Mark-resight model
We used mark-resight models to estimate each species’ abundance within each plot, while accounting for variation in detection probability across species, time of day, and observer. The model shared information about detection and abundance across sites and species, while allowing for site and species-specific deviations via random intercepts and slopes (Frishkoff et al. 2019). Doing so permits robust abundance and detection estimates of both common and rare species. Species abundance was predicted based on plot canopy openness, measured using a spherical densiometer, and mean annual temperature and precipitation variables. The latter were extracted from the Chelsa dataset (Kargeret al. 2017) using plot latitude and longitude.
Mark-resight models were implemented in a Bayesian framework using JAGS V4.2.0 (Plummer 2003) within the R V3.4.4 environment (Su & Yajima 2014). For every species at every plot the model provided estimates of total abundance, and the effects of parameters related to environmental variables. A full model description is available in the supplement.