Methods

2.6.1 Study site and plot design The study site is the Field Museum (FM) Kusaki, an experimental forest of the Tokyo University of Agriculture and Technology, and a part of the Japan Long-Term Ecological Research Network (JaLTER). FM Kusaki is located at the upper Watarase river basin, in Midori City, Gunma Prefecture, central Japan (36˚32'N, 139˚25'E; 640 m to 1140 m a.s.l., Fig. 1). This site belongs to a cool temperate region (warmth index, 69.6 ˚C; coldness index 21.7 ˚C). The mean annual temperature and mean annual precipitation from 1997 to 2013 were 9.0 ˚C and 1807 mm, respectively (Koyanagi et al., 2003; Urakawa et al., 2015). Precipitation increases in mid-summer and decreases in winter, and snow precipitation is at most approximately 300 mm (Urakawa et al., 2015). The bed rock in this area is mainly composed of sandstone and clayslate. Sika deer (Cervus nippon) population has been increasing since 2000 in Gunma Prefecture, and was estimated to be more than 1500 in Midori City (c.a. 208 km2) in 2012 (Gunma Prefecture, 2015). In FM Kusaki, herbivory by Sika deer is markedly increasing yearly. A 1-ha permanent plot (100 m × 100 m) for elucidating long-term forest dynamics was established at a deciduous broad-leaved forest in a wide valley bottom of FM Kusaki in 2005 shown in Fig. 1. This plot was subdivided into horizontal 10 m × 10 m grid cells (Watanabe et al., 2007), according to methods of the Monitoring Sites 1000 Project in Japan (Ministry of the Environment, 2011). To include vegetation from the valley bottom to the ridge condition, we expanded the plot by increasing the number of grid cells (10 m × 10 m), resulting in a total of c.a. 3-ha plot in the forest, although the dataset only currently includes data from the 1 ha plot. There are active channels (less than 1 m width) and micro-topographies such as alluvial fans, talus slopes, foot slopes, head hollows, and channel walls, leading to various ground disturbances in this plot. In addition, deer fences were built with 2-m wooden poles and iron wires or nylon nets at 6 grid cells in 2009 for elucidating effects of deer grazing on vegetation as shown in Fig. 1. Fig. 1. Location of FM Kusaki in Japan (upper left) and its geographical range presented by a thick line (upper right). An array of the 10 m × 10 m grid cells and the 1-ha permanent plot (grey square) in the contour map (bottom). The black quadrats are exclosures with deer-proof fences, gray quadrats are control and dots are the points of litter traps. The dominant tree species were Fraxinus platypoda on the valley bottom, and Quercus crispula and Castanea crenata on the side slope in the 1 ha permanent plot (Fukamachi et al., 2011). On the western exterior of the plot, there was a conifer plantation. This oak and chestnut secondary forest may have been used for cutting fuelwood because there was a trace of charcoal grill kiln at the valley bottom. The forest stand including the plot, belonged to the national forest and the last record of forest operations was in 1893. 2.6.2 Field method a. Tree census Our dataset was based on the standardized methods of the Monitoring Sites 1000 Project in Japan (Ministry of the Environment, 2011). All living trees including lianas with girth at breast height (GBH) > 15 cm in the 1-ha plot were attached with aluminum numbering tags (Racetrack Aluminum Tags, Forestry Supplies Inc.), and were identified at species level. The position of each tree base was measured from one corner of the grid cell as the X and Y coordinates in the grid cells. A tree census was conducted at this plot in August and September of each year since 2005. The census included GBH, tree height, death, tree vigor and recruitment (individuals growing GBH > 15 cm in the survey year). About “tree vigor”, we have checked visually and recorded in remarks in the table such as condition of tree tip, presence of rot and damages by nearby tree and deer. b. Litter fall census According to the manual of the Monitoring Sites 1000 Project (Ministry of the Environment, 2010), 25 conical litter-fall traps with a circular collection area of 0.5 m2 were installed in each of the 20 m × 20 m grid cells to avoid big exposed rocks, steep slope and mountain road in the plot since 2005 as shown in Fig. 1. The litter samples were collected every month, and every week in autumn. The traps were not placed between January and March to prevent them from snow damage in winter. The collected litter samples were air-dried in the laboratory, and then sorted into leaves, branches, reproductive organs, and others. The litter was further grouped as follows: original shape retained, half of shape retained, small pieces with matching texture and properties, and other organ material. These samples were identified at species level as much as possible. In the data, the “First identification” column has a question mark “?” which indicates when complete identification cannot be guaranteed, and in the “Last identification” column, there was no question mark. If absolute data with reliable identification is required, the former can be chosen except for the litter with a question mark, and if the overall trend is required, the latter can be chosen. The sorted samples were dried at 70 ˚C for 72 hours and weighed. The litter sample that retained original shape (large litter in the table) was measured on the 2 decimal place scale, and small pieces of organs, insect droppings, and any other unknown tissues (small litter in the table) were measured on the 4 decimal place scale. c. Vegetation census The percentage cover of vascular plants was recorded at each vertical stratification (tree layer, sub-tree layer, shrub layer, and herb layer) according to Londo’s scale (Londo, 1976), in August to September 2008, April to May 2009 and July 2014 in each 10 m × 10 m grid cell of the plot. Londo’s scale is suitable for census in permanent sites because it has a scale finer (13 levels) than the Braun-Blanquet scale (6 levels). It clearly divides coverage and amount by expressing with not only numbers but also alphabets (Hoshino, 1995). One level was added in Londo’s scale, when coverage was 0.5 % or less we recorded “+” with alphabets (see Supporting Information Table S1).