Determination of winter sleep periods with change-point analysis
Following the activity patterns of nocturnal and secretive wild mammals is often logistically difficult. GPS tracking is replacing conventional radio tracking and visual observations of, e.g., snow tracks as the principal method for monitoring winter activity, but all these procedures are economically challenging (expensive, labour-intensive, etc.). We developed here a new low-cost, low-effort application of CPA to determine the timing and duration of alternating active and passive periods of overwintering using the raccoon dog as the model species. CPA has been scarcely used in physiological research even though some studies have utilized this method [27, 28]. The present experiment is the first to show that CPA can be successfully applied to analyze Tb recordings of a passively wintering wild mammal in order to unravel its foraging patterns with potential applications to other northern carnivores.
The positive correlation between the activity scores and Tb of the raccoon dogs gives an opportunity to draw conclusions on the timing of passive periods by using the Tb data alone. To validate the method, the periods of passivity were determined by both behavioral data and by CPA, and the timing of these periods was compared, i.e., if both methods were able to find the same periods of winter sleep. In individuals that could be tracked for most of the winter, the approximations by CPA and behavioral data shared 91% similarity. As the transitions between activity and passivity are not instant physiological processes, short periods of winter sleep are hard to classify. In the case of more easily defined, longer passive periods of ≥2 days, the similarity was slightly higher (93%). Most of the differences were detected in the transitory days between active and passive wintering. In addition, behavioral data recordings were more sensitive to detect periods of passivity/activity lasting for approximately 1 day.
Although the interrelatedness between the Tb and activity scores was clear, they did not correlate perfectly. This may partly result from the method of collecting activity data, as the measurement was not continuous but lasted for 2 min 8 times a day. During data analyses, it was occasionally noted from the GPS fixes that a raccoon dog was away from the den with a relatively high Tb but the activity values were close to zero. In these cases, the animal remained fairly immobile for the duration of the activity measurement, although the GPS data indicated that it was not displaying winter sleep at the moment. Due to this, the >90% homology between the two evaluating methods can be considered most satisfactory. Based on the time periods defined by CPA, the Tb of the raccoon dogs decreased by 1.0°C from active to passive wintering. The average Tb during the seasonal rest (36.5°C) was similar to previous reports on Finnish  and Japanese raccoon dogs, N. p. albus.
In a previous follow-up , the duration of (mostly uninterrupted) wintertime passivity determined by radio telemetry was highly variable in eastern Finland (a few days–9 weeks), and the present results enforce the observation that the length of winter sleep can vary prominently depending on the individual. When comparing the data to other species, the average sum of passive days in the present study (53 days) resembles previous estimates on Japanese badgers (Meles meles anakuma; ) but is less than in Eurasian badgers (M. meles) in eastern and northern Europe and in Central Asia . It was also noted in the present study that the raccoon dogs with the highest sums of passive days had also the highest total numbers of passive periods and the longest single passive periods. Previously, it was difficult to distinguish the short passive bouts before and after the principal period of winter sleep as only the latter was clearly visible in the Tb curve. CPA seems to be more efficient in pointing out these shorter periods.
Previous studies have indicated that the different types of carnivorean lethargy are highly variable and species-specific. The passivity can be intermittent or continuous, the 24-h rhythm in Tb can be maintained or disappear, and the duration and depth of passive bouts can differ significantly. For instance, the Tb of bears decreases by 4–5°C, and its circadian rhythmicity is maintained . In Eurasian badgers, the decrease in Tb is also pronounced and continuous, but Tb does not show a 24-h rhythm . Moreover, the American badger (Taxidea taxus; ) and the striped skunk (Mephitis mephitis; ) experience torpor bouts with clear decreases in Tb. Based on the present and previous studies on raccoon dogs [5, 6], the winter rest of the species seems to be the most superficial among these northern carnivores and it is characterized by regular 24-h oscillations of Tb.
The fluctuations in the Tb (see also ) and activity scores followed closely the changes in the Ta. With global warming, winter sleep of the raccoon dog could be interrupted by more frequent warm spells leading to both shorter periods of passivity and decreased total duration of winter rest . The interrelationship of Tb/activity and Ta is observed in several mammals [2, 29, 30, 35–37], and similarly, Tb and activity usually increase with day length [29, 32, 36]. For this reason, the negative covariance between the activity levels and day length in the present study was surprising, and, according to Kauhala et al. , day length correlates positively with seasonal activity of raccoon dogs. In the present experiment, the animals were tracked only in winter, and the passive periods occurred mainly in Jan–March when day length was already increasing (Figure 3). This presumably led to the significant negative covariance between these parameters.
