to be less variable than others, so we Forested extent (%) was calculated within a 9.77-km radius
determined variability in individual selection strategies by (300-km2 area) from the center of each site. Data on forest
calculating resource-selection indexes (RSI: % habitat use 2 % cover were obtained from the LCM2000 digital land-use layer
habitat available) from nonpooled data (Thomas and Taylor (Fuller et al. 2005). We included land classified as clear-cut to
1990). We used t-tests to determine sex-related differences in account for changes to forest cover between the year that this
home-range size and body mass. We also used t-tests to data layer was created (2000) and when marten data were
evaluate intersexual differences in the distance travelled by collected among studies. We then calculated the density of
individual martens into the matrix after measuring the mean and forest edges per hectare of forest (m/ha) as an index of forest
maximum distance of locations outside forest habitats. A chi- fragmentation within each landscape. General linear models
square test was used to compare the frequency with which each were used to examine relationships between measures of forest
sex was located outside forest habitats. cover and fragmentation with small mammal consumption and
Forest fragmentation, home-range size, and diet of martens home-range sizes of martens among landscapes. For each
across Scotland.We examined the diet and home-range size analysis a quadratic function of forest cover or edge density
of martens in relation to forest cover and forest edge density was included in the starting model to assess whether this
and extent among Scottish landscapes. We used home-range provided a better fit than a linear function, but was removed if
data from studies that have investigated spatial ecology for nonsignificant. All statistical tests were conducted in SPSS
martens in Scotland (Balharry 1993; Bright and Smithson 17.0 for Windows (SPSS, Armonk, New York) and Microsoft
1997; Halliwell 1997), each of which provided dietary data Excel 2007 (Microsoft Corporation, Redmond, Washington).

FIG. 1.Left) Outline of Scotland showing pine marten distribution (shaded) and locations of 6 landscapes used to examine differences in
fragmentation, diet, and population density of martens: 1) Kinlochewe (Balharry 1993), 2) Strathglass (Balharry 1993), 3) Novar (Haliwell
1997), 4) Morangie (present study; Caryl 2008), 5) Glen Trool (Bright and Smithson 1997), 6) Minnoch (Bright and Smithson 1997). Right)
Detail of forest fragmentation shown as forest (black) against matrix (white). Each landscape covers a circular area of 300 km2.

Data were transformed where necessary to achieve normality home-range scale were universal to all females (Fig. 2a). Males
and equal variance. All averages and coefficients are presented were more variable in their selection patterns, yet agricultural
as means with standard errors unless otherwise stated. land was universally avoided (Fig. 2b). No universal prefer-
ences were apparent at the home-range level except that of
female selection for tussock grassland.
RESULTS There were few significant preferences at the location level,
We captured and collared 16 adult pine martens (8 males suggesting that once habitats were selected from the landscape
and 8 females), of which sufficient data were collected for 4 they were generally used in accordance to their availability
males and 7 females to allow home-range estimation. A total (Table 3). An exception was that stationary females avoided
of 594 locations were obtained, with 50 6 4 locations per heath within home ranges, whereas stationary males avoided
female and 54 6 8 locations per male. Home ranges for deciduous woodland and active males avoided tussock
females reached an asymptote after 29 6 3 locations and for grassland. Individual selection strategies concur there was
males after 40 6 7 locations. Unbuffered home ranges for much less selection by active martens than at the home-range
females (95% minimum convex polygon) were significantly level, with most resource-selection indexes lying near zero
smaller than those of males, measuring 50 6 1 ha and 353 6 (Fig. 2a). However, there does appear to be an indication of
7 ha, respectively (t6 5 24.405, P 5 0.026). There was selection by stationary martens, with females having positive a
considerable body size dimorphism between the sexes; males positive resource-selection index for mature forest and scrub
weighed 1.88 6 0.7 kg (range 5 1.602.30 kg), almost 1.4 and males having a positive resource-selection index for
times heavier than females (1.37 6 0.4 kg, range 5 1.21 mature forest and closed-canopy forest (Fig. 2).
1.55 kg: t14 5 6.195, P , 0.001). There were intersexual differences in habitat utilization for
Habitat selection.Martens clearly selected home ranges all habitats at the home-range level except scrub and tussock
from the landscape in a nonrandom pattern (Table 3; Fig. 2). grassland (Table 3). Scrub and tussock grassland were equally
Both sexes selected the same 3 habitats at the home-range level selected by both sexes despite low availability within the study
(mature forest, tussock grassland, and scrub), while avoiding area. Use of all other habitats appeared to follow a general
the same 2 habitats (closed-canopy forest and agriculture; pattern in which females made greater use of forested habitats
Table 3). In addition, females strongly avoided heath moorland than did males (63.5% 6 5.0% of home range forested),
at the home-range level, whereas males avoided deciduous whereas males made greater use of matrix habitats than did
woodland. Individual selection strategies suggested female females (47.2% 6 10.5% of home range forested). For
aversions to closed-canopy forest, heath, and agriculture at the example, females utilized mature forest and deciduous

