Estimation of habitat selection for central-place foraging animals

Analyses of habitat use for individuals occupying discrete home ranges are typically based on comparison with null models that implicitly assume no spatial context for habitat use within the home range. For species that regularly return to a central place, a more appropriate null model for estimation of habitat selection may be that of a declining expectation of resource use with distance from the central place, such as a nest site. When this null expectation is ignored and a uniform-use expectation is used for central-place foragers, we predicted (1) positive bias of selection for habitat types near the central place, and (2) bias will increase with the degree to which habitat types are spatially correlated to the central place. We explored these predictions with simulated data, using a range of selection intensities and spatial correlations. Results from the simulations confirmed our predictions: biases were large and positive for those habitat types proximal to the central place. To correct for these biases, we included distance from the central place as an explanatory variable in habitat selection models of simulated central-place foraging, and we found that including distance as a linear factor successfully reduced these biases. We then applied these models to field data from northern spotted owls (Strix occidentalis caurina) and red-cockaded woodpeckers (Picoides borealis). For both species, distancebased models performed better than the nonspatial (uniform) model: the models were both statistically superior and produced results more in concordance with our biological understandings. Estimates of selection for habitat types that were disproportionately located near the central place were lower in the distance-based models than in the uniform model, corroborating the results from the simulations. The simple distance-based models we used provide a reasonable means to estimate foraging habitat selection for animals for which a central place can be identified.