ALCES VOL. 44, 2008 GILLINGHAM AND PARKER – HABITAT SELECTION BY MOOSE AND ELK 41 DIFFERENTIAL HABITAT SELECTION BY MOOSE AND ELK IN THE BESA-PROPHET AREA OF NORTHERN BRITISH COLUMBIA Michael P. Gillingham and Katherine L. Parker Natural Resources and Environmental Studies Institute, University of Northern British Columbia, 3333 University Way, Prince George, British Columbia, Canada V2N 4Z9, email: michael@unbc.ca ABSTRACT: Elk (Cervus elaphus) populations are increasing in the Besa-Prophet area of northern British Columbia, coinciding with the use of prescribed burns to increase quality of habitat for ungu- lates. Moose (Alces alces) and elk are now the 2 large-biomass species in this multi-ungulate, multi- predator system. Using global positioning satellite (GPS) collars on 14 female moose and 13 female elk, remote-sensing imagery of vegetation, and assessments of predation risk for wolves (Canis lupus) and grizzly bears (Ursus arctos), we examined habitat use and selection. Seasonal ranges were typi- cally smallest for moose during calving and for elk during winter and late winter. Both species used largest ranges in summer. Moose and elk moved to lower elevations from winter to late winter, but subsequent calving strategies differed. During calving, moose moved to lowest elevations of the year, whereas elk moved back to higher elevations. Moose generally selected for mid-elevations and against steep slopes; for Stunted spruce habitat in late winter; for Pine-spruce in summer; and for Subalpine during fall and winter. Most recorded moose locations were in Pine-spruce during late winter, calv- ing, and summer, and in Subalpine during fall and winter. Elk selected for mid-elevations except in summer and for steep slopes in late winter. Use and selection of 3 habitat classes were prominent for elk: Deciduous and Elymus burns, and Subalpine. Highest overlap between moose and elk occurred during fall and winter when both species used and strongly selected for Subalpine habitat. Neither elk nor moose selected areas to minimize the risk of wolf predation, but elk selected areas with lower risk of predation by grizzly bears and higher vegetation quality during calving and summer. ALCES VOL. 44: 41-63 (2008) Key words: Alces alces, Cervus elaphus, elevation, habitat selection, habitat use, home range, indi- vidual variation, movement rates, resource selection Moose (Alces alces) and elk (Cervus ela- phus) often provide the majority of prey bio- mass for large predators in complex predator- prey systems of North America. As examples, elk support wolf (Canis lupus) populations in Yellowstone and Yukon (Hayes and Harestad 2000, Smith et al. 2003), moose are common prey for wolves in northern coniferous forests (e.g., Post et al. 2002, Vucetich et al. 2002), and both moose and elk provide a prey base for wolves and grizzly bears (Ursus arctos) in northern British Columbia (Bergerud et. al. 1983, Bergerud and Elliott 1998, Parker and Milakovic 2007). Moose and elk are relatively profitable prey types in comparison to smaller ungulates or alternative prey, and given sufficient densities, can sustain large predator populations. Hence, moose and elk are keystone species in the functioning of large-scale large-mammal systems. In multi- ungulate, multi-predator systems, however, they are not studied as commonly as other species because they are less susceptible to disturbance than some species (e.g., Stone’s sheep, Ovis dalli stonei), use smaller areas and, therefore, are not as subjected to landscape disturbance as other species (e.g., woodland caribou, Rangifer tarandus caribou), or have large populations that are less vulnerable to, and can better accommodate change. In ad- dition, the requirements of moose and elk are assumed to be relatively well known (e.g., HABITAT SELECTION BY MOOSE AND ELK – GILLINGHAM AND PARKER ALCES VOL. 44, 2008 42 Franzman and Schwartz 1998, Toweill and Thomas 2002). Because moose and elk are highly visible species with strong social and ecological values, and have the benefit of be- ing high-profile game species, they are often managed to maintain or increase numbers. Yet there are relatively few published studies that have examined the concurrent resource use by these 2 species (e.g., Jenkins and Wright 1988). Moose have long occurred in northern British Columbia (Kelsall 1987), but elk herds are expanding into new areas in response to habitat fragmentation and management, and in some cases translocations (Spalding 1992). Prescribed burning has traditionally been used to create and maintain elk habitat in portions of northeastern British Columbia (Peck and Peek 1991). Fires temporarily result in shrub- and herb-dominated commu- nities and increases in forage biomass, often with higher nutritional value. Burning and its impacts on vegetative communities have been linked to the increase and expansion of elk herds (e.g., Luckhurst 1973, Silver 1976, Parminter 1983). Following fire, elk winter primarily on younger post-burn vegetation dominated by grasses or shrubs, except during severe winter conditions when there is higher use of conifer stands (Peck and Peek 1991). Moose also frequent fire-associated habitats (Peek 1998), but their use of burned habitats can depend on their past exposure to burned areas (Gasaway et al. 1989). Throughout most of their range, moose are primarily browsers and are associated with habitats containing a high proportion of preferred shrubs (Boer 1998). Elk on the other hand, are classified as grazers (Cook 2002, Stewart et al. 2002), but their food selection can shift in response to food availability and they are successful using browsing strategies (Houston 1982). Because of their flexible foraging habits, ability to use a wide variety of terrain types and high fecundity, elk could compete with moose for food (Flook 1964) although competition between the 2 species is thought to be low (Miller 2002). In complex predator-prey systems, high numbers of high-biomass ungulates may alter predator populations, and in turn other species in the same system. The overall goal of this study was to provide an initial description of habitat use and selection by moose and an expanding elk population in the Besa-Prophet area of northern British Columbia. Specifi- cally, we asked whether there was overlap in use and selection by moose and elk that may have implications to the system, and if there might be potential impacts through predation on other species. To do this we used global positioning satellite (GPS) radio-locations, remote-sensing imagery of vegetative com- munities, assessments of predation risk from concurrent studies on grizzly bears and wolves in the same area, and habitat selection model- ing. These data and analyses help characterize the ungulate-predator landscape of the Besa- Prophet watershed and contribute to better land-use planning. STUDY AREA The Muskwa-Kechika Management Area (MKMA) in northern British Columbia is known for its abundance of large ungulates (moose, elk, caribou, Stone’s sheep, a few mountain goats (Oreamnos americanus) and deer (Odocoileus spp.)) and large predators (wolves, grizzly bears, black bears (U. ameri- canus), coyotes (Canis latrans), wolverines (Gulo gulo), and a few cougars (Puma concol- or)). The Besa-Prophet study area is within the Besa-Prophet Pre-tenure Planning area (Fig. 1), one of several pre-tenure areas within the MKMA requiring specific wildlife planning prior to resource extraction or development. It is a highly diverse area including the Besa River south of the Prophet River, and cover- ing ~740,887 ha. Located within the Muskwa Ranges and Rocky Mountain Foothills be- tween 57°11’ and 57°15’ N, and 121°51’ and 124°31’ W, the Besa-Prophet is characterized ALCES VOL. 44, 2008 GILLINGHAM AND PARKER – HABITAT SELECTION BY MOOSE AND ELK 43 by numerous east-west drainages and south- facing slopes (Fig. 1) that provide benefits to wintering species because they are often blown free of deep snows. Other than several permanent outfitter camps and 1 government- designated all-terrain vehicle trail, there is relatively little access into the Besa-Prophet. This activity occurs mostly during late summer and fall during hunting seasons, with some snowmobile use in winter. Valleys at ~800-1300 m are commonly lined with white spruce (Picea glauca), some lodgepole pine (Pinus contorta) and trembling aspen (Populus tremuloides) on dry sites, and black spruce (P. mariana), willow-birch (Salix spp., Betula glandulosa) communities on poorly drained sites (Meidinger and Pojar 1991). There also are slopes that have been burned by the British Columbia Ministry of Environment and local guide outfitters to en- hance ungulate populations, especially Stone’s sheep. The subalpine area is characterized by an abundance of willow and scrub birch, as well as some balsam fir (Abies lasiocarpa) and white spruce often in krummholz form, and various grasses, sedges and fescues (Festuca spp.). Treeline occurs between ~1,450-1,600 m. Alpine tundra above ~1600 m consists of permanent snowfields, rock, mat vegetation, and grasslands (Demarchi 1996). Anecdotal evidence in the Besa-Prophet suggests that elk populations are expanding, enabled by prescribed burns that are conducted primarily for the enhancement of Stone’s sheep populations. In this area, prescribed fire has been officially managed since the early 