TABLE 3.Marten habitat selection preferences indicated with Bonferroni-adjusted 90% confidence intervals (90% CIs) for female (F) and
male (M) martens at different spatial scales: selection of home ranges (HRs) from the study area (SA); selection of active and static locations
from SA and HR. Significant selection and avoidance (P , 0.1) are indicated by +++ and 222, respectively. Significant intersexual differences
are indicated with inequality signs. NS 5 not significant.

Home range Active locations Static locations

Habitat Available (%) (90% CI) Versus SA Intersex (90% CI) Versus HR Intersex (90% CI) Versus HR Intersex
Mature 17.3 38.548.5 +++ F.M 34.752.0 NS NS 33.666.4 NS NS
Deciduous 3.7 4.08.9 +++ F.M 4.514.9 NS NS 0.015.6 NS NS
Closed 27.3 10.117.0 222 F,M 5.316.1 NS NS 0.015.6 NS NS
Scrub 8.1 8.715.2 +++ NS 6.918.6 NS NS 9.437.0 NS NS
Tussock 9.8 14.522.3 +++ NS 15.129.8 NS F.M 0.018.3 NS NS
Heath 12.9 1.14.3 222 F,M 0.01.8 NS F,M 0.00.0 222 NS
Agricultural 20.9 1.75.3 222 F,M 0.01.8 NS NS 0.09.6 NS NS
Mature 17.3 21.126.2 +++ M,F 13.836.4 NS NS 22.365.2 NS NS
Deciduous 3.7 0.82.2 222 M,F 0.04.8 NS NS 0.00.0 222 NS
Closed 27.3 19.524.5 222 M.F 3.320.1 NS NS 6.243.8 NS NS
Scrub 8.1 13.918.4 +++ NS 2.819.1 NS NS 0.010.7 NS NS
Tussock 9.8 12.016.1 +++ NS 0.010.2 222 M,F 0.016.7 NS NS
Heath 12.9 11.315.4 NS M.F 5.023.2 NS M.F 0.016.7 NS NS
Agricultural 20.9 7.611.0 222 M.F 1.215.7 NS NS 0.031.4 NS NS

