Spatio-temporal variation in nocturnal distributions and relative abundancesnext section
We recorded 650 bats (2.2 ± 0.18; range, 1-20) in 298 encounters during 120 transect-nights of field censuses. A single bat was observed in 66.8% of encounters, followed by pairs of bats (16.8%), and in only 3% of cases were bats detected in groups larger than 10 (χ2 = 426.5, d.f. = 4, P < 0.001). On average, we had 6.2 ± 0.53 (range, 0-14) encounters of 13.4 ± 1.43 bats (range, 0-43) per night; or 2.7 ± 0.24 encounters (range, 0-12) and 5.5 ± 0.57 bats (range, 0-32) per transect night, at an overall encounter probability of 75%. Mostly, we recorded five or fewer (72.5%), and rarely over ten (9.2%), bats per km of transect. Bat abundances were not randomly distributed among transect nights (χ2 = 75.75, d.f. = 4, P < 0.001) and were not correlated with the length of the transect (r = 0.135, F 1, 26 = 0.35, P < 0.6). After incorporating the length of each transect, our data translated into a mean linear density of 2.5 ± 0.6 bats/km of transect/night, with a ratio of the variance to the mean of 3.3 (χ2 = 89.3, d.f. = 27, P < 0.001).
Mean density of bats were similar among the 4 periods (2.3 ± 0.66, 2.0 ± 0.54, 2.9 ± 0.76, and 2.3 ± 0.73 bats/km of transect/night) interrupted by three typhoons (F 3, 116 = 0.16, P > 0.5). Among transects, the outskirt trails of Shirahama, Otomi, Midara, and Komi contributed the index values of highest relative importance, followed by Hoshidate Village and Toyohara trail. Each one led in either of the frequency of occurrence, the total relative abundance, or both, and together they accounted for over 54% of the RI value (Table 1).
Table 1. Relative frequency of occurrence, abundance, importance value, and mean linear density (± SE) per kilometer of transect, of flying foxes at different sites on Iriomote.
These two measures were correlated to each other (R = 0.795, F 1, 26 = 44.65, P < 0.001). The highest mean linear density, however, occurred mostly at villages, i.e., Hoshidate, Mihara, Shirahama, Sonai, and Urauchi-Sumiyoshi (Table 1). All villages with higher bat densities, except Mihara, are located on the west coast (W), whereas five villages and three outskirt trails contributed no bat records, seven of which are along the east coast (E). Villages in the east and west did not differ in mean area (ha; E, 12.8 ± 3.3; W, 10.2 ± 0.9), number of residences (E, 78.2 ± 25.2; W, 87.6 ± 10.6), population size (E, 157.5 ± 51.5; W, 163 ± 20.7), population/ha of village land (E, 17.1 ± 3.3; W, 16.4 ± 2.1), or residences/ha of village land (E, 8.8 ± 2.0; W, 8.8 ± 1.1; MANOVA: Wilks’ λ = 0.353, F 5, 7 = 2.57, P = 0.125). Each village in the east, however, had over a four-fold larger surrounding land area (82.93 ± 19.8 ha) used for agriculture or husbandry activities than in the west (20.21 ± 5.03 ha; t = 3.07, d.f. = 10, P < 0.05). Sites in the west also had higher fruit-tree density (47.14 ± 11.13/ha; t = 2.11, d.f. = 12, P < 0.05) and heterogeneity of tree composition (5.39-6.99) than eastern sites (22.15 ± 4.13/ha; 1.5-4.52, except 7.51 for Komi).