1980s, although there also have been natural burns and locally initiated fires before and since that time. Concerns regarding the implica- tions of a rapidly increasing elk population on ecosystem dynamics were the impetus for the comparisons in our study. Densities of moose and elk populations in the Besa-Prophet are not well documented; very rough estimates are approximately 2000 moose and 500 elk (J.P. Elliott, BC Ministry of Environment, Fort St John, personal communication). It is important to note, however, that habitat use and selection will vary as a consequence of population density (Boyce et al. 2003). METhoDS Fifteen adult female moose and 14 adult female elk were fitted with GPS (global po- sitioning satellite) collars (GTX, Advanced Telemetry Systems, Isanti, MN) between March 2003 and January 2005. Collars were programmed to record locations 4 times daily at 6-h intervals with a range of start times between midnight and 0500 hr. Data were retrieved when collars were recovered at the end of a 1-year sampling period. We defined 5 seasons distinguished by biological and ecological characteristics for our analyses of range use and movements, and habitat use and selection: winter (1 November – 28 February) corresponding with the formation Fig. 1. Location of the study area within the Besa-Prophet Pre-tenure Planning area (inset) in northeastern British Columbia. Contour lines (200-m intervals) illustrate the predominance of east-west valleys within the study area. HABITAT SELECTION BY MOOSE AND ELK – GILLINGHAM AND PARKER ALCES VOL. 44, 2008 44 of sex-specific groups following rut; late winter (1 March – 15 May) when movement rates were lowest; calving (16 May – 15 June) during which parturient females became solitary and the onset of plant greening occurred; summer (16 June – 15 August) from plant green-up through peak vegetation biomass to the start of plant senescence; and fall (16 August – 31 October) when senescence of vegetation oc- curred, males and females formed mixed sex groups, and females came into estrus. Seasonal Movements and home Ranges To identify seasonal movement rates of moose and elk, we averaged movement rates (m/h) per individual using GPS locations ob- tained from consecutive 6-h fixes by season, and then averaged across individuals for each season by species (n = number of individu- als). We compared movement rates between moose and elk within seasons using a repeated measures analysis of variance (ANOVA), with differences identified following Bonferroni adjustment of confidence intervals. We used the same approach to compare seasonal eleva- tions used by both species. We determined sizes of annual and sea- sonal home ranges using 100% minimum convex polygons (MCP, Jennrich and Turner 1969) around GPS locations, as well as by fixed-kernel analysis (Worton 1989) for each individual. The MCPs, calculated by connect- ing the outer locations of all animal-use points, tend to overestimate range sizes for animals that have infrequent movements away from a centralized area, but they provide a relative comparison of the extents of the landscape used by moose and elk. In fixed-kernel analyses, kernels are calculated from the 95% prob- ability density of all locations and delineate areas of higher use (core areas) within the home range. Depending on the arrangement of animal locations, the fixed-kernel analysis may identify multiple core areas. Hereafter, we use the term ‘kernel area’ to refer to the total area identified by the fixed-kernel analysis. Because kernels are mathematical interpola- tions, however, they may exclude some areas where movements take place between core areas, and may include substantial ‘buffer’ areas with no animal locations around high- density locations, particularly with small numbers of locations (Seaman et al. 1999). For comparison with other studies on moose and elk, however, we used the Animal Move- ment Extension (Hooge and Eichenlaub 2000) in ArcView (ESRI 2002) to calculate both MCP and fixed-kernel range sizes. We used ANOVA to compare annual and seasonal range sizes between species within season. We used SpatialViewer (M. Gillingham, unpublished Visual Basic program) to examine movement patterns of individual animals. habitat Use and Availability To compare seasonal use of habitat classes between moose and elk, we determined the proportion of GPS locations within each class by individual within season. To index resources available to each collared individual, we defined availability at the scale of seasonal movement, within Johnson’s (1980) third order of selection. Seasonal movement is an animal’s movement potential within a season (e.g., Gus- tine et al. 2006b), or a circle around each use point with a radius equivalent to the distance traveled at each individual’s 95th percentile movement rate from 6-h GPS fixes. Within that area of movement potential, we selected 5 random points for availability locations. We used a raster geographic information system (GIS; Imageworks XPACE; PCI Geomatics Corp. 2001) to query habitat classes for used and available points. We ensured that no 2 points were used twice and that there was no overlap between used and available points (Manly et al. 2002). For both moose and elk within season, we then averaged the propor- tions of habitat classes that were used by, and available to, each individual to eliminate effects of uneven sample sizes among indi- viduals (SE was based on number of collared ALCES VOL. 44, 2008 GILLINGHAM AND PARKER – HABITAT SELECTION BY MOOSE AND ELK 45 individuals). We visually compared use to availability of different habitat classes, but then determined resource selection for combina- tions of additional variables because habitat use occurs in response to multiple variables and not to vegetation class alone. We defined 10 habitat classes for the Besa- Prophet based on a vegetation classification system with 25-m resolution for the area, developed using Landsat TM and Enhanced Thematic Mapper (ETM) remote-sensing im- agery (Lay 2005) (Table 1). Two burn classes (Elymus burn and Deciduous burn) may have included small, but unknown amounts of other disturbed areas such as avalanche chutes, which could not be distinguished separately with remote-sensing imagery. Avalanche chutes were relatively rare in the areas used by moose and elk. Resource Selection We used a suite of GIS layers to extract the attributes for defining resource selection by moose and elk by season, using all GPS use locations and available locations as de- fined above. These layers included habitat class, vegetation biomass, vegetation quality, slope, aspect, elevation, and risk of predation by grizzly bears and wolves. habitat class and vegetation indices – In addition to defining habitat class as we did in analyses for habitat use and availability, we used the same TM (4 June and 22 July 2001) and ETM (15 August 2001) imagery to index vegetation biomass during June, July, and Au- gust using a normalized difference vegetation index (NDVI) that is related to leaf area and plant biomass (Tucker and Sellers 1986, Ruimy et al. 1994). We also developed an index to vegetation quality during the calving and sum- mer seasons by calculating the change in NDVI (subtraction of individual pixel values) between June and July, and July and August images. A positive change in NDVI during the growing season corresponds with growth of new tissue (Groten and Ocatre 2002) and highest rates of green-up are likely the most digestible forage (Griffith et al. 2002, Oindo 2002). Habitat Class Description Non-vegetated Rock and rock habitats; permanent snowfields or glaciers and water bodies. Elymus burn Recently burned and open disturbed sites dominated by Elymus innovatus. Deciduous burn Older burned and disturbed areas containing Populus tremuloides and Populus balsamifera shrubs (<2 m) and trees (≥2 m). Subalpine Deciduous shrubs ≥1600 m in elevation; and spruce-shrub transition zone at middle to upper elevations with white and hybrid spruce (Picea glauca and P. glauca x engelmanni), and dominated by birch and willow. Stunted spruce Low productivity sites often on north-facing slopes with Picea glauca of limited tree height and percent cover. Pine-spruce White and hybrid spruce-dominated communities; and lodgepole pine dominated communities. Riparian Low-elevation, wet areas with black (Picea mariana) and hybrid spruce; often with standing water in spring and summer; exposed gravel bars adjacent to rivers and creeks. Alpine Dry alpine tundra habitat ≥1600 m characterized by Dryas spp.; and wet alpine tundra habitat ≥1600 m dominated by Cassiope spp. and sedge (Carex spp.) meadows. Low shrub Deciduous shrubs <1600 m dominated by birch and willow. Carex Wetland meadows dominated by sedges (Carex spp.) at elevations <1600 m, with intermittent Salix shrubs. Table 1. Description of the 10 habitat classes, derived from Landsat TM and Enhanced Thematic Mapper remote-sensing imagery, used to describe habitat use and selection by moose and elk in the Besa-Prophet area of northern British Columbia. HABITAT SELECTION BY MOOSE AND ELK – GILLINGHAM AND PARKER ALCES VOL. 44, 2008 46 Slope, aspect and elevation – We ob- tained elevation, slope, and aspect from a digital elevation model (DEM) in the 1:20,000 British Columbia Terrain and Resource Inven- tory Management program (British Columbia Ministry of Crown