Model selection results. Abundance was modelled with habitat and status as site-level covariates.

K = number of parameters used.
Delta AICc = difference between lowest AICc model and model AICc.
AICc weight = model probability among all candidate models.
Detection probability was modelled with observation-level covariates: cloud = cloud percentage cover; habitat = building, non-palm or palm; status = presence or absence of P. grandis; and temp = temperature. We used the corrected Akaike Information Criterion (AICc) to determine the best supported model.