Skip to main content

Feast to famine: Sympatric predators respond differently to seasonal prey scarcity on the low-Arctic tundra

Cite this dataset

Warret Rodrigues, Chloé; Roth, James D. (2023). Feast to famine: Sympatric predators respond differently to seasonal prey scarcity on the low-Arctic tundra [Dataset]. Dryad.


Resource fluctuation is a major driver of animal movement, influencing strategic choices such as residency vs nomadism, or social dynamics. The Arctic tundra is characterized by strong seasonality: resources are abundant during the short summers but scarce in winters. Therefore, expansion of boreal-forest species onto the tundra raises questions on how they cope with winter-resource scarcity. We examined a recent incursion by red foxes (Vulpes vulpes) onto the coastal tundra of northern Manitoba, an area historically occupied by Arctic foxes (Vulpes lagopus) that lacks access to anthropogenic foods, and compared seasonal shifts in space use of the two species. We used 4 years of telemetry data following 8 red foxes and 11 Arctic foxes to test the hypothesis that the movement tactics of both species are primarily driven by temporal variability of resources. We also predicted that the harsh tundra conditions in winter would drive red foxes to disperse more often and maintain larger home ranges year-round than Arctic foxes, which are adapted to this environment. Dispersal was the most frequent winter movement tactic in both fox species, despite its association with high mortality (winter mortality was 9.4 times higher in dispersers than residents). Red foxes consistently dispersed towards the boreal forest, whereas Arctic foxes primarily used sea ice to disperse. Home range size of red and Arctic foxes did not differ in summer, but resident red foxes substantially increased their home range size in winter, whereas home range size of resident Arctic foxes did not change seasonally. As climate changes, abiotic constraints on some species may relax, but associated declines in prey communities may lead to local extirpation of many predators, notably by favoring dispersal during resource scarcity.


Capture and satellite telemetry. – Between 2017 and 2019, we captured 10 red foxes and 13 Arctic foxes using Tomahawk (Model 208, Tomahawk Live Trap Co., WI) and padded leghold traps (Softcatch # 1.5, Oneida Victor Ltd, USA). Traps were placed on active dens or by protruding features (e.g., driftwood or spruce islets) and remained open continuously for up to one week. We checked the traps every 4–6 hours and closed them during extreme weather conditions (e.g., blizzard or temperatures below -25°C). We captured adult foxes from March to May when snow still covered the ground and facilitated travel over large distances, except for two adult foxes caught near our field camps in June 2018. We did not anesthetize the foxes, which were easily handled without chemical restraint. Foxes were first wrapped in a blanket and released from the traps, then we assessed sex and body condition, deployed an Iridium satellite collar (#4170 or 4270, Telonics, Mesa, Arizona, USA; ~100g, i.e., 2–4% of a fox body mass), and released them at the site of capture. All handling procedures were approved by the University of Manitoba Animal Care Committee (Protocol F17-012), and the research was conducted under Parks Canada Research and Collection Permits WAP-2017-25781 and WAP-2018-27938, and Manitoba Wildlife Scientific Permits WB20226 and WB21856.

Movement analysis. – Our GPS collars used different schedules throughout the year (see Table S1), so we thinned all the tracks by randomly selecting 1 location per day (the lowest fix frequency). We defined two relevant contrasting periods based on goose phenology. The season of abundant resources (hereafter summer) thus extended from May 15, the approximate date of nest initiation, to the end of October, the last month during which geese can be considered alternative prey for the foxes of this area (Andersen et al. 2010; McDonald et al. 2017). The resource-scarcity period (hereafter winter) extended from November 1 to May 14, when geese are absent and foxes mostly rely on arvicoline rodents.

