Introduction

Interspecific interactions influence species abundance and community composition (Morin 1999). Changes in the abundance of some species or the introduction of exotic species influence changes in species abundance and community composition (Schmitz et al. 2000; Courchamp et al. 2003; Kagata and Ohgushi 2006).

Deer (Cervidae) have recently expanded their range and increased in abundance worldwide, and their impact on natural ecosystems has been dramatic (Côté et al. 2004). Heavy foraging by deer contributes to the modification of vegetation, which also affects animals in various taxonomic groups (Flowerdew and Ellwood 2001; Russell et al. 2001; Rooney and Waller 2003; Côté et al. 2004; Ripple and Beschta 2006; Gill and Fuller 2007; Takatsuki 2009; Ripple and Beschta 2012). In addition, deer affect medium- to large-sized mammals belonging to higher trophic levels in several ways. For example, some carnivores take advantage of scavenging deer carcasses (Jędrzejewski et al. 1989; Selva et al. 2005). Conversely, the black bear (Ursus americanus) population on Anticosti Island may decline because of over browsing of berries by white-tailed deer (Odocoileus virgininianus) (Côté 2005). Interspecific interactions among overabundant deer and other species (particularly mammals at higher trophic levels which often occur at lower densities) can inform effective management practices to preclude potential causes of a decline in these species.

The red fox (Vulpes vulpes), which is distributed across the northern hemisphere, has the widest geographical range of any member of the order Carnivora (Macdonald and Reynolds 2004). Although red foxes are widely distributed in Japan, several populations of the species have been designated as endangered, vulnerable, or near threatened on the Red List at the regional level due to anthropogenic habitat modification (Tsukada 2016). It has also been reported that the number of red foxes decreased due to sarcoptic mange in some regions of Japan (Tsukada et al. 1999; Uraguchi 2008). Furthermore, increasing populations of sika deer (Cervus nippon) may have effects on red foxes in Japan. Small rodent populations, which are important food for red foxes (Koganezawa and Kurokawa 1983; Yamamoto 1994), have declined due to the decrease in understory vegetation and leaf litter caused by sika deer grazing (Suda et al. 2003; Tanaka et al. 2006). Nishio et al. (2013) pointed out that deer density could negatively affect the population of red foxes through a decline in the prey abundance of foxes caused by the modification of habitat structure by deer grazing. Understanding how red foxes respond to changes in food availability and habitat structure due to deer grazing would contribute to effective management plans, which are essential for the conservation of threatened red fox populations in Japan.

In Oku-Nikko, the population of sika deer has been increasing since 1984 and habitat structure has been substantially modified by deer grazing (Koganezawa and Satake 1996; Hasegawa 2008), and a large deer-proof fence was constructed in 2001 to protect vegetation. Similar to other areas, the number of rodents in Oku-Nikko was reported to decrease outside the fence (Koganezawa et al. 2013). However, Seki and Koganezawa (2010, 2013) demonstrated an increasing abundance of earthworms and insects (such as dung beetles) outside of the fence, probably due to the modification of understory vegetation and/or deposition of dung from the increasing sika deer population. As these invertebrates and rodents are primary food resources for red foxes in Japan (Uraguchi 2009), we would expect that the red fox diet would vary accordingly. Furthermore, since red foxes scavenge the carcasses of sika deer (Koganezawa and Kurokawa 1983; Yamamoto 1994; Tsukada and Nonaka 1996; Inagaki et al. 2020), increasing this food source should increase the consumption of deer carcasses by red foxes.

In this study, it was hypothesized that red foxes outside a deer-proof fence would feed more on sika deer, earthworms, and ground-dwelling insects, but less on rodents than inside the fence. To test this hypothesis, the diet of red foxes inside and outside the fence was compared using fecal analysis. In addition, to assess the influence of differences in resource abundance on the population of red foxes, we compared the relative densities of red foxes computed via spotlight counts inside and outside the fence.

