Study area and species
Iriomote (24°15′ ~ 25′N, 123°40′ ~ 55′E) is located at the southernmost tip of the Ryukyu Islands. It is the largest among the nine islets of the Yaeyama Islands, with an area of 289 km2 and its highest peak at 470 m above sea level. This subtropical island is typified by hot and humid summers and warm but windy winters. Temperatures generally rise above 25 °C from May to October, peaking at ca. 29 °C in July, and descend to 17 ~ 18 °C in January, with a mean annual rainfall of over 2300 mm (Japan Meteorological Agency data).
Intact primary or secondary broadleaf forests cover about 80% of this mountainous inland and are characterized by Castanopsis spp., Quercus spp., Schima spp. and various figs (Ficus spp.). Hernanadia nymphaeifolia (Presl) Kubitzki and horsetail pine Casuarina equisetifolia L. are common in coastal forests, whereas Bruguiera gymnorrhiza (L.) Lam, Rhizophora mucronata, and R. stylosa mangroves prevail along estuaries and rivers. Ficus, such as cedar fig F. superba Miq. and white fig F. virgata Blume, are mixed with Cerbera manghas L., Heritiera littoralis, hanging-flower checkerboard foot Barringtonia racemosa (L.) Blume ex DC, and screwpine Pandanus odoratissimus L. f. in swampy wetland forests [34]. Human residences and cultivated areas are located in lowlands below 100 m along the shoreline and are more prominent on the eastern and northern coasts.
The near-threatened Ryukyu flying-fox (Pteropus dasymallus Temminck, 1825) is among the most northerly distributed pteropodids [35, 36]. Its five subspecies are each narrowly distributed but collectively cover a broad latitudinal range along the West Pacific island chain from Tokara Islands, Okinawa Island, Yaeyama Islands, and small islets of the Nansei Islands in southern Japan to Batan and other northern Philippine islets [35, 37]. Among the subspecies, P. d. yayeyamae occurs on most of the Yaeyama Islands [35]. In Iriomote, F. septica Burm. f. and F. variegata Blume are their predominant foods, followed by F. benguetensis Merr., banyan (F. microcarpa L. f.), and at least 35 other fig and non-fig species [17].
Forest structure and tree composition
We conducted field surveys adopting the point-centered quarter method [38, 39] to assess forest structure in July, 2012, along 12 forest transects 1-km in length each. Four transects were established on each of the west, north, and east sides of the island. These transect sites had been previously surveyed for flying-fox abundance and activity [17], making comparisons between the two assessments possible. We did not survey the southern-most side of the island where forest transects were limited by rough terrain, and the southwestern-most corner, Funauki, due to its inaccessibility by land.
In each transect site, we randomly picked six 100-m long sections and six random points within each section to collect measurements using 2-digit random numbers from 01 to 99 in increasing order. We specifically kept the difference of any two successive numbers greater than five, so each pair of random points selected were at least 5 m apart and individual trees would not be measured repeatedly [40, 41]. Our sampling proceeded from the coast toward the forest interior, and the smallest random number selected was set as the initial point along a transect line. At each sampling point, we determined four quarters divided by the transect line and its perpendicular line. In each quarter, the nearest tree to the sampling point with at least 4 cm in diameter was located. The quarter, distance from the nearest tree to the sampling point, species, and CCH (circumference at chest height of 130 cm [42]) were recorded, and fruiting or blooming trees were noted. We calculated density and basal area (BA) using distance and CCH data [41] for total trees, total fig trees, fruiting trees, and fruiting fig trees at each site, and obtained the relative density and relative BA of total fig trees, fruiting trees, and fruiting fig trees in relation to that of total trees [41]. We calculated the relative frequency of occurrence (FO) and relative abundance (RA) for each tree species at each forest site and for the entire sampling across sites. We further adopted and modified Curtis and McIntosh’s [43] use of the arithmetic mean of these standardized measures to obtain an estimate of the relative importance (RI) of each species in samples [17]. The converted Simpson index, 1 − D = 1 − Σ (p
2
i
), was used to assess the heterogeneity (SH) in species composition [39], where p
i
is the relative abundance of particular species i (i = 1 to s, s being the total number of species in a sample).
Foraging bat dispersion and abundance
From 28 June to 22 September, 2012, we conducted 50 bat surveys along the 12 transect sites (mean 4.2 ± 0.11 nights per site) where forest tree composition was assessed. Transect lines followed previously adopted outskirt routes leading toward inland forests [17], except that Aira-gawa and Nanama-gawa forest sites replaced two routes where no traces of bats were recorded. Any two proximate transects were roughly 2 ~ 4 km apart, and collectively they covered areas ranging up to 5 km from the coast. In each survey run, we alternated among the western, northern, and eastern sites of the island and randomly picked transects until all transects were assessed over a period of 2 ~ 3 weeks. We also alternated the nightly proceeding direction within any transect so that no point was ever visited at the same time.
