Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
Nina E. Eide, Norwegian Institute for Nature Research, Fakkelgården, N-2624 Lillehammer, Norway, Tel: +47 95 70 43 83; Fax: +47 61 22 22 15; E-mail: firstname.lastname@example.org
1Home range size, spatial organization and territoriality of reproductive Arctic foxes were studied during the summer. The influence of spatial distribution and availability of the main prey was investigated in order to evaluate whether the spatial organization of Arctic foxes was coherent with key predictions of the resource dispersion hypotheses (RDH). The RDH includes the spatial characteristics of resource abundance, while there is also growing attention to the importance of the temporal characteristics of resource abundance. Hence the role of temporal and spatio-temporal predictability of prey explaining carnivore spatial organization was also investigated in this study.
2The study was conducted on Svalbard; a simple High-Arctic terrestrial ecosystem which allowed unique estimates of prey abundance. The main prey of the Arctic fox (Svalbard reindeer Rangifer tarandus platyrhynchus, seabirds Alcidae and Procellariidae and geese Anseridae) was surveyed systematically. These surveys revealed highly contrasting patterns in prey abundance within the terrestrial ecosystem.
3Arctic fox summer home ranges varied in size (4–60 km2), as well as in overlap (17–76%). The diverse spatial organization covaried with spatial and temporal patterns in prey abundance. Small home ranges (10 ± 5·6 km2) with large overlap (76 ± 19·6%) were characteristic for coastal areas where prey was concentrated in small patches and predictable both in space and time. Medium home ranges (23 ± 4·2 km2) and overlap (50 ± 6·6%) occurred inland where prey was clumped in larger patches and less predictable. Large home ranges (52 ± 8·4 km2) with little overlap (17 ± 3·5%) occurred inland where prey was widely scattered and unpredictable.
4Spatial dispersion and richness of prey resources explained most of the variation in Arctic fox spatial organization. The RDH framework could be used to explain the presence of relaxed territoriality found in this study. We suggest that the observed absence of more permanent social groups is due to the extremely severe winter conditions which force juvenile individuals to disperse from the natal area during the first winter.
5Predictability of resources was an additional significant factor affecting both home-range size and overlap. Resource predictability captures the degree to which an animal can depend on its environment to offer suitable and secure living conditions over time.
6This study emphasize the need to incorporate both spatial and temporal characteristics of resource distribution in order to fully understand the diversity of spatial arrangements among carnivores.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Distribution and abundance of resources have been common factors in numerous attempts to explain the spatial organization and dynamics of animals (see, e.g. MacArthur & Pianka 1966; Macdonald 1983; Carr & Macdonald 1986; Maher & Lott 2000). Macdonald (1983) reviewed several studies of medium-sized carnivores and formulated the resource dispersion hypothesis (RDH). According to the RDH, home-range size is determined by the spatial distribution of resource patches: size increases with increasing patch dispersion. The number of individuals that are allowed within a home range is, however, constrained by total abundance of resources (richness). The RDH hypothesizes that because resources have a patchy distribution, the minimum number of patches required to sustain a breeding pair will sometimes support additional individuals and hence allow social groups to form (Carr & Macdonald 1986). While the RDH was developed originally to explain carnivore social behaviour and presence of group living, it may also explain patterns in spacing behaviour among non-social carnivores (Carr & Macdonald 1986; Johnson et al. 2002).
The RDH takes into account both the distribution and abundance of resources, but it does not consider explicitly resource predictability. It is suggested, however, that predictability of resources could be an important factor affecting animal spatial organization (see review in Maher & Lott 2000). In highly stochastic environments food availability often varies considerably seasonally and/or interannually. Although the influence of resource variability has been discussed with regard to spatial organization in mammalian carnivores (see von Schantz 1984; Carr & Macdonald 1986), few empirical studies are available. Maher & Lott (2000) hypothesize that for most species, the benefits of territorial behaviour should increase with increased predictability of resources, whereas for food-caching species they hypothesized the opposite; the extent of territorial behaviour should decrease with increased predictability. In this study we distinguish between temporal predictability and spatio-temporal predictability of resources. Temporal predictability captures the pure between-year variation in resource availability, while spatio-temporal predictability includes the variation in spatial arrangements of resources.
