Although interspecific killing among carnivores can drive populations toward extinction, it is generally unknown how these intraguild interactions vary among populations, and whether the threat for vulnerable species can be mitigated. We studied imperiled populations of swift foxes (Vulpes velox) in Canada and kit foxes (Vulpes macrotis) in Mexico to determine potential differences in survival or predator-avoidance strategies. Survival rates were significantly lower in Canada than in Mexico because of mortality caused by coyotes (Canis latrans) and golden eagles (Aquila chrysaëtos), and the potential for population recovery is likely higher for the Mexican fox population. Differences in body size between coyotes and foxes, diet, group sizes, intraspecific home-range overlap, home-range sizes of coyotes, and movements of coyotes relative to foxes were similar among study areas. However, Canadian foxes had home ranges that were approximately 3 times larger than those in Mexico, and Canadian foxes were most frequently killed on their home-range peripheries. Home ranges of kit foxes decreased in size as the availability of black-tailed prairie dog (Cynomys ludovicianus) colonies increased and associated refuge holes, which foxes could use to escape predation, were significantly more abundant in Mexico than in Canada. Small home ranges of foxes probably reduced encounters with coyotes in Mexico, and a high availability of refuges likely allowed foxes to elude predators when such encounters did occur. Differences in survival of foxes relative to mortality caused by coyotes demonstrate that interactions between carnivores can -vary greatly between populations and that, in some situations, vulnerable species may be able to coexist with dominant carnivores despite a lack of large-scale habitat partitioning.
Interspecific killing and, in cases where the victim is consumed, intraguild predation is increasingly recognized as widespread among carnivores (Arim and Marquet 2004). Palomares and Caro (1999) identified 97 pairwise combinations of interspecific killing in carnivores, involving 27 killing and 54 victim species. Interspecific killing has most frequently involved canids and felids as the aggressors and canids, felids, or mustelids as the victims. The body mass of interspecific killers is positively correlated to that of their victims and group-living carnivores kill larger victims than do solitary species (Palomares and Caro 1999). The published literature likely underestimates the true frequency of interspecific killing. In Africa, for example, distribution, diet, and body size suggest that carnivores may be killed by an average of 14.8 and eaten by an average of 8.6 other species of carnivores, but relatively few of these interactions have been documented (Caro and Stoner 2003).
Habitat partitioning allows some populations of carnivores to persist despite interspecific killing. For example, in southwestern Spain common genets (Genetta genetta) and Egyptian mongooses (Herpestes ichneumon) avoided Matasgordas habitat that was preferred by Iberian lynx (Lynx pardinus— Palomares et al. 1996). Red foxes (Vulpes vulpes) avoided coyotes (Canis latrans) in the Yukon Territory in Canada by using brush habitats on the peripheries of the home ranges of coyotes (Theberge and Wedeles 1989). Gray foxes (Urocyon cinereoargenteus) that eluded predation by coyotes and bobcats (Lynx rufus) in the Santa Monica Mountains in California also were restricted to brush areas (Fedriani et al. 2000).
The potential to elude aggressive carnivores through habitat partitioning decreases in open, homogeneous environments and subsequent interspecific killing can have dire consequences for imperiled species. Spotted hyenas (Crocuta crocuta) and lions (Panthera leo) frequently kill endangered African wild dogs (Lycaon pictus—Creel et al. 2001). An increase in lions also has caused a significant decrease in the number of surviving young of cheetahs (Acinonyx jubatus) in the Serengeti, and the sustainability of this population of cheetahs has been in question (Kelly et al. 1998). Similarly, red foxes are driving arctic foxes (Vulpes lagopus) to extinction in Scandinavia through the assimilation of dens and killing of pups (Hersteinsson and Macdonald 1982; Tannerfeldt et al. 2002, 2003). Although sympatry of such competing carnivores may lead to serious consequences for vulnerable species, high rates of interspecific killing might not necessarily occur in every population. It is generally unknown how behavioral and demographic characteristics of dominant and imperiled carnivores vary across populations, and how these relative differences correspond to the persistence of vulnerable species.
To address these gaps in knowledge, we chose to focus on a system where an imperiled carnivore would have limited opportunities for habitat partitioning, where mortality rates could be high, and where comparisons could be made between distinct populations. Coyotes killing swift foxes (Vulpes velox) and kit foxes (Vulpes macrotis) on the arid grasslands of North America is the strongest known example of interspecific killing among carnivores (Palomares and Caro 1999). The eradication of wolves (Canis lupus) from the great plains of North America allowed for an increase in densities of coyotes, which are now responsible for up to 100% of swift fox and 87% of kit fox mortalities (Kamler et al. 2003c; White and Garrott 1997). Increases in survival rates of swift foxes after intensive coyote control in Texas suggest that coyotes may limit populations of swift foxes (Kamler et al. 2003b).
Swift and kit foxes are genetically distinct species (Mercure et al. 1993), but they interbreed regionally, display common ecological traits, and share similar conservation needs (Moehrenschlager et al. 2004). These foxes have suffered widespread population reductions primarily due to the loss or degradation of native prairie habitat (List and Cypher 2004; Moehrenschlager and Sovada 2004). Although most swift and kit fox populations in the United States are considered to be stable within a core range that has been drastically reduced from historic levels (Sovada and Scheick 2000), it is unclear what the mortality causes and impacts of interspecific killing are on vulnerable populations along the northern and southern peripheries of the ranges of these foxes.
