The model chosen as the best model is only the best model among the six models ran in SAS.
Therefore, it is necessary to look at the biological meaning of the model and use logic to
conclude if the model is a good predictor of rendezvous sites. The average frequency of land
cover for presence and absence points reveals very little difference in the percent frequency of
used versus avoided (figure 1). The ‘best’ model may not be a biologically meaningful model. A
much more thorough search will follow using the same methods of deriving variables using
ArcMap and GIS layers. Other variables that might help predict rendezvous sites include: slope,
elevation, aspect, distance to water, distance to roads, distance to areas of high human density,
elk habitat, and land ownership. More logistic regression models will be run and the variables
that significantly contribute to the prediction will be kept. The final model can be input into
ArcGrid Workstation or Raster Calculator to develop a resource selection map to predict wolf
rendezvous sites.