Modelling individual migration patterns using a Bayesian nonparametric approach for capture–recapture data

Abstract

We present a Bayesian nonparametric approach for modelling wildlife migration patterns using capture–recapture (CR) data. Arrival times of individuals are modelled in continuous time and assumed to be drawn from a Poisson process with unknown intensity function, which is modelled via a flexible nonparametric mixture model. The proposed CR framework allows us to estimate the following: (i) the total number of individuals that arrived at the site, (ii) their times of arrival and departure, and hence their stopover duration, and (iii) the density of arrival times, providing a smooth representation of the arrival pattern of the individuals at the site. We apply the model to data on breeding great crested newts (Triturus cristatus) and on migrating reed warblers (Acrocephalus scirpaceus). For the former, the results demonstrate the staggered arrival of individuals at the breeding ponds and suggest that males tend to arrive earlier than females. For the latter, they demonstrate the arrival of migrating flocks at the stopover site and highlight the considerable difference in stopover duration between caught and not-caught individuals.