PRELIMINARY FINDINGS OF SWIFT FOX STUDIES IN MONTANA

Amy L. Zimmerman, Dept. of Fish and Wildlife Management, Montana State University, Bozeman, MT 59717 (406-994-3032).

Brian Giddings, Montana Department of Fish, Wildlife, and Parks, P.O. Box 200701, Helena, MT 59620-0701 (406-444-0042; fax: 406-444-4952; e-mail: bgiddings@mt.gov).

ABSTRACT

Swift fox (Vulpes velox) occurrence and distribution were investigated during a graduate research study in northcentral Montana between August 1996 and December 1997. This state report is an edited summary of that study, which is presented fully in a thesis by Zimmerman (1998). Recent information has suggested that swift fox from a reintroduced population in Canada have been dispersing into northcentral Montana to recolonize additional areas. This study focused on northern Blaine county, Montana, with limited surveys conducted in northern Phillips and Valley counties, using a systematic grid trapping design to determine presence and to radio-collar any captured swift foxes. Data was collected to determine home range size, food habits, and survival rates of marked swift fox. Habitat comparisons, land ownership patterns, and the degree of topographical roughness, for each sampled township, were also investigated.

A total of 16 swift fox were trapped in 1996 and 1997 with this sample comprised of five juveniles and 11 adults. During the spring of 1997, radio-collared foxes were used to locate natal den sites which resulted in three known litters being produced in northern Blaine county. The average home range size of marked swift fox was estimated to be between 10.4 km² (Minimum Convex Polygon) to 12.3 km² (Adaptive Kernal). Survivorship was estimated at 51.6%. Food habits determined through scat analysis indicated mammal material (Microtus spp.) was highest in percent occurrence. No relationships were found in habitat composition, land ownership patterns, or topographical roughness in townships with and without swift foxes. This study documented that the swift fox has reestablished a resident population in northcentral Montana, and field evidence suggests that the population is surviving and reproducing. A wider range of habitats in northcentral Montana need to be surveyed to determine statewide swift fox distribution and population size.

INTRODUCTION

The swift fox was declared extinct in Montana in 1969 as a consequence of a 16-year absence of swift fox in fur harvest records (Hoffman et al. 1969). No swift foxes were reported in Montana until 1978, when a male fox was incidentally trapped in Custer county (Moore and Martin 1980).

This animal, however, may have been dispersing from northeastern Wyoming or southwestern South Dakota. In Canada, the last known swift foxes in Saskatchewan and Alberta were collected in 1928 and 1938, respectively (Soper 1964, Banfield 1974). The species was designated as extirpated by the Committee on the Status of Endangered Wildlife in Canada in 1978 due to the lack of observations since 1938 (Brechtel et al. 1996).

A reintroduction program began in southern Alberta and Saskatchewan in 1983-1984, with swift foxes being released annually through 1998 (Brechtel et al. 1996). Objectives of the national recovery plan for the swift fox include: 1) establishing two geographically distinct, yet genetically connected core populations with an average spring-season density of five adult foxes per township; 2) ensuring the long-term security of prairie habitat in two main regions of the Canadian prairie; and 3) to establish swift fox in at least 50 percent of the suitable habitat that remains on the Canadian prairie (Brechtel et al. 1996). The plan’s overall goal is to establish a target population of 420 foxes by the year 2000.

Three primary release sites were established in Alberta and Saskatchewan which include the Milk River ridge area in southern Alberta, the Cypress Hills area along the Alberta-Saskatchewan border, and the Wood Mountain Plateau site south of Mankota, Saskatchewan. The swift fox population located along the Alberta/Saskatchewan border is currently the largest. Approximately 70 to 100 swift fox have been reintroduced annually since 1983, with releases comprised of Canadian captive-reared foxes and Wyoming captured wild foxes (Brechtel et al. 1996). In 1993, the wildlife directors of Alberta, Saskatchewan and Canadian Wildlife Service decided to continue the reintroduction program for an additional five year period (Brechtel et al. 1996). A population assessment, conducted during the winter of 1996-1997, estmated the Canadian swift fox population at 289 animals (95% C.I.; range 179-412) (Cotterill 1997).

