Counting wildlife by voice – uncertainty-aware population estimates from bioacoustics

Conservation management needs reliable information about how many animals a population holds and whether that number is changing over time. Yet for most threatened species such estimates are often unavailable or imprecise, either because the species is cryptic and hard to detect by visual monitoring or because current methods struggle to separate individuals from one another. Passive acoustic monitoring, i.e. networks of autonomous recorders that capture the soundscape of vocal animals, is already deployed at scales no field survey can match, and AI now identifies which species is calling with high reliability. What it cannot yet do reliably is estimating how many animals of a given species are present at a certain location and time. This interdisciplinary PhD aims to close that gap by combining recent advances in acoustic individual identification (AIID) with state-of-the-art acoustic spatial capture-recapture (SCR) modelling. Together, these will move acoustic monitoring from species detection to within-species population inference.