Study area

The study was conducted in Oku-Nikko (36° 45′ N, 139° 25′ E) in Nikko National Park, central Honshu, Japan, which is located in a cool temperate zone. The altitude in the study area ranged from 1270 to 1668 m. The density of sika deer, estimated using the block count method during the summer, ranged from 0.0 to 0.4/km2 in 1981–1982 (Mizuno et al. 1982) and 11.5–37.4/km2 in 1995 (Koganezawa and Satake 1996). To protect vegetation, the 17-km deer-proof fence was constructed by the Ministry of the Environment of the Government of Japan in 2001, enclosing an area of 9.8 km2. The mean ± standard deviation (SD) sika deer density, estimated by using the block count method during the study period (2006–2011), was 8.8 ± 5.3/km2 during the summer outside the fence (Tochigi 2014) and 2.6 ± 2.4/km2 during the autumn inside the fence (computed from the data of the Ministry of the Environment of the Government of Japan (2014)). Although the mean density of sika deer outside the fence did not change significantly from 2006 (11.04/km2) to 2011 (10.17/km2), the density inside the fence decreased from 7.2 to 1.2/km2 during the same period (Ministry of the Environment of the Government of Japan 2014; Tochigi 2014).

Although the trapping sites of rodents were different between years, the trapping rates (the number of trapped individuals/trapping nights) of Japanese field voles (Microtus montebelli) during summer inside and outside the fence respectively were 0.138 (124/900) and 0.011 (4/360) in 2009 and 0.088 (57/650) and 0.003 (4/1300) in 2011; those of Japanese field mice (large Japanese field mice, Apodemus speciosus, and small Japanese field mice, A. argenteus) were 0.114 (103/900) and 0.086 (31/360) in 2009 and 0.102 (66/650) and 0.058 (75/1300) in 2011 (Suda and Morita 2014; KO and YS, personal observations).

The understory inside the fence was dominated by dwarf bamboo (Sasa nipponica), an important forage plant for sika deer (Takatsuki 1983, 1986; Yokoyama and Shibata 1998). Outside the fence, the understory was dominated by plants unpalatable to sika deer (e.g., Aster ageratoides ssp. leiophyllus) and bare ground (Koganezawa and Satake 1996; Hasegawa 2008). The vegetative canopy was dominated by deciduous broad-leaved trees and conifers both inside and outside the fence (Koganezawa and Satake 1996). According to data collected between 1981 and 2010 by the Nikko Weather Station (1292 m), the mean annual temperature was 6.9 °C, ranging from a mean of − 4.1 °C in January to a mean of 18.7 °C in August, and the mean annual precipitation was 2176 mm. The ground usually was covered with snow from mid-December until mid-April.

Methods

Collection of fecal samples

A 15.6-km route inside the fence and an 11.0-km route outside the fence were patrolled for fox fecal samples at least once a month from June 2009 to November 2010 and from May to September 2011. Fresh red fox fecal samples found during the sampling of raccoon dog feces from June 2006 to June 2010 (Seki 2011) were also included. The feces of red foxes were discriminated against from those of other carnivores by the size, shape, color, and smell (Yamamoto 1994; Koike et al. 2012).

The area was not completely fenced, with seven non-fenced points at the road and river crossings. Although a section of the sampling route was located within 500 m of non-fenced points, red fox fecal samples were not collected within this range to obviate the possibility that foxes fed inside and then defecated outside the fence or vice versa. The mean ± SD distance from fecal collection points to the non-fenced points was 1015.1 ± 307.2 m (n = 129) inside the fence and 1979.7 ± 1009.5 m (n = 124) outside the fence (50 fecal samples found from May to September 2011 were excluded because the locations of fecal samples were recorded as only inside or outside the fences). The mean core area of the red fox home range in Honshu, Japan (the same subspecies as in the current study), was 1.27 ± 1.02 km2 (SD) in Mt. Nyugasa and 1.40 ± 0.90 km2 in Mt. Ashio (Takeuchi and Koganezawa 1992; Yamamoto 1996); assuming the shape of their core area was circular, the mean radii of the core area were presumed to be 594.6 ± 232.0 m (SD) and 667.9 ± 241.3 m, respectively, which were lower than the mean distances between the fecal sample points and non-fenced points in this study. The 50 excluded fecal samples were > 500 m from the non-fenced points, and the possibility that red foxes had fed inside the fence and defecated outside the fence, and vice versa, was considered to be low.