woodland significantly more than males at the home-range greater part of the martens diet (F1,4 5 56.77, P 5 0.002, b
level, whereas males used heath and agriculture significantly 5 0.007 6 0.001, R2 adj 5 91.8%; Fig. 3A). A strong
more than females (Table 3). Within home ranges, active negative correlation existed between the extent of forest cover
females utilized tussock grassland more than did males, and and forest edge-density within landscapes (Pearson correlation
active males utilized heath more than did females. There were r 5 20.83, P 5 0.042, n 5 6). Henceforth we only present
no significant differences in habitat use at stationary locations. results in reference to the latter, because edge-density had a
Greater proclivity for open spaces by males was reflected by greater functional significance in relation to the consumption
significant intersexual differences in both the mean and of small mammals. Home ranges of males were logged prior to
maximum distance travelled outside forest habitats by individ- analysis. As predicted, the relationship between fragmentation
uals (t9 5 22.601, P , 0.05 and t9 5 22.405, P , 0.05, and home-range size of female martens was nonlinear (edge
respectively). Males typically travelled 75.1 6 18.6 m outside density F1,3 5 25.53, P 5 0.015; edge density2 F1,3 5 26.23,
forest habitats, up to an average maximum of 199.6 6 46.3 m, P 5 0.014, R2 adj 5 82.9%; n 5 6; Fig. 3B) as was that of
whereas females typically travelled just 30.4 m into the matrix, males (edge density F1,3 5 13.43, P 5 0.035; edge density2
up to an average maximum distance of 93.7 6 21.1 m. Males also F1,3 5 14.06, P 5 0.035, R2 adj 5 71.1%, n 5 6; Fig. 3C).
travelled outside forested habitats more frequently than females; Home ranges of martens were smallest at intermediate levels
46% of male locations occurred outside forests compared with of landscape fragmentation (where the forested extent lay
just 33% of female locations (x215 8.786, P , 0.005). between 25% and 30%), and increased in size at higher and
Forest fragmentation, home-range size, and marten diet lower levels of fragmentation. However, we acknowledge that
across Scotland.There was considerable variation in the because of the small sample size these patterns are sensitive to
amount of forest cover among landscapes across Scotland, individual points, particularly at the lowest and highest edge
with an order of magnitude between the least (4%) and most densities, and should be viewed with caution.
(47%) forested, which corresponded to an equally variable
degree of fragmentation among landscapes (Table 2). Simi-
larly, there was an order of magnitude difference between DISCUSSION
largest (33 km2) and smallest (3 km2) home ranges of male Our results add support to the suggestion that European pine
martens. The frequency of consumption of small mammals martens, like other marten species (e.g., Hearn et al. 2010),
varied from 0.34 to 0.93 occurrences per scat. Small mammal may be less forest-dependent than previously believed
remains were not identified to species at Minnoch and Glen (Mortelliti et al. 2010; Pereboom et al. 2008). Although
Trool (Bright and Smithson 1997), but M. agrestis formed an martens demonstrated a strong preference for mature forest,
average 80% (SE 5 5%) of small mammals in the diet within both sexes showed strong selective preferences for 2 matrix
the other 4 landscapes. The relationship between small habitats within their home ranges: tussock grassland and
mammal consumption and forest fragmentation demonstrated scrub. Consistent selection of these habitats suggests the
a strong positive correlation, indicating that as landscapes matrix is not only penetrable but utilized by martens.
became more fragmented, small mammals constituted a Furthermore, matrix habitats may provide key resources that