Kit foxes in Mexico are nationally listed as vulnerable and swift foxes in Canada, which were extirpated from Canada by 1938, are now classified as endangered with a reintroduced population that is small and isolated (Moehrenschlager and Moehrenschlager 2001). The Mexican Janos-Casas Grandes grasslands are rich with potential prey of canids because they harbor the largest remaining population of black-tailed prairie dogs (Cynomys ludovicianus) in the world. Comparatively, swift foxes are on the northern limit of their range in Canada, where a lack of prey availability may have contributed to their extirpation by the 1930s (Herrero et al. 1991).
Our objective was to determine previously unknown characteristics of these populations such as the availability of refuges for foxes, annual survival rates of foxes, prey use of foxes and coyotes, and the interactive spatial dynamics of these canids. We hypothesized that comparisons between Canada and Mexico might show large interpopulation variation, because these regions represent the maximum distance and potentially the maximum habitat differentiation for swift and kit foxes in North America. We used similar field protocols and standardized data analyses to compare the interaction of concurrently radiotracked foxes and coyotes in both regions.
Materials and Methods
Study areas.—Locations of populations of swift foxes and coyotes in the Canadian study area spanned the borders of Alberta, Saskatchewan, and Montana (48°53.2′—49°28.8′N, 109°44.2′—110°36.0′W). The Mexican study area (30°57.8′—30°37.5′N, 108°12.5′—108°40.3′W) was in the Chihuahuan Desert ecoregion of northwestern Chihuahua. Elevations ranged from 850 to 1,050 m above sea level in Canada and 1,400 to 1,600 m above sea level in Mexico. Native short-grass prairie, primarily composed of grasses and forbs, characterized both locations, but mesquite scrub (Prosopis glandulosa) also was present on the edge of the Mexican study area. Annual precipitation ranged from 280 to 465 mm in Canada and 193 to 252 mm in Mexico. Cattle ranching and limited crop agriculture were the primary land uses in the Mexican and Canadian study areas.
Captures of animals.—Swift foxes on our Canadian study site were trapped from January 1995 until February 1998 using 109 × 39 × 39-cm Tomahawk double-door (Tomahawk Live Trap Co., Tomahawk, Wisconsin), 83 × 31 × 31-cm Tomahawk single door, and 83 × 26 × 26-cm Havahart (Wood-stream Corp., Lititz, Pennsylvania) traps that were modified to reduce capture injuries (Moehrenschlager et al. 2003). Between July 1994 and February 1996, kit foxes in Mexico were captured with Victor soft-catch leg-hold traps (Woodstream Corp.), Aldrich foot snares (Canadian Trading Post, Burlington, Ontario, Canada), and Tomahawk box traps (Tomahawk Live Trap Co.). All foxes were manually restrained, measured, and weighed, and their sex was determined. Age classifications were based on recaptures of previously marked pups and on the size, color, and wear of teeth. Captured foxes were classified as juveniles (6–9.5 months of age) or adults (>15 months of age). No trapping was conducted during the breeding, gestation, and pup-rearing seasons. Foxes were fitted with mortality-sensor radiocollars (Canada and Mexico: Lotek Inc., Newmarket, Ontario, Canada; Mexico: Wildlife Materials Inc., Carbondale, Illinois).
Five foxes on the Canadian site that were used in home-range analyses but not in survival comparisons had been translocated from Wyoming to supplement the reintroduced population. For these individuals, telemetry data were excluded for the first 60 days postrelease, after which the behavior of translocated animals was similar to that of resident foxes (Moehrenschlager and Macdonald 2003), and body weights from their 1st postrelease capture were used for comparisons between species and countries.
In December 1994, Mexican coyotes were trapped with Victor soft-catch leg-hold traps nos. 1.5 and 3 (Woodstream Corp.) in sets baited with Hawbaker's long distance call lure, coyote food lure, and gray fox food lure (Hawbaker's, Fort Loudon, Pennsylvania). During February 1996, coyotes in Canada were pursued with snowmobiles and captured with a salmon landing net. One coyote in a culvert was net-captured opportunistically by a landowner. Coyotes in both countries were manually restrained, aged, and fitted with mortality-sensor radiocollars (Lotek Inc.). Coyotes and foxes in both countries were released without injury after 10–35 min of handling. The methods employed in Canada were compliant with guidelines of the American Society of Mammalogists (Animal Care and Use Committee 1998) and Canadian Council for Animal Care. The Mexican study was conducted under an animal care permit granted by the Instituto de Ecología, Universidad Nacional Autónoma de México.
Telemetry.—Daily relocations in Canada were obtained with 3-element Yagi null-peak telemetry antennas mounted on 4 × 4 trucks. Nocturnal dynamic-interaction radiotracking was conducted by taking simultaneous bearings from 2 vehicles at 15-min intervals (Doncaster 1990) to determine movement speeds of foxes and coyotes. Aerial locations were obtained every 2–3 weeks. In Mexico, radiolocations were obtained through triangulation from mobile antenna stations using 4× 4 trucks. Bearings also were obtained simultaneously from fixed stations located on hilltops utilizing null-peak systems consisting of 2 parallel 4-element Yagi antennas. Relocations were recorded for each individual 1–8 times per 6-h tracking sessions, with a 15-min minimum time interval. Similar to radio-tracking in Canada, relocations were opportunistically obtained during diurnal periods, but the primary emphasis was on nocturnal sessions when coyotes and foxes were most active. Triangulation bearings were converted to location estimates using LOCATE II (Pacer, Truro, Nova Scotia, Canada) in Canada and Wildtrak (Todd 1993) in Mexico.