Lands 1990). We modeled aspect as 2 continuous variables (i.e., north- ness and eastness; Roberts 1986) to avoid introducing additional categorical variables. Northness (the cosine of aspect) values of 1.00 and -1.00 suggest selection for north and south aspects, respectively, whereas values near 0.00 suggest selection for east and west aspects. Eastness (the sine of aspect) values show selection for east (i.e., 1.00) and west (i.e., -1.00) aspects; values of 0.00 show selec- tion for northern/southern exposures (Palmer 1993). We did not assign an aspect to pixels with a slope ≤1°. Predation risk – We defined potential risk of predation using logistic regression models by season from GPS-collared wolves and grizzly bears, which are assumed to be the most significant large mammal predators in the Besa-Prophet area (Bergerud and Elliott 1998). Details of these predator models are in Gustine et al. (2006a, b) and Walker et al. (2007). The predation-risk models included slope, aspect, elevation, habitat class, frag- mentation (an index of vegetation diversity), and distance to linear features (e.g., seismic lines). From these models, we generated a risk surface as a GIS layer that defined selection value to grizzly bears or wolves in each season by applying the coefficients from models to each 25 x 25-m pixel, based on its topographic and vegetation features. We scaled values from 0 to 1 to standardize selection surfaces, and then assumed that the risk of predation to moose and elk by grizzly bears and wolves was directly related to selection values of the predators. Model Calving Summer Fall Winter Late Winter Elevation1+Aspect+Habitat2      Elevation+Slope+Aspect+Habitat      Wolf 3+Habitat      Elevation+Slope+Aspect+Wolf+Bear3+Biomass+Habitat    Elevation+Slope+Aspect+Wolf+Bear+Quality+Habitat   Aspect+Wolf+Bear+Biomass+Habitat    Aspect+Wolf+Bear+Quality+Habitat   Elevation+Slope+Aspect+Wolf+Biomass+Habitat    Elevation+Slope+Aspect+Wolf+Quality+Habitat   Aspect+Wolf+Biomass+Habitat    Aspect+Wolf+Quality+Habitat   Table 2. Candidate models developed a priori to describe resource selection by moose and elk by season in the Besa-Prophet area of northern British Columbia. Vegetation biomass during calving, summer, and fall was based on NDVI measures for June, July, and August, respectively. Vegetation quality, assessed by the change in NDVI between summer months, was only used in calving and summer models. No risk of predation by grizzly bears was included during hibernation (winter and late winter seasons). 1 Elevation was modeled as a quadratic with both a linear and squared term. 2 Habitat as defined by classes in Table 1. 3 Wolf and Bear represent risk of predation by wolves and grizzly bears, respectively; see text for details. ALCES VOL. 44, 2008 GILLINGHAM AND PARKER – HABITAT SELECTION BY MOOSE AND ELK 47 Selection models – We developed 11 a priori, ecologically plausible models (Table 2) from the previously described attributes to define resource selection by moose and elk by season. We used logistic regression with these parameters (K) to characterize differences be- tween use and availability for each individual, and ranked the suite of models using Akaike’s Information Criterion (AIC) values corrected for small sample size (AICc) when n/K < 40 (Burnham and Anderson 2002). Deviation coding was used for categorical variables (Hendrickx 1999). To avoid inflated coeffi- cients and inflated error terms in the models, we used tolerance scores to assess model inputs for collinearity and multicollinearity. If tolerance scores were <0.20, covariates were not included in the same model (Menard 2002). Because logistic regression models do not provide reliable estimates if there is complete or near-complete separation in levels of categorical variables (Menard 2002), we dropped both used and available points for those habitats in which either use or available points in a habitat were ≤4 (Gillingham and Parker 2008). We validated the top models using k- fold, cross-validation (Boyce et al. 2002), and an averaged Spearman’s rank correlation coefficient. Within each model set (i.e., by individual and season), we calculated Akaike weights (wi). If there was not a single model for which this relative weight of evidence, wi, was ≥0.95, we considered competing models to be those for which the sum of wi was ≥0.95 (Burnham and Anderson 2002). For each model set for each individual animal, we averaged the selection coefficients (β) in com- peting models based on their relative wi. We evaluated the importance of specific resources to moose and elk in general after developing pooled models from these averaged individual models by averaging coefficients from the individual models within species during each of the 5 seasons. We assumed significance of all tests at α = 0.05. We used STATA for all statistical and modeling procedures (version 9.2; StataCorp 2007). All means are presented as X ± 1 SE unless noted otherwise. RESUlTS We retrieved 14,534 GPS locations from 14 of the collared moose and 14,870 loca- tions from 13 collared elk. The number of GPS locations recorded as a percentage of the number of attempted GPS locations was 76.7 ± 0.03% (X ± SE) for moose and 82 ± 1.6% for elk. Seasonal Movements and home Ranges Distances moved between consecutive 6-h GPS fixes ranged from <1 m to 14.5 km (straight-line distance) by moose and from <1 m to 10.3 km by elk. Both species moved at lowest rates during winter and late winter (35-41 m/h), and then increased movements to highest rates in summer (>100 m/h) (Fig. 2). Moose usually tended to move at rates slightly lower than elk (repeated-measures ANOVA, P = 0.049), but these rates were sig- nificantly lower only during the calving season (moose = 59 ± 21 m/h, elk = 93 ± 27 m/h). The range sizes estimated by fixed-kernel analyses were always smaller, as expected, than those determined by MCP for both Fig. 2. Comparative differences in the movement rates (m/hr, X ± SE) of adult female moose (n = 14) and elk (n = 13), averaged by individual and then averaged across individuals, by season in the Besa-Prophet area of northern British Columbia. SeasonL ate W inte r Ca lvin g Su mm er Fal l Wi nte r M ov em en t R at e (m /h ) 0 20 40 60 80 100 120 140 Moose Elk HABITAT SELECTION BY MOOSE AND ELK – GILLINGHAM AND PARKER ALCES VOL. 44, 2008 48 species (Table 3). Kernel areas were most comparable to MCPs (~85-90 % of MCP size) during the calving season for moose, and during calving and fall for elk. They were less than half the size of MCPs during fall for moose, and during winter and late winter for elk. Annual home ranges by MCP for moose averaged 195 km2, but were highly variable among individuals, ranging from a minimum of 39 to a maximum of 899 km2 (kernel area = 14-124 km2). Seasonal ranges for moose were typically smallest during the calving season (18 km2), and more than 7 times larger during summer. Annual ranges of MCP for elk averaged 191 km2 and were not different than those of moose (Table 3), and also were highly variable among individuals (MCP range = 50-1000 km2; kernel area = 10- 107 km2). Excluding 1 animal that made a large circular excursion in July away from its other seasonal use areas, annual ranges for elk aver- aged 123 ± 19 km2 (range = 50-250 km2; kernel area = 10-107 km2). In contrast to moose, seasonal ranges for elk were smallest during the winter and late winter seasons (16-20 km2). Similar to moose, seasonal ranges for elk were largest during summer (Table 3). Compared to elk, the seasonal ranges of moose (by MCP) were more than twice as large during winter and late winter, but less than half as large during calving. habitat Use and Availability Moose and elk used elevations on the landscape differently among seasons (repeat- ed-measures ANOVA, P < 0.001) (Fig. 3). During calving and summer (May–August) and winter (November–February), moose used lower elevations than elk (moose = 1333 ± 81 m in calving, 1397 ± 53 in summer, 1519 ± 88 in winter; elk = 1551 ± 49 m in calving, 1671 ± 40 m in summer, 1624 ± 51 m in winter) (all P <0.016). Both species moved to lower elevations from winter to late winter, however, calving strategies appeared to differ between the 2 species. In June, moose were at lowest elevations of the year, and after the calving season moved gradually upslope dur- Season Estimate Moose1 Elk P Mean SE Mean SE Annual MCP 195.28 59.71 190.81 69.61 0.961 Kernel 57.04 8.81 45.64 8.19 0.355 Calving MCP 17.59 4.09 38.00 4.05 0.002 Kernel 15.05 4.28 34.09 5.82 0.013 Summer MCP 132.80 60.25 118.38 63.66 0.871 Kernel 99.42 52.81 57.76 28.70 0.504 Fall MCP 46.85 10.96 60.22 12.86 0.434 Kernel 26.86 6.69 51.02 13.16 0.107 Winter MCP 45.91 9.83 20.40 2.24 0.015 Kernel 32.16 8.88 8.82 1.14 0.012 Late Winter MCP 30.49 4.60 15.52 4.60 0.030 Kernel 25.49 6.39 6.04 1.60 0.009 Table 3. Comparison of annual and seasonal home-range sizes (km2) for 14 female moose and 13 female elk based on 100% minimum convex polygons (MCP) and 95% fixed-kernel (Kernel) es- timates. P-values are from one-way ANOVAs comparing home-range sizes between species for a given technique and season. 