Methods

We counted deer in the daytime in February or March, when the visibility was more acceptable to find deer since deciduous leaves fell and the ground was covered by snow, once per year between 1980 and 2023, except three years, i.e., 41 times (Fig. 2). The counts in 1981, 1985, and 2022 were canceled because of another field work on the island, stormy weather, and the pandemic of coronavirus disease 2019 (COVID-19), respectively. We divided the study area into five blocks; 1) West Peak, 2) North Peak and Northeast Point, 3) East Peak, 4) the two small islets, and 5) South Ridge (Fig 3). Around ten drivers lined up at the highest ridge of each block and walked down to the lakefront, keeping the distance between adjacent drivers for 50-100 m, while an observer on a boat for each block counted deer using a binocular and recorded the number and sex-age class (fawn, i.e., juvenile, or adult, i.e., yearling and older) of deer in each deer group, depending on the body sizes and antlers. However, the distances between the observers and most of deer groups were too far to identify the sex-age class, so we presented only the number of counted deer in the data. The boat moved in the west side of West Peak and the east side of East Peak to search deer before the drivers lined up on the edges in Block 1 and 3, respectively. A few drivers counted deer in Block 4, small islets, without an observer on a boat. Drivers recorded deer not found from the boats. 16 drivers and one or two observers counted deer of all blocks in consecutive two days per year until 1984 (Kaji et al., 1988). The two-day counts during 1980-1983 might be unreliable, comparing to the count in 1984 which was carried out when 81 deer were kept in a corral trap for the first culling as above and more than 100 deer stayed around a feeder (Kaji et al., 1988). Since 1985, around 30 drivers and two or three observers did the same way in a day (Takeshita et al., 2016), i.e., Block 1 and 2 simultaneously in the morning and Block 3, 4, and 5 in the afternoon. Two vantage observers (three only in 2017) were located on and beside South Ridge since 1999 in order to find deer in blind spots from the boats and the drivers in the south face of North Peak, the upper stream between South Ridge and East Peak, and the west face of South Ridge (Fig. 3). Five (or six) and one time-lapse camera traps (Ltl-Acorn 6210; Ltl- Acorn Outdoors, Green Bay, Wisconsin, USA), with one still image at every 10-second intervals, were set in the upper stream between West Peak and South Ridge and the lakefront in Northeast Point, respectively, since 2017 in the same reason as the vantage observers (Fig. 3). Drivers and observers always discussed to exclude double counting just after the fieldwork on the day.