Dead foxes in Canada were briefly examined in the field for external signs of trauma and kill-sites were searched for signs of predators. Detailed necropsies were performed by a pathologist at the Calgary Zoo, Calgary, Alberta, Canada, where causes of death were differentiated from scavenging by assessing areas of hemorrhaging. In Mexico, the cause of mortality of 1 uncollared kit fox that was killed by a predator while caught in a leg-hold trap was determined by field staff.
Diet.—Fox and coyote scats were collected fresh monthly in Canada and twice monthly in Mexico. Scats were oven-dried in Canada, air-dried in Mexico, macerated, and partitioned into prey components. Macroscopic samples were identified using reference collections. Microscopic items such as invertebrate remains, feathers, and hair were examined using a compound microscope. Hair was identified to genus using the color, medullar and cuticular patterns, length, and thickness (Teerink 1991). Relative and absolute percentages of occurrence (Ciucci et al. 1996; White et al. 1996) were determined for coyotes and foxes in each country. Given differences of prey species between countries, prey occurrences were classified into 5 broad categories for comparative analyses: insects, rodents, birds, lagomorphs, and large mammals (ungulates and livestock).
Refuges—The density of holes that would be large enough for foxes to use to escape larger predators was compared between Canadian and Mexican habitats. Holes that foxes used ranged from 10 cm to 30 cm in diameter and were too small for entry by coyotes. We deemed sampled holes suitable as fox refuges if the diameter of the hole was >10 cm and if a continuous tunnel of at least 2 m led underground. Density of escape holes in Canada was sampled at 54 random points using three 200 × 10-m belt transects that radiated outwards and were spaced 120° apart. Ten systematic 1,000 × 3-m transects were sampled in prairie dog towns in Mexico and 9 sites of equivalent dimension were surveyed in Mexican grassland regions that did not contain prairie dogs. Statistical comparisons of densities of holes between regions were made on a transect level.
Data analyses—Kaplan-Meier staggered-entry survival rates were calculated using Egret, version 2.0.31 (Cytel Software Corporation, Cambridge, Massachusetts). Survival rates were compared using Z-tests (Pollock et al. 1989) between foxes in Mexico and foxes in Canada incorporating all mortalities, and between foxes in Mexico and foxes in Canada using only coyote-caused mortalities.
Niche breadth of foxes and coyotes based on categories of prey in the diet were compared using the Shannon-Wiener function (Colwell and Futuyma 1971). Horn's index (Horn 1966) was calculated to compare niche overlap between foxes and coyotes between countries. Differences in prey items in the diets were tested between species and countries using contingency table G-tests.
Comparisons of home-range data between studies are frequently problematic because tracking regimes, autocorrelation among fixes, number of fixes, duration of study, home-range estimators, and calculations with differing software packages are highly variable (Laundre and Keller 1984; Lawson and Rodgers 1997; Swihart and Slade 1985). Independence of fixes can be achieved if the animal theoretically has sufficient time to traverse its home range between sampling intervals (Swihart and Slade 1985). Home ranges for foxes in Mexico have been estimated previously (List and Macdonald 2003) but needed to · be recalculated for consistency. We calculated species- and country-specific times to independence for subsequent comparisons of home-range size by determining the time required for animals to traverse their ranges. Maximum travel distances were calculated by determining the longest trajectory of 95% minimum convex polygons using Ranges 6 (Kenward et al. 2003). Speeds were calculated by averaging the time and distance traveled of animals that were relocated at least twice within 1 h. Necessary time intervals for independence of fixes were calculated by dividing the maximum travel distance by the average speed of coyotes and foxes in Canada and Mexico, respectively. Fixed-kernel and adaptive-kernel home-range sizes (Seaman and Powell 1996; Worton 1989) were determined using fixes separated by at least the time thus calculated for independence, the Ranges 6 default smoothing factor (Kenward et al. 2003), and only animals that had a minimum of 20 independent fixes (representing generally >60 nonindependent fixes). To control for potentially confounding differences in duration or frequency of tracking between countries, general linear model comparisons of home-range sizes included the number of independent fixes and total radiotracking duration in the model. Because of larger sample sizes for foxes, age and sex were incorporated into models for foxes but not into those for coyotes. In Mexico, prairie dog towns were mapped and the percentage of the area of home ranges of kit foxes consisting of prairie dog towns was calculated using ArcGIS8 (ESRI, Redlands, California). Home-range overlap was determined for conspecifics that were tracked concurrently for at least 1 month by averaging the reciprocal proportion of 95% fixed-kernel home ranges. Differences in overlap were compared between countries for coyotes, between pairs of mated foxes, and between fox neighbors.
Fixed-kernel 50% and 99% utilization areas were determined using all available fixes for foxes in Canada that were tracked sufficiently for calculations of home-range size to determine the relative location of mortalities caused by coyotes and eagles in cores and peripheries of home ranges. To compare the relative attraction to or avoidance of fox dens by coyotes between countries, 95% fixed-kernel utilization areas were constructed for 10 coyotes in Canada and 7 coyotes in Mexico using all radiofixes. Within these contours, numbers of fox dens and movements of coyotes relative to den sites were compared between countries. Within each area, the average distance of 500 random points to the closest fox den was determined to yield the expected distance of coyotes to fox dens if coyotes moved randomly. The observed distance was determined by averaging the distances of individual coyote relocations to the closest fox den within respective contours.