Fecal analysis

The fecal analysis was done as previously described (Yamamoto 1994; Kaneko et al. 2009; Seki et al. 2014), which involved a procedure for detecting the presence of earthworms. The fecal samples were stored in a freezer at − 20 °C until analysis. Defrosted fecal samples were washed through a 0.5-mm mesh sieve with 1000 mL of water, and materials remaining on the sieve were preserved in 70% ethyl alcohol for identification. Food items were identified to the fullest taxonomic level possible. In addition, the rinsing water was collected into a container (height approximately 85 mm; length 130 mm; and width 140 mm) and left undisturbed for ≥ 15 min. Five milliliters of the sediment was withdrawn using a pipette, diluted with 10 mL of water, and examined for the presence of earthworm chaetae.

Data analysis

Fecal collection dates were divided into winter (January–March), spring (April–June), summer (July–September), and autumn (October–December). The frequency of occurrence (FO) was calculated for each food item in the feces: FO = 100 × number of fecal samples containing a specific food item/total number of fecal samples. Of food items occurring in red fox fecal samples, we targeted Japanese field voles, mice (including large Japanese field mice and/or small Japanese field mice), insects (Scarabaeidae, Carabidae, and Rhaphidophoridae), and earthworms, which were analyzed by Koganezawa et al. (2013) and Seki and Koganezawa (2013), and sika deer.

To assess the effect of location (inside or outside the fence) on the occurrence of each food item, a generalized linear mixed model (GLMM) was used, assuming a binomial error distribution and using a logit link function. Separate models for each type of food item (voles, mice, deer, Scarabaeidae, Carabidae, Rhaphidophoridae, and earthworms) were created. We also created a model of increased insects, including Scarabaeidae and Rhaphidophoridae, which were reported to increase outside the fence in the study area (Seki and Koganezawa 2013). These included the presence or absence of each food item in the feces as a response variable, location (inside or outside the fence) as a fixed effect, and month and year as random effects to correct for repeated sampling. Analyses were conducted in R version 3.2.0 (R Core Team 2015) using the “glmer” function of the “lme4” package.

To assess the fixed effects, the best model was selected by comparing Akaike’s information criterion (AIC) between models with and without location (full and null models, respectively). If the difference in AIC values between the full and null models was less than 2, we considered the two models as competing models (Burnham and Anderson 2002). In addition, odds ratios and their associated 95% confidence intervals (CIs) were estimated in the full models to assess the extent of the fixed effects. Here, odds are represented as the ratio of the probability that the event of interest occurs to the probability that it does not. Odds ratios are the ratios between the odds of two groups; the odds ratio is 1 when there is no relationship (Bland and Altman 2000; Kasuya 2012). For example, an odds ratio > 1 indicates an increased occurrence of food outside the fence.

In this study, the sample sizes in each season were quite different among years. It is well documented that voles can have very dynamic population cycles in Hokkaido, northern Japan (Saito et al. 1998; Stenseth et al. 2003). We also observed some differences in the trapping rates of both voles and mice between years (see the Study area section). However, the ratios of the number of their scat samples to sample sizes during each season were not significantly different among years (Fisher’s exact test, all seasons for voles and mice, P > 0.268; winter could not be analyzed due to small sample sizes). Thus, there could be few biases in the FOs of rodents caused by yearly differences in the sample sizes.

