Volume 88, Issue 4 e22569
RESEARCH ARTICLE
Open Access

Phase-dependent red fox expansion into the tundra: implications for management

Caitlin Wilkinson

Caitlin Wilkinson

Department of Zoology, Stockholm University, 106 91, Stockholm, Sweden

Search for more papers by this author
Jan Vigués

Jan Vigués

Department of Zoology, Stockholm University, 106 91, Stockholm, Sweden

Search for more papers by this author
Marianne Stoessel

Marianne Stoessel

Department of Physical Geography, Stockholm University, 106 91, Stockholm, Sweden

Search for more papers by this author
Mikael Vinka

Mikael Vinka

Geunja 1, 924 95 Ammarnäs, Sweden

Search for more papers by this author
Anders Angerbjörn

Anders Angerbjörn

Department of Zoology, Stockholm University, 106 91, Stockholm, Sweden

Search for more papers by this author
Karin Norén

Corresponding Author

Karin Norén

Department of Zoology, Stockholm University, 106 91, Stockholm, Sweden

Correspondence Karin Norén, Department of Zoology, Stockholm University, 106 91 Stockholm, Sweden.

Email: [email protected]

Search for more papers by this author
First published: 29 February 2024

Abstract

Expansion of boreal species into tundra ecosystems is a consequence of climate change and human exploitation that threatens local species through increased predation, competition, and pathogen transmission. Under these circumstances, efficient control of expanding boreal species may be necessary, but the efficiency of such action depends on understanding the ecological influences of expansion. The red fox (Vulpes vulpes) is expanding into the tundra across the Arctic. In Scandinavia, red foxes threaten local tundra species and communities including the endangered Arctic fox (V. lagopus). The ecological dynamics in the tundra are influenced by small rodent cycles (classified into different phases based on seasonal abundance fluctuations), which can affect red fox expansion, distribution, and abundance. We used a 17-year (2004–2020) dataset from the tundra in Sweden, consisting of raw snow track data, to test how cyclic prey influenced red fox distribution and abundance, and subsequently red fox control. The winter abundance of red fox was influenced by small rodent phase, with higher abundance during high prey availability (i.e., increased number of prey numbers) with no support for a time lag between red fox and small rodent abundance. This suggests that high prey availability attracts red foxes to the tundra and that higher immigration from the boreal zone can be expected in response to increased prey abundances. There was no relationship between red fox control and small rodent availability, but control was influenced by red fox abundance during the previous year, which highlights an opportunistic control strategy. We recommend an adaptive management strategy where authorities include small rodent dynamics in the planning and execution of red fox control.

Over the past century, climate change has influenced the structure and dynamics of tundra ecosystems (Ims and Fuglei 2005, Post et al. 2009, Gilg et al. 2012). Many boreal species have undergone population increases and extended their distribution through range expansions at higher latitudes in response to a milder climate, longer productive periods, and increased availability of anthropogenic subsidies (Hersteinsson and Macdonald 1992, Elmhagen et al. 2015, Tape et al. 2016b, Gallant et al. 2020). Several boreal invasions have been recorded in tundra ecosystems across a range of different Arctic mammal species (Hope et al. 2015). Further evidence has been seen in red foxes (Vulpes vulpes; Elmhagen et al. 2015), moose (Alces alces; Tape et al. 2016a), snowshoe hares (Lepus americanus; Tape et al. 2016b), and beavers (Castor fiber; Tape et al. 2018).

Even though climate change is often pinpointed as a cause of boreal expansions, there are top-down and bottom-up influences that affect the expansion process. Individuals expand their range when they can no longer predict resources or avoid harsh weather conditions, competitors, and predators (Jonzen et al. 2011, Marneweck et al 2019). Apex predators have a structured effect on lower trophic levels and can suppress meso-predator populations through direct predation or by creating a landscape of fear (Ritchie and Johnson 2009). Additionally, a loss in top predators or competitors can cause increases in meso-predator abundances and range expansions (i.e., meso-predator release; Hellmann et al. 2008, Trewby et al. 2008, Ritchie and Johnson 2009). Further, these influences interact with the climate where anthropogenic-induced release of meso-predators can enhance predation pressure on their prey causing subsequent effects on ecosystems (Prugh et al. 2009, Conner and Morris. 2015, Roos et al. 2018). Bottom-up influences (i.e., resource availability) affect species distributions and expansions, where high prey abundances caused by an increasing number of prey species and size of prey populations attract predators to new habitats (Svenning et al. 2014), which is important in marginal ecosystems such as the transition to Arctic tundra (Neubert et al. 2000, Rød-Eriksen et al. 2022).