Distances between foxes and coyotes that were located within 15 min during dynamic interaction tracking were calculated for foxes and coyotes that had overlapping home ranges. Simultaneous relocations of both species were consistently obtained when separation distances were less than 2 km. The relative frequency of 0–1 km and 1–2 km fox-coyote separation distances was compared between countries. Group sizes of coyotes were compared for sightings in Canada and Mexico. Variance values reported in the “Results” are means ± SE.
Captures and body weights of foxes and coyotes—In Canada, 47 foxes (14 adult males, 15 adult females, 6 juvenile males, and 12 juvenile females) and 11 coyotes (5 adult males, 4 adult females, and 2 juvenile females), and in Mexico 11 foxes (7 adult males, 3 adult females, and 1 juvenile female) and 8 coyotes (4 adult males, 2 adult females, and 2 juvenile females) were captured and radiocollared. One juvenile female coyote in Mexico was located for only 1 month and 1 juvenile female coyote in Canada was never relocated, presumably because of a failed radiocollar; these individuals were excluded from subsequent analyses. Body weights of adult foxes did not differ significantly between sexes (males: 2.3 ±0.1 kg, females: 2.1 ± 0.1 kg; F = 1.8, d.f. = 1, 39, P = 0.19) or countries (Canada: 2.2 ± 0.1 kg, Mexico: 2.1 ± 0.1 kg; F = 1.2, d.f. = 1, 39, P = 0.28). Body weights of adult coyotes were greater for males than females (males: 15.4 ± 0.8 kg, females: 10.9 ± 0.7 kg; F = 14.4, d.f. =1, 15, P < 0.01) and were similar between countries (Canada: 13.3 ± 1.2 kg, Mexico: 12.9 ± 0.5 kg, F = 2.6, d.f. = 1, 15, P< 0.13). Body weights of combined sexes of coyotes were 6.3 times greater in Canada and 5.9 times greater in Mexico than those of foxes, indicating that body-size parameters determining potential interspecific resource competition were similar between countries.
Survival of foxes in Canada and Mexico—Survival rates of foxes were lower in Canada than Mexico. After 548 days of radiotracking (i.e., the total duration of the Mexican study), none of the 11 kit foxes in Mexico had died, but 51.1% (24/47) of foxes in Canada had died. At this point, survival of foxes was significantly lower in Canada than Mexico for foxes dying of any cause (Z = 16.6, d.f. =1,P< 0.0001). Survival rates of foxes were also lower in Canada than Mexico if only mortalities due to coyotes were considered (Z = 4.9, d.f. =1,P < 0.0001; Fig. 1).
A coyote killed and consumed an uncollared kit fox that had been caught in a leg-hold trap. Although the population of golden eagles at Janos-Casas Grandes is the largest in Mexico (Manzano-Fischer et al. 2006), no foxes were killed by this predator. However, 1 eagle was observed on the ground, attempting to reach a kit fox that had been caught in a leg-hold trap within a den entrance. One year after radiotracking in Mexico ended, 5 marked foxes were seen that had consequently survived for a minimum average of 694.6 days since initial collaring (range: 591–1,051 days). Of 23 foxes in Canada that were radiocollared in January 1995, 1 could not be relocated after September 1997 and the remaining 22 were known to be dead by February 1998. Over the duration of the Canadian study, a staggered entry Cox proportional hazards model incorporating age and sex of the 47 tracked foxes was not significant (likelihood ratio χ2 = 1.0, d.f = 2, P = 0.58), with each of the single factors producing no significant effects. Adult survival rates varied from 0.38 to 0.52 among years (Table 1).
|Year||Age||At risk||Survival rate||95% confidence interval|
|11 January 1995–||Adult||18||0.38||0.15–0.60|
|10 January 1996||Juvenile||8||0.50||0.11–0.80|
|11 January 1996–||Adult||21||0.52||0.29–0.72|
|10 January 1997||Juvenile||8||0.63||0.23–0.86|
|11 January 1997– 10 January 1998||Adult Juvenile||19 3||0.43 0||0.20–0.64 0|
|Year||Age||At risk||Survival rate||95% confidence interval|
|11 January 1995–||Adult||18||0.38||0.15–0.60|
|10 January 1996||Juvenile||8||0.50||0.11–0.80|
|11 January 1996–||Adult||21||0.52||0.29–0.72|
|10 January 1997||Juvenile||8||0.63||0.23–0.86|
|11 January 1997– 10 January 1998||Adult Juvenile||19 3||0.43 0||0.20–0.64 0|
Of the 47 foxes in Canada, 8 died of unknown causes and 8 survived until the end of the study. Of the remaining 31 deaths, the proportion of mortalities due to coyotes was 0.56 in 1995, 0.44 in 1996, and 0.17 in 1997. The proportion of mortalities of foxes caused by golden eagles was 0.22 in 1995, 0.22 in 1996, and then increased dramatically to 0.75 in 1997 (Table 2). With 1 possible exception, coyotes apparently did not consume the swift foxes that they had killed. For coyote-caused mortalities, the carcass was completely intact except for the area of injury in 8 cases (6 retrieved within 7 days, and 2 retrieved within 22 and 19 days, respectively, since the last relocation when the foxes were alive). The fox that may have been scavenged was consumed less than 30% within 3 days of the relocation when it was recorded alive, and 2 others had partially decomposed soft tissues. For 4 eagle-caused mortalities, the carcass appeared intact externally, but talon punctures were found during necropsy (retrieved within 1, 1, 2, and 9 days since the last relocation when the fox was alive). For 5 eagle-caused mortalities, the body was primarily intact but the intestines had been dragged out of the body cavity and partially eaten (retrieved within 0, 1, 1,9, and 17 days since the last relocation when the fox was alive). For 4 eagle-killed foxes, consumption of intestines and the rest of the body was evident (retrieved after 7, 12, 15, and 17 days since the last relocation when the fox was alive).