We also analyzed data from spotlight counts (n = 176) conducted between May and November during 2006–2011 by the Laboratory of Wildlife Management of Utsunomiya University to compare the relative densities of red foxes inside and outside the fence. The spotlight surveys were conducted over a fixed 9.0-km route (3.4 km inside the fence and 5.6 km outside the fence; see more details for the method in Seki et al. 2014). To minimize the effects of differences in vegetation on the probability of detecting red foxes between inside and outside the fence, we compared the number of red foxes observed on the road per 10 km (sighting rates) in each survey. As the perpendicular distance of red foxes from the road was not recorded two times (for three individuals) inside the fence and three times (for four individuals) outside the fence, we excluded these data from the analysis. A significant difference, which was set at P < 0.05, was determined between inside and outside the fence using the Wilcoxon signed-rank test in R version 3.2.0 (R Core Team 2015).

Results

A total of 303 fecal samples (159 samples inside the fence and 144 outside the fence) were analyzed (Table 1). Animal remains (mammals, birds, reptiles, fish, insects, Acari, and earthworms), plant remains (seeds, grasses, Musci, and small pieces of leaves, stems, and roots), and artificial remains (wrapping foil and elastic) were found (Table 2). In June 2011, a hoof of a sika deer fawn was found in one fecal sample. Based on the mean FOs of the four seasons, the most frequently occurring animals were Japanese field voles (63%) followed by Rhaphidophoridae (10%), sika deer (10%), Carabidae (8%), and birds (8%) inside the fence, and Japanese field voles (45%), sika deer (29%), Rhaphidophoridae (14%), Carabidae (13%), earthworms (13%), birds (8%), and Scarabaeidae (8%) outside the fence (Table 2). In addition, other frequently occurring food materials inside and outside the fence (respectively) were fruits such as Actinidia spp. (12% and 12%), Vitis coignetiae (8% and 7%), and Malus toringo (7% and 13%), grasses (11% and 5%), and artificial materials (12% and 9%) (Table 2). Although FOs of other small plant materials were relatively high (22% and 6%), these may have been accidentally ingested by red foxes while feeding. The FO of sika deer was relatively higher in spring than in other seasons; the FO for each month during spring was 60% (3/5) in April, 43% (6/14) in May, and 63% (10/16) in June.

Table 1 Number of fecal samples of red foxes collected inside and outside a deer-proof fence from 2006 to 2011 in Oku-Nikko, Japan
Full size table
Table 2 Frequency of occurrence of each food item in the fecal samples of red foxes inside and outside a deer-proof fence per season in Oku-Nikko, Japan
Full size table

The FOs of target food items in each season are shown in Fig. 1. The AIC values for all full models except one mouse model were lower than those for null models (Table 3). The 95% CIs of odds ratios were > 1 for deer (6.68–64.10), increased insect (1.14–3.87), Carabidae (1.09–4.31), and earthworm (1.99–13.67) models (Fig. 2), indicating that their occurrence tended to be higher outside than inside the fence (Fig. 3). However, for the Scarabaeidae and Rhaphidophoridae models, ∆AIC was less than 2 between full and null models and the 95% CIs of odds ratios included 1, while the proportions of values > 1 were relatively large (0.81–4.65 and 0.87–3.48, in Scarabaeidae and Rhaphidophoridae models, respectively; Fig. 2). On the other hand, the 95% CIs of odds ratios for the vole model were lower than 1 (0.38–0.96; Fig. 2), indicating that their occurrence was lower outside than inside the fence (Fig. 3).

