The study site (3638 N, 13847 E) was located on Mt. Asama, central Japan and covered approximately 210 ha (elevation range: 1180-1590 m above sea level, Fig. 1). The area was in the cool temperate zone and covered by snow (approximately 30-50 cm in depth) between December and March. The overstory primarily comprised coniferous forests (87%) of Japanese larch (Larix kaempferi) and Japanese red pine (Pinus densiflora), and partly comprised broad-leaved forests (13%) of Mongolian oak (Quercus crispula). The understory mostly comprised dwarf bamboo (Sasa nippon- ica) and deciduous broad-leaved shrub thickets. The popula- tion density of serows was 4.5 individuals/km2 and six adult serows were identified during the study periods (Takada et al. 2020b). Wild predators were absent, because wolves have been extinct in the area for more than 100 years, but serows in this site have exhibited vigilance toward humans, indicating that they perceive humans as a potential danger (Takada et al. 2019).

According to Takada et al. (2020a), vegetation types were classified as "shrub rich" and "shrub poor" based on under- story vegetation (0.5-2.0 m, within reach of serows), which 57.5% and 42.5% of the study area, respectively. Shrub-rich habitat provides five times more deciduous broad-leaved trees than shrub-poor habitats, whereas shrub-poor habitat provides more forbs or graminoids than shrub-rich habitat (Takada et al. 2020a). We generated a slope map for the study site from a 10-m digital elevation model (Geographi- cal Information Authority of Japan) using QGIS 2.18 (QGIS Development Team 2012). The mean slope was 20.8 (range: 4.1-51.0). We defined four seasons, spring (March-May), summer (June-August), autumn (September-November), and winter (December-February), based on plant phenol- ogy in the study site.

We captured four adult serows (three males and one female) using a dart rifle in March 2012 and fitted each with a VHF radio collar (ATS, USA). Their home range locations are stable throughout the year and inter-annually (Takada et al. 2020b). We located the collared serows by standard triangu- lation using VHF receivers (YAESU, Japan) and a 6-element Yagi antenna at 2-h intervals 2-6 days per month during Jun 2013-May 2014 (total: 12 months, Table S1). We attempted to locate individuals from at least three receiving locations as soon as possible to obtain the most accurate telemetry locations. The time period was defined as day (sunrise to sunset) and night (sunset to sunrise), and we recorded the time period in which the serow was located. Home ranges of serows for each season were estimated using the 100% mini- mum convex polygon method (MCP100) to analyze habitat selection.

To assess the habitat selection of serows, we compared serow telemetry locations with random locations using generalized linear mixed models (GLMM) with binomial distributions and a logit-link function (Table S3). We gener- ated random locations within the seasonal home ranges (as determined by MCP100) of each radio-collared individual for each season. For each individual, each season, and each time period, we used the same number of random locations as there were telemetry locations. We set the dependent variable as the probability of serow locations (1 = serow location, 0 = random location). Fixed effects included the vegetation type, slope, and following interactions: season vegetation, season slope, time period vegeta- tion, and time period slope. Individual identity was set as a random-effect factor to minimize variation among indi- viduals. To select the model that best explained the observed pattern of serow habitat selection, Akaike's information cri- terion corrected for small sample size (AICc; Burnham and Anderson 2002) was used. Models with the lowest AICc value were considered the best models; any pair of models with a difference in AICc values (AICc) less than 2 was considered to have similar support. We conducted model averaging on competing models with a AICc less than 2 to obtain model-averaged conditional beta estimates and p-values of the effect of predictor variables.

Between Jun 2013 and May 2014, we obtained a total of 1176 telemetry locations (Table S4). Serows used habitats in a non-random manner. The ranking of models predicting serow locations showed that the best models (AICc

In this study, we preliminary present the first data on habi- tat selection of the Japanese serow in their typical habi- tat. Serows selected shrub-rich habitat and steep slopes throughout the year. Serows at our site fed mainly on deciduous broad-leaved trees throughout the year (warm season: leaves; winter: twigs and buds: Takada 2018), and the foraging efficiency (bite/min) was higher in shrub-rich habitats than in shrub-poor habitats (Takada 2018). In addition, serows used shrub thickets and steep slopes for concealment and as escape terrain, respectively, against potential danger (Takada et al. 2018, 2019). Therefore, serows may have selected these habitats providing both abundant food and shelter. The snow depth of our site was relatively low (30-50 cm) and the distribution of main food hardly changed seasonally. Therefore, the habitat type selected by serows seemed to hardly change season- ally. Conversely, the avoidance of shrub-poor habitats was more pronounced in winter. Although shrub-poor habitats supply forbs in the warm seasons, snow cover severely reduces their supply in winter (Takada 2018), and this may be why serows tended to avoid shrub-poor habitats more in winter than in other seasons. In the present study, although there were no substantial predators for serows, serows selected safer habitats (i.e., steep slopes). This habitat selection seems to support the idea of 'ghosts of predators past', that is, prey's behaviors have coevolved with their predators over tens of thousands of years, and an absence of the predators for a few cen- turies is not sufficient to remove them (Byers 1997). In fact, it has been suggested that the habitat use of serows at a landscape scale prioritizes security over food access (Takada 2020). Many of the studies on habitat selection of ungulates have shown trade-offs between access to foraging sites and exposure to predation risk (Houston et al. 1993). For exam- ple, for mountain ungulates, cliff-adjacent habitats provide security, whereas the most productive habitats typically occur on gentle slopes that are more easily accessible to a predator (Grignolio et al. 2007; Hamel and Ct 2007). In contrast, our results suggest that serows have few such trade-offs to make: shrub-rich habitats with steep slopes seemed to be adequate habitat, providing both good forag- ing conditions and shelter. The habitat selection of serows showed little variation with time of day. This also differs from some ungulates; ungulates prefer the most productive habitats at night when they actively forage and are less vis- ible to predators or humans, whereas they prefer habitats with shelter during daytime when they are resting (Godvik et al. 2009; Bjrneraas et al. 2011). As described above, in this study there appeared to have been an adequate habitat for serows. Therefore, it seemed to be suitable for both feeding and resting (i.e., all times of day), with the result that time of day may have little effect on habitat selection. In fact, both feeding and resting were frequently observed in shrub-rich habitats with steep slopes, while in shrub- poor habitats with gentle slopes, only movement was fre- quently observed,

Answer the following:

-Explain the purpose of the study and briefly describe how and where it was conducted.

Summarize the key findings of the study.

Explain how these results may be applied to management or conservation of the species.