Regional differences in winter diets of bobcats in their northern range

Abstract When generalist predators have wide geographic ranges, diets may differ dramatically, largely as a result of differing prey communities. Bobcats (Lynx rufus) are widely distributed across southern North America, with their northern range edge occurring in southern Canada and in the northern US states. Within this northern range, bobcats are exposed to cold and snowy winters and a limited number of prey species, conditions that are atypical for most of the range of bobcats. We examined winter diets of bobcats in high elevation and very snowy forests in northwest Montana to determine how these generalist predators managed in these harsh conditions in comparison with elsewhere in the northern range. Bobcats consumed five major prey types: Red squirrels (Tamiasciurus hudsonicus) and Cricetid rodents comprised >78% of the dietary biomass, whereas the larger snowshoe hares (Lepus americanus), deer (Odocoileus spp.), and grouse were consumed much less often. The standardized niche breadth of bobcat diets was 0.29; bobcats from across the northern range also routinely ate multiple prey species, although Eastern bobcats appear to consume more lagomorphs than do Western bobcats. These results indicate that bobcats remain generalists in difficult winter conditions while preying primarily on small‐bodied prey, although bobcats have highly variable diets across their northern range.

gion with limited winter prey and snow conditions that would seem to favor lynx over bobcats, the query becomes how bobcats manage the challenges of limited prey and the presence of a specialist congener. In these winter forests, the ~1,400-g hares offer substantially more calories than do the ~200-g red squirrels (Tamiasciurus hudsonicus) and <40-g small mammals that are available. There are three basic ways the dietary flexibility of bobcats could manifest in this setting: (a) via bobcats "becoming lynx-like" by acting as facultative specialists and preying primarily on the energetically rich snowshoe hares; (b) by eating a suite of species, including hares; or (c) by focusing on prey other than hares. Previous evidence is mixed; bobcats in Eastern North America consumed >50% hares in winter (Litvaitis & Harrison, 1989;Litvaitis, Clark, & Hunt, 1986;Matlack & Evans, 1992;Pollack, 1951), whereas bobcats in Idaho consumed only 1.5% hares (Koehler & Hornocker, 1989).
We thus have two research objectives. First, we characterize bobcat diets to assess how specialized their winter diets are in a region of Montana that is higher elevation and much snowier than study areas used in previous work on bobcat diets in their northern range. Second, we compare the diets of these montane bobcats in northwestern Montana (hereafter "Montana bobcats") to bobcats from elsewhere in the northern range, to assess how flexible bobcats are in their diets across areas that experience prolonged snowy winters. For this objective, we determined dietary niche breadths of northern bobcats after a thorough literature search for data on bobcat diets in northern latitudes. For both objectives, we are particularly interested in how prevalent snowshoe hares are in bobcat diets, as these prey do not occur in the southern range of bobcats and because hares are the primary prey of Canada lynx.

| MATERIAL S AND ME THODS
Our study area was the Tally Lake Ranger District of the Flathead National Forest, northwestern Montana, USA (48°30′0″N, 114°45′0″W), located in the center of the Salish Range. The Salish Mountains (48°12′N, 114°48′W) encompass 10,684 km 2 , with >30 peaks over 1,828 m, of which 10 peaks were located in our study area. TLRD encompasses 1,137 km 2 , with elevations ranging from 945 to 2,008 m. Annual temperatures range from −42 to 38°C and mean annual precipitation is 58 cm at 975 m in Olney, Montana, on the northeast edge of the TLRD (NOAA, 2017). Winter temperatures range from −42 to 7°C, and annual snowfall typically exceeds 300 cm at elevations >1,300 m and can exceed 700 cm at elevations >2,000 m (NOAA, 2017).  Alces alces) may also be available to bobcats.