be equally unsuitable. Yet evidence increasingly suggests that

it is the structural elements within forests rather than forest
composition or age that most influences marten habitat
selection (Brainerd et al. 1994; Hearn et al. 2010; Payer and
Harrison 2003; Poole et al. 2004). Many of the structural
features associated with old-growth forests (e.g., multilayered
tree canopies, large-diameter trees, standing and fallen
deadwood, and dense understory vegetation) are required by
martens because they help reduce predation threats (Storch et
al. 1990), increase foraging efficiency (Andruskiw et al. 2008;
Thompson and Harestad 1994), and offer thermally insulated
resting sites (Zalewski 1997). However, these structural
legacies are often not available in regions of the martens
geographic range in which much of the original forest cover
was removed many centuries ago. For instance, the majority
(76%) of British forests are secondary, that is, they occupy
ground that has not been wooded until recently, and so possess
few of the structural features of old-growth forest (Humphrey
2005). In the absence of suitable structural features it seems
that martens in our study area were supplementing the
resources required to avoid predation, den, and forage from
complementary, low-contrasting habitats within the matrix.
Predator avoidance is an important determinant of habitat
use by martens (Helldin 1998), and structure near the ground
is known to provide protective cover for martens (e.g.,
Slauson et al. 2007). Female martens are smaller than males
and subject to greater predation risk and energetic costs, and
are thus likely to be more selective in their habitat choice. We
found females were more reluctant to leave forest habitats
than males, and typically moved no more than 100 m outside
forest edges. Martens have 2 extant predators within Scot-
landred foxes (Vulpes vulpes) and golden eagles (Aquila
chrysaetosHelldin 1998)both of which were present
FIG. 2.Resource-selection indexes (RSI: % use 2 % availability) within the study area (F. M. Caryl, pers. obs.). One similarity
for habitats at the home-range level (white bars), at active locations among those habitats strongly avoided by female martens
(pale gray bars), and at static locations (dark gray bars) by a) female (i.e., closed-canopy forest, heath moorland, and agricultural
and b) male pine martens from radiotracking data collected in fields) was that cover near the ground was either absent or
Morangie Forest, Scotland, 20062007. A reference line (y 5 0) greatly reduced in stature. The high shading beneath closed-
indicates habitats are used according to their availability (i.e., no
canopy stands restricts understory vegetation resulting in a
selection), data below this line indicate avoidance; data above
ground cover dominated by bryophytes and needle litter (Hill
indicate preference. Boxes show the median (Q2) and interquartile
range (IQR: Q3 2 Q1). Whiskers indicate the smallest observation 1986); exposure and heavy grazing of upland heath often
within 1.5 IQR of Q1 and the largest observation within 1.5 IQR of results in compact, low-lying vegetation (Grant and Hunter
Q3. Points lying outside of this range are considered to be outliers 1962), whereas continuous grazing has obvious impacts on
(circle) and extreme outliers (star). agricultural pastures. By contrast, ground vegetation was well
established in habitats that were preferred by female martens.
Thinning to reduce stem density within mature stands (one of
are unavailable within forest habitats; the most notable of our defining criteria) increases irradiance to the ground
which being Microtus, which are absent from forest interiors resulting in a dense understory similar to that found in
(Caryl 2008), but are the primary prey of martens (Balharry seminatural forests (Ferris et al. 2000; Wallace and Good
1993; Caryl 2008; Halliwell 1997; Lockie 1961). 1995). Scrub habitats provide essential structural features for
Matrix habitats that have a high degree of structural contrast many small carnivore species within the agriculturally
with remnant vegetation are less penetrable to species than dominated landscapes of western Europe (Matos et al.
low-contrast habitats (e.g., Forman 1995). Martens are 2009; Pereboom et al. 2008; Rondinini and Boitani 2002;
commonly described as specialists of old-growth forest Santos and Santos-Reis 2010; Virgos 2001), and so it is not
(Brainerd 1990; Lindstrom 1989; Storch et al. 1990), and as surprising that this habitat is important for martens within our
such all nonforested habitats have typically been considered to study area. Although lacking in understory and canopy layers,

the HolcusDeschampsia communities that are typical of

ungrazed tussock grassland form characteristic tussocks up to
a meter in height (Rodwell 1998), thus providing some
degree of ground cover in addition to foraging resources (see
Resource-limited animals may expose themselves to greater
risks in order to acquire resources across habitat boundaries
(Belisle 2005). That martens utilized 2 matrix habitats in
addition to mature forest suggests that matrix habitats may
provide resources that are unavailable within mature forests
alone. The most important prey of Scottish martens are
Microtus voles (Balharry 1993; Caryl 2008), which is unusual
given that Myodes are numerically dominant within home
ranges of martens (Balharry 1993; Caryl 2008) and form the
bulk of their diet throughout the martens range (Zalewski
2004). Microtus are denizens of ungrazed tussock grassland
(Caryl 2008), which undoubtedly accounts for the selection of
this habitat by martens within our study area. Interestingly,
females made significantly greater use of this habitat while
active than did males, which may be indicative of resource
partitioning between the sexes. A dietary preference for
Microtus may indicate a behavioral trait that has adapted in
response to Scotlands highly fragmented past during which
forest-dwelling Myodes would have been scarce. Alternative-
ly, low volumes of coarse woody debris within plantations
may mean that Myodes remain unavailable to martens despite
their abundance, because coarse woody debris is needed to
provide sensory cues to martens as to the location of forest
rodents (Andruskiw et al. 2008; Thompson and Harestad
1994). Despite management steps to retain coarse woody
debris within plantations, even volumes of coarse woody
debris within mature forest stands remain low relative to old-
growth forests (Humphrey 2005; Kirby et al. 1998). It is
unclear whether further restoration of coarse woody debris
within plantations will change the vole preference of Scottish
martens, but at present they are supplementing their prey
resources with subsidies from the matrix. Martens in our study
area also seemed to be supplementing den sites from the
matrix. Plantation forests have few of the arboreal cavities that
are commonly used as den sites by martens in old-growth
forests to provide protection from foxes (Birks et al. 2005).
Instead, we found that scrub habitat was being used more at
female resting locations (23% 6 6%) than at active locations
(13% 6 8%), which suggests that it may provide important
resting habitat. Ground-level dens are unusual for maternal
dens (J. Birks, The Vincent Wildlife Trust, Ledbury, United
Kingdom, pers. comm.), yet we observed 4 of the 7 females