Mean linear densities were positively correlated to frequencies of occurrence of bats both among census nights (R = 0.837, F 1, 40 = 89.03, P < 0.001; Fig. 2a), and also among habitat types (R = 0.772, F 1, 3 = 4.44, P < 0.15; Fig. 2b) but with a larger variation. Occurrence and spatial distribution of bats were affected by both the site location (Wilks’ λ = 0.161, F 4, 70 = 2.82, P < 0.05) and the associated habitat types (Wilks’ λ = 0.147, F 4, 70 = 2.69, P < 0.05). The mean total abundance (Fisher’s LSD, P < 0.05) was higher in villages than in other habitats except inland forests, and the mean linear density (P < 0.05) was higher in villages than in mangroves and cultivated land (Fig. 3). Bats perched at lower heights in cultivated areas than in inland forests (P < 0.05), at lower heights on the east coast than on the west (P < 0.05), and were further distances from trails when occurring in mangrove forests than in inland and costal forests (P < 0.05; Fig. 3). Among the five types of habitats, bottom canopy height and ground coverage were positively correlated, but shrub height was negatively correlated, with the numbers of bats occurring (multiple regression R = 0.49, F 6, 170 = 8.93, P < 0.001; Table 2).
On average, bat perches in cultivated land had lower top- and bottom-canopy heights, lower shrub height, and less canopy coverage than those of most other habitat types, especially inland forests. While top canopy heights were higher in villages than in inland and coastal forests, perches in inland forests had greater canopy coverage, and ground coverage of bat perches was lower in coastal forests than in inland forests and cultivated land (Table 2). The heterogeneity of trees (TH) at the < 3 m range was higher in cultivated fields (6.55) and villages (6.19) than in other forested habitats (3.43-4.39). Yet, cultivated fields had the lowest TH at ranges of 3-5 (4.50) and > 5 m (3.79) compared to all other habitats (3-5 m, 5.26-5.78; > 5 m, 4.61-6.19), whereas that of inland forests was highest at both levels (3-5 m, 5.78; > 5 m, 6.19). The bat densities were correlated both with heterogeneity of tree composition (r = 0.655, F 1, 12 = 9.04, P < 0.05; Fig. 4a), and strongly with the density of major fruiting trees (R = 0.893, F 1, 12 = 47.12, P < 0.001; Fig. 4b).
Fig. 4. Relationships of mean (± SE) linear density (bat/km of transect) of flying foxes with (a) the heterogeneity of tree composition and (b) the density of major fruiting trees.
Diets and patterns of resource use
Yaeyama fruit bats fed on at least 31 species of fruits, 13 species of flowers, and seven species of leaves, of a total of 39 species, 21 families, and 28 genera of plants (Table 3). We found feces more frequently (360 locations, 5.1 ± 0.71 locations /transect-day, N = 70) than rejecta pellets (153, 2.5 ± 0.29, N = 62) and dropped fruits (17, 1.4 ± 0.19, N = 12) (χ2 = 63.7, d.f. = 2, P < 0.001). In contrast, the mean number of samples per locations recorded varied, with greater numbers of pellets (8.1 ± 1.25) and dropped fruit remains (5.7 ± 3.11), but lower numbers of feces (2.9 ± 0.17) at each location. We found 53 piles of feces or traces of food remains repeatedly or periodically directly beneath or very near trees of 16 species. The happiness tree Garcinia subelliptica and Ficus septica had the greatest proportions, followed by the large-leaved banyan F. superba, guava Psidium guajava,and charcoal tree Trema orientalis; leaves of the latter were most frequently consumed (Table 3).
Plants dominated the food composition, with seven cultivated species (17.9%). Animal items occurred in lower frequencies and included mostly insects but also a vertebrate (Table 4). Figs were the major items, but with various importance values among the types of samples. Ficus septica and F. variegata were the most dominant items in both fecal and dropped samples. They accounted for over 63% of the FO and 81% of the PM in fecal/pellet samples, followed mostly by other species of figs, mulberries Morus spp., happiness trees,and charcoal trees in lower ranks. In dropped samples, F. septica and F. variegata also led in FO (37%) and PM (61.3%), but were followed, especially in mass proportion, by more non-Moraceae plants in higher proportions, such as chinaberry Melia azedarach, guava, and ebony Diospyros egbert-walkeri (Table 4).