Summarizing, Ta seems to be an important external factor regulating Tb and physical activity levels of the raccoon dog. As Ta is liable to the climate change, the duration of winter sleep could be susceptible to global warming with a reduction in the total length of winter sleep and/or a more intermittent pattern of passivity. With the help of CPA, i) the timing of passive periods of overwintering can be determined by using Tb recordings alone without the burden of VHF and GPS tracking and activity recordings. ii) Significantly more individual animals can be recruited by using the simple methods of data loggers and ear tags. iii) Hunting can be used as a tool to return ear-tagged individuals with data loggers, allowing iv) the extension of follow-up studies to several years with moderate costs. v) Applying the Tb and CPA methods to other palearctic or nearctic carnivores that are abundant and actively-hunted species would be logistically practical.
As the gastrointestinal tract analysis is limited to the last meals, the FAS of sc fat were also analyzed and compared to the contents of the alimentary canals to obtain data on the more long-term dietary habits . The majority of the animals did consume food in Nov−April, as 84% of the stomachs and 95% of the intestines contained digestible material. The stomach and intestinal compositions were relatively uniform, and the average volume of stomach contents (71 ml) corresponded quite well to previous data (53 ml; ). In Jan−Feb, the average food volumes decreased, and the animals relied more on their fat stores. The selective mobilization of FA confirmed our previous results with C12–17 SFA, C14–17 MUFA, and C18–20 n-3 PUFA as the most preferably utilized FA during negative energy balance . Also the diversity index reduced in Jan−Feb as observed previously in different parts of Europe from summer to winter [20, 38–42]. Earlier fasting experiments on pigs (Sus scrofa domestica) showed that food deprivation of the host decreases the fecundity of roundworms  and, thus, the prolonged negative energy balance could have contributed to the observed decrease of intestinal helminths. In the CBC, there were no indications of deleterious health effects caused by wintertime loss of body condition.
As an opportunist, the raccoon dog is very versatile in its selection of food, which in this study consisted of mammals, plants, birds, fish, and invertebrates in this order of importance. Generally, the species can be considered an urban wild animal, as it sometimes lives near or in the habitats humans create. Our raccoon dogs frequently visited farms, yards, gardens, composts, and waste heaps indicated by the occurrence of household waste (e.g., boiled meat and potatoes, imported fruits) and undigestible, man-made material in their gastrointestinal tracts. Also previous studies reported that raccoon dogs utilized compost piles and refuse dumps and that this behavior could be emphasized in winter [20, 38, 44–47]. In the use of anthropogenic food resources, the raccoon dog resembles several omnivorous canids and mustelids . According to South Korean studies , raccoon dogs with insufficient autumnal fattening could leave their principal habitats to search for food in villages increasing the transmission risk of rabies to domestic animals. Not dissimilarly, the raccoon dogs of the present study also utilized human resources. These observations suggest that, if the periods of passivity become shorter in the future , the frequency of interactions between raccoon dogs, other wildlife, domestic animals, and humans could increase due to the climate change.
The most common prey items of mammalian origin were small rodents and shrews together with hares, cervids, and raccoon dogs—the latter probably consumed as carrion. The utilization of mammals was the highest in early winter, when it was presumably easier to catch voles and shrews as the snow cover was not yet deep. It is also probable that the vole densities were higher during this period compared to late winter. In earlier studies, voles were important food items for overwintering European raccoon dogs [38, 39, 41, 47, 50, 51], but the relative proportions of Myodes and Microtus spp. were highly variable [38, 40, 41, 44]. In the present study, the occurrences were equal in the stomachs but Microtus spp. dominated in the intestines, suggesting that Microtus could be an even more important genus in the area. Based on the continuous trapping by the Finnish Forest Research Institute during the last 20 years, approximately 60% of the voles have been M. glareolus and 40% M. agrestis in the study area (Otso Huitu, pers. comm.).