Observations of swift fox have been reported recently in northcentral and southeastern Montana. An increase in the frequency and intensity of reports since 1992 suggests that swift fox may be recolonizing northcentral Montana as a result of the Canadian reintroduction program (Giddings and Knowles 1995). Brechtel et al. (1993) documented instances of Canadian released swift fox dispersing into northcentral Montana between 1989 through 1992. It was unclear, however, if these dispersing individuals were surviving to establishing a resident population in Montana.

The goal of this study was to verify and document evidence of swift fox presence in northcentral Montana. One primary study objective was to locate natal den sites as a means of documenting that breeding and reproduction is occurring in Montana. Additional study objectives included an investigation of relative distribution, home range size, reproductive success, survival rates and food habits of radio-collared swift foxes. Habitat composition, land ownership/management, and the degree of topographical roughness of areas that swift foxes currently occupy were also analyzed to determine if these attributes affected the probibility of swift fox presence. The null hypothesis was that there would be no difference in the probability of capturing swift fox in areas of different habitat compositions, land ownership/management, or topographical roughness.

STUDY AREA

The study area was portions of northcentral Montana, with the primary study site located in northern Blaine county, approximately 25 miles northeast of Havre, Montana. Secondary study sites were located in northern Phillips and northern Valley counties. The landscape consists of large blocks of gently rolling, glaciated, short and midgrass prairie habitat with agricultural areas interspersed on upland sites. Moderate relief is created from creek drainages that disect the prairie grassland. Annual mean temperature ranges from -1.78^o C to 13.9^o C. Mean annual precipitation is 32 cm and ranges from 19 cm to 52 cm.

Dominant grasses on the study area were: western wheatgrass (Agropyron smithii), prairie junegrass (Koeleria pyramidata), blue gramma (Bouteloua gracilis), buffalo grass (Buchloe dactyloides), and Sandberg bluegrass (Poa secunda). Some localized areas have been seeded to crested wheatgrass (Agropyron cristatum). The main shrub occurring on the prairie is silver sagebrush (Artemisia cana). Dominant forbs include: American vetch (Astragalus spp.), scarlet globemallow (Sphaeralcea coccinea), fringed sagewort (Artemisia frigida), cudweed sagewort (Artemisia ludoviciana), toadflax (Linaria spp.), plains clubmoss (Selaginella spp.), and prickly pear (Opuntia spp.). Upland sites in the study area were characterized by loam or clay loam soil surfaces and clay or clay loam subsoil types (National Cooperative Soil Survey 1976). Elevation on the study area ranged from ranges from 670 m to 2,103 m. Twenty-eight percent of the study area was managed by the Bureau of Land Management (BLM) and six percent was controlled by the state of Montana. Sixty-six percent of the study area was in private land ownership.

METHODS

Fourteen townships that comprised 92 km² were systematically trapped using Tomahawk live- traps placed in a block-grid sampling design (Smith et al. 1975). Cage trap dimensions were 42x12x12 cm and 32x10x12 cm. Traps were placed at corners of the nine, four mi² blocks, in each township. The base number of traps per township grid therefore would equal 16 traps with each trap located approximately two miles apart. The basic block design for trap placement was modified according to road and trail accessibility within each township sampled so that traps were placed as close to a corner as existing roads or trails would allow. This method was used to increase the efficiency and ease of surveying a large area. Loy (1981) found that placement of traps along roads or trails increased the ease of setting and checking traps and proved more successful for capturing swift fox than a random placement of traps over the landscape. Each sampled township was trapped for 3 to 7 days with traps checked daily between the hours of 05:30 to 10:00. Trapping occurred from 12 August to 29 November, 1996 and from 7 July to 15 August, 1997. A short-term trapping effort also occurred from 12 October to 18 October, 1997. Trapping success was recorded as the number of foxes captured per 100 trap-nights of trapping effort.

Standard and modified (Zimmerman 1998) methods were used to capture, handle and monitor swift fox on the study area. Radio-collars were manufactured by Wildlife Materials (WM) and Advanced Telemetry Systems (ATS) which weighed approximately 40-50 grams and contained a mortality sensor that was activated after 4 hours (WM) or 8 hours (ATS) of inactivity.

