Latent ID Acoustic Capture-Recapture

Speaker: Paul Van Dam-Bates (University of St Andrews)

Abstract

Spatial capture-recapture can be applied to acoustic monitoring using a grid of microphones where the call is “recaptured” across the grid. However, currently it is only able to estimate “call” and not “caller” density due to the uncertainty in individual ID between calls. For species that stay relatively close to the same location throughout the study, we observe clusters of calls around their activity centre. Using this spatial information can help differentiate the unique individuals in the population. I will review some modern methods that use this type of information for assigning latent capture histories when individuals are not identified. Finally, I will introduce some new ideas using Dirichlet Process Mixture Models to partition the set of latent capture histories to the most likely group in the context of terrestrial acoustics.