The red fox is a widespread opportunistic meso-predator occupying ecosystems globally (Schipper et al. 2008), and it is described as one of the most invasive carnivores out of its original range (Lowe et al. 2000). Northward establishment of the red fox is one of the earliest described and documented examples of boreal expansion, with a poleward range shift of >1,700 km into the Arctic tundra (Hersteinsson and Macdonald 1992, Elmhagen et al. 2017a, Gallant et al. 2020). In Scandinavia, top-down factors connected to anthropogenic activity influence red fox population expansion and abundance (Gomo et al. 2021). These can be direct or indirect and can cause higher resource abundance mediated by increased anthropogenic resource subsidies, reduced top predators, and milder climates (Hersteinsson and Macdonald 1992, Gilg et al. 2012, Elmhagen et al. 2017a). The extirpation of apex mammal predators has altered several tundra predator communities (Prugh et al. 2009, Ritchie and Johnson 2009, Estes et al. 2011), with current densities of large carnivores in Scandinavia being too low to regulate meso-predators (Rød-Eriksen et al. 2022).

The red fox is a generalist predator, with a diet consisting of a range of mammal and bird species, with a large proportion of their diet being microtine rodents (Elmhagen et al. 2002, Warret Rodrigues 2022, Wilkinson et al. 2022). The abundance of tundra red foxes varies across years and seasons, with abundance and distribution of prey being proposed as an important factor influencing their dynamics during winter (Stoessel et al. 2019). Theoretical simulations have demonstrated that fluctuating prey populations have a particularly strong effect on predator expansion rates (Neubert et al. 2000). Addressing the effect of predator-prey interactions in relation to predator range expansion and fluctuating prey abundance is thus of particular importance in the tundra ecosystem, which typically has strong multiannual cyclical dynamics of microtine rodent populations (Elton 1924, Henden et al. 2009). Lemmings (Lemmus spp.) and voles (Myodes spp., Microtus spp.) are key prey species and influence the dynamics of other species in the tundra community (Elton 1924, Ims and Fuglei 2005). These rodent cycles have a duration of 3-5 years and 25-200-fold amplitude (Stenseth and Ims 1993, Henden et al. 2009, Ehrich et al. 2020). During a rodent cycle there are distinct phases classified based on seasonal rodent abundances, which are obtained through bi-annual systematic snap-trapping (Ecke and Hörnfeldt 2020) or winter snow track data. The fluctuating prey abundance can be classified into low (a distinct decline in rodent abundance), increase (higher rodent abundance than low phase but a lower abundance than peak), peak (highest rodent abundance over the cycle), and crash and post-peak (rodent abundance declined after peak but remained relatively high). These may cause a phase-dependent expansion pattern influencing the expansion and establishment of boreal predators (Henden et al. 2020). Understanding the relationship between small rodent dynamics and boreal expansion may be of particular importance under climate change, which is expected to dampen the amplitude of small rodent cycles (Ims and Fuglei 2005, Ims et al. 2008).

Both top-down and bottom-up influences affect Arctic fox (Vulpes lagopus) and red fox distribution and sympatry, and red fox population dynamics have been reported to be independent of small rodent abundance (Rød-Eriksen et al. 2022), contrasting with previous studies (Henden et al. 2010). This may have resulted from prey switching to available alternative food sources such as reindeer (Rangifer tarandus) carrion in winter or the population may be sustained by immigration from source populations (Rød-Eriksen et al. 2022). The boreal forest surrounding the tundra has higher red fox density, which can function as source populations of emigrating foxes (Norén et al. 2017, Elmhagen et al. 2017a). According to population genetic analyses, the expansion process occurs in parallel from multiple boreal sources towards the tundra (Norén et al. 2015), which acts as a meeting point for red foxes from different boreal regions (Norén et al. 2017). But it is unclear whether red fox populations in the Arctic tundra are continuously immigrating from surrounding source populations or if they have established stationary populations. One possible mechanism of boreal predator immigration is areas shifting between boreal and tundra ecosystems, triggered by high small rodent abundance (Henden et al. 2010). In the late increase and peak phase of the rodent cycle, when resources are super-abundant in the tundra, the red fox performs functional habitat switching (Mysterud and Ims 1998, Henden et al. 2010), and invades this the tundra ecosystem (Elmhagen 2003). For instance, red fox individuals, without permanent territories, regularly migrate long distances (40–70 km) into adjacent ecosystems (Butler 1951, Jedrzejewski and Jedrzejewski 1998, Tolhurst et al. 2016). This movement has been described as direct prey tracking (Korpimäki 1994), which may occur with a time lag (Gilg et al. 2006), where we would expect the highest red fox abundance 1 year after a small rodent peak.

