Modelling individual variability in habitat selection and movement using integrated step‐selection analysis

Abstract

Abstract Integrated step‐selection analysis (ISSA) is frequently used to study habitat selection using animal movement data. Methods for incorporating random effects in ISSA have been developed, making it possible to quantify variability among animals in their space‐use patterns. Although it is possible to model variability in both habitat selection and movement parameters, applications to date have focused on the former despite the widely acknowledged and important role that movement plays in determining ecological processes from the individual to ecosystem level. One potential explanation for this omission is the absence of readily available software or examples demonstrating methods for estimating movement parameters in ISSA with random effects. We demonstrated methods for characterizing among‐individual variability in both movement and habitat‐selection parameters using a simulated data set and by fitting two models to an acoustic telemetry data set containing locations of 35 red snapper (Lutjanus campechanus). Movement kernels were assumed to depend on either the type of benthic reef habitat in which the fish was located (model 1) or the distance between the fish's current location and the nearest edge habitat (model 2). In both models, we also quantified habitat selection for different benthic habitat classes and distance to edge habitat, and we allowed for individual variability in movement and habitat‐selection parameters using random effects. The simulation example highlights the benefits of a mixed‐effects specification, namely, we can increase precision when estimating individual‐specific movement parameters by borrowing information across like individuals. In our applied example, we found substantial among‐individual variability in both habitat selection and movement parameters. Nonetheless, most red snapper selected for hardbottom habitat and for locations nearer to edge habitat. They also moved less when in hardbottom habitat. Turn angles were frequently near ± π but were more dispersed when fish were far away from edge habitat. We provide code templates and functions for quantifying variability in movement and habitat‐selection parameters when implementing ISSA with random effects. In doing so, we hope to encourage ecologists conducting ISSA to take full advantage of their ability to model among‐individual variability in both habitat‐selection and movement patterns.