|Cause of mortality||1995||1996||1997||1998||Total|
|Cause of mortality||1995||1996||1997||1998||Total|
Diet of foxes and coyotes.—Diet assessments were based on 384 fox scats from Canada (538 prey occurrences), 174 coyote scats from Canada (289 prey occurrences), 303 kit fox scats from Mexico (345 prey occurrences), and 76 coyote scats from Mexico (87 prey occurrences; Table 3). Dietary diversity between foxes and coyotes was similar between countries (Horn's indexCan = 1.08; Horn's indexMex = 0.97) and niche breadth was similar for foxes (Shannon's indexCan = 0.67, Shannon's indexMex = 0.76) and coyotes (Shannon's indexCan = 0.77, Shannon's indexMex = 0.77). The country × species × prey interaction was significant (χ2 = 24.2, d.f. =4, P = 0.0001) because occurrences of prey items differed between countries and between species. Two-way interactions showed that occurrences of prey types differed between countries for foxes (χ2 = 18.5, d.f. = 4, P < 0.001) and coyotes (χ2 = 30.0, d.f. = 4, P < 0.0001). Rodents were the most common prey item for foxes and coyotes in both countries, making up more than 65% of prey occurrences, and frequencies of rodent occurrence were similar between species in Canada and Mexico. Insects were more common in fox than coyote scats, whereas birds occurred more frequently in coyote than fox scats of both countries. Foxes consumed lagomorphs more frequently than did coyotes in Canada, but relative occurrences of lagomorphs in scats were similar between species in Mexico.
|Swift Fox (n = 384)||Coyote (n =174)||Kit fox (n =303)||Coyote (n =76)|
|Swift Fox (n = 384)||Coyote (n =174)||Kit fox (n =303)||Coyote (n =76)|
Movement speeds, home-range sizes, and overlaps.—The movement speed of foxes averaged 31.1 ± 1.7 m/min in Canada and 28.6 ±1.1 m/min in Mexico. Coyotes in Canada averaged 47.8 ± 6.7 m/min compared to coyotes in Mexico, which moved at 49.3 ±11.2 m/min. Maximum trajectories of 95% minimum convex home-range polygons differed between species (F = 39.8, d.f. = 1, 64, P < 0.001) and countries (F = 9.2, d.f = 1, 64, P < 0.01), but not in the interaction term (F = 2.7, d.f = 1, 64, P = 0.1). Trajectory distances for foxes spanned 11.9 ± 0.9 km in Canada (n = 36) and 5.2 ± 0.3 km in Mexico (n = 11; t = 7.5, d.f. = 40.7, P < 0.001), whereas those of coyotes were similar between countries (Canada: 18.5 ± 2.0 km, Mexico: 16.6 ± 1.0; t = 0.8, d.f. = 15, P = 0.46). Using convex polygon trajectories and speed averages, independent fix times for subsequent home-range size comparisons were 6.4 h and 3.1 h for foxes in Canada and Mexico, respectively, whereas those of coyotes in Canada and Mexico were 6.5 h and 5.6 h, respectively. Thirty-six foxes in Canada (121.9 ± 17.7 independent fixes), 10 coyotes in Canada (69.2 ± 13.3 independent fixes), 10 foxes in Mexico (83.7 ± 17.5 independent fixes), and 4 coyotes in Mexico (39.8 ± 8.9 independent fixes) were sampled sufficiently for subsequent home-range size comparisons.
Foxes were unable to partition their habitats from coyotes in either country. Home-range overlaps between foxes and coyotes were extensive in Canada and Mexico, and home ranges of foxes in the centers of respective study areas were completely enveloped by home ranges of coyotes (Fig. 2). Home-range sizes of foxes in Canada and Mexico were influenced positively by duration of tracking (fixed kernel: F =6.1, d.f. = 1, 47, P = 0.02; adaptive kernel: F = 5.2, d.f. = 1, 47, P = 0.03) and number of fixes (fixed kernel: F = 3.7, d.f. = 1, 47, P = 0.06; adaptive kernel: F = 4.4, d.f = 1, 47, P = 0.04), but not sex (fixed kernel: F = 0.2, d.f = 1,47, P = 0.66; adaptive kernel: F = 0.2, d.f. = 1, 47, P = 0.66) or age (fixed kernel: F = 0.32, d.f. = 1, 47, P = 0.6; adaptive kernel: F = 0.1, d.f. = 1, 47, P =0.74). Home-range sizes were significantly larger in Canada (fixed kernel: 31.9 ± 4.8 km2, adaptive kernel: 40.8 ±6.1 km2) than Mexico (fixed kernel: 10.4 ± 1.0 km2, adaptive kernel: 11.9 ± 1.0 km2) for foxes (fixed kernel: F = 4.2, d.f =1, 47, P = 0.048, adaptive kernel: F = 5.0, d.f. = 1, 47, P = 0.038; Fig. 3). Foxes in Canada were twice as susceptible to predation in the 51–99% kernel utilization area, where 10 foxes were killed by eagles or coyotes, than in the 50% fixed-kernel core area, where only 5 foxes were killed by these predators. In Mexico, home-range size of foxes was correlated to abundance of prairie dogs. As the percentage of the area of home ranges composed of prairie dog towns increased, home-range size estimated through fixed kernels (F =13.6, d.f = 1, 9, P < 0.01, r2 = 0.63) or adaptive kernels (F = 16.5, d.f = 1, 9, P < 0.01, r2 = 0.67) decreased significantly (Fig. 4).
