Design and analysis of wildlife habitat selection studies typically do not assess the effect of spatial pattern on the habitat selection process. Effects of landscape scale pattern on habitat selection cannot be accomplished without replicate study areas, because pattern is a single, albeit multifaceted, attribute of an area. For a single area, however, the influence of pattern-related characteristics, such as shape and edge shared with adjacent patches, can be estimated by using GLIM (McCullough and Nelder 1983) procedures to model patch-specific frequency counts of animal use as a function of these parameters. This approach is evaluated and illustrated with simulated breeding-bird counts in a South Carolina study area for which a GIS land cover classification is available. A related technique for evaluating whether movement from patch to patch is selective is developed and illustrated for designs that involve collection of trajectory data from monitored individuals. These designs and analyses are feasible given current GIS and GPS technology. Statistical inferences from habitat selection studies should be interpreted within the context of a range of scales at which animals differentiate between patch attributes.

Key Words
Geographical Information Systems; GLIM models; Poisson regression.

David L. Otis is Unit Leader, South Carolina Cooperative Fish and Wildlife Research Unit, USGS/BRD, Clemson University, Clemson, SC 29634.