The red fox has global relevance for conservation and wildlife management and is now subject to management actions in many places across the globe (Fletcher et al. 20102013, Kämmerle and Storch 2019, Marolla et al. 2019). For >2 decades, red fox culling has been implemented to conserve tundra populations (Elmhagen et al. 2015). Increased abundance of red foxes has wider implications, including increased intra-guild predation and competition with the Arctic fox (Hersteinsson and Macdonald 1992, Ims et al. 2017, Elmhagen et al. 2017a), changes in Arctic fox distribution (Hersteinsson and Macdonald 1992, Dalén et al. 2004, Rodnikova et al. 2011), and behavior (Frafjord et al. 1989). Other effects have been reported from introduction of zoonotic diseases, such as sarcoptic mange (Sarcoptes scabiei; Mörner 1992), rabies virus (Rabies lyssavirus; Steck and Wandeler 1980), and tape worm (Echinococcus multilocularis; Lind et al. 2011). Additional effects on prey communities include dampening of rodent cycles (Hanski et al. 1991) and community wide decline in northern European bird populations (Svensson et al. 1999, Lehikoinen et al. 2014, Breisjøberget et al. 2018, Marolla et al. 2019). Implementing efficient red fox control on the tundra is important to minimize these detrimental effects of expansion. Currently, control of tundra red foxes is implemented by authorities in Sweden, Finland, and Norway with the purpose of minimizing effects on native tundra species (Angerbjörn et al. 2013, Marolla et al. 2019, Henden et al. 2020). In Sweden, control for the purpose of Arctic fox conservation is established, but the intensity varies between different areas and over time (Angerbjörn et al. 2013). At present, there is no clear strategy behind red fox control in Sweden, with control intensity being influenced solely by general workload, economic resources, and visual red fox observations (S. Almroth, County Administration Board, personal communication).

To have an ecological effect, red fox control comes with several challenges such as the number of removed individuals required for biological effects (Bomford and O'Brien 1995, Harding et al. 2001), the seasonal and phase-dependent timing of removal, and the economy and effort needed from the responsible authorities. Researchers using theoretical models concluded that red fox control should be intensified before an expected small rodent peak (Henden et al. 2009). Despite being a long-term action, the intensity of red fox control in the tundra in Sweden is often operated opportunistically, which may limit the ecological and economic efficiency because of the time delay.

The objective of this study was to investigate red fox establishment on the tundra, inter-annual changes in red fox abundance, and how this is related to cyclic prey availability. Based on previous studies, we predicted that small rodent abundance affects the expansion process, and that a high prey abundance will attract red foxes from the boreal forest to the tundra, showing a phase-dependent expansion pattern. If red foxes are established on the tundra and have a self-sufficient stationary population, we expected that tundra red fox abundance would show a time lag in relation to small rodent abundance. If on the other hand there is constant immigration from the boreal forest, we expected a higher population abundance of red foxes during the peak phase of the small rodent cycle. This understanding of establishment should influence required intensity of control actions, which thus far has not been considered by authorities. Furthermore, we address how small rodent cycles and variable red fox abundance are related to red fox control actions to produce recommendations for management.

STUDY AREA

The study area is situated in Vindelfjällen nature reserve, Västerbotten county, Sweden (66°00N, 16°00E). The study area is comprised of 5,600 km2 of land ranging from 500–1,700 m above sea level. The study period was between 2004–2020, during winter and spring, with a September to June snow season. The local weather is characterized as classical tundra conditions, with strong seasonality, short growth season (120 days) and temperatures ranging from an average of −16.6°C in February to 19°C in July. Land is covered by dry heath, grass heath, dry fen, rocks and fern with typical tundra species present in the plant community (Stoessel et al. 2019). Birch (Betula spp.) and forested valleys cross the mountain tundra below the tree line, positioned at 750 m (Elmhagen 2003). Above the treeline the different vegetation species present include wetland areas dominated by water sedge (Carex aquatilis), white cotton grass (Eriophorum scheuchzeri), meadow-grasses (Poa spp.) and Semaphore grass (Pleuropogon sabinei); mesic areas containing saxifrages (Saxifraga spp.), cinquefoils (Potentilla spp.), buttercups (Ranunculus spp.), graminoids (families Poaceae, Juncaceae, and Cyperaceae), shrubs (Salix spp.), and mosses (Bryophyta spp.); and dry heath areas composed of bilberry (Vaccinium myrtillus), crowberry (Empetrum nigrum), and lichens (Lichenes spp.; Le Vaillant et al. 2018). The focal study species are the red fox and microtine rodents, as community dynamics in the study are influenced by the small rodent cycles (Elton 1924, Korpimäki and Krebs 1996, Ims and Fugeli 2005). Four species of microtine rodents are present in the area: Norwegian lemming (Lemmus lemmus), grey-sided vole (Myodes rufocanus), field vole (Microtus agrestis), and bank vole (Myodes glareolus; Elmhagen 2003). Voles mostly follow the demographic pattern of lemmings, with some inter-annual and seasonal variations (Ecke and Hörnfeldt 2020). Additional herbivorous prey present in the area, include mountain hare (Lepus timidus), reindeer (Rangifer tarandus), rock ptarmigan (Lagopus muta), willow ptarmigan (Lagopus lagopus), passerines (Passerinformes), ducks (Anseriformes), gulls, and waders (Charadriiformes; Elmhagen 2003, Stoessel et al. 2019). Other known predator species include the Arctic fox, stoat (Mustela erminea), least weasel (Mustela nivalis), pine marten (Martes martes), wolverine (Gulo gulo), golden eagle (Aquila chrysaetos), white-tailed eagle (Haliaeetus albicilla), rough-legged buzzard (Buteo lagopus), gyrfalcon (Falco rusticolus), kestrel (Falco tinnunculus), long-tailed skua (Stercorarius longiccaudus), and raven (Corvus corax; Svensson et al. 1999, Stoessel et al. 2019). The area is used for reindeer husbandry by local Sami villages. Furthermore, it is a population area for tourism (e.g., hiking and fishing).