We arrived at a site at least 30 min before sunset to observe bat arrivals or passing until sunset. Assessments began within 30 min after sunset and ended usually within 2 h, and transects were walked at roughly 1 km/h. A group of 2 ~ 3 workers searched for bats with binoculars (Leica 10 × 42 BN, Solms, Germany), aided with head- or spotlights and by bat sounds while feeding or interacting with each other. Upon each encounter, we tallied the number of bats present and recorded the species of trees where bat perched or searched for fruits, perch heights, and bats’ behaviors (e.g., moving, searching, feeding, individual interactions). We restricted our searches to a strip of 30 m on either side and assumed a complete census. Our prior tests indicated that this is a suitable distance in most habitats, but in inner forests we acknowledge that observations were more limited. After each transect survey, we remained following and monitoring bats present until past the midnight or an hour since the last bat observed left a site, whichever came first.
Fig phenology, fruit sampling, and feeding traces
Along with each bat survey at each transect, we assessed tree phenology in the early afternoon, sampled bat feeding traces after each bat survey, and sampled fresh mature fruits at dawn. We estimated crop size in the mature edible stages (i.e. the post-floral phase defined in [44]) of each fruiting monoecious fig tree and for dioecious female figs that produced seed-carrying syconia. We visually estimated crop sizes for trees with few or small amounts of fruits [45], but applied a stratified sampling method [26] when fruits were too many to count reliably. For each tree, we divided branches into three classes and defined the largest fig-bearing branches as the 3rd class, which merged into the 2nd class and then further merged into the 1st class, often bifurcating directly from the main trunk. We randomly selected 30 tertiary branches, tallied the number of figs on each branch, and then obtained the mean value (n). The number of the 1st class branches was counted (b
1
), then six 1st class branches were randomly selected to estimate the mean number of 2nd class branches per 1st class branch (b
2
) and the mean number of the 3rd class branches per 2nd class branch (b
3
). Total crop size was estimated as crop size = n × b
1
× b
2
× b
3
[25].
We randomly collected accessible mature figs (the stage E [44]) directly from each fig tree where bat feeding was observed. Ryukyu flying-foxes usually drop ejecta pellets beneath the feeding tree, only occasionally mouth-carrying a large fruit to a feeding perch [46 YFL unpubl. data], thus we searched feeding traces underneath each feeding fig tree after our nightly observations, mostly discarded pellets but also culled fruits and fecal samples. Fresh mass (fm) and volume size of intact fruits and pellets were measured, and the numbers of seed they contained were tallied. We obtained dry mass (dm) of intact fruits and pellets after oven drying overnight at 50 °C. Water mass (wm, g) was determined as the difference between fresh fruit mass and dry fruit mass and water content (%) was calculated as WC = wm/fm × 100 [47]. It was not possible to count the actual number of seeds of each fig fed on by bats, particularly in field conditions. Thus we used the difference in mean seed numbers between sampled fresh intact fruits and ejecta pellets to obtain an approximate estimate of the proportions of seeds being removed away from different species of fig fruits after bat feeding.
Data analyses
Data are presented as the mean ± standard error (SE) unless otherwise noted. Statistical tests were conducted using Statistica 12.0 (StatSoft, Tulsa, USA) with the significance level set at α = 0.05. Proportional data were arcsine-transformed to meet the normality requirement [48]. We assessed the correlation (Pearson’s r) between relative frequency of occurrence (FO) and relative abundance (RA) for dominant tree species, among tree density, basal area (BA) coverage, and heterogeneity for forest transect sites, and between bat densities estimated over sites between the 2005 and 2012 assessments. A χ2 test was used to determine if the frequency distribution was random among abundance levels. We used analysis of variance (ANOVA) to examine the effects of site location on variances in bat density. Multivariate analysis of variance (MANOVA) was also used to examine the effects of site location on variances in tree density, BA coverage, and heterogeneity, and that in mean volume size, seed number, dry-pulp mass, water mass, and water content of fresh fruits and pellets among different fig species. We used additional multiple-range comparisons (HSD for unequal sample sizes) to locate the differences when significant differences were observed. We performed multiple regression analysis to examine the relationship of bat abundance with density, BA coverage, and heterogeneity of total trees among forest sites. The same analysis was conducted with relative density and relative BA for fruiting trees, total fig trees, fruiting fig trees, and selected predominant food fig species (i.e., F. septica and F. variegata combined), respectively. Linear regression was also conducted to examine the respective relationships of bat density and pellet number with canopy volume [48].