FIG. 3.The relationship between forest edge density (m/ha) as a

measure of fragmentation and A) an index of small mammal r
consumption by pine martens among six 300-km2 landscapes. The
relationship between forest edge density and mean home-range sizes was given as the frequency of occurrence of small mammals in scats
of B) female and C) male martens is shown for the same 6 (FO) divided by the number of scats sampled (N). Nonlinear
landscapes. Data on the diet and home-range sizes of martens were trendlines are shown in B and C because they fit the data better
taken from 4 studies conducted across Scotland between 1988 and than linear trendlines. Because of the small sample size, however,
2007 (Balharry 1993; Bright and Smithson 1997; Caryl 2008; these patterns are highly sensitive to individual points and should be
Halliwell 1997). A standardized index of small mammal consumption viewed with caution.

we tracked (3 of which had young) denning within scrub ACKNOWLEDGMENTS

habitat; specifically 2 beneath dense shrubs (U. europaeus and We thank Scottish Natural Heritage, Forest Research, and the University
R. ponticum) and 2 beneath wind-thrown trees or clear-cut of Stirling for providing funding for this work; Forest Enterprise Scotland
debris, such as piles of branches. for logistical support; numerous volunteers who assisted with fieldwork; A.
Spatially subsidized predators (i.e., those that gain resources Peace for statistical advice; and R. Raynor, R. Coope, and D. Balharry for
across habitat boundaries) often have abundances that exceed invaluable advice and discussion. The manuscript was greatly improved by
what local resources would otherwise support (e.g., Andren suggestions from 2 anonymous reviewers.
1994). The strong positive correlation we found between the
consumption of small mammals and forest edge density among LITERATURE CITED
Scottish landscapes suggests that foraging opportunities for ANDREN, H. 1994. Effects of habitat fragmentation on birds and
small mammal prey increase where fragmentation results in mammals in landscapes with different proportions of suitable
increased availability of edge habitat. As predicted, this initially habitat: a review. Oikos 71:355366.
supported higher population densities of martens, as evidenced ANDRUSKIW, M., J. M. FRYXELL, I. D. THOMPSON, AND J. A. BAKER.
by smaller home-range sizes. Once forest fragmentation passed 2008. Habitat-mediated variation in predation risk by the American
a threshold level, however, population densities of martens marten. Ecology 89:22732280.
decreased (i.e., their home ranges expanded). Unfortunately, BALHARRY, D. 1993. Factors affecting the distribution and population
because of the small sample size of studies used in our synthesis density of pine martens (Martes martes) in Scotland. Ph.D.
dissertation, University of Aberdeen, Aberdeen, Scotland.
we were not able to investigate the effect of different land uses
BELISLE, M. 2005. Measuring landscape connectivity: the challenge of
within the matrix. However, the composition and configuration
behavioral landscape ecology. Ecology 86:19881995.
of the matrix will undoubtedly influence processes within BIRKS, J. D. S., J. E. MESSENGER, AND E. C. HALLIWELL. 2005. Diversity
landscapes. Previous studies into the effects of forest fragmen- of den sites used by pine martens Martes martes: a response to the
tation on population density of martens in Fennoscandia (Kurki scarcity of arboreal cavities? Mammal Review 35:313320.
et al. 1998) and North America (Chapin et al. 1997; Hargis et al. BRAINERD, S. M. 1990. The pine marten and forest fragmentation: a
1999) also have not accounted for variation within the matrix. review and general hypothesis. Pp. 421434 in Transactions of the