1 Only 12 animals were used in the winter models because of collar failure. ALCES VOL. 44, 2008 GILLINGHAM AND PARKER – HABITAT SELECTION BY MOOSE AND ELK 49 ing summer and fall. Elk, in contrast, moved from their late wintering areas to higher eleva- tions to calve in June, continuing upslope in July, and descending to the same elevations used by moose in the fall (Fig. 3). Predominant use of specific habitat classes differed between moose and elk, coinciding with some of the elevational differences be- tween the species. During late winter, calving, and summer, most locations for moose (28- 36% across animals) were in the Pine-spruce habitat class (Fig. 4). This contrasts to the fall and winter periods, when most locations were in Subalpine vegetation (33-39%). There was relatively little use by moose during any season of Riparian (<6 % of locations; distinct from Low shrub and Carex), Alpine (<4%), or Non-vegetated (<1%) habitat classes. Low shrub vegetation was used least (10 %) by moose during the calving season, and most during winter (22%). Use of the Deciduous burn class by moose was relatively consistent at 13-16% throughout the year. For elk, the use of 3 habitat classes was prominent: Elymus burn, Deciduous burn, and Subalpine (Fig. 5). Seasonally, the 3 classes always totaled between 59 and 83% of use locations. Highest use by elk occurred in the Subalpine in all seasons (33.0% of locations during calving; 29.5% during fall, and 40.5% in winter) except late winter (13.1% of loca- tions), when they increased use of both burned habitat classes (~70% of locations). During summer when elk used the Subalpine more than any other season (64% of locations), they spent less time in Elymus and Deciduous burn habitats. Compared to moose, elk used the Pine-spruce habitat class very little (<5% of locations) in any season except fall (17 ± 2%). Resource Selection There were relatively few variables that were selected for or against by moose because of variation in individual selection strategies (Table 4). In general, moose selected for mid-elevations in all seasons but late winter. During calving, summer, and fall they avoided Non-vegetated areas and steep slopes. In the calving season, most moose rarely used (<4 locations) Subalpine (10/14 animals), Alpine (13/14), or Carex (12/14) habitat classes even though avoidance of these classes was not in- dicated by the pooled calving model. Moose selected strongly for Pine-spruce habitat in summer, Subalpine and Low shrub in fall, Subalpine in winter, and Stunted spruce in late winter. In no seasons were vegetation biomass and quality important to overall resource selection (although some individual moose selected for or against these variables). Risk of predation was an important factor for some moose, but was only important in the pooled selection models during fall, when moose locations were in areas of relatively higher predation risk than that present in the area around them. Elk exhibited selection by season for and against more variables than did moose (Table 5). Mid-elevations were selected in all seasons but summer, when higher elevations were more important. Elk selected against steep slopes from summer through winter, but for them in late winter. They consistently selected against northness and Alpine habitat in all seasons, Fig. 3. Comparative differences in elevations (X ± SE) used by GPS-collared female moose (n = 14) and elk (n = 13), averaged by indi- vidual and then averaged across individuals, by season in the Besa-Prophet area, northern British Columbia. SeasonL ate W inte r Ca lvin g Su mm er Fal l Wi nte r E le va tio n (m ) 0 1200 1400 1600 1800 Elk Moose HABITAT SELECTION BY MOOSE AND ELK – GILLINGHAM AND PARKER ALCES VOL. 44, 2008 50 and Non-vegetated habitat in all seasons but calving. During calving, elk selected Elymus burns and Stunted spruce (although <2% of locations across animals were in this class) and avoided the Subalpine habitat class (even though it comprised 32 ± 5% of use locations). In summer, they selected for Subalpine as well as Stunted spruce. In both calving and sum- mer, elk selected for areas with high vegetation quality (as measured by change in NDVI) even though the risk of wolf predation was relatively high in those areas, and against the risk of bear predation (although coefficients were very small; Table 5) and Low shrub habitat. From calving through fall, elk selected against areas with high vegetation biomass (as indexed by NDVI). From fall through late winter, there was strong selection for Elymus Fig. 4. Comparison of proportional use versus availability (+ SE) of habitat classes for female moose in the Besa-Prophet area of northern Brit- ish Columbia. Standard errors were determined from averages for each individual by season. P ro po rt io n Summer P ro po rt io n 0.0 0.1 0.2 0.3 0.4 0.5 Winter Late Winter 0.0 0.1 0.2 0.3 0.4 0.5 No n- ve ge ta te d El ym us b ur n De cid uo us b ur n Su ba lp in e St un te d sp ru ce Pi ne s pr uc e Ri pa ria n Al pi ne Lo w sh ru b Ca re x P ro po rt io n 0.0 0.1 0.2 0.3 0.4 0.5 No n- ve ge ta te d El ym us b ur n De cid uo us b ur n Su ba lp in e St un te d sp ru ce Pi ne s pr uc e Ri pa ria n Al pi ne Lo w sh ru b Ca re x No n- ve ge ta te d El ym us b ur n De cid uo us b ur n Su ba lp in e St un te d sp ru ce Pi ne s pr uc e Ri pa ria n Al pi ne Lo w sh ru b Ca re x Calving 0.0 0.1 0.2 0.3 0.4 0.5 P ro po rt io n No n- ve ge ta te d El ym us b ur n De cid uo us b ur n Su ba lp in e St un te d sp ru ce Pi ne s pr uc e Ri pa ria n Al pi ne Lo w sh ru b Ca re x P ro po rt io n Fall 0.0 0.1 0.2 0.3 0.4 0.5 Used Available No n- ve ge ta te d El ym us b ur n De cid uo us b ur n Su ba lp in e St un te d sp ru ce Pi ne s pr uc e Ri pa ria n Al pi ne Lo w sh ru b Ca re x ALCES VOL. 44, 2008 GILLINGHAM AND PARKER – HABITAT SELECTION BY MOOSE AND ELK 51 Fig. 5. Comparison of proportional use versus availability (+ SE) of habitat classes for female elk in the Besa-Prophet area of northern British Columbia. Standard errors were determined from averages for each individual by season. Late Winter 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Summer 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Winter 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 P ro po rt io n No n- ve ge ta te d El ym us b ur n De cid uo us b ur n Su ba lp in e St un te d sp ru ce Pi ne s pr uc e Ri pa ria n Al pi ne Lo w sh ru b Ca re x P ro po rt io n P ro po rt io n No n- ve ge ta te d El ym us b ur n De cid uo us b ur n Su ba lp in e St un te d sp ru ce Pi ne s pr uc e Ri pa ria n Al pi ne Lo w sh ru b Ca re x No n- ve ge ta te d El ym us b ur n De cid uo us b ur n Su ba lp in e St un te d sp ru ce Pi ne s pr uc e Ri pa ria n Al pi ne Lo w sh ru b Ca re x Calving 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Fall 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Used Available P ro po rt io n P ro po rt io n No n- ve ge ta te d El ym us b ur n De cid uo us b ur n Su ba lp in e St un te d sp ru ce Pi ne s pr uc e Ri pa ria n Al pi ne Lo w sh ru b Ca re x No n- ve ge ta te d El ym us b ur n De cid uo us b ur n Su ba lp in e St un te d sp ru ce Pi ne s pr uc e Ri pa ria n Al pi ne Lo w sh ru b Ca re x and Deciduous burns and Subalpine habitat. The pooled selection models indicated that the Carex habitat class was selected by elk in all seasons, but <1% of used locations were in this class and the majority of individuals (n = 7-11, depending on season) rarely or never used the class. HABITAT SELECTION BY MOOSE AND ELK – GILLINGHAM AND PARKER ALCES VOL. 44, 2008 52 Parameter Calving Summer Coef SE P + - Coef SE P + - Elevation 78.94 40.01 0.048 7 1 40.74 15.75 0.010 10 0 Elevation2 -30.0 15.19 0.048 1 7 -14.4 5.59 0.010 0 10 Slope -0.05 0.02 0.029 1 6 -0.04 0.01 0.007 1 8 Northness <0.01 0.18 0.994 0 1 -0.06 0.14 0.672 0 1 Eastness 0.01 0.19 0.962 1 0 -0.01 0.13 0.968 0 0 Wolf Risk -1.91 2.11 0.365 1 2 0.46 0.97 0.639 2 0 Bear Risk 0.06 1.30 0.962 1 1 0.48 0.99 0.624 0 0 Biomass -0.49 0.88 0.574 0 1 -1.27 1.00 0.206 1 3 Quality 1.28 1.61 0.428 2 1 0.78 0.77 0.308 3 0 Nonveg -1.22 0.15 <0.001 0 1 -2.47 0.50 <0.001 0 4 Elymus burn 0.25 0.37 0.487 1 1 0.23 0.51 0.648 2 1 Deciduous burn 0.24 0.45 0.594 3 1 0.62 0.35 0.077 5 0 Subalpine -0.17 0.28 0.554 0 0 0.59 0.50 0.234 5 1 Stunt 0.07 0.39 0.854 1 0 0.57 0.41 0.167 5 0 Pine/Spruce -0.19 0.51 0.711 2 2 0.80 0.33 0.016 5 0 Riparian -0.45 0.53 0.398 1 1 0.50 0.51 0.328 3 1 Alpine 0.18 0.09 0.055 0 0 -0.13 0.59 0.831 0 0 Low Shrub 0.51 0.51 0.321 1 0 0.08 0.40 0.834 3 4 Carex 0.11 0.14 0.434 0 0 -0.80 0.37 0.032 1 1 Parameter Fall Winter1 Late Winter Coef SE P + - Coef SE P + - Coef SE P + - Elevation 54.99 23.93 0.022 7 0 37.65 13.67 0.006 6 2 58.26 33.70 0.084 7 0 Elevation2 -18.6 8.15 0.023 0 7 -12.3 4.41 0.005 1 7 -21.7 12.84 0.092 0 7 Slope -0.03 0.01 0.005 0 8 -0.01 0.01 0.235 2 4 -0.01 0.01 0.305 2 4 Northness 0.01 0.12 0.946 1 0 <0.01 0.08 0.987 0 1 0.03 0.10 0.748 0 0 Eastness -0.02 0.13 0.879 0 1 -0.02 0.08 0.838 0 0 -0.02 0.11 0.836 0 1 Wolf Risk 0.94 0.44 0.033 4 0 0.47 0.62 0.443 2 0 -0.21 0.76 0.787 1 2 Bear Risk -0.06 1.35 0.966 2 2 Biomass -1.16 1.25 0.351 1 5 Quality <0.01 0 0 Nonveg -1.39 0.51 0.006 1 3 -0.74 0.46 0.107 0 4 -0.46 0.47 0.330 0 2 Elymus burn 0.30 0.37 0.413 4 0 -0.10 0.34 0.779 1 1 0.00 0.36 0.989 2 0 Deciduous burn 0.65 0.36 0.068 8 0 0.35 0.18 0.055 6 1 0.29 0.24 0.220 2 0 Subalpine 1.12 0.30 <0.001 11 0 0.53 0.23 0.021 9 0 0.09 0.27 0.738 2 0 Stunt -0.36 0.44 0.404 0 1 0.06 0.23 0.782 1 0 0.42 0.22 0.050 6 1 Pine/Spruce 0.01 0.32 0.966 1 3 -0.22 0.20 0.275 2 6 0.08 0.22 0.733 2 1 Riparian -0.30 0.38 0.430 0 2 0.08 0.26 0.774 1 0 -0.09 0.41 0.822 1 0 Alpine -0.94 0.67 0.157 0 5 -0.40 0.40 0.310 0 4 -0.27 0.33 0.414 0 1 Low Shrub 0.87 0.35 0.012 9 0 0.44 0.23 0.054 6 0 -0.06 0.37 0.882 2 0 Carex 0.03 0.21 0.888 0 0 -0.01 0.25 0.970 1 0 -0.01 0.20 0.979 0 0 Table 4: Comparison of significant selection coefficients by season from averaged resource selection models for 14 female moose in the Besa-Prophet area of northern British Colum- bia. For each season, the number under the + indicates the number of individual final models that showed selection for that parameter; the num- ber under the – indicates the number of individuals that avoided that attri- bute. Coefficients, SE, and P-values refer to the pooled models. 