METHODS

Rodent and red fox monitoring and control

Västerbotten County has monitored abundance of predators yearly through snow track surveys since 2004 (Stoessel et al. 2019); we used data from 2004–2020. This monitoring follows the wildlife triangle scheme (Lindén et al. 1996). Within the study area, there are 28 randomly distributed triangular transects, which are 12 km in length (i.e., 4 km on each side of the triangle), permanent, and are distributed across the nature reserve (total area of nature reserve is 565,000 ha; Stoessel et al. 2019). Only 2 out of 28 triangles were located below the treeline. Fieldworkers on snowmobiles conducted monitoring in March-April, only on days after new snowfall or wind to ensure detection of recent tracks (Linden et al. 1996). For each snow track present, trained fieldworkers recorded a global positioning system point and identified the focal species based on the overall shape and pace of the tracks (Stoessel et al. 2019). Snow tracking can mainly detect lemming tracks and we therefore use this as a proxy for overall small rodent density. To estimate both red fox and small rodent abundance indexes, we selected 24 of the triangles based on the position and relevant temporal data available, with an average of 20.53 (range = 10 – 24) triangles monitored each year (Figure 1). We calculated an annual abundance index separately for foxes and lemmings based on the number of tracks/km; we also calculated the proportion of surveyed triangles with identified red fox or lemming tracks.

Details are in the caption following the image
The Vindelfjällen nature reserve in northern Sweden, with 3 distinct regions: A) Björkfjället, B) Ammarfjället, and C) Guvertfjället. The triangle transects represent the 24 triangles monitored for red fox and Norwegian lemming snow tracks in winter 2004–2020. The background map was downloaded from the Swedish Mapping Authority (using LM Open Data WMTS plugin on QGIS 3.8.0), with the lighter areas representing treeless alpine tundra and green areas representing forested valleys. Black lines represent the core areas within each region.

There were 5 small rodent cycles from 2004–2020. To classify the phase of the small rodent cycle at the time of monitoring, we used snow tracking data of lemmings. Phases were classified as low, increasing, peak, and post-peak following Meijer et al. (2013; Table 1). A low phase was identified when the previous year was a peak and the present year had a density decline. An increase phase was defined as a density increase relative to the previous year (that displayed a density decline). A peak phase was defined as a year when density increased, following a previous year of increase. There was only 1 post-peak phase in our study: in 2012 lemming densities had declined since 2011 but remained relatively high. This was followed by a rapid crash and low abundance in June 2012 (Ecke and Hörnfeldt 2020). The data set contained 6 low phases, 5 increase phases, 5 peak phases, and 1 post-peak phase (Table 1; Figure 2). Voles mostly follow the demographic pattern of lemmings, with some inter-annual and seasonal variations (Ecke and Hörnfeldt 2020).

Table 1. The phase, description, and corresponding year of the microtine rodent cycle (Norwegian lemming and grey-sided, field, and bank vole) in Vindelfjällen nature reserve in northern Sweden, 2004–2020, and average number of tracks/km for red fox and Norwegian lemmings.
Phase Classification Study year Red fox tracks/km SD Norwegian lemming tracks/km SD
Low A distinct decline in small rodent abundance relative to the other phases 2006, 2009, 2013, 2016, 2017, 2020 0.459 0.429 0.008 0.040
Increasing Small rodent abundance is higher than the low phase but lower than the peak phase 2004, 2007, 2010, 2014, 2018 0.444 0.508 0.052 0.113
Peak The highest small rodent abundance over the cycle, appearing after the increase phase and before the low phase 2005, 2008, 2011, 2015, 2019 0.809 0.710 0.216 0.337
Post-peak Small rodent densities had declined but remained relatively high 2012 0.642 0.516 0.125 0.158
Details are in the caption following the image
Average number of red fox tracks/km (black dots; ±SE) between 2004–2020 in Vindelfjällen nature reserve in northern Sweden. Grey bars show average number of Norwegian lemming tracks/km. The rodent phase for every year was classified into low (L), increasing (I), peak (P), and post-peak (PP).