Fig. 1
figure 1

Seasonal changes in the frequency of occurrence of the primary food resources of red foxes inside and outside a deer-proof fence in Oku-Nikko, Japan. Fecal samples were collected during the period June 2006 to September 2011. Figures in parentheses give the number of years of fecal sample collecting. Error bars represent the standard error of the mean

Full size image
Table 3 Estimation of a fixed effect (location outside a deer-proof fence) and Akaike’s information criterion (AIC) values obtained by generalized linear mixed model
Full size table
Fig. 2
figure 2

Odds ratios and 95% confidence interval for estimation of a fixed effect (location outside a deer-proof fence) in generalized linear mixed models. Increased insects include Scarabaeidae and Rhaphidophoridae

Full size image
Fig. 3
figure 3

Estimation of the frequency of occurrence (FO) of the primary food resources of red foxes and their 95% confidence interval in generalized linear mixed models inside and outside a deer-proof fence in Oku-Nikko, Japan. The FO values were back-transformed to the 0–100 scale. Increased insects include Scarabaeidae and Rhaphidophoridae

Full size image

The observed numbers of red foxes on the road were 5 inside the fence and 20 outside the fence. Sighting rates, the mean of which was 0.084 ± 0.037 (the standard error of the mean) inside the fence and 0.203 ± 0.047 outside the fence, were not significantly different between inside and outside the fence (Wilcoxon signed-rank test, V = 95, P = 0.178).

Discussion

Red fox diet

Red foxes are adaptable and opportunistic omnivores, with a diet ranging from invertebrates to mammals, birds, and fruit (Macdonald and Reynolds 2004), with small mammals as their main food resource (e.g., Uraguchi 2009; Murdoch et al. 2010; Días-Ruiz et al. 2013; Spencer et al. 2014; Tsukada et al. 2014). We found that red foxes ate small mammals, sika deer, insects, earthworms, and fruits in Oku-Nikko (Table 2). Red foxes mainly consumed Japanese field voles, whereas mice were not fed on frequently (Table 2). In Oku-Nikko, the trapping rates of mice were not lower compared to those of the voles (see the Study area section), which does not necessarily mean that there were few mice in the study area, but that red foxes prefer Japanese field voles. This preference for voles is reported not only in Japan but in other countries too (Abe 1975; Macdonald 1977; Misawa 1979; Yoneda 1979; Kondo 1980; Koganezawa and Kurokawa 1983; Doncaster et al. 1990; Jȩdrzejewski and Jȩdrzejewska 1992; Yamamoto 1994; Sidorovich et al. 2006; Panzacchi et al. 2008b; Kondo and Shiraki 2012, 2013; Meisner et al. 2014; Lanszki et al. 2020). The low consumption of mice by red foxes, even in areas with abundant mice, has been shown in other areas, including Japan (Macdonald 1977; Misawa 1979; Yoneda 1979; Doncaster et al. 1990; Panzacchi et al. 2008b; Kondo and Shiraki 2012, 2013).