Animal relocations used for home range and movement data were sampled approximately four times per week during the primary field seasons (fall 1996; summer 1997). Radio locations were sampled most frequently by ground between the hours of 20:30 to 02:00 and from 03:00 to 05:30. Hines and Case (1991) reported that 90 to 96 percent of swift fox activity occurred between the hours of 20:00 and 04:00. Locations were also obtained from fixedwing aircraft. An H antenna was mounted to the wing strut of a Bellanca Scout 53819. Flights were made during the day, from 07:00 to 14:00 or between 18:00 to 19:30. During the winter of 1996-1997, regular flights were made every 10 to 14 days to document movements and survival. Flights were also used to describe dispersal distances of radio-collared juvenile swift fox.

In the spring of 1997, potential natal den sites were located and observed. Swift fox pups emerge from the den at 3 to 4 weeks of age (Kilgore 1969). Dens were observed several times per week in the morning or evening beginning in June, until pups were seen outside the den. Natal dens were then observed twice a week to count the number of pups present. All natal dens were described and measured to record the number of entrances, dimensions of openings, opening exposure, slope, distance to the nearest road, and dominant vegetation.

Home ranges were estimated using the triangulation program LOCATE II (Nams 1990) and home ranges were estimated with the program CALHOME (Kie et al. 1994). Home ranges are reported for the Minimum Convex Polygon (MCP) method (Mohr 1947) and the Adaptive Kernel method (Worton 1989).

Annual survivorship was estimated by hand, using the Kaplan-Meier, or Product Limit procedure, modified for a staggered entry design (Pollock et al. 1989). This method allows for new animals to be added after the study has begun and lends itself well to a study where the exact number of individuals to be captured is not known at the beginning of the study. Mortalities were indicated from the mortality signals in the radio-collars being activated and retrieval of the carcass and/or radio-collar was then initiated.

Scat analysis was used to investigate swift fox food habits in northcentral Montana. Swift fox scats were collected in each township trapped during regular field activities. Searching occurred along fence lines, trails and gates. Scat was also collected from captured animals or at den sites when available. Each scat was placed into a separate, plastic zip-lock bag, labeled with the date, location, swift fox frequency number, and immediately frozen. Scat was later placed into a strip of nylon hosiery and washed by hand in a 5 gallon bucket with a tablespoon of laundry detergent. Scats were rinsed until the rinse water appeared clear and were air dried. Dried scats were placed on a sieve and separated into seven categories: hair, bones, teeth, insect, feathers, vegetation, and other. Teeth were used to identify small mammals to genus. Insects were identified to order; feathers were simply classified as "bird"; and plant material was labeled as vegetation. Fecal items were reported in percent occurrence.

Habitat types occurring in a 200 m radius around each trap site were recorded during trapping efforts. Habitat types were classified as rangeland, cropland, Conservation Reserve Program (CRP) land, or a mixture of these. Therefore the 200 m radius was either 100% range, 100% crop, 100% CRP, or contained a mixture of these habitat types. Trapping success for each habitat type was recorded as the number of foxes captured per 100 trap-nights of effort.

The habitat composition and pattern of land ownership/management of townships where foxes were captured were compared with townships having no captures using a Mann-Whitney t-test (Ott 1988). Habitat attributes consisted of classified habitat types, land ownership/management category, and topographical roughness factor.

RESULTS

Trapping Success - Fourteen townships were block trapped over the course of the study. Swift foxes were captured in six of the 14 townships (Fig. 1). A total of 1,205 trap nights resulted in nine swift fox captures in 1996. Trapping success for 1996 was 0.75 foxes/100 trap-nights. Trapping efforts in 1997 resulted in seven captures in 511 trap-nights, for a trapping success of 1.37 foxes/100 trap-nights. The overall trapping success for the study was 0.93 foxes/100 trap-nights.