Distribution and abundance of food in space and time is difficult to measure. Consequently, most studies of carnivore spacing behaviour use habitat (e.g. vegetation types, terrain) as an indirect measurement of both food dispersion and richness, assuming that patches of preferred habitat reflect both the distribution and abundance of food resources (see, e.g. Kruuk & Parish 1982; Macdonald 1983; Woodroffe & Macdonald 1993). Prey distribution may, however, be highly dynamic within similar habitats and veil animal habitat use (Jepsen et al. 2002). This reduces the credibility of habitat as a factor structuring animal spatial organization. In most cases, the distribution of habitats per se is constant, which also limits estimates of temporal patterns in availability of food resources. This might explain why resource predictability has rarely been considered in empirical studies attempting to explain spatial organization of carnivore populations.
The High-Arctic terrestrial ecosystem found on Svalbard has a simple community structure and a surveyable landscape, highly suitable for studying predator–prey relationships. The Arctic fox (Alopex lagopus L.), together with the glaucous gull (Larus hyperboreus), are the only predators, and their prey base is limited to a few prey species such as various colony-nesting seabirds, geese and reindeer. In addition, highly contrasting patterns in spatial and temporal distribution of prey over relatively short gradients make the area suitable for studying and testing predictions related to intraspecific spatial organization of carnivores.
In this study we examined the spatial organization of reproducing Arctic foxes during summer, exploring in particular how the size and overlap of home ranges related to the spatial distribution, availability and predictability of prey resources. We predict that home-range sizes of Arctic foxes will vary with the spatial distribution of prey − with small home ranges occurring where prey sources are clumped and large home ranges where prey sources are dispersed. Furthermore, we predict that home-range overlap should increase with increasing prey abundance. Moving beyond the RDH, we also explore the role of temporal and spatio-temporal predictability of prey. Arctic foxes are highly dependent on cached food stores; hence, following Maher & Lott (2000), benefits from defending a territory should increase with decreasing predictability. Thus, we predict that home-range overlap will decrease with decreasing predictability in prey availability, whereas home-range sizes will increase.
Materials and methods
study area and species
The Svalbard Archipelago (62·700 km2) is located in the High Arctic (74°−81°N, 10°−30°E). The study area (1000 km2) was located on the west coast of Spitsbergen, one of the most productive regions of the whole archipelago (Fig. 1). Two major U-shaped valleys, Adventdalen and Sassendalen, cut through the study area (Fig. 1). The landscape is mountainous (< 1000 m) and moderately glaciated.
Several large colonies of sea birds dominated by fulmar (Fulmarus glacialis), Brünnich's guillemot (Uria lomvia), little auk (Alle alle) and puffins (Fratercula arctica) are found in the study area (see Fig. 1). Pink-footed geese (Anser brachyrhynchus) and barnacle geese (Branta leucopsis) breed in large numbers in inland river canyons in the eastern part of the study area (Sassendalen). The diversity and density of other birds is otherwise low, restricted to snow buntings (Plectrophenax nivalis), purple sandpiper (Calidris maritima) and a few other waders (Calidris spp.). All avian species except the Svalbard rock ptarmigan (Lagopus mutus hyperboreus) leave the archipelago in winter. The terrestrial mammalian fauna consists of the Arctic fox and the endemic Svalbard reindeer (Rangifer tarandus platyrhynchus). Except for a small population of introduced voles (Microtus rossiaemeridionalis) living outside the study area, rodents are absent on Svalbard.
A total of 15 adult Arctic foxes (10 females and five males) were tracked during two study periods: 1987–88 and 1997–2000 (Table 1). Of these, 14 had cubs. Based on recorded numbers of litters and litter sizes in the study area, and data on the proportion of reproducing foxes and juvenile foxes in a trapped material of foxes, the density of foxes in the study area in late summer has been estimated to be 1–1·5 foxes per 10 km2 (Prestrud 1992d). The density in the second study period seems to have been in the same order of magnitude (Nina Eide, unpublished data). The number of reproducing adults constituted between 12 and 26 per year. Details on trapping and radio tracking of three foxes captured during 1987–88 are reported in Prestrud (1992a); these data are included and reanalysed in this paper. Seven of the 12 animals captured in 1997–2000 were caught using a net released by a remote-control device. The remaining animals were trapped using a Swiss snare system (Breitenmoser et al. 1993; Nybakk et al. 1996), which consists of a spring-mounted, plastic-coated foot snare. Both trapping methods demanded continuous monitoring from a distance, and foxes were removed from traps in < 5 min. Foxes were marked with coloured ear-tags and fitted with VHF radio-collars (150 Mhz, Telonics MOD-125 or MOD-305, Telonics Inc., AZ, USA) that had an expected longevity of 12 months.
