The gray wolf (Canis lupus) started recolonizing northern Wisconsin in the late 1970s, 15 years after extirpation from the state. The population in the state has increased from less than 50 in 1980 to over 500 in 2007 and spread spatially as well. As populations recover and optimal habitat is occupied, the factors influencing habitat selection may change. Mladenoff et al (1995) used a logistic regression model to predict habitat suitability based on pack presence in 1993. They found that road density, an indication of human contact, was the best predictor of wolf presence on the landscape at that time. The model was reassessed using newly established packs between 1993 and 1997 (Mladenoff et al. 1999) and continued to perform well. We calculated a new habitat selection model for the gray wolf after 30 years of recovery in northern Wisconsin. We also compared this model to our original (1995) model to assess how habitat selection has changed. We used logistic regression to model wolf presence in the current landscape using 95 randomly selected pack areas from 2007 surveys and 95 randomly located non-pack areas. Predictor variables included land cover classes from the 2001 National Land Cover Data, road density, highway density, distance to highway, and stream density.

Results/Conclusions

The stepwise logistic regression procedure converged on a new model that explains wolf presence based on a negative relationship with open water and agriculture and a positive association with non-forested wetlands. To compare this new model with the original roads-only model, we calculated a model based on the new wolf pack data using roads alone. This model is still significant, though with less explanatory power than the new model. Regressing road density on the landcover variables from the new model also shows that road density is significantly related to these landcover variables. Our findings show that as the most preferred habitat, based on the original roads model, has become occupied, factors associated with human contact (roads, agricultural areas) remain as negative influences on wolf selection on the landscape, but become more complex in interesting ways. As the delisting in the western Great Lakes is complete and the gray wolf is afforded less protection, these factors may again shift. As we suggested in our previous work, predictive models calculated on dynamic, colonizing populations should not be considered to reflect permanent relationships.