| Sample collection
Bobcat scats were collected throughout the study area during winter (December-February, 2009 when encountered along snowmobile tracks or while backtracking a bobcat. Appearance of the scat and the presence of bobcat tracks were used to confirm the scat was from a bobcat. Scats were also collected from live-trapped bobcats ( Figure 1) (Newbury, 2013). Scats collected from traps were assumed to be from the bobcat's meal prior to ingesting trap bait (deer), and indeed, no scats contained deer. Any fur from trap bait that was frozen or stuck to the outside of scats was removed. Live-trapping adhered to strict protocols for trapping and handling and permits from Montana State Fish, Wildlife, and Parks (2009-059, 2010-002, 2011 do not know whether all scats were from separate individuals. We combined stomach, colon, and scat samples in our analysis. Stomach samples are less digested than colon samples, so feathers, fur, and bones were often more identifiable than in scats; however, in both stomach and scat/colon samples, we could not always separate small mammals to species or genus (deer, grouse, squirrel, and hare remained identifiable). These samples thus provide comparable information, and we do not think there is a bias from combining sample types. Samples were stored at −23°C until 24-48 hr prior to analysis, when they were thawed at room temperature.

| Sample analysis and prey identification
All scat and colon samples were oven-dried until sample mass remained constant. Sample contents were analyzed following Reynolds and Aebischer (1991). Dry mass of each scat or colon sample was recorded; then, samples were broken down in water and rinsed through a 0.5-mm sieve to separate microscopic from macroscopic fragments. After thawing, stomach samples were immediately rinsed through a 0.5-mm sieve (Litvaitis, Stevens, & Mautz, 1984). The 0.5mm mesh captured even the smallest rodent bones and teeth. Each sample was sorted into categories such as fur, bone, feather, and incidental ingestion (e.g. pine needles) and then air-dried prior to identifying species.
Prey items were identified to species by using diagnostic hair, teeth, and bones. Bones and fur present in samples were compared to specimens in the Philip L. Wright Zoological Museum for species confirmation. When no diagnostic teeth or bones were present, hairs were identified by using a compound microscope, reference hairs, and a key to mammalian guard hairs (Moore, Spence, & Dugnolle, 1974). This approach was often necessary for mice and voles, although sometimes we were able to identify only to subfamily or family for rodents because of severe degradation of hair and bone in samples.
We excluded probable trap bait in two ways. First, deer tracks were rarely located on our study site in winter, but we used roadkilled deer to bait live traps. We found no deer in scats collected from live-trapped animals or from scats found along tracks and roads in the trapping area. Second, to account for trap bait in stomach and colon samples from bobcat carcasses, we sent surveys to trappers who had turned in bobcat carcasses. When we received a trapper's response (~50%), we removed that bait type from prey remains in the gut. None of the trappers who responded had used red squirrel or snowshoe hare as bait. We also excluded items such as domestic chicken that were very likely to be bait. However, we did not exclude deer from samples where the trapper did not respond, because for some samples for which a trapper did respond, deer fur/meat was contained in the sample, but the trapper had not used deer as bait.
After prey species were identified, the volume of each sample composed of that species was visually estimated following Reynolds and Aebischer (1991). Most samples (83%) were composed of one prey species. We were not able to quantify the number of individuals in each sample, given the degraded quality of bone and fur. This decision may underestimate individual Cricetid rodents consumed, but is unlikely to underestimate the number of larger prey.

| Statistical analyses
We calculated absolute frequency of occurrence (AFO) of each prey species found (number of occurrences of a given prey type/ total number of samples; Wright, 2010), and relative frequency of occurrence (RFO; number of occurrences of a given prey type/total number of prey species occurrences). RFO accounts for more than one prey type being found in some samples (Ackerman, Lindzey, & Hemker, 1984).
We estimated the biomass consumed of each prey species from Baker's , Warren, and James (1993) regression equation for bobcats that relates dry mass of each prey type in the scat to the fresh consumed biomass. Following Baker et al.'s (1993) regression equation y = 16.63 + 4.09x, where x is the average mass of each prey type (Table 1) and y is the biomass consumed, we calculated conversion F I G U R E 1 An adult male bobcat (Lynx rufus pallescens), M1, that was captured and radio-collared as part of this study on the Tally Lake Ranger District, Flathead National Forest, northwest Montana. M1 weighed ~15 kg when collared on 12 December 2009. In this photograph, M1 was recaptured on 25 January 2010 and released without handling factors for each prey type except deer. For deer, we used the empirical results from Baker et al. (1993), that is, a conversion factor of 27.0.
Dry masses per prey type in stomach samples were not determined because stomach samples were not dried prior to analysis. To incorporate stomach samples into biomass estimates, we used the average dry weight of each prey type from the scat and colon samples; for example, each stomach sample that contained deer was assigned a value of 5.5 g of deer. We summed the total dry mass per prey type in our samples, multiplied by the conversion factor, and then divided by total mass summed across all prey types to determine percent biomass consumed of each prey type (Baker et al., 1993).
We then compared diets of bobcats in our study area to diets of bobcats from similar northern latitudes but across a wide longitudinal gradient. We searched Web of Science and the ProQuest database of theses and dissertations for bobcat winter dietary research in the northern range. We lumped the data into Eastern and Western states/provinces, because finer geographic subdivision resulted in very uneven sample sizes; some studies also lumped data from several states. Although we present data from midwestern populations, we do not compare these analytically because of the low sample size of studies. In studies from which absolute frequency of occurrence data could be extracted, we grouped diet into six categories: Cervidae, Lagomorpha, Sciuridae, Rodentia, Aves, and Other.
We then used a G test for independence with a significance value of p < 0.05 to compare average proportions of each prey category reported in studies of winter bobcat diets.