Table 1. Mates, sex, age, litter size, mass, dates of capture and relocation of 15 reproducing Arctic foxes radio-tracked on Svalbard, Norway and in 1987–88 and 1997–2000
Foxes were radio-tracked on foot or by boat using portable receivers and hand-held antennae. In 1987–88, the three foxes were located approximately once every 2–3 h in a subset of intensive periods (Prestrud 1992a). During 1997–2000 we attempted to locate the animals every 15 min during 24-h periods over 2–3 consecutive days. The summer season was defined as 16 June–15 October, We located foxes by taking cross-bearings or triangulations. Locations were classified according to accuracy, and locations based on a bearing-angle less than 30° or more than 150° were not included in the analysis. Through field tests in representative topography we determined the accuracy of locations to be ±300 m. We used the ranges v computer package (Kenward & Hodder 1996) to estimate home ranges defined by 95% minimum convex polygon (MCP). More sophisticated measures (e.g. kernel or harmonic mean) might be more realistic representations of the area actually used by an animal, but they are difficult to compare between studies because the specific settings (e.g. grid resolution, smoothing parameter) under which they are calculated are rarely given. As long as MCP home ranges are based on a sufficient sample to reach an asymptote size, this measure is the least sensitive to choice of software and initial settings (Lawson & Rodgers 1997). However, for comparison we also present 100% MCP and harmonic mean estimates. The home-range calculations were based on locations separated by a minimum of 4 h. Locations separated by > 4 h were independent (calculated by the autocorrelation option in ranges v). In accordance with the mean error of tracking positions, the resolution option in ranges v was set to 300 m, as recommended by Kenward & Hodder (1996). Only home ranges (n = 15) based on 30 fixes or more were included in the statistical analysis. No individuals caught during the first study period (1987–89) were present during the second period (1997–2000). When an individual was followed for > 1 year, as was the case for a few individuals during the second study period, data from both years were included only if this individual moved to a different territory between years. We performed statistical analysis only on MCP95 home ranges.
Only reindeer, seabirds and geese were considered in analyses of prey distribution because in conjunction they constituted more than 80% of the summer diet of foxes in this study area (reindeer: 32%, seabirds: 33% and geese: 15%; Eide 2002). Although foxes use rock ptarmigan as prey, especially during winter (Prestrud 1992b), scat analyses show that this species occurs in < 3% of the scats of both coastal and inland foxes during summer (Eide 2002). Other prey items such as snow bunting, waders, ducks, gulls and seals occur in less than 5% of the remains in both stomach and scat analysis (Prestrud 1992b; Frafjord 1993; Eide 2002).
Intensive prey surveys were conducted only at < 400 m a.s.l. because more elevated areas were not considered to be fox habitat (Prestrud 1992c; Eide, Nellemann & Prestrud 2001). We recorded all prey observations on a 0·5 × 0·5 km grid map (n = 2528 grid cells). However, because the resolution of the aerial surveys used to estimate reindeer distribution (see below) was judged to be slightly coarser than this, we adopted a common spatial scale for all surveys; 2 × 2 km quadrats (n = 240 grid cells).
Live adult reindeer cannot be killed by foxes. Consequently, only reindeer carcasses contribute to available prey biomass. In Adventdalen, systematic ground surveys (total counts) have been conducted annually along contiguous transects separated by 1 km during 7 days in mid-July since 1980 (Nicholas Tyler, personal communication). As part of this survey all reindeer carcasses (natural deaths) are located and classified according to sex and age. The population fluctuates between 400 and 800 individuals (including juveniles), with total annual mortality rates ranging from 1 to 35% (Tyler & Øritsland 1999). In the entire study area positions of reindeer carcasses within fox home ranges were determined by global positioning system (GPS) during systematic ground surveys along transects separated by 200 m. During the reindeer hunt (August–September), remains from 70 to 90 hunter-killed reindeer constitute a large, temporary food source for Arctic foxes in parts of the study area (note: Adventdalen is closed to hunting). Surveys were made during and after the hunt to determine the locations of these kills. Carcasses of reindeer calves are consumed rapidly or transported back to the breeding dens and are therefore rarely found on the tundra. Minimum numbers of dead reindeer calves were therefore estimated based on the numbers of calf-leg bones present at fox dens.