1 Only 12 animals were used in the winter models because of collar failure. ALCES VOL. 44, 2008 GILLINGHAM AND PARKER – HABITAT SELECTION BY MOOSE AND ELK 53 Parameter Calving Summer Coef SE P + - Coef SE P + - Elevation 37.31 4.82 <0.001 8 7.81 3.65 0.032 7 3 Elevation2 -11.9 1.55 <0.001 1 9 -2.02 1.09 0.065 3 7 Slope 0.00 0.01 0.350 4 5 -0.02 0.00 <0.001 2 9 Northness -0.31 0.07 <0.001 4 8 0.20 0.05 <0.001 8 3 Eastness -1.74 0.12 <0.001 13 -0.71 0.06 <0.001 12 Wolf Risk 0.98 0.48 0.042 6 2 1.22 0.35 <0.001 6 4 BearRisk -0.00 0.00 <0.001 10 2 -0.00 0.00 <0.001 4 5 Biomass -2.68 0.30 <0.001 1 7 -1.51 0.13 <0.001 9 Quality 5.46 0.88 <0.001 10 1 5.62 0.67 <0.001 12 Nonveg 0.06 0.28 0.824 1 6 -0.91 0.29 0.002 2 6 Elymus burn 0.34 0.16 0.036 6 4 -0.09 0.20 0.660 3 4 Deciduous burn 0.14 0.17 0.415 8 3 0.25 0.19 0.200 6 3 Subalpine -0.63 0.21 0.003 4 7 0.39 0.20 0.045 6 2 Stunt 0.48 0.18 0.006 4 2 0.59 0.18 0.001 5 1 Pine/Spruce 0.05 0.23 0.814 3 6 0.11 0.24 0.655 6 4 Riparian 0.35 0.21 0.097 4 3 0.56 0.18 0.002 5 1 Alpine -0.84 0.19 <0.001 1 7 -1.25 0.18 <0.001 11 Low Shrub -1.02 0.24 <0.001 3 6 -0.89 0.23 <0.001 2 10 Carex 1.07 0.14 <0.001 4 1.23 0.17 <0.001 5 Parameter Fall Winter Late Winter Coef SE P + - Coef SE P + - Coef SE P + - Elevation 8.58 2.46 <0.001 7 4 11.50 2.63 <0.001 5 6 54.31 6.19 <0.001 12 1 Elevation2 -2.55 0.85 0.003 5 7 -2.49 0.83 0.003 6 5 -18.1 2.13 <0.001 1 12 Slope -0.02 0.00 <0.001 2 9 -0.04 0.00 <0.001 12 0.02 0.01 <0.001 8 3 Northness 0.02 0.04 0.649 6 5 0.06 0.03 0.059 6 4 -0.09 0.07 0.197 3 6 Eastness -0.45 0.06 <0.001 1 11 -0.81 0.05 <0.001 13 -1.90 0.16 <0.001 13 Wolf Risk -0.03 0.14 0.841 3 3 0.58 0.29 0.047 5 5 -0.22 0.40 0.587 3 6 BearRisk -0.28 0.44 0.526 4 6 Biomass -2.92 0.47 <0.001 2 11 Quality Nonveg -2.63 0.33 <0.001 12 -1.52 0.18 <0.001 12 -0.84 0.23 <0.001 1 7 Elymus burn 0.91 0.15 <0.001 12 1 0.24 0.08 0.005 8 2 0.45 0.12 <0.001 8 Deciduous burn 1.39 0.13 <0.001 13 0.37 0.07 <0.001 10 2 0.59 0.11 <0.001 11 2 Subalpine 1.01 0.10 <0.001 11 1 0.51 0.07 <0.001 8 1 0.34 0.16 0.034 6 2 Stunt -0.40 0.19 0.032 4 6 0.10 0.11 0.386 5 3 -0.69 0.29 0.018 7 Pine/Spruce 0.00 0.09 0.985 5 4 0.01 0.10 0.940 5 5 -0.66 0.11 <0.001 7 Riparian -0.52 0.17 0.002 4 6 0.23 0.18 0.183 6 2 0.26 0.16 0.111 4 1 Alpine -0.71 0.14 <0.001 1 9 -0.30 0.09 0.001 2 7 -0.32 0.15 0.040 2 6 Low Shrub 0.51 0.14 <0.001 7 1 0.20 0.11 0.077 6 4 0.37 0.17 0.029 5 1 Carex 0.44 0.16 0.007 3 0.18 0.03 <0.001 2 0.50 0.11 <0.001 3 Table 5: Comparison of significant selection coef- ficients by season from averaged resource selec- tion models for 13 female elk in the Besa-Prophet area of northern British Columbia. For each sea- son, the number under the + indicates the number of individual final models that showed selection for that parameter; the number under the – indicates the number of individuals that avoided that attribute. Coefficients, SE, and P- values refer to the pooled models. HABITAT SELECTION BY MOOSE AND ELK – GILLINGHAM AND PARKER ALCES VOL. 44, 2008 54 DiScUSSion Resource partitioning by ungulates typi- cally occurs relative to habitat, dietary, and special niche requirements (Bowyer and Kie 2004). Both moose (Miquelle et al. 1992, Bowyer 2004) and elk (Peek and Lovaas 1968, Weckerly 1998) sexually segregate to a greater or lesser extent throughout their range. Consequently, we make inferences only to female moose and elk in the Besa-Prophet, with the recognition that our findings provide preliminary comparative insights based on only 1 year of data for each species. Further, although locations of animals were based on 4 fixes per day distributed throughout noctural and diurnal periods, these fixes that were 6 h apart do not allow us to examine habitat selection at finer feeding-patch scales. Seasonal Movements and home Ranges Home-range estimates are difficult to compare among studies because of differ- ences in methodologies (e.g., definitions of seasons), analyses (Lawson et al. 1997), and available resources. In this study, most fixed-kernel home ranges had multiple core areas and, particularly for moose, often did not include large portions of the valley bot- toms that were used regularly for moving back and forth between core areas within a season. Consequently, defining availability of resources within just the core areas would have greatly underestimated available habitat during a given season. We observed large seasonal variation in moose and elk home ranges whether we com- pared the extent of use (MCP) or kernel areas. Sizes of annual home ranges for moose in the Besa-Prophet probably reflect the relatively open, mountainous terrain of the study area and were more similar in size to MCP home ranges of female moose reported for the Ke- nai Peninsula of Alaska (Hundertmark 1998: 128 km2), south-central Alaska (Ballard et al. 1991: 290 km2), and the Mackenzie Valley of Northwest Territories (Stenhouse et al. 1994: 174 km2) than to the smaller ranges from more southern latitudes (Hundertmark 1998). In summer, when moose exhibited the longest average movements between 6-h fixes, the extent of their ranges (MCP) was also com- parable to the large home ranges reported for south-central Alaska (Hundertmark 1998). In winter, the seasonal ranges of moose were comparable in size to MCP home ranges at more similar latitudes (e.g., 42-47 km2 in north-central Alberta; Lynch and Morgantini 1984). Cederlund and Sand (1994) observed that female moose with calves had larger home ranges than those without calves. Although we do not know the reproductive status of the animals in our study, home ranges for moose during calving were the smallest of all seasons, an observation that would be consistent with an ungulate with a hiding reproductive strategy. Elk also hide their young, and their home ranges during calving were small, although they were not the smallest by season. Home-range size can be related to the abundance of important resources because a seasonal home range must meet an ani- mal’s energy and nutritional requirements (Anderson et al. 2005). For example, the home-range sizes of elk in summer and win- ter in Alberta and Wisconsin were inversely related to mean forage biomass; elk increased home-range sizes in winter when quality of forage was reduced and snow cover reduced forage biomass (Anderson et al. 2005). In our study, elk did not have larger range sizes in winter or late winter than summer. Rather, home ranges were smallest during winter and late winter. This may indicate that snow was more limiting in our system, but perhaps more likely that food was not limiting, particularly on the wind-blown south-facing slopes in the Besa-Prophet. Elk home ranges during summer are reported to be highly variable even within the same study area (Strohmeyer and Peek 1996: MCP = 79-593 km2) because of the juxtaposition of habitat components within ALCES VOL. 44, 2008 GILLINGHAM AND PARKER – HABITAT SELECTION BY MOOSE AND ELK 55 the area. Summer was also the season with the most variability in home ranges for elk in the Besa-Prophet. In our study, elk exhibited the highest 6-h movement rates during sum- mer (as much as 10 km in 6 h) and summer home ranges based on fixed-kernel estimates were comparable to those reported for large- herd, migratory elk in Yellowstone National Park (Boyce 1991). We recorded a very long movement of a minimum of 138 km (MCP area around those locations was ~800 km2) over 20 days in July by 1 animal that returned to its original starting point. Edge et al. (1986) found that only 2-3% of marked female elk dispersed and most other studies have reported strong philopatry by female elk for seasonal and annual home ranges (e.g., Craighead et al. 1972). Resource Selection Models Calculations of use, based on proportions of GPS locations, give some indication of areas that are most frequented, and therefore, important to managers. Selection of locations by animals on the landscape, however, is usu- ally a response to multiple variables and not simply to elevation or habitat alone. Selection models allow the quantification of tradeoffs that animals make, for example, in relation to predation risk or nutritional value. At the same time, selection models require that numerous assumptions be met (e.g., Thomas and Taylor 2006), sometimes resulting in the exclusion of habitats that are rarely used or completely avoided (Gillingham and Parker 2008). In our analyses, we presented seasonal selection models for moose and elk after pool- ing individual models instead of developing a seasonal global model across individuals. Rather than unequally weight individual ani- mals with more GPS locations, delete known observations from some