Yet as our results demonstrate, the nonforested matrix is not 19th International Union of Game Biologists Congress (S.
homogenous from the perspective of a pine marten. In fact, Myrberget, ed.). Trondheim, Norway.
complementary resources from the matrix may be essential for BRAINERD, S. M., J. O. HELLDIN, E. LINDSTROM, AND J. ROLSTAD. 1994.
the completion of their life phases depending on local Eurasian pine martens and old industrial forest in southern boreal
Scandinavia. Pp. 343354 in Martens, sables and fishers: biology and
conditions. Fragmentation effects in landscapes composed of
conservation (S. W. Buskirk, A. S. Harestad, M. G. Raphael, and R. A.
matrix habitats that contrast strongly with remnant vegetation Powell, eds.). Cornell University Press, London, United Kingdom.
are likely to be worse than those in which the matrix is BRIGHT, P. W. 2000. Lessons from lean beasts: conservation biology
structurally similar (Forman 1995). Therefore, to gain greater of the mustelids. Mammal Review 30:217226.
insight into the landscape requirements of martens (and indeed BRIGHT, P. W., AND T. J. SMITHSON. 1997. Species recovery
other forest-dwelling species), we need to account for a more programme for the pine marten in England: 19951996. English
complex perception of habitats than forest versus matrix. This is Nature Research Reports 240:164.
particularly so where the length of time since habitat loss has BUSKIRK, S. W., AND R. A. POWELL. 1994. Habitat ecology of fishers
impacted faunal life histories, or where remaining forest and American martens. Pp. 283296 in Martens, sables and fishers:
habitats are secondary. biology and conservation (S. W. Buskirk, A. S. Harestad, M. G.
Raphael, and R. A. Powell, eds.). Cornell University Press,
Much of the forest cover in Scotland today comprises
London, United Kingdom.
afforested plantations in the closed-canopy stage (Malcolm
CARYL, F. M. 2008. Pine marten diet and habitat use within a
et al. 2001; Mason 2007), which make poor-quality habitat for managed coniferous forest. Ph.D. dissertation, University of
martens. In addition, a large proportion of the Scottish landscape Stirling, Stirling, Scotland.
is dominated by inhospitable matrix habitats (i.e., moorland and CHAPIN, T. G., D. J. HARRISON, AND D. M. PHILLIPS. 1997. Seasonal
agriculture). The reluctance of martens to cross large areas habitat selection by marten in an untrapped forest preserve. Journal
without cover means that even relatively nearby forest remnants of Wildlife Management 61:707717.
may not be used if they are isolated by unsuitable intervening CHERRY, S. 1998. Statistical tests in publications of the Wildlife
habitats (Ricketts 2001), and as slow-breeding mammals, Society. Wildlife Society Bulletin 26:947953.
martens are particularly vulnerable to local extinctions should DUNFORD, W., AND K. FREEMARK. 2004. Matrix matters: effects of
functional connectivity be lost (Bright 2000). Strategies for surrounding land uses on forest birds near Ottawa, Canada.
Landscape Ecology 20:497511.
conserving martens in these highly fragmented landscapes should
seek to restore old-growth conditions within existing forest
Relationships between vegetation, site type and stand structure in
stands while managing the surrounding matrix to ensure that coniferous plantations in Britain. Forest Ecology and Management
complementary (i.e., low-contrast) habitats are available. As our 136:3551.
results have shown, the matrix has enormous potential to mitigate FISCHER, J., D. B. LINDENMAYER, AND I. FAZEY. 2004. Appreciating
some of the negative effects of habitat loss and fragmentation that ecological complexity: habitat contours as a conceptual landscape
should not be overlooked. model. Conservation Biology 18:12451253.

