Flexible density surface estimation for spatially explicit capture-recapture surveys

Abstract

1. Existing spatially explicit capture-recapture (SECR) software does not have the ability to fit flexible nonparametric models of animal density. 2. We describe and implement in the R package secrgam, a flexible method for estimating density surfaces from SECR data, using regression splines. 3. Package secrgam is an extension of package secr to implement some models available in the generalised additive model package mvcv. It accommodates density models that are arbitrarily flexible functions of spatially- and temporally-referenced variables. This includes one-dimensional and multi-dimensional smooths of covariates and smooths with interactions. The shape and smoothness of the fitted density surfaces is data-driven and can be determined using AIC or similar criteria. We illustrate use of the package by estimating the density surface from a simulated camera trap survey of leopards. 4. Package secrgam provides a flexible tool for species distribution modelling using SECR data.