animals to equalize sample sizes (Thomas and Taylor 2006), or use a random-effects approach to account for unequal sample sizes (Gillies et al. 2006), we chose to develop models for each individual in each season, dropping those habitats that had complete or near-complete separation (<4 used points). Further, our approach for considering which competing models were averaged was conservative (sum wi ≥ 0.95) (Burnham and Anderson 2002). We then de- termined a pooled model for each season by averaging the final individual models within each season. There are 2 different approaches that can be used when averaging competing models for which estimates of all coefficients are not found in each model (e.g., predation risk was not in all competing models; Burn- ham and Anderson 2002). We chose the more conservative approach of assuming that miss- ing coefficients had the value of 0 in a given model. Our approach ensures that the final models are not unduly weighted by individual animals (Burnham and Anderson 2002), but it has the effect of reducing the magnitude of coefficients that occur in only some models and likely results in models with fewer significant coefficients when compared to the alternative approach of simply combining all individual data to develop seasonal, global models. It is important to understand (but not over-emphasize) the influence of individual variation. In our study, other than elk selecting against eastness (i.e., for western exposures) during calving and for Deciduous burns in fall, all animals did not show the same selec- tion for any attribute on the landscape. Our approach incorporates into each seasonal pooled model any zero selection value by an animal for an attribute with a weighting equal to significant selection coefficients by other animals. As such, the final pooled models do not overemphasize the importance of attributes selected by one or a few individuals. We also recommend presenting selection analyses with observations of use to identify habitat classes (or other categorical variables) that are avoided or rarely used, and to assess the relative magnitude of use for highly selected classes. Findings in our study may be limited by bi- HABITAT SELECTION BY MOOSE AND ELK – GILLINGHAM AND PARKER ALCES VOL. 44, 2008 56 ases attributed to relatively low fix rates (Moen et al. 1996, Frair et al. 2004), particularly for moose. Most locations within each season (55-65%), however, were obtained from habitat classes with forest cover (Pine-spruce, Stunted spruce, Subalpine) that probably had the poorest signal attenuation. Therefore, al- though our interpretation may underestimate the magnitude of use and selection of those classes, the classes were nonetheless noted as important in our analyses despite potential fix bias. habitat Use and Selection Moose were often at lower elevations than elk on the Besa-Prophet landscape. Both species, however, moved down in elevation between winter and late winter. In other areas of British Columbia with high topographical diversity, the greatest single determinant of late winter habitat use by moose was decreasing elevation, which may be a surrogate for snow depth (Poole and Stuart-Smith 2005, 2006). Snow depth is a primary factor affecting late winter distribution of moose populations (Peek 1998), and moose in interior mountain- ous areas typically move to lower elevations throughout the winter (e.g., Pierce and Peek 1984, van Dyke et al. 1995). Elk also move to lower elevations during the winter (e.g., Unsworth et al. 1998, Boyce et al. 2003), possibly to take advantage of increased food availability. Because of the abundance of south-facing, wind-swept slopes in the Besa- Prophet, moose and elk may not be as affected by snow depth as they are in other areas. Moose frequently inhabit shrub commu- nities throughout their range whenever snow depths do not exceed 100 cm (Peek 1974), and select coniferous forests in regions of deeper snow, provided that browse is available (Bun- nell and Eastman 1976, Peek et al. 1982, Pierce and Peek 1984). In the mountainous interior regions of Alaska, shrub-dominated commu- nities above timberline are important moose habitat (Peek 1998) and those dominated by willow species appear to be the most important to moose (Risenhoover 1989). Moose in our study used Subalpine habitats most in fall and winter, and the Pine-spruce habitat class most during late winter, calving, and summer. Elk often occupy south-facing, seral brushfields (Irwin and Peek 1983) or wind- swept, grass-dominated slopes (Knight 1970; Houston 1982) during winter, except when deep or crusted snow causes them to seek timber (Houston 1982). Peck and Peek (1991) reported that elk in northeastern British Co- lumbia wintered primarily in post-fire grass and shrub communities, except during severe weather when conifer stands were used. Tim- bered habitats also have been reported to be important in other seasons. For example, elk in Idaho shifted from using a high proportion of shrub and open timber habitats in spring to using more timbered habitats in fall (Unsworth et al. 1998). The availability of forest cover may help reduce thermal stress and predation risk (Anderson et al. 2005). Habitats most used by elk in the Besa-Prophet were Decidu- ous and Elymus burns during late winter, and Subalpine during all other seasons. According to pooled selection models, when elk selected areas with relatively high wolf risk, they did not select significantly for forest cover to help minimize risk. In Yel- lowstone National Park, when wolf activity was centered around dens and rendezvous sites in summer, elk apparently avoided wolves by selecting higher elevations, less open habitat, burned forest, and, in areas of high wolf density, steeper slopes (Mao et al. 2005). Elk did not spatially separate themselves from wolves in winter, but relied on behavioral anti-predatory strategies such as grouping (Mao et al. 2005). Although we have no data on group sizes, anecdotally elk in the Besa-Prophet tended to group together more frequently than moose during summer through winter. Elk typically select habitats character- ized by early seral stage (Thomas et al. 1979, ALCES VOL. 44, 2008 GILLINGHAM AND PARKER – HABITAT SELECTION BY MOOSE AND ELK 57 Irwin and Peek 1983, Grover and Thompson 1986), which may be facilitated by burns. In the Besa-Prophet, the relative selection by elk for Deciduous and Elymus burns varied by season. Elk showed greater selection for the more open Elymus burns during calving, and for Deciduous burns during fall, winter, and late winter. Moose response to burns also is generally positive throughout their range, although Gasaway et al. (1989) observed that traditional movement patterns by moose ap- parently prevented animals without pre-fire use from finding burns. In the Besa-Prophet, moose did not specifically select for burned areas, but frequented the older Deciduous burns as a small, but consistent part of their habitat use. Potentially related to predation risk and forage quality, the calving strategies of moose and elk appeared to differ in our study. Moose used the lowest elevations of the year (1333 ± 81 m) during calving and had the smallest seasonal use areas. These locations were not typically on valley bottoms (~800-1300 m) per se, but rather on the coniferous side slopes. Use and selection of Pine-spruce by moose was greatest during the subsequent summer season. Poole et al. (2007) reported that moose in southern British Columbia showed 2 elevational strategies during calving related to predation risk. They described ‘climber’ moose, which moved to higher elevations to calve in areas with lower forage quality and quantity and, therefore, farther from grizzly bears found at lower elevations. In contrast, ‘non-climber’ moose calved at low eleva- tions with much higher forage values, but potentially at increased risk of predation. Given that grizzly bears in the Besa-Prophet tend to remain in higher areas during spring (Parker and Milakovic 2007), moose at lower elevations would avoid bears and have access to early green-up of shrub vegetation. This calving strategy, however, would come with the potential risk of wolf predation, given than wolves select for shrub vegetation in spring (Parker and Milakovic 2007). Although the pooled selection models did not indicate that moose selected locations to specifically avoid predation risk during calving or summer, they did, however, avoid risk of wolf predation during the fall. In contrast to moose, elk moved upslope from their late winter locations to calve ~220 m higher than moose. During calving and summer, the higher elevations corresponded with the high use of the Subalpine habitat class. Their selection for vegetation quality was probably facilitated by access to Elymus and Deciduous burns, which typically green up earlier in spring, and then by the spruce-shrub transition zone of the Subalpine in summer. Unlike moose, elk appeared to select calv- ing and summer areas on the landscape that minimized some predation risk by bears given that they used similar elevations during these seasons (Parker and Milakovic 2007). Management implications and Recom- mendations Combining both use and selection in- formation from this study, it appears that the highest potential for overlap between moose and elk may be during fall and winter, when both species used