Spatial Capture-recapture with Partial Identity

TL;DR

This paper introduces a Bayesian spatial capture-recapture model that infers individual identities from partial data, such as separate left- and right-side photos, by leveraging spatial proximity information.

Contribution

It develops a novel inference framework that accounts for partial identities and spatial information, extending existing models to handle incomplete data in capture-recapture studies.

Findings

Effective Bayesian method for partial identity data

Utilizes spatial proximity to improve individual matching

Extends capture-recapture models to incomplete data scenarios

Abstract

We develop an inference framework for spatial capture-recapture data when two methods are used in which individuality cannot generally be reconciled between the two methods. A special case occurs in camera trapping when left-side (method 1) and right-side (method 2) photos are obtained but not simultaneously. We specify a spatially explicit capture-recapture model for the latent "perfect" data set which is conditioned on known identity of individuals between methods. We regard the identity variable which associates individuals of the two data sets as an unknown in the model and we propose a Bayesian analysis strategy for the model in which the identity variable is updated using a Metropolis component algorithm. The work extends previous efforts to deal with incomplete data by recognizing that there is information about individuality in the spatial juxtaposition of captures. Thus, individual records obtained by both sampling methods that are in close proximity are more likely to be the same individual than individuals that are not in close proximity. The model proposed here formalizes this trade-off between spatial proximity and probabilistic determination of individuality using spatially explicit capture-recapture models.