All seabird colonies, except for the largest, were surveyed annually from 1988 to 2000 as part of a seabird monitoring program (SCRIB 1998). The surveys followed standard field methodology as outlined in Walsh et al. (1995). Breeding colonies of geese were located by systematic searches along transects separated by approximately 5 km in all areas. The abundance and distribution of breeding geese were estimated in early July by counting nests immediately after the geese had left the breeding colonies.
Based on the contrasting patterns in distribution and availability of these three most important prey species, the study area was partitioned into three resource landscapes: (1) inland areas with only reindeer present (Adventdalen); (2) inland areas with both geese, reindeer and reindeer slaughter remains present (Sassendalen); and (3) coastal areas dominated by seabird colonies with reindeer and some breeding geese present.
spatial distribution of prey
A prey dispersion index (SD) was calculated from counts of prey in each 2 × 2 km quadrat. Quadrats > 400 m altitude were assigned zero values for all species. To identify clustering of prey, we conducted a block-size analysis of variance (anova: two-term local quadrat variance analysis (ttlqv) after Ludwig & Reynolds 1988). This program groups contiguous quadrats in increasingly larger patches (two and two quadrats, four and four quadrats, eight and eight quadrats, etc.) and calculates a variance for each patch size. Plotting variance against patch size provides information on scale and pattern (see Ludwig & Reynolds 1988); peaks in variance indicate clustering of quadrats with abundant prey at that patch size. We compared block-size anovas for reindeer carcasses, reindeer slaughter remains, nesting geese and seabird colonies to determine degree of covariance with prey spatial distribution patterns and Arctic fox home-range sizes. In order to perform further statistical tests on the relationship between prey spatial pattern and Arctic fox spatial organization, a continuous variable, based on prey peak variances, was calculated for each home range. Peak variance reflects intensity of spatial pattern, i.e. the degree of clumping (Ludwig & Reynolds 1988). For foxes with several prey species available within their home ranges, a ‘total-prey peak variance’ was calculated by summing up prey peak variances (log-transformed), scaled by their relative contribution to total prey abundance (see below) within each home range.
Estimates of available prey biomass (kg) were based on estimates of counts of prey species present within each Arctic fox home range (from surveys described above). For reindeer, estimates of abundance were based upon the number of dead reindeer (carcasses, slaughter remains and dead calves) only. The number of avian prey (adults, eggs and chicks) available to the predator community was estimated from the total count of seabirds and geese present in each home range, which in turn was calculated by the percentage of breeding and non-breeding birds, species-specific reproduction and mortality data according to the following equation:
where (nadb) is the number of breeding adults, (nadnb) is the number of non-breeding adults (qadsm) is ‘at site’ adult mortality, (njuv) is the number of eggs laid per breeding pair and (qjuv) is the breeding mortality, or the mortality from egg until fledged chick. Data values and literature references used in eqn 1 are shown in Table 2. Owing to similarities in life-history parameters, the small number of puffins and little auks were added to the Brünnich's guillemot data.
Table 2. Life history data used as input in eqn 1, to estimate the available number of avian prey
Input in formula
Mortality until fledging from nest (hence a minimum estimate of juvenile seasonal mortality).
2 Barnacle geese and pink-footed geese were treated as one species.
As biomass represents more clearly the importance of each prey type the number of the different available prey types was recalculated in relation to their average mass (kg) as follows: reindeer carcass 20 kg, reindeer slaughter remains 10 kg, reindeer calves 4 kg, geese 1·5 kg, seabirds 0·5 kg and eggs 0·1 kg (Nicholas Tyler, personal communication; Mehlum & Gabrielsen 1995; Tombre et al. 1996). Biomass values were multiplied by the number of the different prey and summed to give total available biomass (AB; hereafter prey abundance) for each home range. Density of prey abundance (D) is AB divided by the individual Arctic fox home-range size.
