Spatial turnover analysis
Non-metric multidimensional scaling analyses (NMDS) based on presence-absence data using the Jaccard similarity index were performed for four major spatial components (i.e., forest types, elevational zones, localities, and habitats). For this purpose, each single trap was assigned for a particular spatial component (Table S1). Forest types included four entities: a) evergreen wet lowland forests, b) evergreen dry lowland forests, c) sub-montane forests, and d) montane forests. Elevation had five units: EZ1: 0–500 m, EZ2: 501–1000 m, EZ3: 1001–1500 m, EZ4: 1501–2000 m, and EZ5: 2001–2500 m. The entity ‘locality’ included the 11 individual sampling localities. Habitat comprised seven types: abandoned plantation, grassland, rock outcrop, hilltop, forest edges, central forest, and disturbed forest. NMDS ordination was performed on the full data set. Entities of spatial components (i.e., forest types, elevational zones, localities, and habitats) were subsequently mapped on the ordination results. Spatial turnover analysis as well as a regression between qualitative species composition similarity and geographic distances of sampling sites were performed for the full assemblage and for assemblages partitioned by body-size and phylogenetic lineage membership (see above).