A novel method for quantifying habitat selection and predicting habitat use

Abstract

1 The study of habitat selection and habitat use are crucial for understanding the biological requirements of animals and the strategies they use to fulfil their needs. A variety of statistical techniques are available to quantify habitat selection, most of them based on the comparison of habitat attributes in sites used by the animals and in unused, but available, sites. Because of the difficulties in defining what is available from an animal's perspective for many species, statistical approaches such as first-passage time (FPT) analysis have been developed, which explore habitat-use intensities through the areas used by animals. 2 In this study, we expand on that approach by using FPT-derived data within the framework of mixed-effects Cox proportional hazards models (CPH models) to quantify habitat selection. By modelling FPTs, CPH models evaluate which habitat attributes best explain the ‘risk’ or ‘hazard’ of the animal leaving an area of a given radius. For quantitative interpretations, the coefficients of the CPH model can be used to calculate hazard ratios that estimate relative habitat preferences: the lower the hazard ratio (i.e. the lower the risk of leaving), the higher the preference. Accounting for individual variability in habitat use by means of a random-effect term added to CPH models provides inferences that more appropriately reflect the typical hierarchical structure of telemetric data on animal space use. 3 Synthesis and applications. In addition to providing inferences about habitat selection based on the estimated parameters, survival functions of fixed-effects CPH models can be used to construct spatial predictions, for instance maps of population-level space use. Such predictions can be particularly useful for applied purposes, for example as a basis for species conservation plans and reserve selection. The method is illustrated using two data sets from marine mammal species: ringed seals Phoca hispida and white whales Delphinapterus leucas, but it is broadly applicable to habitat selection and prediction studies of other highly mobile animals in marine or terrestrial systems.