predictability of prey
The temporal predictability (TP) of prey describes the between-year variation in total prey abundance within a fox home range. The spatio-temporal predictability (STP) describes the between-year variation in the spatial distribution of prey abundance within the home range. TP was calculated based on the prey abundance in each home range in each year. Numbers were log-transformed and the coefficient of variation (CV) was used as an index of temporal predictability of prey. When several prey types were present in a single home range an average index was calculated by summing the values for each prey species, scaled by their relative contribution to total prey abundance within that particular home range. Thus if estimates of total prey abundance in a given home range vary little between years, temporal predictability is high. STP was based on counts of the different prey species (reindeer carcasses, geese and seabirds) in each 2 × 2 km quadrat over 3 years, 1998–2000, except for the largest seabird colony counted only in 1988 (Scrib 1998). Only quadrats with complete censuses of reindeer carcasses during all 3 years were used in the calculations. The between-year correlation coefficient (rs) was calculated for each prey type and the average rs over all 3 years was used as an index of spatio-temporal predictability of prey. A high STP value thus indicates that good squares (high prey abundance) are consistently good between years, while poor squares (low prey abundance) are consistently poor. When several prey types were available in a single home range, an average index was calculated by summing the values for each prey species, scaled by their relative contribution to total prey abundance within that particular home range.
We performed the statistical analysis in minitab (release 13·0). anova and t-statistics compared home-range sizes in the three resource landscapes and between sexes. Spearman's rank order correlation tests estimated spatial predictability of prey. To evaluate the relative importance of variables related to resource abundance (AB and D), resource dispersion (SD) and resource predictability (TP and STP) for home-range size and overlap, we performed linear regression analysis. The five variables – prey spatial distribution (SD), total prey abundance (AB), prey density (D), temporal predictability (TP) and spatio-temporal predictability of prey (STP) – were correlated to a varying degree (Table 3). Hence we evaluated only single parameter models. Models were compared using Akaike's information criterion corrected for small sample sizes (AICc) and the strength of evidence for each model was determined using Akaike weights.
Table 3. Pearson correlation coefficient (r) for the five variables used to describe prey availability and distribution. Prey spatial distribution (SD), total available prey biomass (AB), prey biomass density biomass/km2 (D), temporal predictability of prey (TP) and spatio-temporal predictability of prey (STP). Bold-type figures are significant at α = 0·05
Home range size data (n = 15)
Home range overlap data (n = 6)
Male and female home ranges did not differ in size: Area 2: t = 0·00, P = 1·0 (n = 3 females and 3 males); Area 3: t = 1·24, P = 0·30 (n = 3 females and 2 males). Hence in the following analyses all data are pooled. Reproducing Arctic foxes in the study area occupied summer home ranges (MCP95) varying from 4 to 60 km2 (Table 4). Mean home-range sizes were 52 ± 8·4 km2 (n = 4) in Area 1, 23 ± 4·2 km2 (n = 6) in Area 2 and 10 ± 5·6 km2 (n = 5) in Area 3. Home-range size differed significantly among the three resource areas (F[2,12] = 52·9, P < 0·001, n = 15).
Table 4. Reproducing Arctic fox home-range (HR) estimates (km2) and movement distances from den site (km) during summer on Svalbard, Norway (100% and 95% minimum convex polygons (MCP) and harmonic mean estimates (HM)). Bold numbers refer to data used in the statistical analyses. Only fixes separated by > 4 h were used in home-range calculations
spatial dispersion, abundance and predictability of prey
Seabirds, and to a lesser extent geese, exhibited a clumped spatial distribution, numerous in a few small areas (Fig. 2). Offal from the reindeer hunt also had a clumped distribution pattern, whereas carcasses resulting from natural mortality had a dispersed distribution. Both seabirds and breeding geese occurred in clumped distributions of small patches of 4 and 16 km2, respectively. Breeding geese were also clumped in larger patches of 36 km2. Reindeer slaughter remains occurred in larger patches with variance peaking at 44 km2 (Fig. 2). Home-range sizes concurred with their prey-distribution patterns (see shaded areas in Fig. 2): small patch sizes → small home ranges; large patch sizes → large home ranges; even distribution → very large home-range sizes.
Home-range size decreased with an increased degree of clumping (Fig. 3a) and increasing prey density, but not with total prey abundance (Fig. 3b,c). Home-range size decreased both with increasing temporal and spatio-temporal predictability of prey (Fig. 3d,e). The regression analysis (Table 5) indicated that resource dispersion was the most important variable determining home-range size followed by variables related to resource predictability.
