Incorporating Utilization Distributions into the Study of Resource Selection: Dealing With Spatial Autocorrelation

Abstract

An accurate estimate of resource use by an animal can be summarized by regressing local- and landscape level resources on an individual animal’s or population’s utilization distribution (UD) in a spatially explicit way. The resulting equation is termed a Resource Utilization Function and the regression coefficients indicate the intensity, direction, and consistency of resource use. However, using the UD as a response variable requires sampling individual pixels within the UD, which introduces spatial autocorrelation into the data, potentially affecting results and conclusions regarding resource use by the study animal. We discuss methods to remove the spatial autocorrelation so that the significance of regression coefficients for each resource can be evaluated for each animal. By using this methodology, we are able to quantify the individualistic nature of resource selection and test for consistency in use of resources by a population. We demonstrate the effects of removing spatial autocorrelation on individual parameter estimates of resource use by individual animals using radio telemetry data for 25 radiotagged Steller’s jays (Cyanocitta stelleri) in western Washington. The technique of removing spatial autocorrelation from our samples allows the use of pixel-level sampling of the UD in future resource selection studies.