the Subalpine habitat class more than other classes and selected strongly for it. In winter, there may be some elevational separation between the 2 species, given that elk locations were on average 100 m higher than those of moose. In fall, however, the elevations used by moose and elk were not different. Both species also selected for Low shrub habitat during fall, where elk undoubtedly consume higher amounts of forbs and grass that are not yet senescent (Stevens 1970) compared to the more browse-dominated diets of moose. Elk are generalist feeders that maximize their food intake through mechanisms of habitat selection rather than selection of specific foods (Irwin and Peek 1983). Consequently, they can successfully shift from herbaceous species to browse (Houston 1982), and may HABITAT SELECTION BY MOOSE AND ELK – GILLINGHAM AND PARKER ALCES VOL. 44, 2008 58 be efficient competitors with moose when resources are limited (Flook 1964). During periods when resources are limited, overlap in resource use between the 2 species could result in temporary interspecific competition (Jenkins and Wright 1988). If prescribed burns that are conducted primarily to enhance Stone’s sheep popula- tions are enabling increases in non-target elk populations in the Besa-Prophet, there also may be potential for competition between elk and Stone’s sheep during some times of the year (Walker et al. 2007). Further, it is likely that with an expanding elk population, predator numbers will increase in response to the increased prey base. Higher wolf numbers would be expected to affect predator-prey dynamics by expanding into adjacent areas, particularly via burned slopes to upper eleva- tions, and by increasingly preying on Stone’s sheep and caribou. Prescribed burning also may provide additional opportunities for grizzly bears that select for burned vegetation classes (i.e., Elymus burn and Deciduous burn) throughout the non-denning period (Parker and Milakovic 2007), thereby augmenting predation risks to moose and elk. The man- agement action of prescribed fire may help to sustain some of the diversity and abundance of large mammals for which the Besa-Prophet area is known, but it could also shift the prey base for predators and change the dynamics of the system. Additional studies involving population estimates and animal distributions should be specifically designed to determine how intensity, frequency, and locations of pre- scribed burns affect habitat use by ungulates (principally elk, moose, and Stone’s sheep) and subsequently predators (e.g., wolves and grizzly bears) and their movements relative to ungulate prey. Our findings suggest that there is currently some seasonal overlap between elk and moose in the Besa-Prophet, and that expanding elk numbers will affect other spe- cies in the system. AcKnowlEDgEMEnTS We thank the Muskwa-Kechika Trust Fund, the British Columbia Ministry of Envi- ronment and the University of Northern Brit- ish Columbia for their support. G. Williams helped familiarize us with the Besa-Prophet landscape. R. B. Woods, with the BC Ministry of Environment, captured and collared all of the animals monitored in this study. We also acknowledge J. B. Ayotte, S. G. Emmons, D. D. Gustine, R. J. Lay, B. Milakovic, and A. B. D. Walker who assisted in aspects of data visualization and analyses. REfEREncES Anderson, d. P., J. d. Forester, M. G. turner, J. L. FrAir, e. H. MerriLL, d. Fortin, J. s. MAo, and M. s. Boyce. 2005. Factors influencing female home range sizes in elk (Cervus elaphus) in North American landscapes. Landscape Ecology 20:257- 271. BALLArd, W. B., J. s. WHitMAn, and d. J. reed. 1991. Population dynamics of moose in south-central Alaska. Wildlife Monograph 114. BerGerud, A. t., and J. P. eLLiott. 1998. Wolf predation in a multiple-ungulate system in northern British Columbia. Canadian Journal of Zoology. 76:1551-1569. _____, W. Wyett, and J. B. snider. 1983. The role of a wolf population in limiting a moose population. Journal of Wildlife Management 47:977-988. Boer, A. H. 1998. Interspecific relationships. Pages 337-349 in A. W. Franzmann and C. C. Schwartz, editors. Ecology and Management of North American Moose. Smithsonian Institution Press, Washing- ton, D.C., U.S.A. BoWyer, r. t. 2004. Sexual segregation in ruminants: definitions, hypotheses, and implications for conservation and management. Journal of Mammalogy 85:1039-1052. _____, and J. G. Kie. 2004. Effects of foraging ALCES VOL. 44, 2008 GILLINGHAM AND PARKER – HABITAT SELECTION BY MOOSE AND ELK 59 activity on sexual segregation in mule deer. Journal of Mammalogy 85:498-504. Boyce, M. S. 1991. Migratory behavior and management of elk (Cervus elaphus). Applied Animal Behaviour Science 29:239-250. _____, P. r. Vernier, s. e. nieLsen, and F. K. scHMieGeLoW. 2002. Evaluating resource selection functions. Ecological Modelling 157:281-300. _____, M. s., J. s. MAo, e. H. MerriLL, d. Fortin, M. G. turner, J. FryxeLL, and P. turcHin. 2003. Scale and heterogeneity in habitat selection by elk in Yellowstone National Park. Ecoscience 10:421-423. BritisH coLuMBiA Ministry oF croWn LAnds. 1990. Terrain and resource inventory management in British Columbia speci- fications and guidelines for geomatics: digital baseline mapping at 1:20,000. Ministry of Crown Lands for the Govern- ment of British Columbia, Victoria, British Columbia, Canada. BurnHAM, K. P., and d. r. Anderson. 2002. Model selection and multi-model infer- ence: a practical information–theoretic ap- proach. Second edition. Springer-Verlag, New York, New York, U.S.A. BunneLL, F. L., and D. S. eAstMAn. 1976. Effects of forest management practices on wildlife in the forests of British Columbia. Pages 631-676 in Proceedings XVI IU- FRO World Congress, Oslo, Norway. cederLund, G., and H. sAnd. 1994. Home- range Size in relation to age and sex in moose. Journal of Mammalogy 75:1005- 1012. cooK, J. G. 2002. Nutrition and food. Pages 259-349 in D. E. Toweill, and J. W. Thom- as, editors. North American Elk: Ecology and Management. Smithsonian Institution Press, Washington, D.C., U.S.A. crAiGHeAd, J. J., G. AtWeLL, and B. W. o’GArA. 1972. Elk migrations in and near Yellowstone National Park. Wildlife Monograph 29. deMArcHi, d. A. 1996. Introduction to the ecoregions of British Columbia. British Columbia Wildlife Branch, Ministry of Environment, Lands and Parks, Victoria, British Columbia, Canada. edGe, W. d., c. L. MArcuM, s. L. oLson, and J. F. LeHMKuHL. 1986. Nonmigra- tory cow elk herd ranges as management units. Journal of Wildlife Management 50:660-663. ESRI 2002. ArcView GIS, version 3.3. En- vironmental Systems Research Institute, Redlands, California, U.S.A. FLooK, d. r. 1964. Range relationships of some ungulates native to Banff and Jasper National Parks, Alberta. Pages 119-128 in D. J. Crip, editor. Grazing in Terrestrial and Marine Environments. Blackwell Scientific Publications, Oxford, U.K. FrAir, J. L., s. e. nieLsen, e. H. MerriLL, s. r. LeLe, M s. Boyce, r. H. M. Munro, G. B. stenHouse, and H. L. Beyer. 2004. Removing GPS collar bias in habitat selec- tion studies. Journal of Applied Ecology 41:201-212. FrAnzMAn, A. W., and c. c. scHWArtz. 1998. Ecology and Management of the North American Moose. Smithsonian Institute Press, Washington, D.C., U.S.A. GAsAWAy, W. c., s. d. duBois, r. d. BoertJe, d. J. reed, and d. t. siMPson. 1989. Re- sponse of radio-collared moose to a large burn in central Alaska. Canadian Journal of Zoology 67:325-329. GiLLies, c. s., M. HeBBLeWHite, s. e. nieLsen, M. A. KrAWcHuK, c. L. ALdridGe, J. L. FrAir, d. J. sAHer, c. e. steVens, and c. L. Jerde. 2006. Application of random effects to the study of resource selection by animals. Journal of Animal Ecology 75:887-898. GiLLinGHAM, M. P., and K. L. PArKer. 2008. The importance of individual variation in defining habitat selection by moose in northern British Columbia. Alces 44: 7-20. HABITAT SELECTION BY MOOSE AND ELK – GILLINGHAM AND PARKER ALCES VOL. 44, 2008 60 GriFFitH, B., d. c. douGLAs, n. e. WALsH, d. d. younG, t. r. MccABe, d. e. russeLL, r. G. WHite, r. d. cAMeron, and K. r. WHitten. 2002. The Porcupine caribou herd. Pages 8-37 in Arctic Refuge Coastal Plain Terrestrial Wildlife Research Sum- maries. U.S. Geological Survey, Bio- logical Resources Division, Biological Science Report 2002-0001, Anchorage, Alaska, U.S.A. Groten, s. M. e., and r. ocAtre. 2002. Monitoring the length of the growing season with NOAA. International Journal of Remote Sensing 23:2797-2815. GroVer, K. E., and M. J. THoMPson. 1986. Factors influencing spring feeding site selection by elk in the Elkhorn Mountains, Montana. Journal of Wildlife Manage- ment 50:466-470. Gustine, d. d., K. L. PArKer, r. J. LAy, M. P. GiLLinGHAM, and D. C. HeArd. 2006a. Calf survival of woodland caribou in a multi-predator ecosystem. Wildlife Monograph 165. _____, _____, _____, _____, and _____. 2006b. Interpreting resource selection between scales among individual wood- land caribou in winter. Journal of Wildlife Management 70:1601-1614. HAyes, r. d., and A. s. HArestAd. 2000. Wolf functional response and regulation of moose in the Yukon. Canadian Journal of Zoology 78:60-66. HendricKx, J. 1999. Using categorical variables in STATA. STATA Technical Bulletin 52:2-8. HooGe, P. n., and B. eicHenLAuB. 