Table 5. Linear regression models to evaluate the relative importance of variables related to resource dispersion (SD), resource abundance (AB and D) and resource predictability (TP and STP). The strongest evidence is in favour of the model with the highest Akaike weight
HR size = SD
HR size = AB
HR size = D
HR size = TP
HR size = STP
HR overlap = SD
HR overlap = AB
HR overlap = D
HR overlap = TP
HR overlap = STP
Home ranges of mated pairs of foxes (see Table 4) overlapped by 61 ± 8·7% (n = 8). Home ranges of neighbouring reproducing foxes overlapped by 17 ± 3·5% (n = 2), 50 ± 6·6% (n = 2), 76 ± 19·6% (n = 2) for Areas 1, 2 and 3, respectively. Home-range overlap between reproducing fox neighbours increased significantly with increasing degree of clumping of prey resources (Fig. 4a), with increasing total prey abundance (Fig. 4b), increasing density of prey (Fig. 4c) and increasing spatio-temporal predictability of prey (Fig. 4e). The regression analysis (Table 5) showed that spatio-temporal predictability (STP) was the most important variable determining home-range overlap, followed by resource density (D).
Home-range sizes of reproducing Arctic foxes in this study varied by a factor of 10 along a coastal–inland gradient. This variation was caused mainly by considerable differences in distribution and predictability of prey. Along the coast, where home ranges were smallest, prey appeared in highly clumped distributions at high densities. Hence the likelihood of finding rich food sources on the same spot between years was high. In inland areas close to the coast, where seabird cliffs were not present, geese and reindeer slaughter remains occurred in clumped distributions at lower densities. Here food availability was less predictable than along the coast and home ranges were larger. In inland areas further from the coast without breeding geese or bird cliffs, the distribution and availability of reindeer carcasses varied considerably from year to year and consequently food availability was both temporally and spatio-temporally unpredictable. Here, home ranges were largest.
Predictability of prey resources was also an important factor in explaining the variation in Arctic fox home-range sizes. Few empirical studies have considered the predictability of prey availability as a factor explaining spatial organization in terrestrial carnivores. Experiments on lower taxa suggest that predictability of food availability is an important factor structuring spatial organization of animal populations (see review in Maher & Lott 2000). Carr & Macdonald (1986) have, however, explored the influence of temporal variation in resource availability in theoretical models of uncertainty structuring the spatial organization of medium-sized canids. According to Carr & Macdonald (1986), a certain number of resource patches must be included in a territory to ensure a critical probability that sufficient food is available over time. In an environment where prey availability may vary considerably in space and time, additional food or habitat patches must be included in the territory in order to reach this critical probability. With some precaution, due to the small sample size, this appears to be the situation in the inland parts of our study area (Adventdalen: area 1), where the overwinter mortality of reindeer may change annually from below 10 up to 200 animals (Tyler & Øritsland 1999). Even though prey abundance varied by a factor of 25 between individual inland Arctic fox home ranges, the home-range sizes were approximately similar. The large home ranges found in the inland areas where prey abundance is unpredictable are also in agreement with the ‘obstinate strategy’ (von Schantz 1984). The ‘obstinate strategy’ states that an individual who experiences resource fluctuation over a prolonged period should maintain a territory size sufficient to meet its needs during years of resource scarcity. Our results supports the idea that foxes maintain home ranges large enough to provide sufficient food also in years of food scarcity. Lack of responses in home-range size to considerable changes in local prey availability has also been reported for other carnivores (Kruuk & Parish 1982; Doncaster & Woodroffe 1993). The ‘obstinate strategy’ does not conflict with the RDH. It does, however, contribute to explain why there was no relationship between total prey abundance and home-range size (Fig. 3b).
Home ranges of neighbouring Arctic foxes overlapped significantly in coastal areas, but were almost exclusive in inland areas. This indicates weaker territorial behaviour among foxes living in areas where prey resources are clumped, highly abundant and predictable. Based on restricted overlap between home ranges, other telemetry-based studies have concluded that Arctic foxes are territorial (Eberhardt et al. 1982; Hersteinsson & Macdonald 1982; Angerbjörn et al. 1997; Anthony 1997; Strand et al. 2000). Our study revealed that degree of territorial behaviour could be highly variable and largely related to spatial and temporal availability of prey.