Red foxes also frequently consumed sika deer outside the fence especially between winter and spring (Table 2, Fig. 2). Although the FOs of sika deer during winter seemed to be almost the same between inside and outside the fence, this could be due to the small sample size or temporal distribution of the samples. All samples collected inside the fence were derived from one year (2010), whereas samples outside the fence were from 4 years (2006–2010) (Table 1); sika deer remains were found in one fecal sample in 2007 and in two samples from 2010 outside the fence. According to the Nikko Weather Station, snowfall during winter was 247 cm in 2010, which was greater than that in 2007–2009 (117 cm, 228 cm, and 108 cm, respectively). Thus, we inferred that deer mortality could have been higher in 2010 with severe weather conditions, which contributed to higher FO inside the fence during winter. The mortality of sika deer older than 1 year is highest during winter and spring (Minami et al. 2009). The consumption of deer carcasses by red foxes (also observed by Seki 2017) was probably opportunistic. Such seasonal patterns for the consumption of ungulate carcasses by red foxes have been reported in Japan and in other countries (Koganezawa and Kurokawa 1983; Cavallini and Volpi 1996; Padial et al. 2002; Cagnacci et al. 2003; Sidorovich et al. 2006; Giuliano et al. 2019). The red fox is known to be the major predator of roe deer (Capreolus capreolus) fawns (Borg 1962; Aanes and Andersen 1996; Kjellander and Nordström 2003; Jarnemo et al. 2004; Jarnemo and Liberg 2005; Panzacchi et al. 2008a, b; Melis et al. 2013). The roe deer is quite small as compared to sika deer: the adult weight of roe deer is ≤ 35 kg (Macdonald and Barrett 1993), although that of the subspecies of sika deer (C. n. centralis) ranges from 37.6 to 100.0 kg (Jiang 2015). Thus, it is considered to be more difficult for red foxes to predate on sika deer fawns. However, red fox predation on sika deer fawns is believed to occur quite often. During spring, sika deer remained in a higher proportion each month. As only a few sika deer older than 1 year die during the period between May and September (Minami et al. 2009), the availability of their carcasses was probably low in May–June. Thus, we inferred that the higher consumption of sika deer by red foxes during the period was not derived from the carcasses but from predation on deer fawns since the calving season of sika deer in the study area was estimated to be from May to August, with a peak in the middle of May (Iwamoto et al. 2009). In this study, a deer fawn hoof was found in a fecal sample in June. Tsukada and Nonaka (1996) also reported that red foxes fed on sika deer fawns in June. These results indicate that sika deer fawns may be an important food resource for red foxes. Chasing a sika deer under a year old (presumably approximately 10 months) by a red fox was also observed on March 27, 2007, in the study area (Y.S., personal observation). Red fox predation on fawns of white-tailed deer, similar sized to sika deer (Macdonald and Barrett 1993), was also observed in South Carolina (Epstein et al. 1983).

Invertebrates such as insects and earthworms were also eaten by red foxes inside and outside the fence (Table 2). Red foxes ate Coleoptera (Scarabaeidae and Carabidae) from spring to summer and Orthoptera (Rhaphidophoridae) from summer to autumn (Table 2, Fig. 1). In Japan, red foxes eat insects (Misawa 1979; Yoneda 1979; Koganezawa and Kurokawa 1983; Ueuma et al. 2005). These studies did not identify insects to order and family levels, but where this was done, foxes were reported to mainly eat Coleoptera and Orthoptera (Yamamoto 1994; Koike et al. 2012; Kondo and Shiraki 2013). A similarly high occurrence of Coleoptera or Orthoptera among insects in diets of red foxes has also been observed in many other countries (Doncaster et al. 1990; Serafini and Lovari 1993; Lucherini and Crema 1994; Cagnacci et al. 2003; Baker et al. 2006; Lanszki et al. 2007; Panzacchi et al. 2008b; Murdoch et al. 2010; Remonti et al. 2012; Drygala et al. 2013; Giuliano et al. 2019; O’Connor et al. 2020), indicating that these insect groups are the primary food resource from among insects in the fox diet. Earthworms are also frequently hunted by red foxes in some areas (Macdonald 1980). In England, for example, earthworms constituted up to 38% of the red fox diet in Wiltshire (Baker et al. 2006) and up to 45% in Oxford (Doncaster et al. 1990). One possible explanation for the higher consumption of earthworms by red foxes is prey availability, because foraging for earthworms by red foxes has been shown to be correlated with earthworm abundance (Macdonald 1980). In Finland and Germany, however, the consumption of earthworms by red foxes was low, whereas other carnivores frequently consumed earthworms in the same area (Kauhala et al. 1998; Drygala et al. 2013). Such differences in the consumption of earthworms by red foxes are probably caused by differences in the abundance of their primary prey, that is, small mammals, as Weber (1996) has shown that red foxes consumed a higher proportion of earthworms in the years when voles became scarce.