| Dietary niche breadth and overlap among bobcat populations
Winter niche breadth for Montana bobcats and other bobcats in northern latitudes was calculated from AFO in Levins (1968) ∑ p 2 j , where B = niche breadth and p j = fraction of items in the diet that are of food category j. We converted it to a standardized dietary breadth on a scale of 0-1 following Hurlbert's (1978) measure: BA = (B − 1)/(n − 1), where BA = standardized niche breadth, B = niche breadth, and n = number of possible resources (Krebs, 1998).
We examined dietary overlap among northern bobcat populations to see whether bobcat diets differed regionally despite all of our comparisons occurring in areas where bobcats experience snowy winters (in contrast to the southern United States and northern Mexico) and a similar prey base (deer, hares, squirrels, grouse, and small mammals were the main prey available in winter in the regions we compared). Dietary niche overlap for bobcat populations in Western and Eastern regions was calculated in EcoSim Professional v1.2d (Entsminger, 2014) using Pianka's (1974) where p i is the proportion of prey type i in the diet of the first group and q i is the proportion of the same prey type in the diet of the second group. The index ranges from 0 to 1, from no overlap to complete overlap. We ran 1,000 randomized simulations within EcoSim to determine whether the probability of observed overlaps was greater or less than expected by chance.

Reithrodontomys megalotis
Western harvest mouse a We used median mass for deer to account for differences between sex and age classes; adult deer have higher average biomass than shown here. b This average mass was used for all Cricetidae, except muskrats and woodrats. c Myodes gapperi and Microtus spp. are most common on the study site. Ondatra zibethicus are also common in the area, and were easy to identify in remains compared to the smaller arvicolids. d Neotoma cinerea and Peromyscus maniculatus were most common on the study area and were easy to distinguish from one another in remains.
samples, respectively ( Across the northern range, bobcats displayed diets that varied substantially, with most of the variation arising from differences between Eastern and Western groups (Table 3). Western bobcats consumed far more squirrels and rodents but fewer lagomorphs than did Eastern bobcats (Figure 2). Bobcats in Eastern locations consumed more lagomorphs and cervids in their diets than did Western bobcats. We located only two studies that addressed bobcat diets from the Great Lakes states (Gilbert, 2000;Rollings, 1945); in this region, bobcats ate primarily deer and snowshoe hares.
Winter dietary niche breadth of bobcats ranged from a low of Maine study, where bobcats still consumed over 50% of their winter diet as hares and cottontails. However, bobcat niche breadth among regions was very similar, and indicated generalized diet although the prey composition of regional diets was highly variable.
Dietary overlap was variable in Western bobcat populations (Table 5) (Croteau, Heist, Nielsen, Hutchinson, & Hellgren, 2012;Reding, Bronikowski, Johnson, & Clark, 2012), so there could potentially be subtle behavioral or phenotypic variation between these groups as well that would affect hunting success of prey selection.