In 1996 there were three adult females, 3 adult males and 3 juvenile males captured in the study area. During 1997, a new adult male and female were captured in northern Blaine county and an adult male and female and two female pups were captured in Phillips county, just north of Whitewater, Montana. An adult male swift fox was captured in northwest Valley county, over 30 miles east of Whitewater, Montana. An ear tattoo indicated that this swift fox originated from the captive breeding program at the Valley Zoo in Edmonton, Alberta, was born in May of 1995, and was released in the fall into Grasslands National Park (A. Moehrenschlager, pers. comm.).

Standard Measurements - Captured males weighed slightly more than females, with the average weight of males being 2.9 kg compared to 2.5 kg for females. Average ear lengths and hind foot lengths were equal for males and females (5.8 cm and 12.0 cm, respectively). Males had longer average tail length and total body length compared to females. Average tail length for males was 26.9 cm and 24.9 cm for females while average total body length was 72.1 cm for males and 67.9 cm for females. Juvenile body measurements fell in between adult male and female measurements.

Home Range - Annual home range estimates were calculated for five radio-collared swift foxes (Table 1). The number of radio locations per fox ranged from 28 to 72. Using the MCP method, home range size ranged from 7.3 km² to 16.9 km², with an average size of 10.4 km². Using the AK method, home range size ranged from 8.7 km² to 20.3 km², with an average size of 12.3 km². A sample of thirty known locations from several radio-transmitters were used to determine the degree of error in ground triangulation relocations. The overall standard deviation of the error angle was 12.87^o with a 95% confidence interval of -9.03^o to 34.77^o (n=30, df=28).

Population Dynamics: Survival - Of the nine swift fox captured in the fall of 1996, five of these foxes died over the winter and four foxes survived to breed and reproduce in the spring of 1997. Of the surviving four, there were two males and two females. The average number of days survived in 1 year for the nine radio-collared foxes was 249 days, with a range of between 77 and 365 days (Table 2). Time was divided into 2 week intervals, since flights during the winter were done on a biweekly basis. The Kaplan-Meier survival function is based on the probability of an individual animal surviving x unit of time from the beginning of the study. In this study, this time interval was 2 weeks. Mean survivorship over the 1 year period was 51.6% with a standard deviation of 37.0%. The staggered entry survival curve for nine swift foxes in northcentral Montana is illustrated in Fig. 2.

Den Characteristics and Reproduction - Eleven separate natal dens were documented during this study. Two of the swift fox pairs had four different natal dens through the summer, and the third pair occupied three separate dens. All natal dens were located less than 500 m from a dirt road or two-track trail. All natal dens were located in areas with slopes less than 15^o and 73% were located in areas with slopes of 0-5^o. Over half (54%) of den openings were exposed to a south or southeast direction. A north or northeast exposure was the next most common opening at 36%. One den site had an opening toward the southwest, but no openings were exposed to the west or northwest. The average dimension for den openings was 24.6 cm x 42.0 cm (n=34, sd=6.8 cm x 17.8 cm). Each of the two pairs of radio-collared swift fox that survived the winter of 1996-1997 produced offspring. One pair produced a litter of five, the other a litter of three. The swift fox pair captured in July of 1997 produced a litter of seven. Also, two pups were captured in northern Phillips county, indicating that reproduction had occurred in this area.

Food Habits - Scat or fecal analysis indicated that small mammals were the most common item found in the diet of swift fox. Mammal remains were found in 91% (n=59) of the 65 scats analyzed. The second most common food item was insects, which occurred in 66% (n=43) of the scats. Vegetation occurred in 48% (n=31) of the scats, while bird remains occurred in 32% (n=21) of the scats. Seven of the 65 scats analyzed contained unidentified mammal material (11%).

The most common mammals occurring in scats were Microtus spp.(68%), followed by Peromyscus spp. (18%), Spermophilus richarsonii (9%), Perognathus spp. (3%), and Sorex spp. (2%). Microtus, Peromyscus, and Perognathus often occurred in combination while Spermophilus richardsonii was usuall found alone in a scat sample. Only one instance of Sorex spp. was encountered in the 65 samples analyzed.