Red foxes also frequently ate plant materials such as fruits, particularly Actinidia spp., Vitis coignetiae, and Malus toringo during autumn (Table 2). The higher fruit consumption of red foxes has also been reported in other countries (Calisti et al. 1990; Cagnacci et al. 2003; Remonti et al. 2012; Bakaloudis et al. 2015; Lanszki et al. 2019; O’Connor et al. 2020). In Japan, the higher FO of Actinidia spp. has also been reported in other areas (Misawa 1979; Yamamoto 1994; Tsukada and Nonaka 1996; Ueuma et al. 2005; Kondo and Shiraki 2013); V. coignetiae was also found in red fox fecal samples (Misawa 1979; Yamamoto 1994; Ueuma et al. 2005), but the FO of fruit was not high. It has been suggested that the red fox prefers A. arguta over V. coignetiae (Kondo and Shiraki 2012, 2013). In Hungary, however, red foxes frequently consumed V. vinifera (Lanszki et al. 2019), suggesting that the fruit selection of the red fox can differ even though the fruits are from the same genus. Furthermore, frequently occurring fruits in other areas of Japan were different, i.e., Rubus vernus in Mt. Hakusan (Ueuma et al. 2005) and Akebia spp. in Mt. Nyugasa (Yamamoto 1994), indicating that fruit consumption by red foxes may depend on the availability of fruit (Tsukada and Nonaka 1996). Grasses were found with relatively high FOs but are presumably to improve occasional intestinal disorders which occur in carnivores (Imaizumi 2007).

Effects of sika deer density on red fox diet and population

Voles, sika deer, insects, and earthworms were the principal food resources of red foxes in the current study. The FOs of voles were higher inside than outside the fence, with an odds ratio < 1, indicating that red foxes fed more on Japanese field voles inside the fence (Table 3, Figs. 2–3). For sika deer, Carabidae, increased insects (including Scarabaeidae and Rhaphidophoridae), and earthworms, the FOs were lower inside the fence and the odds ratio were > 1, indicating that these food resources were eaten more outside the fence than inside (Table 3, Figs. 2–3). We propose several possible explanations for these results, which are not necessarily mutually exclusive.

The densities of sika deer were different inside and outside the fences (see Study area section). Populations of Japanese field voles and Japanese field mice decreased and insect (Scarabaeidae and Rhaphidophoridae) and earthworm populations increased following modification of the understory vegetation and/or defecation by sika deer in the study area (Seki and Koganezawa 2010, 2013; Koganezawa et al. 2013). Consequently, differences in the feeding pattern of red foxes inside and outside the fence were probably motivated by differences in resource abundance (Abe 1975; Jȩdrzejewski and Jȩdrzejewska 1992). Although the abundance of Japanese field mice was higher inside than outside the fence (Koganezawa et al. 2013; see also the Study area section), the current study did not show greater consumption of mice by red foxes inside the fence, which was probably due to their low proclivity for mice. Carabidae abundance was not significantly different inside and outside the fence (Seki and Koganezawa 2013), but the insects were consumed more outside the fence by red foxes, indicating that other factors other than insect abundance were driving the consumption of the insect by red foxes.

Consumption of certain prey by red foxes can be driven by foraging opportunity. In general, greater vegetative cover can provide increased protection from predators of small-bodied animals (Birney et al. 1976; Peles and Barrett 1996). It has been reported that the consumption of voles tended to decrease with increasing forest cover (Yoneda 1983; Kidawa and Kowalczyk 2011). The mean herbaceous layer height and coverage were reported to be higher inside than outside the fence in the present study area (Okuda et al. 2012; Okuda 2013). Lower understory coverage outside the fence is expected to increase fox foraging opportunities, which could contribute to the higher consumption of invertebrates outside the fence. However, even though there is higher coverage inside the fence, the consumption of voles by red foxes was still higher inside the fence. This indicates that the consumption of voles by red foxes is more likely to be affected by changes in prey abundance than changes in understory coverage.

