Ecologists studying emerging wildlife diseases need to confront the realism of imperfect pathogen detection across heterogeneous habitats to aid in conservation decisions. For example, spatial risk assessments of amphibian disease caused by Batrachochytrium dendrobatidis (Bd) has largely ignored imperfect pathogen detection across sampling sites. Because changes in pathogenicity and host susceptibility could trigger recurrent population declines, it is imperative to understand how pathogen prevalence and occupancy vary across environmental gradients. Here, we assessed how Bd occurrence, prevalence, and infection intensity in a diverse Neotropical landscape vary across streams in relation to abiotic and biotic predictors using a hierarchical Bayesian model that accounts for imperfect Bd detection caused by qPCR error. Our model indicated that the number of streams harboring Bd-infected frogs is higher than observed, with Bd likely being present at ~ 43% more streams than it was detected. We found that terrestrial-breeders captured along streams had higher Bd prevalence, but lower infection intensity, than aquatic-breeding species. We found a positive relationship between Bd occupancy probability and stream density, and a negative relationship between Bd occupancy probability and amphibian local richness. Forest cover was a weak predictor of Bd occurrence and infection intensity. Finally, we provide estimates for the minimum number of amphibian captures needed to determine the presence of Bd at a given site where Bd occurs, thus, providing guidence for cost-effective disease risk monitoring programs.
Keywords: Amphibian disease; Atlantic forest; Batrachochytrium dendrobatidis; Bayesian hierarchical model; Tropical streams.