Home-range overlap between foxes was similar between countries (F = 0.1, d.f = 1, 54, P = 0.77), but differed between mated pairs and neighbors (F = 41.3, d.f =1, 54, P < 0.001); the interaction term was not significant (F = 0.8, d.f. = 1, 54, P =0.38). Mated pairs shared larger proportions of home ranges than neighbors in Canada (mated pairs: 74.3% ± 4.2%, n = 12; neighbors: 28.8% ± 3.8%, n = 31) and Mexico (mated pairs: 84.0% ± 2.9%, n = 2; neighbors: 23.9% ± 5.9%, n = 9).
Differences in home-range size among coyotes could not be explained by duration of tracking (fixed kernel: F = 0.02, d.f = 1, 14, P = 0.89; adaptive kernel: F = 0.1, d.f. =1, 14, P = 0.79) or number of fixes (fixed kernel: F = 1.2, d.f. =1, 14, P = 0.3, adaptive kernel: F = 1.0, d.f. = 1, 14, P = 0.34). Home-range sizes were similar between coyotes in Canada (fixed kernel: 88.9 ± 25.8 km2, adaptive kernel: 109.7 ± 29.9 km2) and coyotes in Mexico (fixed kernel: 80.2 ± 17.3 km2, adaptive kernel: 99.1 ± 18.2 km2; fixed kernel: F = 0.1, d.f = 1, 14, P = 0.80; adaptive kernel: F =0.1, d.f. = 1, 14, P =0.78; Fig. 3). In Canada, 3 radiocollared females and 1 radiocollared male coyote had litters, but their mates were uncollared individuals. In both countries, collared coyotes were not observed hunting together or sharing a den; consequently documented home-range overlaps of coyotes were not between mates. Home-range overlap between coyotes was similar (t = 0.3, d.f =27, P = 0.78) between Canada (30.8% ± 4.2%, n = 21) and Mexico (33.0% ± 5.3%, n = 8).
Movements of coyotes relative to foxes and fox dens—In Canada and Mexico, all utilization areas of coyotes overlapped fox dens. Although in both countries the number of fox dens within utilization areas of coyotes depended upon the proximity of the home ranges of coyotes to the areas in which foxes were studied most intensively, the number of known fox dens overlapped by individual coyotes did not differ between countries (Canada: 38.3 ± 9.3, Mexico: 42.6 ± 11.3; t = 0.29, d.f. =15,P =0.77). Utilization-area sizes did not influence the absolute (F = 1.4, d.f = 1, 15, P = 0.25) or proportional differences (F = 0.4, d.f = 1, 15, P = 0.54) between the observed and expected mean distances of locations of coyotes from fox dens. Using paired Wilcoxon signed-rank tests, observed and expected mean distances of locations of coyotes from fox dens were similar in Canada (Z = 1.5, n =10, P = 0.14) and in Mexico (Z = 1.7, n = 7, P = 0.09), indicating that coyotes in both countries did not seem to target or avoid fox dens. Proportional differences between observed and expected distances were similar between countries (Mann-Whitney U =31, nCan = 10, nMex = 7; P = 0.70), indicating that the movements of coyotes relative to fox dens were similar in Canada and Mexico.
Because foxes and coyotes with overlapping home ranges were frequently several kilometers apart, the relative frequency that foxes and coyotes were in close proximity could not be compared between Canada and Mexico. However, when coyotes and foxes were located within 15 min and within 2 km of one another, distances of less than 1 km for these paired locations were more frequent in Mexico. In Mexico 65.0% of 60 paired locations were within 1 km compared to 38.5% of 244 paired locations in Canada (likelihood-ratio χ2 = 13.7, d.f. = 1, P < 0.001). Foxes in Mexico likely faced a greater number of coyotes during encounters than did foxes in Canada. Coyotes in Mexico were more frequently observed in pairs or groups of at least 3 than coyotes in Canada (Canadian sightings, n = 430): 1 coyote, 0.76; 2 coyotes, 0.19, ≥3 coyotes: 0.05; Mexican sightings, n = 58: 1 coyote, 0.45; 2 coyotes, 0.31, ≥3 coyotes: 0.24; likelihood-ratio χ2 = 27.4, d.f. = 1, P < 0.001).
Foxes in Mexico had a greater density of potential escape holes than did foxes in Canada. Densities of suitable escape holes differed between prairie dog towns in Mexico (57.0 ± 6.1 holes/ha), Mexican grassland areas that were dominated by kangaroo rats (3.7 ±1.4 holes/ha), and Canadian grassland areas (0.9 ± 0.3 holes/ha; Kruskal-Wallis χ2 = 35.8, d.f. = 1, P < 0.001). Although the density of holes in Canadian habitats differed most extensively from prairie dog towns in Mexico, densities of holes were also significantly different from Mexican grassland sites that did not contain prairie dogs (Mann-Whitney U = 174, nCan = 54, nMex = 10, P = 0.02).
Although 48–62% of swift foxes in Canada died annually, no mortalities occurred among radiocollared foxes in Mexico over 1.5 years. Coyotes accounted for up to 56% of annual mortalities of foxes in Canada, whereas survival of foxes in Mexico was not affected by coyotes. This finding illustrates that the outcome of interactions between dominant and imperiled carnivores can vary greatly between populations. Differences in survival rates suggest that kit fox populations in Mexico may have a higher recovery potential than swift fox populations in Canada. Understanding how such differences can arise despite a lack of habitat partitioning may be crucial to facilitate the persistence or recovery planning of other vulnerable carnivore populations.