We plotted all fox tracks in ArcGIS 10.3 (ESRI 2017, Redland, CA, USA) to remove possible major erroneous locations and identify movement strategies: residency and long-range movements. We labelled a fox as a resident only if it maintained a home range (i.e., showed non-directional movements within a geographically circumscribed area) from the start of a given season until the end of that season or until its death, if it occurred after the area resulting from movement analysis had reached an asymptote. We never included the season of capture in movement tactic and home range comparisons, since we could not know if a fox was dispersing earlier that season. Using a subset of 16 individuals with 111 to 187 locations each, we determined that home range areas reached an asymptote with 38 locations on average. All our resident foxes exceeded this threshold with at least 61 locations. All foxes that underwent long-range movements (hereafter dispersals) were considered dispersers since none returned to their departure area (they either died dispersing or settled elsewhere). The dispersal events we used to compare movement tactics were not natal dispersal because we only included adults (at least 1.5 years old), unlike the track descriptions, which included all available tracks.

For each dispersal (including those initiated during the season of capture that were not included in any other analysis) we calculated the cumulative distance travelled (i.e., sum of straight-line distances between successive daily relocations), the duration (starting with the last position within the home range boundaries), the cumulative to straight-line distance ratio (a proxy for fox behavior during dispersal), the cardinal direction (the angle of the vector between first and last locations, degrees from due North), the main substrate used for movement (sea ice or land), and the average daily speed. We considered that the dispersal started with the last location in a home range prior to dispersal initiation, or at the point of capture if a fox did not exhibit residency prior to dispersal (and thus was likely captured while already dispersing), and ended with the first location associated with a settlement of >7 days in a new delimited ranging area (on land, not ice) or with the death of the fox. Although foxes can exhibit staged dispersal, exploring delimited areas for a temporary period ranging from a few days up to a few weeks (e.g., Walton et al. 2018), we never observed clear staging behavior.

We estimated residents’ home ranges and core areas, defined as the 95% and the 50% utilization distribution isopleths, respectively, with local convex hulls (LoCoH) using the package T-LoCoH v.1.40.07 in R (Lyons et al. 2013). LoCoH are nonparametric estimates of utilization distributions and perform better than parametric kernel methods to identify boundaries (such as coastlines) and unused areas (Getz et al. 2007; Stark et al. 2017). As such, they are well-suited for our main objective to determine if red foxes were using the sea ice. We were not specifically interested in the temporal partition of space within seasons since we modelled space use using only one location per day. We, therefore, set the user-defined parameter s to 0, which entailed that the time-scaled distance was equivalent to the Euclidian distance (Getz et al. 2007) to heterogeneous location densities, we used the adaptive method (a-LoCoH). We selected the a value for each animal using the graph tools provided in the T-LoCoH package and following the recommendations to minimize the risks of both excluding used areas and including unused areas. To estimate seasonal home-range shifts in each fox, we measured summer and winter home range overlaps using the package v. 1.34.00/r12 and the distance between their centroids estimated in ArcGIS 10.3 (ESRI 2017). All home ranges and core areas are displayed in Supplemental file 1. Based on the same dataset, we also estimated home ranges (95% utilization distribution) using a classic bivariate kernel density estimator (KDE) with a reference bandwidth, with R package adehabitatHR v.0.4.20 (Calenge 2006). Although we decided not to use kernel density methods in this study, we provide the areas resulting from the KDE in Table S2, for comparison purposes.

Many residents undertook short-distance and short-duration trips outside the boundaries of their home range, either on land or on the sea ice. We defined excursions as any exploratory movement <7 days unusually far away from the current center of activity followed by a return to the home range. Home-range borders include areas that are already peripheral to the center of activity. Therefore, to avoid making arbitrary decisions on a distance threshold to the border, we differentiated excursions from other movements near the home range border, based on the distribution of the distances between a location and the home-range centroid. Locations that appeared to be outliers using a one-sided Hampel filter (upper bound = median (Tukey-transformed distance) + 3 median absolute deviations) were considered excursions. If a trip outside the boundaries of the LoCoH home range estimate consisted of multiple consecutive locations, we used the farthest away of the consecutive locations to determine if that trip was an excursion. Finally, we called “commuting trip” any excursion on the sea ice (Lai et al. 2017).


Natural Sciences and Engineering Research Council

National Geographic Society

Polar Continental

University of Manitoba Fieldwork Support Program

Churchill Northern Studies Centre Northern Research Fund

Oakes-Riewe Aboriginal-Environmental Studies Research Award