Red fox control is implemented in the study area by the County Administration Board of Västerbotten and the Environmental Protection Agency focused on red fox removal in tundra through shooting (Angerbjörn et al. 2013, Elmhagen et al. 2017b). Control measures have been carried out by Västerbotten county, during winter and spring, since the early 2000s, with varying intensity of control across the area because of limited personnel and economic resources (S. Almroth, County Administration Board, personal communication).

Data analysis

We conducted all statistical analyses in R 4.0.4 (R Core Team 2013) and the R Commander 2.7 package (Fox 2017). To test the relationship between small rodent and red fox abundance, we used a linear mixed model (LMM) where red fox tracks/km of each triangle was the response variable, rodent phase (classified into categorical low, increasing, peak, and post-peak) was the fixed effect and triangle identity (geographic site) and year were random effects to control for repeated measurement and effects of specific years or localities. We log transformed the data, inspected model residuals to control for goodness of fit, and examined the mixed model with Tukey's post hoc test.

To assess the occurrence of a time lag in the response of red foxes to rodent abundance, we used 2 different approaches. First, we followed the procedure of Smedshaug et al. (1999) where we cross correlated average number of small rodent snow tracks/km and average number of red fox tracks/km using Spearman's correlation test in R 4.0.4 (R Core Team 2013) and the R Commander 2.7 package (Fox 2017). We plotted correlation coefficients for red fox abundance and small rodent abundance with no time lag (t = 0) and 1-year time lags (t = −1, red fox abundance shows a response to previous year's rodent abundance; and t = 1, red fox abundance shows a response to the following year's rodent abundance). Second, we used an LMM, where red fox tracks/km was the response variable, small rodent tracks/km was the fixed effect and triangle identity (geographic site) and year were random effects to control for repeated measurement and effects of specific years. We ran analyses for no time lag (t = 0) and 1-year time lags (t = −1 and t = 1).

To assess whether rodent phase influenced the number of shot red foxes during control measures, we used a Kruskal-Wallis test. To further assess the relationship between red fox occurrence and number of shot red foxes, we used an LMM with average number of red fox tracks/km or proportion of triangles with red fox tracks as fixed effects and area (geographic site) and year as random effects. We used red fox indices for the specific mountain range within the Vindelfjällen nature reserve, where red fox control is reported separately: Björkfjället (n = 10 triangles), Guvertfjället (n = 7 triangles), and Ammarfjället (n = 4 triangles). To account for a 1-year time lag between the number of shot red foxes in relation to red fox abundance the previous year, we repeated the same model using average red fox tracks/km as an estimate of previous year's abundance (t = −1). To account for differences in control strategies and available funding for the action, we included only data from 2010–2020 in the statistical analyses. We interpreted significance values using a gradual notion of evidence described by Muff et al. (2022).

RESULTS

Red fox tracks fluctuated across years, ranging from 0.22 tracks/km ± 0.218 (SD) in 2006 to 1.33 ± 0.993 tracks/km in 2015. The peak phase had twice many red fox tracks (0.809 ± 0.710 tracks/km) as the low (0.459 ± 0.429 tracks/km) and increase (0.444 ± 0.508 tracks/km) phases (Figure 2). During the post-peak phase (2012), 0.642 ± 0.516 tracks/km were recorded. This was 80% of the number of tracks recorded at the peak phase but 40% higher than the number of tracks recorded at the low and increase phases. The proportion of triangles with red fox tracks also varied between years, from 60% in 2004 to 100% in 2012, with an average proportion of 84.7% across years. The peak phase had an average of 91% of monitored triangles (20/24 triangles with red fox tracks, range = 15–23) with red fox tracks, whereas red fox tracks were present in 75% of the monitored triangles during the increase phase (18.2/24 triangles with red fox tracks, range = 10–23), and 85% during the low phase (20.5/24 triangles with red fox tracks, range = 17–23).