The frequency of mortalities of foxes caused by golden eagles was far greater in the Canadian study area than in other studies of swift or kit foxes where mortalities caused by golden eagles have been rare (Moehrenschlager et al. 2004). Although golden eagles can potentially drive populations of small canids such as the endangered island fox (Urocyon littoralis) toward extinction (Roemer et al. 2001), the magnitude of such impacts was only evident in Canada in 1997, when eagles were responsible for 75% of mortalities of swift foxes. The fact that mortalities of foxes caused by coyotes were relatively rare during this year may suggest that foxes were more susceptible to eagle predation in 1 year, or that coyote and eagle predation of swift foxes is largely compensatory. If the killing of swift foxes is compensatory, coyote-caused mortalities would have been more frequent in the absence of eagle predation in 1997. Because foxes in Canada that were killed by coyotes were almost never consumed, such killing likely reflects interference competition instead of intraguild predation; in contrast, a greater frequency of consumption of carcasses by golden eagles suggests that such mortalities are primarily predatory.
Survival rates of adult swift foxes in Canada were similar to the 0.40–0.75 range of annual survival rates of adult foxes in the United States (Andersen et al. 2003; Kamler et al. 2003b; Kitchen et al. 1999; Olson and Lindzey 2002; Schauster et al. 2002; Sovada et al. 1998), but the lack of mortalities among radiotracked kit foxes in Mexico was unique compared to 0.44–0.61 annual survival rates of endangered San Joaquin kit foxes (V. macrotis mutica) in California (Cypher et al. 2000; Ralls and White 1995; Spiegel 1996; Standley et al. 1992). Although we do not have density estimates for golden eagles, they were likely more common in the Mexican study area than in Canada because of the high availability of prairie dogs (Ceballos et al. 1999). Two breeding pairs and up to 16 different transient golden eagles have been recorded in a single day within the ranges of kit foxes in the Janos-Casas Grandes grasslands (National Audubon Society 1999). Furthermore, golden eagles are abundant in these areas throughout the year (Manzano-Fischer et al. 1999), whereas they are rare on the Canadian prairies in winter. Although kit foxes were primarily nocturnal, radio-tracking data, observations of dens, the observation of an attempt by an eagle to kill a captured kit fox, and a high proportion of prairie dogs, which are diurnal, in the diet of kit foxes indicate that kit foxes in Mexico were vulnerable to golden eagles, which primarily hunt by day. Because coyotes were consistently observed in the home ranges of kit foxes during spotlight surveys (List and Macdonald 1998), coyotes occasionally hunted within 1 km of simultaneously tracked foxes, and a captured kit fox was killed by a coyote, kit foxes also were vulnerable to being killed by coyotes. This information raises the question of how kit foxes in Mexico were able to elude golden eagles and coyotes.
Higher survival rates of foxes in Mexico than in Canada could not be explained by differences in relative body size compared to coyotes, prey use, intraspecific overlap in home ranges, group sizes, home-range sizes of coyotes, hunting strategies of coyotes, or movement rates of either canid. Because a large proportion of the home ranges of foxes were entirely overlapped by those of coyotes, foxes did not, and probably could not, partition their habitat relative to coyotes in either country. Within these overlap zones, coyotes in both countries did not appear to divert their movements toward simultaneously tracked foxes or their dens. In Colorado, interspecific separation distances between swift foxes and coyotes, which caused 80% of identifiable mortalities of swift foxes, also did not differ from those expected by chance (Kitchen et al. 1999), and the separation distances of San Joaquin kit foxes from coyotes, which caused 65% of mortalities of foxes (Ralls and White 1995), were random in California. These results suggest that the killing of swift and kit foxes may primarily depend on the likelihood of coyote encounters.
Home-range sizes of swift foxes were approximately 3 times larger than those of kit foxes. Home ranges of swift foxes in Canada were larger than those documented elsewhere (Kamler et al. 2003a; Kitchen et al. 1999; Pechacek et al. 2000; Sovada et al. 2001; Zimmerman 1998), but home ranges of kit foxes in Mexico were similar to those recorded in California and Arizona (White and Ralls 1993; Zoellick and Smith 1992). Home ranges of kit foxes in Mexico, which were approximately 8 times smaller than those of sympatic coyotes, were likely overlapped by fewer home ranges of coyotes than those of swift foxes in Canada, which were only 3 times smaller than those of coyotes. Because intraspecific spacing patterns were similar, the likelihood of foxes encountering coyotes was probably higher in Canada than Mexico. Because Canadian foxes were twice as likely to be preyed upon in the outer half of their home ranges than in the home-range core, the wandering associated with larger home ranges also may have placed foxes in Canada at greater risk of coyote and eagle predation than kit foxes in Mexico.
