Geographic Information Systems (GIS) allow practical new approaches for the study of habitat/resource selection. In this article, we illustrate several statistical techniques that can be used in resource selection studies involving GIS with data from a study of winter habitat selection by moose (Alces alces) on the Innoko National Wildlife Refuge in Alaska. We identify statistical issues to consider when analyzing similar data. The habitat data used is a land cover map derived from Landsat TM (Thematic Mapper). Two scales of selection are considered. Maps are developed that illustrate the relative probability a resource unit is selected by moose. Bootstrapping procedures are used to estimate the final resource selection model coefficients and associated variances by using the moose locations as the basic experimental unit.

Key Words
Geographic Information System; Logistic regression; Moose; Resource selection.

Wallace P. Erickson is Biometrician, and Trent L. McDonald is Biometrician, Western EcoSystems Technology, Inc., 2003 Central Ave, Cheyenne, WY 82001. Robert Skinner is Wildlife Biologist, U.S. Fish and Wildlife Service, Innoko National Wildlife Refuge, P.O. Box 69, McGrath, AK 99627.