Red fox tracks/km was affected by the phase of the small rodent cycle (Figure 3A; χ 3 2 ${\chi }_{3}^{2}$  = 15.681, P = 0.001). A further examination showed very strong evidence that the peak phase was different from the low phase (Tukey's post hoc test P < 0.001), and strong evidence that it differed from the increase phase (Tukey's post hoc test P < 0.01) but no evidence that it differed from the post-peak phase (Figure 3A). Occurrence of red fox and lemmings, measured as average number of tracks/km, displayed weak evidence for a correlation (LMMt=0: t = 1.728, χ 1 2 ${\chi }_{1}^{2}$ = 2.986, P = 0.08). There was no evidence for a negative (t − 1) or positive (t + 1) time lag between red fox and small rodent abundance (LMMt−1: t = −1.170, χ 1 2 ${\chi }_{1}^{2}$  = 1.369, P = 0.242, and LMMt+1: t = 0.624, χ 1 2 ${\chi }_{1}^{2}$  = 0.390, P = 0.532). In agreement with this, red fox abundance showed no evidence for a negative or positive time lag to small rodent abundance according to the cross-correlation test (Figure 3B).

Details are in the caption following the image
A) Average red fox tracks/km and small rodent phase in Vindelfjällen nature reserve in northern Sweden, 2004–2020. Error bars show standard error and asterisks represent significance values (***P < 0.001, **P < 0.01). B) Cross correlation analyses (Spearman correlation coefficient) for average red fox tracks in relation to average lemming tracks for a time lag of −1, 0, and 1 year.

Rangers shot 268 foxes within the reserve from 2004–2020. The yearly number of shot red foxes on the tundra varied between 0 and 81 with an average of 15.7 ± 20.65 (SD) foxes/year between 2004–2020 (Figure 4A). There was no evidence of lemming phase on number of shot foxes ( χ 2 2 ${\chi }_{2}^{2}$  = 1.357, P = 0.507; Figure 4B). The number of shot foxes was comparable between low (13.8 ± 13.4), increase (6.40 ± 6.34), and peak (17.20 ± 15.80) phases (Figure 4B). For the only post-peak phase in 2012 (n = 1), the number of shot red foxes was considerably higher (n = 81; Figure 4B) than the peak phase (4.76 times), the increase phase (12.65 times), and the low phase (5.87 times). Furthermore, there was no evidence of an effect of red fox abundance, measured as average number of red fox tracks/km (LMM: t = 0.551, χ 1 2 ${\chi }_{1}^{2}$  = 0.309, P = 0.582) or proportion of triangles with red fox tracks (LMM: t = 0.925, χ 1 2 ${\chi }_{1}^{2}$  = 0.856, P = 0.355) on shot red foxes. There was moderate evidence for a relationship between number of shot red foxes and red fox abundance the previous year (LMM: t = 1.990, χ 1 2 ${\chi }_{1}^{2}$  = 3.967, P = 0.047), suggesting that the number of shot red foxes during year t + 1 increases with a higher red fox abundance during year t.

Details are in the caption following the image
A) Number of shot red foxes by rangers in Vindelfjällen nature reserve in northern Sweden, 2004–2020, with small rodent phase categorized as low (L), increasing (I), peak (P), and post-peak (PP). B) Average number of shot red foxes in relation to small rodent phase during 2010–2020.

DISCUSSION

The aim of this paper was to investigate if temporal variation in red fox abundance is connected to the cyclic prey availability on the tundra. A fluctuating pattern of winter red fox abundance followed the small rodent cycle, and both red fox and lemming abundance index peaked during 2008, 2011, and 2015. In line with this, there was a greater occurrence of red fox tracks during the peak phase compared to the low and increase phases. Furthermore, our analyses suggested that a direct response to changes in lemming abundance was more likely than a time lag (Korpimäki 1994). The shift of red foxes from boreal forest to tundra occurred during peaking lemming abundance (Mysterud and Ims 1998, Henden et al. 2010) without a time lag and is likely a direct response induced by high prey availability. Henden et al. (2010) suggested that such functional habitat switching is an important mechanism explaining red fox expansion and secondary consequences for other tundra species.

There was almost twice as many red fox tracks/km at the peak phase compared to the low and increase phases, which supports the mechanism of red foxes being attracted to the tundra by high prey availability (Mysterud and Ims 1998, Henden et al. 2010). The greater occurrence of red foxes in winter could also be due to higher survival of red fox juveniles born on the tundra, rather than phase-dependent expansion patterns from the boreal forest. Based on summer monitoring data of all known fox dens in the study area during the same time, Wilkinson (2020) reported that there was no clear difference in occupied red fox dens, or red fox reproductions between different phases. Red fox breeding activities on the tundra are considerably more common in areas without red fox control (Kaikusalo and Angerbjörn 1995, Ims et al. 2017), which means that the red fox control implemented in the area likely limits the number of breeding red foxes. Even though this needs to be more thoroughly evaluated, we suggest that the increased winter abundance of red foxes during the peak phase is mainly influenced by increased expansion rates from the boreal forest, which is supported by other studies from tundra regions (Henden et al. 2010, Gomo et al. 2021). There are limitations to making conclusions based on snow tracking data alone. Snow tracking is an established approach to estimate abundance of wildlife species and this approach is robust in terms of comparing changes in local abundances over time (Stoessel et al. 2019). It should not be regarded as absolute estimates of population density. However, an evaluation of different non-invasive approaches in estimating carnivore abundances in northeastern North America reported that snow tracking is the most efficient approach to detect winter-active species (Gompper et al. 2006). Because lemmings spend a considerable amount of time in the subnivean space, snow tracking cannot be used for estimating absolute abundance (Stoessel et al. 2019), but it can be used to estimate relative abundance and temporal changes.