Arctic fox home-range overlap increased with increasing prey density, indicating a higher tolerance among foxes when food is abundant. This is in accordance with the prediction of the RDH, and suggests that the mechanism that, in the RDH framework, allows formation of social groups in non-cooperative carnivores might in other cases lead to relaxed territoriality. The Prey Renewal Hypothesis (PRH; Waser 1981) makes a similar prediction. According to the PRH, rapid renewal rates of prey may mean that additional individuals can be tolerated within a territory at little expense to the territory owner. In the system studied here, however, the only prey type likely to have a very high renewal rate is also the one that occurs with a very high abundance and predictability (cliff nesting birds). Effects of abundance vs. renewal rates will thus be indistinguishable. Incidental congregations of highly tolerant adult foxes around temporary prey resources, as was found in this study, emphasize the flexibility in Arctic fox spatial organization. Relaxed territoriality has also been found in other carnivores (red fox: Tsukada 1997; Macdonald et al. 1999; bat-eared foxes Lotocyon megalotis: Lamprecht 1979).
Arctic fox caching behaviour suggests that they are dependent on cached food stores for winter survival. When prey is highly abundant, resident foxes should therefore invest in caching food rather than defending food resources, even more so if the high abundance is temporally restricted (such as during the fledgling of Brünnich's guillemot chicks). This might explain why coastal foxes had very high tolerance to non-resident foxes. Non-resident foxes also cache food within the territory of the residents (personal observation), hence the resident foxes might even benefit from allowing non-residents inside their home ranges during events of vast prey abundance. By forcing non-residents out later, resident foxes could obtain exclusive access to more cached food than they would have been able to store alone. Red foxes display a similar response with increased territorial defence following abrupt changes in prey abundance (Tsukada 1997).
Referring to the RDH framework, the question remains why Arctic foxes on Svalbard do not form more permanent social groups. Despite several hundred hours of observation at den sites, social groups of adult Arctic foxes were never observed in this study. Juveniles of both sexes from the previous year were seen within the parent's territory on a few occasions, but never for any length of time. The presence of additional reproducing females within the home ranges of the radio-collared foxes could be excluded. The highest biomass estimates within Arctic fox home ranges were, however, most probably more than sufficient to sustain additional individuals. In other parts of the species’ range, adult Arctic foxes have appeared in social groups, with several females occurring inside a male home range (Hersteinsson & Macdonald 1982; Frafjord 1991; Strand et al. 2000). What separates Svalbard from the areas where social group formation has been documented are the extremely severe winter conditions. Most probably the bottleneck of resource availability is too narrow for juvenile foxes to remain even in the richest natal territories over the winter. By the time patch richness increases (when the birds return to the cliff in late winter/early spring) juvenile offspring have already dispersed. Surplus food can thus only be utilized by neighbouring resident individuals, resulting in overlapping home ranges rather than true social groups.
This study has shown that both spatial and temporal characteristics of prey abundance must be incorporated in order to fully understand the diversity of spatial arrangements among carnivores. Spatial dispersion and richness of prey resources explained most of the variation in Arctic fox spatial organization, consistent with key predictions in the RDH. The mechanisms in the RDH also proved to be useful explaining the presence of relaxed territoriality found among Arctic foxes in this study area. The RDH does not, however, consider explicitly the predictability of prey. Resource predictability captures the degree to which an animal can depend on its environment to offer suitable and secure living conditions over time. This is clearly an important component to consider in future attempts to disentangle the external factors shaping carnivore social organization.
Assistance from numerous volunteers, students and friends was essential to the success of this study. Special thanks to C. Nellemann, A. Derocher, J. E. Swenson, A. Angerbjörn and J. D. C. Linnell for constructive input and discussion that significantly improved the manuscript and the analysis. S. M. Brainerd and Lynn P. Nygaard kindly revised the language. R. A. Ims, P. Wegge, P. Hersteinsson, Ø. Wiig, T. Coulson and one anonymous referee gave valuable comments on drafts of this manuscript. Funding for this project was provided by the Norwegian Polar Institute, the Agricultural University of Norway, the University Courses on Svalbard (UNIS), the Research Council of Norway and the Directorate for Nature Management. The Governor of Svalbard and the logistic department of the Norwegian Polar Institute provided important logistic assistance. All methods used for trapping and marking animals in this study were approved by the Norwegian Animal Research Authority.