Estimating mammalian species richness and occupancy in tropical forest canopies with arboreal camera traps

Large and medium-bodied rainforest canopy mammals are typically surveyed using line transects, but these are labour intensive and usually ignore nocturnal species. Camera traps have become the preferred tool for assessing terrestrial mammal communities, but have rarely been used for arboreal species. Here, we compare the efficiency of arboreal camera trapping with line transects for inventorying medium and large-sized arboreal mammals, and assess the viability of using camera traps in trees to model habitat occupancy. We installed 42 camera traps, spaced 2km apart, in the canopy of the Maijuna-Kichwa Regional Conservation Area, Peru and walked 2014km of diurnal line transects on 22 trails at the same site. We compared the efficiency of each method using species accumulation curves. We applied a multi-species occupancy model, while examining the effect of camera height on detection probabilities, including the distance from a village and from a river as covariates to examine variability in habitat occupancy. In 3147 camera days, 18 species of arboreal medium and large-sized mammals were detected by cameras, while 11 species were recorded on line transects. Ten of these species were detected by both methods. Diurnal species were detected more quickly and with less effort using arboreal camera trapping than using diurnal line transects at the same site, although some species were more easily detected during line transects. Habitat occupancy was positively correlated with distance from the village for two species, and negatively correlated with distance from the river for one. Detection probabilities increased modestly with camera height. Practical limitations of arboreal camera trapping include the requirement for specialized climbing techniques, as well as increased potential for false triggers, requiring extended processing time. Arboreal camera trapping is an efficient method for inventorying arboreal mammals and a viable option for studying their distribution relative to environmental or anthropogenic variables when abundance or density estimates are not required.