Cyclic fluctuations of small rodents, and especially lemmings, are characteristic of northern tundra ecosystems and have been a topic of research interest in the last century (Krebs and Myers 1974, Norrdahl 1995, Korpimäki and Krebs 1996). The specific dynamics of a cycle can differ over time (Ehrich et al. 2020), and rodent cycles have been dampening in some parts of Europe (Hörnfeldt et al. 2005) whilst returning in others (Brommer et al. 2010). These alterations may in turn influence boreal expansion patterns (Henden et al. 2020, Gomo et al. 2021). One notable feature in our data set was the high occurrence of red fox tracks during the post-peak phase in 2012. This post-peak phase was characterized by a small rodent decline following the peak in 2011, but prey abundance remained relatively high in 2012. This facilitated a high abundance of red foxes on tundra. The deviant pattern of 2012 is in contrast to the decline in red fox tracks following the small rodent crashes in 2006, 2009, and 2016 when crashes occurred rapidly. This highlights the importance of phase-specific prey dynamics in terms of amplitude, duration, and timing of crashes to fully understand the process of boreal expansion (Henden et al. 2020).

In Scandinavia, there is conflicting evidence regarding the role of small rodent fluctuations and tundra red foxes. Small rodent cycles have influenced red fox scavenging patterns in Norway, with increased number of red fox visits at baits after a small rodent peak (Gomo et al. 2021). Further, a link between small rodent red fox abundance was also reported in Sweden (Stoessel et al. 2019); however, in both cases environmental conditions were important factors influencing fox abundance. Conversely, tundra red fox abundance was not explained by small rodents in Norway; abundance was associated with competing predators and environmental variables (Cano-Martínez et al. 2021) and higher densities caused by high human settlement density and anthropogenic subsidies (i.e., gut piles from moose hunting or reindeer carrion; Jahren et al. 2020, Rød-Eriksen et al. 2022). In similar Canadian systems, the speed and magnitude of red fox invasions (Gallant et al. 2020), movement strategies of red foxes (Warret Rodrigues and Roth 2023), and red fox population growth (Verstege et al. 2023) were influenced by a resource availability (i.e., anthropogenic subsidies or seasonal fluctuations of prey). Winter conditions, including warmer temperatures, snow depth, and snow cover, were also stated as important factors in each case (Gallant et al. 2020, Verstege et al. 2023, Warret Rodrigues and Roth 2023). Thus, resource availability plays an important role in red fox population dynamics, but a contextual awareness and consideration of environmental factors is needed when making conclusions on patterns of expansion.

Resource constraints directly influence red fox populations; however, top-down influences connected to anthropogenic activity also have effect (Gomo et al. 2021). Top-down regulation of red foxes has been reduced in tundra regions because of intense lethal control of large predators and reduced hunting pressure from humans, allowing population growth and spread (Selås and Vik 2006, Pasanen-Mortensen et al. 2013). In Scandinavian tundra regions, apex predator species are either functionally extinct (i.e., grey wolf [Canis lupus]; Elmhagen and Rushton 2007, Rød-Eriksen et al. 2022), have too low densities for structing potential (i.e., wolverine and golden eagle; Rød-Eriksen et al. 2022), or their distribution does not reach the Arctic or sub-Arctic (i.e., lynx; Anon 2015, Breitenmoser 2015). Increased exploitation competition through shared resources or interference competition from negative interactions between species (i.e., intraguild predation) are documented in this context (Ims and Fuglei 2005, Henden et al. 2010, Killengreen et al. 2011, Morin 2011). Interference competition between meso-predators in the tundra favors the larger red fox compared to the Arctic fox, where several cases of intraguild predation have been documented (Tannerfeldt et al. 2002, Pamperin et al. 2006). Although top-down influences can have significant structuring effects on tundra communities in the summer, they have been shown to have weak effects in winter abundance distributions, with bottom-up effects being critical for boreal predator species (Stoessel et al. 2019). Red foxes are flexible and are expected to switch to alternative resources when rodents are scarce, using additional prey and anthropogenic food subsidies (Selås and Vik 2006, Killengreen et al. 2011, Šálek et al. 2015, Carricondo-Sanchez et al. 2016, Willebrand et al. 2017). Although the importance of other food resources (i.e., reindeer carrion) can influence the population dynamics of red foxes in the Arctic tundra (Killengreen et al. 2011, Angerbjörn et al. 2013, Rød-Eriksen et al. 2022), there is currently no report on the abundance or availability of carrion, and scavenging of carrion by red foxes in our study area. The importance of stable food sources, especially in regions with minimal top-down regulation, highlights our recognition of fluctuating prey and resource constraints as influencing red fox populations, and in turn their expansion.