The raccoon dogs, hares, and squirrels were probably consumed as roadkills. The cervids (presumably moose Alces alces) had been shot during the hunt and scavenged as discarded offal (rumen, intestines, spleen, kidneys) and hides, as moose are rarely killed by predators or traffic in the study area. Polish and Belarusian raccoon dogs did not rely on medium-sized mammals in winter [40, 50], and the importance of hares for Polish, Lithuanian, and Finnish raccoon dogs was previously quite small [38, 41, 42, 51]. The occurrence of raccoon dogs in the diet was not surprising, as similar findings were reported also in an earlier study . Moreover, several papers demonstrated that ungulate carcasses are important as alternative food for European raccoon dogs in winter [20, 39–41, 47, 51, 52]. The occurrence of ungulates in the alimentary tracts was not reflected in the FAS, although ruminant fats contain high proportions of SFA and moderate amounts of trans-FA due to the microbial biohydrogenation of unsaturated FA in the rumen [53, 54]. The raccoon dog fat, however, contained traces of 18:2nc 9t 11, which could have derived from moose tissues.
Plants are an important year-round food source for raccoon dogs in all studied geographical regions , but there exist scarcely any previous taxonomical data on the diversity of plant-based food items. In the present study, this was realized by the detailed analysis of seeds in the gastrointestinal contents. In agricultural landscapes of Germany, maize (Zea mays) was the most popular winter food item , whereas in Finland oat was one of the most common species, especially in early winter. Its presence indicated that the animals visited fields, arable lands, grain storages, and/or feeding stations of game animals. This was supported by the observations of common weeds, such as Chenopodium album. The occurrence of cereal grains in the winter food has been established also previously [20, 38, 44].
Bird feeders—a food source used by urban foxes —were also visited by raccoon dogs as, in addition to oat often present in bird feeds, their gastrointestinal tracts contained seeds of H. annuus, A. hypogaea, Sorghum sp., Panicum miliaceum, Setaria italica, and Linum usitatissimum, plant species not cultivated in the study area. Berries rich in carbohydrates were important food items for raccoon dogs during the autumnal fat gain [38, 56]. Also the winter diet contained berries, especially S. aucuparia, V. vitis-idaea, and V. oxycoccos (see also [40, 42]), all of which are well-preserved through winter. Our study animals also consumed dropped or discarded apples together with pears and bananas presumably originating from compost piles.
The occurrence of birds in the winter diet was relatively similar or higher in the present experiment compared to previous European studies [38, 40–42, 44, 47, 50, 51]. The consumed specimens were mainly gallinaceous birds, waterfowl, and corvids [38, 44], and most of the identified species were year-round residents. In addition to Anas sp., only two of them (Tetrastes bonasia, Lyrurus tetrix) could be considered game bird species and, thus, the results suggest that in the study area the raccoon dog, albeit an introduced species, cannot be considered a significant threat to wildfowl in winter (see also [38, 44]).
Ice-fishing is popular in the study area, especially in late winter−early spring, when the occurrence of fish in the diet increased and raccoon dog tracks were observed on ice [5, 38]. Most specimens that could be identified belonged to Percidae and Cyprinidae ( for the snow-free period). Presumably this does not represent any preference of the raccoon dogs but reflects the species that are the most likely to be discarded by fishers. Previous European studies reported low fish consumption by overwintering raccoon dogs [38, 40, 42, 47] but, in some areas, the importance of fish increased in early spring [20, 38]. According to PCA, the occurrence of fish separated the FAS of the animals. It is impossible to know, if the individuals with remains of fish in their alimentary tracts foraged regularly on fish left by fishermen or households, but the FAS analysis suggests that this could be the case. Freshwater fish can contain high amounts of, e.g., n-3 and n-6 PUFA , which could explain the higher proportions of LA and longer-chain PUFA in the fat tissue of these individuals. The DHA/LA ratio, correlating with aquatic food web exploitation , averaged 0.085 in the raccoon dogs—slightly higher than calculated for terrestrial mustelids and felids but approximately at the same level as in semi-aquatic mustelids [58, 59]. This suggests that also the raccoon dog foraged on freshwater food during winter, even though its DHA/LA ratio was below that of the European otter (Lutra lutra).
Invertebrates were of minor importance for the overwintering raccoon dogs, and amphibians or reptiles were not detected in any of the samples. In previous investigations, the occurrences of these food items were similar or higher than in the present study [38, 40–42, 44, 47, 51]. As the winters can be harsh in eastern Finland, it is difficult for the animals to find, e.g., earthworms or anurans under deep snow, ice, and ground frost. In the more temperate climate of Japan, insects are an important food class year-round . It is possible that some insects (and some other dietary items) found in the gastrointestinal tracts in the present study could have been predigested by the prey animals.