State of the worlds forests. Food and Agriculture Organization of Abundances of red fox and pine marten in relation to the
the United Nations, Rome, Italy. composition of boreal forest landscapes. Journal of Animal
FORMAN, R. T. T. 1995. Land mosaics: the ecology of landscapes and Ecology 67:874886.
regions. Cambridge University Press, Cambridge, United Kingdom. LINDSTROM, E. 1989. Food limitation and social regulation in a red fox
FOURNIER-CHAMBRILLON, C., J. CHUSSEAU, J. DUPUCH, C. MAIZERET, AND population. Holarctic Ecology 12:7079.
P. FOURNIER. 2003. Immobilization of free-ranging European mink LOCKIE, J. D. 1961. The food of the pine marten Martes martes in
(Mustela lutreola) and polecat (Mustela putorius) with medetomi- West Ross-shire, Scotland. Proceedings of the Zoological Society
dineketamine and reversal by atipamezole. Journal of Wildlife of London 136:187195.
Diseases 39:393399. MALCOLM, D. C., W. L. MASON, AND G. C. CLARKE. 2001. The
FULLER, R. M., ET AL. 2005. The UK land cover map 2000: planning, transformation of conifer forests in Britainregeneration, gap size
construction and calibration of a remotely sensed, user-oriented and silvicultural systems. Forest Ecology and Management 151:
map of broad habitats. International Journal of Applied Earth 723.
Observation and Geoinformation 7:202216. MASON, W. L. 2007. Changes in the management of British forests
GARSHELIS, D. L. 2000. Delusions in habitat evaluation: measuring between 1945 and 2000 and possible future trends. Ibis 149:4152.
use, selection, and importance. Pp. 111164 in Research MATOS, H. M., M. J. SANTOS, F. PALOMARES, AND M. SANTOS-REIS.
techniques in animal ecology: controversies and consequences 2009. Does riparian habitat condition influence mammalian
(L. Boitani and T. K. Fuller, eds.). Columbia University Press, carnivore abundance in Mediterranean ecosystems? Biodiversity
New York. and Conservation 18:373386.
GEHRING, T. M., AND R. K. SWIHART. 2003. Body size, niche breadth, METOFFICE. 2008. Mapped climate averages for Northern Scotland
and ecologically scaled responses to habitat fragmentation: 19712000. Accessed
mammalian predators in an agricultural landscape. Biological 20 June 2008.
GRANT, S., AND R. HUNTER. 1962. Ecotypic differentiation of Calluna 2010. Experimental design and taxonomic scope of fragmentation
vulgaris (L.) in relation to altitude. New Phytologist 61:4455. studies on European mammals: current status and future priorities.
HAILA, Y. 2002. A conceptual genealogy of fragmentation research: Mammal Review 40:125154.
from island biogeography to landscape ecology. Ecological NAMS, V. O. 1989. Effects of radiotelemetry error on sample size and
Applications 12:321334. bias when testing for habitat selection. Canadian Journal of
HALLIWELL, E. C. 1997. The ecology of red squirrels in Scotland in Zoology 67:16311636.
relation to pine marten predation. Ph.D. dissertation, University of NAMS, V. O. 2006. Locate III. Pacer Computer Software, Tatama-
Aberdeen, Aberdeen, Scotland. gouche, Nova Scotia, Canada.
HANSSON, L. 1978. Small mammal abundance in relation to PAYER, D. C., AND D. J. HARRISON. 2003. Influence of forest structure
environmental variables in three Swedish forest phases. Studia on habitat use by American marten in an industrial forest. Forest
Forestalia Suecica 147:140. Ecology and Management 179:145156.
influence of forest fragmentation and landscape pattern on AND T. LODE. 2008. Movement patterns, habitat selection, and
American marten. Journal of Applied Ecology 36:157172. corridor use of a typical woodland-dweller species, the European
HARRIS, S., W. J. CRESSWELL, P. G. FORDE, W. J. TREWHELLA, T. pine marten (Martes martes), in fragmented landscape. Canadian
WOOLLARD, AND S. WRAY. 1990. Home-range analysis using radio- Journal of Zoology 86:983991.
tracking dataa review of problems and techniques particularly as POOLE, K. G., A. D. PORTER, A. DE VRIES, C. MAUNDRELL, S. D.
applied to the study of mammals. Mammal Review 20:97123. GRINDAL, AND C. Cassady St. CLAIR. 2004. Suitability of a young
HEARN, B. J., D. J. HARRISON, A. K. FULLER, C. G. LUNDRIGAN, AND deciduous-dominated forest for American marten and the effects of
W. J. CURRAN. 2010. Paradigm shifts in habitat ecology of forest removal. Canadian Journal of Zoology 80:423435.
threatened Newfoundland martens. Journal of Wildlife Manage- RICKETTS, T. H. 2001. The matrix matters: effective isolation in
ment 74:719728. fragmented landscapes. American Naturalist 158:8799.
HELLDIN, J. O. 1998. Pine marten (Martes martes) population RODGERS, A. R., AND A. P. CARR. 1998. Home Range Extension HRE
limitation: food, harvesting of predation? Ph.D. dissertation, for ArcView. Ontario Ministry of Natural Resources, Thunder Bay,
Swedish University of Agricultural Sciences, Uppsala, Sweden. Ontario, Canada.
HILL, M. O. 1986. Ground flora and succession in commercial forests. RODWELL, J. S. 1998. British plant communities. Vol. 3: grasslands
Institute of Terrestrial Ecology, Cambridge, United Kingdom. and montane communities. Cambridge University Press, Cam-
HUMPHREY, J. W. 2005. Benefits to biodiversity from developing old- bridge, United Kingdom.
growth conditions in British upland spruce plantations: a review RONDININI, C., AND L. BOITANI. 2002. Habitat use by beech martens in
and recommendations. Forestry 78:3353. a fragmented landscape. Ecography 25:257264.
KIRBY, K. J., C. M. REID, R. C. THOMAS, AND F. B. GOLDSMITH. 1998. SANTOS, M., AND M. SANTOS-REIS. 2010. Stone marten (Martes foina)
Preliminary estimates of fallen dead wood and standing dead trees habitat in a Mediterranean ecosystem: effects of scale, sex, and
in managed and unmanaged forests in Britain. Journal of Applied interspecific interactions. European Journal of Wildlife Research
Ecology 35:148155. 56:275286.
the ocean for the islands: the mediating influence of matrix-based OF THE AMERICAN SOCIETY OF MAMMALOGISTS. 2011. Guidelines of
processes on forest fragmentation. Global Ecology and Biogeog- the American Society of Mammalogists for the use of wild
raphy 15:820. mammals in research. Journal of Mammalogy 92:235253.