The number of red foxes on the tundra is influenced by the small rodent dynamics and the highest number of winter red foxes is expected during a small rodent peak. Red fox control varied considerably over time and we did not observe an apparent relationship to prey abundance. Instead, a trend suggesting a correlation between the number of shot red foxes and previous year's red fox abundance, supports the opportunistic control strategy described by rangers. The post-peak phase in 2012 had a high number of shot foxes. This is likely a consequence of the late and slow small rodent crash following the 2011 peak, which facilitated 2 consecutive years with high red fox abundance on the tundra, with an increased outtake during the second year. Thus, such a time-lagged and opportunistic control strategy may not be sufficient to efficiently control tundra red foxes. Another aspect that should be highlighted is whether red fox removal in the surrounding forest region can influence the abundance of red foxes on the tundra through source-sink dynamics or density-dependent dispersal (Norén et al. 2017). Red fox management is a strongly debated topic (Baker et al. 2002, Leader-Williams et al. 2002, Aebischer et al. 2003) and red fox culling is an ineffective strategy to reduce population numbers in some study areas, where human control was opportunistic and non-selective (Comte et al. 2017). The efficiency of control strategies depends on the method used and the spatio-temporal design of the practice (Comte et al. 2017), with comparable control measures having contrasting results in different regions (Heydon and Reynolds 2000). Efficient management practices focus on juvenile foxes (McLeod and Saunders 2001) during winter (Rushton et al. 2006, Lieury et al. 2015). An evaluation of a red fox control program in Australia demonstrated that ≥50% of the adults and 25% of the juveniles need to be removed to prevent a further population increase; however, for long-term effects, it was necessary to also control red foxes immigrating into the control area (Harding et al. 2001). In any scenario, removal of a substantial proportion of a population across large areas and over long periods is necessary, to generate ecologically meaningful output and prevent rapid recolonization (Newsome et al. 2014, Comte et al. 2017). This is particularly challenging in a cyclic ecosystem where small rodent dynamics and subsequent expansion rates may vary between areas. For local effects, it is therefore necessary to make priorities to maximize the biological output of the action. Sweden has a national small rodent monitoring program, funded by the Swedish Environmental Protection Agency, and conducted by the County Administration Board twice a year (Ecke and Hörnfeldt 2020), which can be used for future management strategies.

MANAGEMENT IMPLICATIONS

We recommend that red fox control efforts in Sweden should take small rodent dynamics into account and intensify actions, especially at high prey abundances, using data from the national small rodent monitoring program to predict the phase of the rodent cycle the following year. Furthermore, control efforts at high prey density should be planned based on the direction of the rodent cycle and executed across administrative boundaries to increase efficiency. Given that the influx of red foxes can be limited through efficient and well-planned control, breeding red foxes are probably best controlled through directed efforts adapted to prey dynamics at specific dens during late spring. When sufficient knowledge on prey abundance is not available, different strategies may be required (i.e., intensive control across large spatial and temporal scales with a focus on juveniles during winter) to have effective reductions on red fox abundances.

ACKNOWLEDGMENTS

We are grateful to S. Almroth and M. König for providing data on red fox control. We thank the editors and anonymous reviewers for their valuable comments, which improved our manuscript. The study was financed by the Wildlife Management Fund (802-0199-18) by the Swedish Environmental Protection Agency, Felles Fjellrev Nord II, and the County Administration Board in Västerbotten.

    CONFLICT OF INTEREST STATEMENT

    The authors declare no conflicts of interest.

    ETHICS STATEMENT

    We carried out data collection in Vindelfjällen nature reserve with a specific permit from the County Administration Board in Västerbotten (521-3191-2014 and 521-4640-2019). We drove snowmobiles off trails with allowance from the County Administration Board in Västerbotten. Lethal red fox control was accomplished by rangers at the County Administration Board in Västerbotten for management purposes.

    DATA AVAILABILITY STATEMENT

    Data is deposited in Dryad: doi:10.5061/dryad.pg4f4qrxg

    • Associate Editor: Francesco Ferretti.