Insect material was mainly composed of the classes Coleoptera (beetles) and Orthoptera (grasshoppers). Coleopterans were found in 49% (n=32) of the samples, and Orthopterans were found in 43% (n=28) of the scats. Seven different families of beetle were found in scats, including: tiger beetle (Cicindelidae), weevil (Curculionidae), pill beetle (Byrrhidae), dung beetle (Scarabaiedae), ground beetle (Carabidae), darkling beetle (Tenebrionidae), and carrion beetle (Silphidae). Two families of Orthoptera were found in scats: Acrididae (grasshopper) and Gryllidae (cricket). Formicidae (ants) from the order Hymenoptera, occurred in 14% of the scats. A few samples contained material from the order Hemiptera (stink bug), Lepidoptera (caterpillar skins) and Odonata (damselfly adults).

Percent occurrence of food items, by season (spring/summer/fall), is presented in Table 3. Mammals were represented evenly throughout the three seasons. Insects occurred in all three seasons, poeaking in the fall season at 75% occurrence. Vegetation occurrence was highest in the summer (65%) but was present in scats in all three seasons. Occurrence of bird parts was highest in the spring (60%), was present in the summer (39%), and dropped off in occurrence in the fall (20%).

Habitat Use - There were a total of 1,716 trap-nights over the course of the study. Of these, 1,394 trap-nights occurred in rangeland, 43 were in cropland, 8 were in CRP land, and 271 were in a mixed habitat. There were 16 individual swift fox captured. Fifteen captures occurred in rangeland, one occurred in mixed habitat, and no captures occurred in cropland or CRP land. Trapping success for rangeland was 1.08 foxes/100 trap-nights while success in mixed habitat was 0.37/100 trap-nights.

Statistical tests indicated no significant difference in the habitat composition of townships with (n=6) and without (n=8) swift foxes. The average amount of rangeland was 9,680 ha in townships with foxes and 7,615 ha in townships without foxes, with the average amounts of cropland being 2,754 ha and 780 ha, respectively. The proportion of rangeland (p=0.747) and cropland (p=0.175) did not differ significantly between the two groups. There was no evidence that the probability of capturing swift foxes was different for townships of different habitat compositions. The average amount of private land in townships with swift foxes was 44,786 ha and 53,630 ha in townships without foxes. The average amount of BLM managed land in townships with swift foxes was 42,253 ha and 29,750 ha in townships without swift foxes. State managed land represented an average of 4,800 ha in townships with foxes and an average of 7,900 ha in townships without foxes. No significant difference was found for the area of private land (p=1.000), BLM managed land (p=0.95), or state owned land (p=0.213) between townships with and without swift foxes. The average roughness factor for townships with foxes was 1.0 and for townships without foxes was 0.9. No significant difference was found in the overall roughness factors for the two groups of townships (P=0.401). Thus, there was no evidence to suggest that the probability of capturing swift foxes would be different for townships of different roughness.

DISCUSSION

A small percentage of the potential available swift fox habitat in northcentral Montana was surveyed through live-trapping in this study. There are approximately 688,000 ha of prairie rangeland in northcentral Montana, primarily located in Blaine, Phillips and Valley counties (Giddings and Knowles 1995, C. Knowles, pers. comm.). The total area surveyed during this study was 129,024 ha or 19% of estimated prairie rangeland available in northcentral Montana. In 1996, trapping success was 0.75 foxes/100 trap-nights and 1.37 foxes/100 trap-nights in 1997. Success rates in Kansas, where swift fox are considered common, were 1.75 fox/100 trap-nights (Matlack 1997). If the success rate observed in this study was representative for Montana, then fox densities in occupied habitats may be comparable to other regions in swift fox range. Physical measurements of captured fox in northcentral Montana appeared to be similar to those in other ecological studies. Scott-Brown et al. (1987) and Kilgore (1969) found adult male swift fox to be slightly larger and heavier than adult females. Overall, Montana study animals appeared to be in good physical condition.