SLAUSON, K. M., W. J. ZIELINSKI, AND J. P. HAYES. 2007. Habitat WALLACE, H., AND J. GOOD. 1995. Effects of afforestation on upland
selection by American martens in coastal California. Journal of plant communities and implications for vegetation management.
Wildlife Management 71:458468. Forest Ecology and Management 79:2946.
STORCH, I., E. LINDSTROM, AND J. DE JOUNGE. 1990. Habitat selection ZALEWSKI, A. 1997. Patterns of resting site use by pine marten Martes
and food habits of the pine marten in relation to competition with martes in Bialowieza National Park (Poland). Acta Theriologica
the red fox. Acta Theriologica 35:311320. 42:153168.
THOMAS, D. L., AND E. J. TAYLOR. 1990. Study designs and tests for ZALEWSKI, A. 2004. Geographical and seasonal variation in food
comparing resource use and availability. Journal of Wildlife habits and prey size of European pine martens. Pp. 7898 in
Management 54:322330. Martens and fishers in human-altered environments: an interna-
THOMAS, D. L., AND E. J. TAYLOR. 2006. Study designs and tests for tional perspective (D. J. Harrison, A. K. Fuller, and G. Proulx,
comparing resource use and availability. Journal of Wildlife eds.). Springer, London, United Kingdom.
Management 70:324336. ZALEWSKI, A., AND W. JEDRZEJEWSKI. 2006. Spatial organisation and
THOMPSON, I. D., AND A. S. HARESTAD. 1994. Effects of logging on dynamics of the pine marten Martes martes population in
American martens and models for habitat management. Pp. 355 Biaowieza Forest (E Poland) compared with other European
367 in Martens, sables and fishers: biology and conservation (S. W. woodlands. Ecography 29:3143.
Buskirk, A. S. Harestad, M. G. Raphael, and R. A. Powell, eds.). ZIMMERMAN, J. W., AND R. A. POWELL. 1995. Radiotelemetry error:
Cornell University Press, London, United Kingdom. location error method compared with error polygons and
VAN HORNE, B. 1983. Density is a misleading indicator of habitat confidence ellipses. Canadian Journal of Zoology 73:11231133.
quality. Journal of Wildlife Management 47:893901.
VIRGOS, E. 2001. Relative value of riparian woodlands in landscapes Submitted 19 April 2011. Accepted 3 October 2011.
with different forest cover for medium-sized Iberian carnivores.
Biodiversity and Conservation 10:10391049. Associate Editor was Samantha M. Wisely.