Multi-scale, multivariate community models improve designation of biodiversity hotspots in the Sunda Islands

Species occur in sympatric assemblages, bound together by ecological relationships and interspecific interactions. Borneo and Sumatra host some of the richest assemblages of biota worldwide. The region, however, faces the highest global deforestation rates, which seriously threaten its unique biodiversity. We used a large camera trap dataset that recorded data for 70 terrestrial species of mammals and birds, to explore the drivers of regional species richness patterns. Using a multi-scale, multivariate modelling framework which quantified the main environmental factors associated with patterns of biodiversity, while simultaneously assessing individual relationships of each species, we determined the ecological drivers of sampled biodiversity, and their contributions to community assemblages. We then mapped predicted species richness, evaluated the effectiveness of protected areas in securing biodiversity hotspots, performed gap analysis to highlight biodiverse areas lacking protection and compared our predictions with species richness maps produced by using IUCN range layers. Finally, we investigated the performance of each species as an indicator of sampled biodiversity. We demonstrate that biodiversity in Borneo and Sumatra is primarily affected by gradients of ecological and anthropogenic factors, and only marginally by topographic and spatial factors. In both islands, species are primarily associated with elevational gradients in vegetation and climate, leading to altitudinal zonation in niche separation as a major factor characterizing the islands' biodiversity. Species richness was highest in north-eastern Borneo and in western Sumatra. We found that most predicted biodiversity hotspots are not formally protected in either island; only 9.2 and 18.2% of the modelled species richness occurred within protected areas in Borneo and Sumatra, respectively. We highlighted that our prediction for Borneo performed better than, and differed drastically from, the IUCN species richness layer, while for Sumatra our modelled species richness layer and the IUCN one were similar, and both showed low predictive power. Our analysis suggests that common and generalist carnivores are the most effective indicators of sampled biodiversity and have high potential as focal, umbrella or indicator species to assist multi-species vertebrate conservation planning. Understanding existing drivers and patterns of biodiversity is critical to support the development of effective community conservation strategies in this rapidly changing region