Home range information was rather limited in this study. The ground telemetry system used was apparently inadequate to locate marked animals consistently due to spatial constraints resulting from difficult and remote road and trail access, which was characteristic of the study area. Often, a fox could not be located during a sampling night because of an inability to get close enough (1.6 km) to receive the signal. Therefore it may be assumed that home range sizes in this study could have been underestimated. Swift fox home range estimates in this study averaged 10.4 km² with a range from 7.3 km² to 20.3 km² compared to Canadian studies which have estimated home range sizes up to 34.1 km² (A. Moehrenschlager, pers. comm.). Hines (1980) found average home range sizes of 17.3 km² for males and 12.4 km² for females in Nebraska, which is more comparable to the average home range size found in this study. More recently, Hines and Case (1991) estimated home range sizes (up to 32.3 km²) in Nebraska and speculated that the large home range sizes may have been the result of low population densities or prey availability.

The estimated survival rate for swift foxes in this study (51.6%) was comparable to reported survival estimates elsewhere. Sharps and Whitcher (1984) reported a 50% mortality rate among 14 radiomarked swift fox over 1 year in South Dakota. Zumbaugh (1984) also reported a 50% mortality rate among adults based on 2 years of harvest data in Kansas. Minimum annual survival rates for wild-born Canadian swift foxes have been estimated at 46% for adults and 36% for juveniles during their first year (Brechtel et al. 1993). Montana estimates were based on all nine foxes captured in the fall of 1996. Of these, six were adults and three were juveniles (young of the year).

Natal den sites in this study were similar to den sites described in other ecological studies in that they were close to roads and trails. Other studies have shown the tendency for den sites to be located 500 m or less from a road (Hines 1980, Hines and Case 1991, Loy 1981, and Jackson (1997) observed den sites less than 1 km from a road. Den sites in northcentral Montana were located in open, level areas, often near the crest of a small rise or ridge. The majority of den sites in this study were located in areas with a slope of less than 15^o which is a similar finding of other studies (Hillman and Sharps 1978, Loy 1981, Jackson 1997). Other common characteristics include selection for a eastern or southern exposure, a loamy soil and areas of sparse vegetation (Kilgore 1969, Hillman and Sharps 1978, Hines and Case 1991, Loy 1981, Uresk and Sharps 1981, Jackson 1997). In this study, the number of den entrances seemed to be tied to the number of pups in the litter. Dens of the litter of three had one or two entrances while the dens with litters of five and seven pups had four den entrances.

One of the primary objectives of this study, to document evidence of swift fox reproduction, was accomplished by identifying natal den sites and observing active litters, which ranged in size from three, five and up to seven animals. Reports of litter sizes in the reintroduced Canadian population have ranged from one to seven pups, with an average litter size of 3.9 (Brechtel et al. 1993). A record litter size of eight pups was recorded during the spring of 1997 (A. Moehrenschlager, pers. comm.). Covell (1992) recorded the average litter size for swift fox at 2.4 pups/pair without a helper and at 4.2 pups/pair with a helper. Larger average litter sizes of four and five were reported for swift foxes by Hillman and Sharps (1978) and Kilgore (1969), respectively.

Mammals were the most frequent food item found in the scats collected during this study. Uresk and Sharps (1981) also found mammals to be the most frequent food item (49%) in the diets of swift fox in South Dakota, followed by insects (27%), vegetation (13%), and birds (6%). Rodents, particularily voles, were found to be the most important food item for the swift fox in northcentral Montana. Hines (1980) found that Cricetidae rodents, especially M. ochrogaster, increased in occurrence in the spring. Hines and Case (1991) also found M. ochrogaster to be the most common mammal species found in 52 scat samples in Nebraska. Cameron (1984) reported that rodents were important food items in spring and summer on the Pawnee National Grassland in Colorado.

Leporidae were not found in any of the 65 scat samples in this study, even though they seemed to be quite abundant on the study area. Cottontails and jackrabbits have been found to be important food items in other studies, especially in winter (Cameron 1984, Hines and Case 1991, Kilgore 1969, Zumbaugh et al. 1985). Hines (1980) and Kilgore (1969) documented that Leporidae species decreased in occurrence in the spring, but were important in winter and early spring. The scats from the Montana studt were collected from late spring, summer and fall. This may explain the absence of Leporidae.

A high percentage of scats examined had at least some insect material present. In a study in Nebraska, Hines and Case (1991) found that 56% of the swift fox scats had insect remains, most of which consisted of Orthoptera (37%) and Coleoptera (25%). This result is quite similar to Montana results of 66% occurrence of insect material in collected scats. In South Dakota, Uresk and Sharps (1981) reported that insects were present in 27% of scats.