To sum up, i) overwintering raccoon dogs are opportunists and active participants in the food web. ii) Mammals (arvicolines, shrews, cadavers) and plants (oat, berries) together with birds and discarded fish comprise the most important winter food items. iii) Raccoon dogs utilize anthropogenic food resources (cereals, garden supply, household waste, spilled bird seeds, carcasses of hunted moose) similar to several other carnivores, and iv) the occurrence of fish in the diet can be detected in the FAS of overwintering raccoon dogs.
Size of winter home ranges and habitat preferences
There is a limited amount of data on comparisons of home range sizes estimated by VHF radio telemetry and GPS tracking. In the present study, the sizes of winter home ranges determined by GPS averaged 5.3 km2 (K95%) and the core areas 1.3 km2. These estimates were similar or higher than previous measurements conducted with VHF radio telemetry in Finland during the snow-free season (K95%: 1.0–3.9 km2; K50/60%: 0.2–0.7 km2; [5, 60–63]) and in winter (K95%: 3.7 km2; K50%: 0.5 km2; ). In a New Zealand study, Recio et al. determined the home range size of feral cats (Felis catus) with GPS and compared it to results obtained by radio telemetry. They concluded that the home range sizes were of a similar magnitude or higher when using GPS. Furthermore, Medri & Mourão  reported for the giant anteater (Myrmecophaga tridactyla) that a home range recorded with GPS over 9 days could be larger than a range obtained with VHF tracking over 252 days. These data support our results showing similar or larger winter home range sizes than estimated previously for raccoon dogs in the same study area with traditional tracking methods .
In Germany, dispersal of juveniles was rare in winter , but our results suggest that they could be more mobile and have larger home ranges than adults during the cold season. The size of the stable postdispersal home range of M5 (K95%: 59 km2) seemed exceptionally large but, in fact, an even higher value was reported for a juvenile German raccoon dog (K100%: >150 km2), which roamed solitarily in search of a suitable habitat and a mate . Supported by data of Åhlén and Dahl , GPS can have better potential than radio telemetry to unravel rapid long-distance dispersions of raccoon dogs, as these can occur unexpectedly towards an unpredictable direction. They can cause the loss of animals from conventional VHF telemetry especially in rugged terrain, where the distance for receiving signals can be fairly short, while dispersion presents no obstacle for GPS.
The study area was sparsely-populated small-scale mosaic with little patches of fields, forests, and gardens. This can have methodological significance, as even small inaccuracies in GPS fixes could switch the position of the relocation to the adjacent habitat. Also, when the density of forest canopy increases, the GPS fix attempts may be less successful , which could cause a bias in habitat selection data. The observation success rates varied greatly between individuals (16–62%), and this could have been partly derived from the differences in the proportion of forest habitats in the home ranges, although the activity levels presumably affected the observation success rates the most. Moreover, the significance of the canopy is presumably lower in winter with no foliage in the deciduous trees. It is not possible to state with certainty why the % of successful GPS fixes was lower than in a previous follow-up of raccoon dogs in Greater Tokyo in Dec and March—98% in a relatively open area and 70% in a mosaic area . Potential explanations include differences in climates (lack of permanent snow cover in the Tokyo region), in underground hiding behavior of raccoon dogs combined with the characteristics of the habitats, and in coverage of GPS.
The raccoon dog is an ecological generalist. The biotopes preferred by the overwintering individuals were gardens, shores, deciduous forests, and sparsely forested areas. These findings fit relatively well with the results of the dietary analyses discussed above. The wintertime habitat choice of the species has been scarcely studied, and the comparisons between experiments can be complicated due to the different classification of habitat types and varying habitat availabilities in other countries. For instance, German raccoon dogs showed almost neutral preference for forests, small woods, maize fields, hedges, reeds, and meadows but avoided open farmland, water surfaces, and human settlement .
