Multivariate models of bobcat habitat selection for Pennsylvania landscapes

An understanding of bobcat-habitat relationships is needed to develop effective bobcat management strategies. I estimated bobcat habitat selection, using radiotelemetry studies, and developed multivariate habitat selection models. Multivariate models of habitat selection, based on Mahalanobis distance calculations, classified 70 percent of the initial study area as suitable for males and 54 percent as suitable for females. Validation of Mahalanobis distance-based models of habitat selection indicated a reclassification success rate of 85.7 percent for males and 86.4 percent for females. Prediction success of Mahalanobis distance models for males, as estimated from independent validation, was 13 percent less than that estimated by crossvalidation. Estimates of predictive success for females were similar. Logistic regression models were developed using cover type, slope, and aspect conditions at radiotelemetry locations and at random points. Independent models were developed using all random points in the study area (complete) and only random points in unsuitable areas as predicted by Mahalanobis distance models (conditional). Parameter estimates were similar, but reclassification success was greater using conditional randomization. Independent validation indicated 78 percent classification success for females and a 71 percent classification success for males. Bobcat home range size was inversely correlated with percent composition of areas classified as suitable habitat by Mahalanobis distance based models and Logistic regression models. Models were extrapolated to produce state-wide maps of bobcat habitat suitability. Total area of 18,564 km2 was classified as suitable for both male and female bobcats, whereas 39,067 km2 was suitable for males but not for females. Suitable male and female habitat was identified in northcentral, central and southcentral regions. Evaluation of 18,770 simulated female home ranges suggested that 4,222 (22.5 percent) contained greater than 25 percent suitable habitat within them. Questionnaire results from 122 Wildlife Conservation Officers and reports of vehicle-caused bobcat mortality supported habitat and home range predictions. State-wide application of multivariate habitat selection models for bobcats in Pennsylvania provides an information source for a habitat-based management decisions and serves as a basis to develop hypotheses concerning local- and landscape-level habitat associations.