Vegetation was found in 48% of the scats, consisting mainly of grass in only trace amounts, suggesting vegetation may have been ingested incidentally. However, three scat samples were composed entirely of vegetation with one of grass, a second was an unidentified moss and the third was comprised entirely of berries, probably currants or gooseberries (Ribes spp.). Other vegetation present in scats included spike moss (Selaginella spp.) and nutlets from the family Borinaceae. Hines and Case (1991) found that 54% of the scats they examined contained vegetation. Uresk and Sharps (1981) described vegetation in 13% of the scats they collected. It is unknown whether swift fox ingest vegetation as they consume prey or if vegetation is consumed purposely. Zumbaugh et al. (1985) examined only one stomach sample that contained enough grass to indicate that it had been purposely ingested. Hines and Case (1991) concluded that vegetation was eaten intentionally in some instances. The three scat samples composed entirely of vegetation collected during this study suggest that vegetation was intentionally consumed.

Birds remains occurred in 32% of the scats examined during this study. Most feathers were found in scats deposited in late spring and early summer. This occurrence is consistent with studies reporting that both bird and insect remains increase from winter to summer as their availability increases (Hines 1980, Hines and Case 1991). Feathers found in scats were usually too small to identify, although mallard (Anas platyrhynchos) and horned lark (Eremophila alpestris) remains were often observed at den sites during the summer of 1997. On a single occasion, a willet (Catoptrophorus semipalmatus) was found lying beside a live-trap that held a juvenile female swift fox south of Whitewater Lake in Phillips county.

Swift fox in northcentral Montana seemed to be the most strongly associated with rangeland habitat, based on the high trapping success in this type when compared to other available habitat types. Den site locations and home range locations supported this association. No den sites were observed in any habitat other than native rangeland and only one out of 291 telemetry locations occurred near cropland habitat. The high proportion of trap sites located in rangeland habitat reflects the dominance of grassland prairie (118,997 ha) over cropland (22,761) in the study area and may also indicate that these foxes are selecting habitats in proportion to availability rather than simply keying in on rangeland habitat.

No significant differences were found in the habitat composition or land ownership patterns and management practices between townships with and without swift fox. It was hypothesized that townships with a greater amount of rangeland may be more attractive to swift fox, and thus, habitat composition would influence where swift fox were captured. It was also hypothesized that the pattern of land ownership may play a role in the presence or absence of swift foxes, considering that there may be certain vegetation cover types that attracted foxes to areas as a consequence of land management practices. However, the study data did not support these hypotheses. There are several possible explanations why land management practices or habitat compostion did not effect the probability of swift fox presence. Trapping was probably not preformed over a wide enough area to detect a gradient of swift fox densities that could be attributed to differences in the amount of rangeland versus cropland habitats across townships (or accordingly, the amount of private, BLM, or state land across townships). Other explanations may be that sampling effort was simply not intense enough and therefore did not have enough power to detect a difference (power ranged from 5%-29%) or that these attributes do not have an effect on whether swift fox are present or absent in a specific area. A final explanation could be that the swift fox population is still colonizing the area and has simply not yet filled all of the most desirable habitats.

Habitat quality was also hypothesized to be related to landscape characteristics such as roughness (coulees and breaks). This factor could influence swift fox presence if rough terrain maintained higher coyote densities (Pyrah 1984) or diminished a foxes ability to chase and capture prey (Samuel and Nelson 1982). Habitat quality and landscape selection by swift fox in northcentral Montana is likely related to habitat that minimizes encounters with coyotes. Pyrah (1984) reported that surveys indicated consistently higher coyote densities in breaks habitats as compared to the adjacent upland prairie habitats in central Montana. Coyotes are the primary cause of direct mortality for swift fox, both in the reintroduced Canadian population (S. Black, pers. comm.) and in other swift fox populations within the U.S. (Covell 1992, Matlack 1997). Statistical tests on the differences in landscape roughness within townships with and without swift fox did not indicate any relationship between large scale topographical complexity or relief and the presence or absence of swift foxes.