2000. Animal movement extension to ArcView. Version 2.0. Alaska Science Center, Bio- logical Sciences Office, U.S. Geological Survey, Anchorage, Alaska, USA. Houston, D. B. 1982. The Northern Yel- lowstone Elk: Ecology and Management. Macmillan Publishing, New York, New York, U.S.A. HundertMArK, K. J. 1998. Home range, dis- persal and migration. Pages 303-335 in A. W. Franzmann and C. C. Schwartz, edi- tors. Ecology and Management of North American Moose. Smithsonian Institution Press, Washington, D.C., U.S.A. IrWin, L. L., and J. M. PeeK. 1983. Elk habitat use relative to forest succession in Idaho. Journal of Wildlife Management 47:664-672. JenKins, K. J., and r. G. WriGHt. 1988. Resource partitioning and competition among cervids in the northern Rocky Mountains. Journal of Applied Ecology 25:11-24. JennricH, r. i., and F. B. turner. 1969. Measurement of non-circular home range. Journal of Theoretical Biology 22:227-37. JoHnson, d. H. 1980. The comparison of usage and availability measurements for evaluating resource preference. Ecology 61:65-71. KeLsALL, J. P. 1987. The distribution and status of moose (Alces alces) in North America. Swedish Wildlife Research (Supplement) 1:1-10. KniGHt, r. r. 1970. The Sun River elk herd. Wildlife Monograph 23. LAWson, e. J. GALLerAni, and A. r. rodGers. 1997. Differences in home-range size computed in commonly used software programs. Wildlife Society Bulletin 25:721-729. LAy, r. J. 2005. Use of Landsat TM and ETM+ to describe intra-season change in vegetation with consideration for wildlife management. M.Sc. Thesis, University of Northern British Columbia, Prince George, British Columbia, Canada. LucKHurst, A. J. 1973. Stone sheep and their habitat in the northern Rocky Mountains foothills of British Columbia. M.Sc. Thesis, University of British Columbia, Vancouver, British Columbia, Canada. LyncH, G. M., and L. e. MorGAntini. 1984. Sex and age differential in seasonal home ALCES VOL. 44, 2008 GILLINGHAM AND PARKER – HABITAT SELECTION BY MOOSE AND ELK 61 range of moose in northwestern Alberta. Alces 20:61-78. MAnLy, B. F., L. L. McdonALd, and d. L. tHoMAs. 2002. Resource Selection by Animals: Statistical Design and Analy- sis for Field Studies. Second edition. Chapman-Hall, London, U.K. MAo, J. s., M. s. Boyce, d. W. sMitH, F. J. sinGer, d. J. VALes, J. M. Vore, and e. H. MerriLL. 2005. Habitat selection by elk before and after wolf reintroduction in Yellowstone National Park, Wyoming. Journal of Wildlife Management 69:1691- 1707. MeidinGer, d., and J. PoJAr. 1991. Ecosystems of British Columbia. British Columbia Ministry of Forests, Victoria, British Columbia, Canada. MenArd, s. 2002. Applied Logistic Regres- sion Analysis. Second edition. Sage, Thousand Oaks, California, U.S.A. MiLLer, W. 2002. Elk interactions with other ungulates. Pages 259-349 in D. E. To- weill, and J. W. Thomas, editors. North American Elk: Ecology and Management. Smithsonian Institution Press, Washing- ton, D.C., U.S.A. MiqueLLe, d. G., J. M. PeeK, and V. VAn BAL- LenBerGHe. 1992. Sexual segregation in Alaskan moose. Wildlife Monograph 122. Moen, r., J. PAstor, y. coHen, and c. c. scHWArtz. 1996. Effects of moose movement and habitat use on GPS collar performance. Journal of Wildlife Man- agement 60:659-668. oindo, B. 2002. Predicting mammal species richness and abundance using multi-tem- poral NDVI. Photogrammetric Engineer- ing and Remote Sensing 68:623-629. PALMer, M. W. 1993. Putting things in even better order: the advances of canonical cor- respondence analysis. Ecology 74:2215- 2230. PArKer, K. L., and B. MiLAKoVic. 2007. Defining the predator landscape in the Besa-Prophet. Part 3 of An Ecosystem Approach to Habitat Capability Model- ling and Cumulative Effects Management. Final Report submitted to the Muskwa- Kechika Advisory Board, Fort St. John, British Columbia, Canada. PArMinter, J. 1983. Fire-ecological relation- ships for the biogeoclimatic zones and subzones of the Fort Nelson timber sup- ply area. Northern Fire Ecology Project, British Columbia Ministry of Forests, Victoria, British Columbia, Canada. Pci GeoMAtics corPorAtion. 2001. PCI Works Version 7.0. Richmond Hill, On- tario, Canada. PecK, V. r., and J. M. PeeK. 1991. Elk, Cervus elaphus, habitat use related to prescribed fire, Tuchodi River, British Columbia. Canadian Field-Naturalist 105:354-362. PeeK, J. M. 1974. On the winter habitats of Shiras moose. Canadian Naturalist 101:131-141. _____. 1998. Habitat relationships. Pages 351-375 in A.W. Franzman, and C.C. Schwartz, editors. Ecology and Man- agement of the North American Moose. Smithsonian Institute Press, Washington, D.C., U.S.A. _____, and A. L. LoVAAs. 1968. Differential distribution of elk by sex and age on the Gallatin winter range, Montana. Journal of Wildlife Management 32:553-557. _____, M. d. scott, L. J. neLson, d. J. Pierce, and L. L. irWin. 1982. Role of cover in habitat management for big game in north- western United States. Transactions of the North American Wildlife and Natural Resources Conference 47:363-373. Pierce, d. J., and J. M. PeeK. 1984. Moose habitat use and selection patterns in north-central Idaho. Journal of Wildlife Management 48:1335-1343. PooLe, K. G., and K. stuArt-sMitH. 2005. Fine-scale winter habitat selection by moose in interior montane forests. Alces 41:1-8. HABITAT SELECTION BY MOOSE AND ELK – GILLINGHAM AND PARKER ALCES VOL. 44, 2008 62 _____, and _____. 2006. Winter habitat selec- tion by female moose in western interior montane forests. Canadian Journal of Zoology 84:1823-1832. _____, r. serrouyA, and K. stuArt-sMitH. 2007. Moose calving strategies in interior montane ecosystems. Journal of Mam- malogy 88:139-150. Post e., n. c. stensetH, r. o. Peterson, J. A. VuceticH, and A. M. eLLis. 2002. Phase dependence and population cycles in a large-mammal predator-prey system. Ecology 83:2997-3002. risenHooVer, K. L. 1989. Composition and quality of moose winter diets in interior Alaska. Journal of Wildlife Management 53:568-577. roBerts, d. W. 1986. Ordination on the basis of fuzzy set theory. Vegetatio 66:123-131. ruiMy, A., G. dedieu, and B. sAuGier. 1994. Methodology for the estimation of terres- trial net primary production from remotely sensed data. Journal of Geophysical Research 99:5263-5283. seAMAn, d. e., J. J. MiLLsPAuGH, B. J. Kerno- HAn, G. c. BrundiGe, K. J. rAedeKe, and r. A. Gitzen. 1999. Effects of sample size on kernel range size estimates. Journal of Wildlife Management 63:739-747. siLVer, r. s. 1976. Ecological features of moose (Alces alces andersoni) winter habitat in the boreal white and black spruce zone of northeastern British Columbia. M.Sc. Thesis, University of British Co- lumbia, Vancouver, British Columbia, Canada. sMitH, d.W., r.o. Peterson, and d.B. Hous- ton. 2003. Yellowstone after wolves. BioScience 53:330-340. sPALdinG, d. J. 1992. The History of Elk (Cervus elaphus) in British Columbia. Publication Number 18, pp. 1-27. Royal British Columbia Museum, Victoria, Brit- ish Columbia, Canada. stAtAcorP. 2007. STATA Version 9.1. Col- lege Station, Texas, U.S.A. stenHouse, G. B., P. B. LAtour, L. Kutny, n. MAcLeAn, and G. GLoVer. 1994. Productivity, survival, and movements of female moose in a low-density population, Northwest Territories, Canada. Arctic 48:57-62. steVens, d. r. 1970. Range relationships of elk and livestock, Crow Creek Drainage, Montana. Journal of Wildlife Manage- ment 30:349-363. steWArt, K. M., r. t. BoWyer, J. G. Kie, n. J. ciMon, and B. K. JoHnson. 2002. Temporospatial distributions of elk, mule deer, and cattle: resource partitioning and competitive displacement. Journal of Mammalogy 83:229-244. stroHMeyer, d. c., and J. M. PeeK. 1996. Wapiti Home Range and Movement Pat- terns in a Sagebrush Desert. Northwest Science 70:79-87. tHoMAs, d. L., and e. J. tAyLor. 2006. Study Designs and Tests for Comparing Resource Use and Availability II. Journal of Wildlife Management 70:324-336. THoMAs, J. W., H. J. BLAcK, R. J. ScHerzinGer, and R. J. Pedersen. 1979. Deer and elk. Pages 104–127 in J. W. Thomas, techni- cal editor. Wildlife Habitats in Managed Forests: the Blue Mountains of Oregon and Washington. U.S. Department of Agriculture Agricultural Handbook 533, Washington, D.C., U.S.A. toWeiLL, d. e., and J. W. tHoMAs. 2002. North American Elk: Ecology and Management. Smithsonian Institution Press, Washington D.C., U.S.A. tucKer, c. J., and P. J. seLLers. 1986. Satel- lite remote sensing of primary production. International Journal of Remote Sensing 7:1395-1416. unsWortH, J. W., L. KucK, e. o. GArton, and B. B. ButterFieLd. 1998. Elk habitat se- lection on the Clearwater national forest, Idaho. Journal of Wildlife Management 62:1255-1263. ALCES VOL. 44, 2008 GILLINGHAM AND PARKER – HABITAT SELECTION BY MOOSE AND ELK 63 VAn dyKe, F., B. L. ProBert, and G. M. VAn VeeK. 1995. Seasonal habitat use characteristics of moose in south-central Montana. Alces 31:15-26. VuceticH, J. A., r. o. Peterson, and c. L. scHAeFer. 2002. The effect of prey and predator densities on wolf predation. Ecology 83:3003-3013. WALKer, A. B. d, K. L. PArKer, M. P. GiLLinG- HAM, d. d. Gustine, and r. J. LAy. 2007. Habitat selection and movements of Stone’s sheep in relation to vegetation, topography and risk of predation. Ecosci- ence 14:55-70. WecKerLy, F. W. 1998. Sexual segregation and competition in Roosevelt elk. North- western Naturalist 79:113-118. Worton, B. J. 1989. Kernel methods for estimating the utilization distribution in home-range studies. Ecology 70:164- 168.