Determining landscape factors influencing tropical amphibians using a multi-species occupancy model

Background/Question/Methods - The International Union for Conservation of Nature has designated nearly one third of amphibian species as threatened or extinct, citing habitat loss as the greatest threat to amphibian persistence. The Brazilian Atlantic Forest (AF) has high diversity and endemism of amphibians, but much of the remaining forest is fragmented. The purpose of this study was to quantify the response of local amphibian communities to landscape-scale characteristics. Specifically, we estimated rates of occupancy and detection for each species, as well as species richness across 50 AF streams in southeastern Brazil across a gradient of forest cover ranging from 11-100%. During 2015–2016, we surveyed amphibians in 100-m stream segments using Automated Acoustic Recorders (ARR) and the Standardized Acoustic and Visual Transect Sampling (SAVTS) methods. We built a hierarchical multi-species occupancy model to investigate the influence of forest cover and catchment area (drainage basins) on the occupancy probability for individual species, while accounting for the imperfect detection of each species using the two survey methods. We also evaluate the influence of landscape covariates on species richness at each site.

Results/Conclusions - We observed 36 amphibian species across the sampling locations, but our multi-species occupancy model estimated richness as 52 species (95% credible interval: 39-71). The community-level effect of catchment area on occupancy was negative, however, only three species showed a significant effect (i.e., credible intervals that did not overlap zero). There was no effect of forest cover on occupancy. Mean posterior richness was not influenced by forest cover or catchment area. We found a significant effect of sample method on detection at the community level, with SAVTS performing better than ARR. However, species-level parameter estimates varied; AAR performance was inferior to SAVTS for only six species, four of which were not calling during the study period and thus could not be recorded via AAR. Species detection probabilities were also influenced by Julian date, with peak community-level detection occurring during the middle of the rainy season, although few species’ detections were significantly influenced by the survey date. Overall, we found that the forest cover and catchment area were not useful predictors of amphibian’ richness, but at the community-level, occupancy probability is highly associated with small headwater streams.