Modeling winter habitat for white-tailed deer in southwestern Virginia

Pellet group surveys were conducted on 21 transects in February-March, 1985, and January-March, 1986, to estimate relative deer densities on eleven study areas on the Jefferson National Forest, southwestern Virginia. Habitat data were collected on the same transects in July-September, 1985. Additional habitat information was measured from aerial photographs and USDA Forest Service compartment maps. These data were used to develop eleven multiple linear regression models and one pattern recognition (PATREC) model for predicting deer winter habitat quality, based on the assumption that relative density of deer is a good indicator of habitat quality. The densities of evergreen broad-leaved shrubs and"Nonforage" shrubs, basal area, mean distance to a field, and percent slope were among the most important variables selected in the regression model building process. Six variables were selected for use in the PATREC model: mean tree diameter, oak basal area, basal area of"Other Winter Forage" tree species, density of"Nonforage" shrubs, mean distance to a gated gravel road, and mean canopy closure. Spearman's rank correlations were used to compare the model outputs with estimated pellet group densities. All of the models had correlation coefficients ≥ 0.60, four had correlation coefficients > 0.80. The models need to be validated, i.e. tested with independent data from areas outside the study sites. These tests will help refine the models and assess their effectiveness in other regions of the southern Appalachian Mountains.