Pairing a user-friendly machine-learning animal sound detector with passive acoustic surveys for occupancy modeling of an endangered primate.

Population declines and range contractions due to habitat loss are pervasive among nonhuman primates, with 60% of species threatened with extinction. However, the extensive vocal activity displayed by many primates makes them excellent candidates for passive acoustic surveys. Passive acoustic survey data is increasingly being used to support occupancy models, which have proven to be an efficient means of estimating both population trends and distributions. Passive acoustic surveys can be conducted relatively quickly and at broad scales, but efficient audio data processing has long proven elusive. The machine learning algorithm BirdNET was originally developed for birds but was recently expanded to include nonavian taxa. We demonstrate that BirdNET can accurately and efficiently identify an endangered primate, the Yucatán black howler monkey (Alouatta pigra), by sound in passive acoustic survey data (collected in southeastern Chiapas, Mexico), enabling us to use a single-season occupancy model to inform further survey efforts. Importantly, we also generated data on up to 286 co-occurring bird species, demonstrating the value of integrated animal sound classification tools for biodiversity surveys. BirdNET is freely available, requires no computer science expertise to use, and can readily be expanded to include more species (e.g., its species list recently tripled to >3000), suggesting that passive acoustic surveys, and thus occupancy modeling, for primate conservation could rapidly become much more accessible. Importantly, the long history of bioacoustics in primate research has yielded a wealth of information about their vocal behavior, which can facilitate appropriate survey design and data interpretation.