Optimising sample sizes for animal distribution analysis using tracking data

Abstract

Knowledge of the spatial distribution of populations is fundamental to management plans for any species. When tracking data are used to describe distributions, it is sometimes assumed that the reported locations of individuals delineate the spatial extent of areas used by the target population. Here we examine existing approaches to validate this assumption, highlight caveats, and propose a new method for a more informative assessment of the number of tracked animals (i.e. sample size) necessary to identify distribution patterns. We show how this assessment can be achieved by considering the heterogeneous use of habitats by a target species using the probabilistic property of a utilisation distribution. Our methods are compiled in the r package SDLfilter. We illustrate and compare the protocols underlying existing and new methods using conceptual models and demonstrate an application of our approach using a large satellite tracking dataset of flatback turtles Natator depressus tagged with accurate Fastloc-GPS tags (n = 69). Our approach has applicability for the post hoc validation of sample sizes required for the robust estimation of distribution patterns across a wide range of taxa, populations and life-history stages of animals.