| D ISCUSS I ON
TA B L E 3 Winter diet of bobcats in the northern United States and southern Canada (1939Canada ( -2005. These results are based on absolute frequency of occurrence  Toweill and Anthony (1988) includes 4% elk, Litvaitis and Harrison (1989) includes 5.9% moose, and Koehler and Hornocker (1989) includes 15.6% bighorn sheep (Ovis canadensis) and 1.5% unknown. b Snowshoe hare and Sylvilagus spp. c Eastern gray squirrel (Sciurus carolinensis), American red squirrel, and northern flying squirrel (Glaucomys sabrinus). d This grouping includes voles, mice, ground squirrels, and mountain beaver (Aplodontia rufa), and the rare report of ground squirrel spp.; hibernating ground squirrels were not available to bobcats in our Montana study. e Other includes large rodents >2 kg, that is, beaver (Castor canadensis), woodchuck (Marmota monax), and marmots (Marmota spp.). Bobcats also consumed raccoons (Procyon lotor), porcupine (Erethizon dorsatum), skunks (Spilogale and Mephitis spp.), and opossum (Didelphis virginiana), lynx, bobcat, mink (Neovison vison), fox (Vulpes vulpes), domestic cat (Felis catus), and otter (Lontra canadensis). Bobcats also consumed fish, vegetation, and berries. f Gilbert (2000) uses different prey categories and presents results in proportion biomass. We present values for deer, hare, and birds by calculating stomachs with that item present divided by total sample size. Gilbert lumps "medium" and "small" mammals, so we could not separate squirrels from other rodents.

| Do Montana bobcats have more specialized winter diets than other bobcats?
Montana bobcats ate significantly more squirrels and small mammals and fewer snowshoe hares than did bobcats from other northern forests. Such dependence on rodents is more similar to bobcat diets across their southern range (McCord & Cardoza, 1982;Anderson, 1987;Rolley, 1987;Larivière and Walton 1997;Tewes, Mock, & Young, 2002), but is atypical compared to bobcats in other northern forests. Squirrels and Cricetid rodents combined comprised on average ~41% of bobcat diet in 12 studies from northern lati- tudes, yet winter diet of bobcats in Montana and in Idaho (Koehler & Hornocker, 1989) was dominated by rodents (~83% and ~90%, respectively), likely due to similarities in regional topography, vegetation, climate, and prey types. Geoffroy's cats (Leopardus geoffroyi, 0.52; Berg, 2007). These studies of felid species' dietary niche breadth take place in warmer, tropical climates, and the slightly lower winter niche breadth of northern bobcats could simply reflect paucity of prey species available during winter months in these areas, as many potential prey species hibernate or migrate seasonally.
Other studies of niche breadth of bobcats have used the nonstandardized niche breadth measure; for example, winter diet of bobcats in California had B = 8.97 (Neale & Sacks, 2001 (Westfall, 1956) 3.01 0.40 Maine (Litvaitis, Clark, et al., 1986;Litvaitis, Sherburne, et al., 1986) 3.19 0.44 Maine (Litvaitis & Harrison, 1989)  because the gut or scat samples were too degraded to be confident of species identity within these groups. Had we been able to identify all prey remains to species, the total number of prey species consumed would be higher and the percentage of diet composed of each species would be lower for these mingled groups. Our estimate of snowshoe hares and red squirrels in the diet would remain the same for absolute frequency of occurrence and biomass, but would drop for relative frequency if smaller prey species had been separated out. When we compare our results to those from other studies, these problems continue, and in some studies, bobcats also ate other lagomorphs. We also note our survey draws together literature from 1945 to present, and given the large changes in climate and land use over this time span, we suspect there is almost certainly large temporal variation of bobcat diets within each region as well. Despite these challenges, bobcats show clear regional differences in winter diets across their northern range, as exemplified by the range in lagomorphs consumed, which ranged from 1.5% to >90% of the diet.
In another congeneric pair, Lovari et al. (2013) report that snow leopards (Panthera uncia) and common leopards (Panthera pardus) used different habitats, but these species showed much higher dietary overlap than we found here. Felid predators have a large suite of behaviors (scent-marking, timing of movements, diet, and habitat selection, inter alia) that are likely employed in areas of sympatry to reduce harmful direct interactions with members of other species (Ramesh, Kalle, Sankar, & Qureshi, 2012). Here, we seem to see dietary separation, and some studies have hinted at fine-scale spatial separation among these species as well Scully, Fisher, Miller, & Thornton, 2018).

AUTH O R S' CO NTR I B UTI O N S
RN and KH jointly developed the idea and study design and cowrote the manuscript. RN conducted the fieldwork, autopsies, scat analyses, and statistics. KH obtained most of the funding for this work.

DATA ACCE SS I B I LIT Y
Upon acceptance for publication, the authors will make data pub-