Larger home-range sizes of foxes in Canada suggest that available prey for foxes may have been less abundant in Canada than in Mexico. In a meta-analysis, home ranges of swift and kit foxes were found to decrease with increasing availability of lagomorph prey (White and Garrott 1997). Similarly, in our study, home ranges of kit foxes in Mexico decreased as the availability of prairie dog towns increased. In the Mexican study area, there was an abundance of 15.6 prairie dogs per hectare, 4.4 banner-tailed kangaroo rats (Dipodomys spectabilis) per hectare, 4.1 Merriam's kangaroo rats (Dipodomys merriami) per hectare, as well as tawny-bellied cotton rats (Sigmodon fulviventer), white-throated woodrats (Neotoma albigula), and 3 species of mice (Pacheco et al. 2000). In the Canadian study area, prairie dogs, kangaroo rats, woodrats, and cotton rats were not present, but Richardson's ground squirrels (Spermophilus richardsonii) were potentially available at a greater abundance than the 0.7 spotted ground squirrels (Spermophilus spilosoma) per hectare we documented in Chihuahua. However, Richardson's ground squirrels hibernate from late October to late February on the Canadian prairies (Michener 1998), making them inaccessible to swift foxes. Swift foxes consequently rely on arvicoline rodents during the winter, but Klausz (1997) only caught 163 small mammals in 9,360 trap nights in our swift fox study area from November 1995 to March 1996. This small abundance along with low species diversity (96% were North American deermice [Peromyscus maniculatus] and 4% were shrews [Sorex]), led Klausz (1997) to hypothesize that the late-winter period could lead to high mortalities of swift foxes. We documented 1 starvation in late winter, and swift foxes were likely more food stressed in Canada than Mexico, which probably led to larger home ranges in Canada.
In addition to being a crucial prey item for kit foxes in Mexico, prairie dogs create refuge holes that allow foxes to potentially escape from other predators; in contrast, foxes in Canada primarily rely on the diggings of American badgers (Taxidea taxus—Herrero et al. 1991), a carnivore that is increasingly imperiled in the region. Availability of refuges from predation was significantly higher in Mexican sites that were dominated by kangaroo rats or prairie dogs than in the Canadian study area. This presents a further disadvantage for foxes in Canada. First, foxes in Canada may have encountered coyotes more frequently than foxes in Mexico, and 2nd, fewer refuge holes were available for foxes in Canada to escape from coyotes or golden eagles when they were encountered.
Although differences in fox home-range sizes and the availability of refuge holes may have contributed to higher survival rates of foxes in Mexico than in Canada, the relative importance of these factors is difficult to tease apart. However, it is clear that foxes in Chihuahua, Mexico, had higher survival rates than swift or kit foxes that have been studied anywhere else, and the protection provided by the prairie dog complex, which has recently been threatened by expansions of cropland agriculture (Ceballos et al. 2005), may be crucial to facilitate the recovery of kit foxes in northern Mexico. Experiments that alter prey availability to assess respective effects on home- range sizes of foxes and coyotes and subsequently test effects of encounter probabilities on survival would be desirable, but may be logistically difficult for these wide-ranging species. Restricting access to existing refuge holes would be problematic because many populations of swift and kit foxes are already imperiled, but experiments that increase availability of refuges may be feasible to evaluate potential improvements in survival of foxes. A potential link between the survival of imperiled species and the abundance of animals that make refuges could have profound conservation consequences, because it might also lead to the protection of refuge-creating species. Temporary refuges have been shown to benefit prey species (Ylönen et al. 2003), but such advantages have not been quantified previously for imperiled carnivores.
Further studies should aim to determine causal links between prey abundance, home-range size, refuge availability, and the survival of imperiled species that are otherwise killed by dominant carnivores. We speculate that the abundance and differential exploitation of prey will determine relative home-range sizes of dominant and vulnerable intraguild competitors (in this case, carnivores). The ratio of the home-range sizes of the dominant carnivore to those of the vulnerable carnivores is predicted to increase as the availability of prey that is best exploited by the vulnerable competitor increases. Conversely, this ratio is expected to decrease as the availability of prey best exploited by the dominant competitor increases. We predict that, by determining interspecific encounter frequencies, the ratio of the home-range size of the dominant species to home-range size of the vulnerable species will be positively correlated to the survival of the inferior competitor. We also predict that residual variation from this relationship will be explained by the group size of the dominant competitor, which will be negatively correlated to survival of the vulnerable competitor, and the abundance of potential refuge micro-habitats, which will be positively correlated to the survival of the vulnerable competitor. We encourage modeling approaches to test these predictions, experimental manipulations of refuge and prey availability, and further standardized field studies of similar interspecific interactions between carnivores in different environments.
We thank M. Doughty, M. Eaton, E. Jiménez, G. Johnson, J. Johnson, A. McKinney, J. Michie, C. Moehrenschlager, C. Philcox, and J. Scharlemann for their dedicated fieldwork. D. Biggins, G. Ceballos, P. Manzano-Fischer, B. Miller, and J. Pacheco provided invaluable support on logistics, fund raising, and fieldwork. We thank S. Black for her skillful necropsies, M. Percy for global information system support, and A. South for software consultation. Preliminary versions of the manuscript benefited from constructive comments by D. Bell, L. Bonesi, P. Ferreras, K. Gibson, P. J. Johnson, P. Honess, J. Kamler, A. Loveridge, G. Mason, and J. Murdoch. This project was undertaken while R. List held scholarships from The British Council and Consejo Nacional de Ciencia y Teconología. Grant support was provided by Alberta Environmental Protection, Alberta Sport Recreation Parks and Wildlife Foundation, Canadian Wildlife Service, Comisión Nacional para el Conocimiento y Uso de la Biodiversidad, Direcióon General de Apoyo al Personal Académico-Universidad Nacional Autónoma de México, Express Pipelines, Green Plan International, Husky Energy, Nature Saskatchewan, Parks Canada, People's Trust for Endangered Species, the Rocky Mountain Elk Foundation, Saskatchewan Environment and Resource Management, Swift Fox Conservation Society, University of Alberta Biodiversity Fund, Wildlife Preservation Trust Canada, United States Agency for International Development, and World Wildlife Fund Canada.