Another possible factor that could cause a higher consumption of certain prey by red foxes outside the fence is a dietary shift towards alternative prey species with a decline in their main prey, that is, voles. Red fox predation is a plausible mechanism determining the dynamics of the small vertebrate community, and their predation pressure on alternative prey species varies in relation to the prey abundance and consumption of voles (Angelstam et al. 1985; Lindström et al. 1994). Dietary shifts in red foxes have been shown not only towards vertebrates such as grouse, hare, deer fawn, and mammalian carrion (Angelstam et al. 1985; Lindström et al. 1994; Kjellander and Nordström 2003; Sidorovich et al. 2006) but also for invertebrates, wild fruits, and rubbish (Ferrari and Weber 1995; Weber 1996). Thus, lower vole availability outside the fence could force red foxes to feed mainly on more deer carcasses in winter, deer fawn and invertebrates in spring, invertebrates in summer, and fruits and invertebrates in autumn (Table 2). If this concept is correct, Carabidae is presumably the main alternative prey among insects because there is no significant difference in their abundance between inside and outside the fence (Seki and Koganezawa 2013).

Días-Ruiz et al. (2013) showed a latitudinal pattern in the diet of the red fox. The current results suggest that the diet of red foxes varies even on a relatively small spatial scale, supporting that red foxes can plastically vary their diet in the area with substantially modified habitat structure by sika deer grazing. The influence of these changes in diet on the population of red foxes was not significant during the study period. The most probable reason for this is that the effect of vole decline caused by sika deer on the fox population has been offset by the increased availability of other prey such as insects, earthworms, and carcasses and fawns of sika deer outside the fence. Such compensation has also been reported in other areas. Weber et al. (1999) reported that the increase in the consumption of anthropogenic food probably prevented a decrease in red fox breeding success. In areas with thick snow cover and thus low availability of small rodents, increased ungulate carcass availability may compensate for the negative effects of severe winters (Selås and Vik 2006). Although the vole abundance decreased outside the fence, the overall effects of sika deer on the red fox population may be non-significant, at least during the study period, due to the presence of other increased or available alternative foods.

In contrast, populations of other carnivores such as raccoon dogs (Nyctereutes procyonoides) and Japanese badgers (Meles anakuma) probably increase when their food (insects and earthworms) increases due to habitat modification by deer (Seki and Koganezawa 2013; Seki et al. 2014). Such differences could be caused by differences in their dietary plasticity. The principal food of Japanese badgers and raccoon dogs is invertebrates and fruits in Japan (Kaneko 2009; Saeki 2009). The proportion of rodent consumption by Japanese badges and raccoon dogs was reported to be very low, whereas the consumption by red foxes was quite high in the same area (Yamamoto 1994). The ability of Japanese badgers and raccoon dogs to hunt rodents may be lower than that of the red fox because the legs of the former carnivores are relatively shorter than those of the fox. Only a few Japanese badgers and raccoon dogs were observed inside the fence (Seki and Koganezawa 2013; Seki et al. 2014), which also suggests that their plasticity against changes in the food abundance caused by sika deer is lower than red fox plasticity.

As the carcasses and fawns of several ungulates are primary food resources for the red fox, as previously mentioned, recent increases in deer abundance worldwide have had a positive direct effect on the red fox population. Red fox predation also plays a key role in contributing to the high mortality of roe deer fawns (Aanes and Andersen 1996; Jarnemo et al. 2004; Jarnemo and Liberg 2005; Panzacchi et al. 2008a). Our results also suggest the possibility of red fox predation on sika deer fawn. If the higher occurrence of sika deer in the scats of red foxes in spring was largely derived from red fox predation on sika deer fawn, conservation of red foxes may potentially contribute to decreasing the population of sika deer. In addition, the effect of deer abundance on populations of small mammals, ground-dwelling insects, and earthworms has been demonstrated in several areas (Gardner et al. 1997; Stewart 2001; Kanda et al. 2005; Karberg and Lilleskov 2009; Buesching et al. 2011), which could also influence the population of the red fox. Interactions between deer and red fox food sources will improve our understanding of the plasticity of red foxes with habitat modification by deer and will facilitate the design and implementation of deer management practices.