Previously, raccoon dogs were observed to prefer gardens and yards during the snow-free season in Finland [60, 63], and they seemed to favor man-made biotopes of this type also in winter. This finding is supported by the gastrointestinal tract contents with fruits (apples), berries (rowanberries, chokeberries), bird feeds, and household waste from compost piles (potatoes, carrots, and their peals, banana skins, pear seeds, etc.). Anthropogenic food resources provide many carnivores an abundant, relatively stable, and highly concentrated food source and, in some cases, larger home ranges in rural areas may be associated with lower availability of these food items compared to urban and suburban sites . Raccoon dogs favored also shores during the snow-free season in Finland [56, 62]. Shores were presumably preferred as they provide food (e.g., frogs), shelter, and quick escape into water when attacked. In winter, the popularity of frozen lake and pond shores may be related to the availability of discarded fish as discussed above. The available area of lake ice was large in the potential home range of some individuals, but this habitat was not among the most preferred, even though >20% of the relocations of M5 were on ice.
When discussing forest types, raccoon dogs were previously documented to favor deciduous and mixed forests and to avoid coniferous forests in the snow-free season in Finland [60–63]. In Lithuania, overwintering raccoon dogs selected spruce/mixed coniferous forests over deciduous/mixed forests, while pine forests with poor food resources were avoided [42, 71]. In the present study, deciduous forests were clearly preferred over mixed or coniferous forests. The unpopularity of low-open coniferous woodland may result from the lack of protective undergrowth, although these sites could provide animals with, e.g., berries. One possibility could also be the difficulty to obtain a GPS fix through the coniferous, often snow-covered canopy in winter. The GPS collared raccoon dogs of the present experiment utilized wetland habitats of the study area hardly at all, unlike in Lithuania and Belarus [40, 42, 71], but cranberries can be available also on pond and lake shores covered by a zone of Sphagnum moss. The cranberries found in the gastrointestinal tracts presumably originated from these sites and from household waste.
One reason for the preference of sparsely forested areas could be the fact that the most frequently used winter den of a raccoon dog pair was situated in this habitat type and, as a result, a major portion of their GPS fixes was located close to the den. Sparsely forested sites were structurally complex as they consisted of, e.g., sapling stands and clear-cuts with abundant undergrowth providing the animals with shelter and small prey. Also lingonberries could be available in clear-cut areas. Furthermore, raccoon dogs commonly rest under lower branches of young spruces Picea abies, which could be one explanation to the popularity of sparsely forested areas. In contrast to the present results, sapling stands and clear-cuts were not preferred during the snow-free season in Finland . Even though almost 20% of the relocations of the present study were in fields, their high areal proportion in the study district led to almost neutral preference, which is similar to observations on the utilization of maize fields in Germany . Fields were among the preferred habitats by Finnish raccoon dogs in the snow-free season [60, 62, 63], and they were probably good sources of cereals and small mammalian prey  for overwintering raccoon dogs, as well.
Many mammalian species are disturbed by human infrastructure such as roads. They have negative ecological effects by forming movement barriers and by causing habitat fragmentation and loss, disturbance, pollution, and mortality . Roadsides can also provide refuges, new habitats, and movement corridors, although many carnivores tend to avoid the presence of roads [73, 74]. Raccoon dogs have high mortality due to traffic , but the use of groomed roads and railroad beds indicates their possible benefits as travel routes due to thinner snow with a more supportive surface. This could decrease the locomotory costs, as raccoon dogs have small footpads and they sink easily into soft, deep snow. Also other mammals use railway tracks to facilitate traveling during periods of deep snow cover .
The emergence of rodents from hibernation became significantly earlier in the Colorado Rocky Mountains from 1976 to 1999 . Similar long-term follow-ups would be useful also in other passively wintering species, such as the raccoon dog, which is likely to benefit from the extending growing season and shorter winters. According to climate change scenarios , Ta in Finland will increase during all seasons but especially in winter. Towards the end of the present century, the number of frost days will decrease by 40–80 days, the period of snow cover will shorten by >1 to 2 months, and snow depth in midwinter will decrease to about 33–70% of that at present. Passivity of the raccoon dog will probably become shorter and more intermittent during milder winters . Abundance of the raccoon dog and other medium-sized predators can increase and their distribution expand to north . This could also influence the occurrence and distribution of zoonotic diseases and parasites they transmit.
Summarizing, i) the wintertime home ranges determined by GPS could give a more accurate estimation of the home range sizes of raccoon dogs compared to traditional methods. ii) The preferred habitat types were gardens, shores, deciduous forests, and sparsely forested areas, while fields had close to neutral preference. These data support the results of the dietary analyses. iii) The raccoon dogs used roads and railroads as wintertime travel routes.