Abstract
White-tailed deer (Odocoileus virginianus) populations have recovered to about 30 million animals in the United States, but land cover has changed during the interval of recovery. To address the relationship between deer densities and current land cover at regional scales, I applied random forests and extreme gradient boosting classifiers to model low and high deer density classes, at two different thresholds (5.8 and 11.6 deer/km(2)), and land classes in three regions during approximately 2003. For low and high deer density classes divided at 5.8 deer/km(2), deciduous broadleaf forest overall was the most influential and positive variable in the central east and central regions and crop and pasture were the most influential and negative variables in the southeast region. Deer density increased with area of deciduous and mixed forests, woody wetlands, and shrub in all regions. Deer density decreased with area of crop, developed open space, and developed low and medium residential density in all regions. For density classes divided at 11.6 deer/km(2), deer density had the strongest relationship with woody wetlands in the central east region, mixed and deciduous forest in the southeast region, and woody wetlands and herbaceous vegetation in the central region. Deer density increased with deciduous and mixed forests, woody wetlands, and shrub in all regions. Conversely, deer density decreased with herbaceous vegetation, crop, and developed low residential densities in all regions. Therefore, at regional scales, deer overall occurred at greater densities in forests and woody wetlands and lower densities in agricultural and residential development, which did not appear to support more deer. Deer preference for forests does result in damage to forest products, but alternatively, some may consider that deer provide important socioeconomic and ecological services by reducing number of small trees, particularly in the absence of other disturbances that historically controlled tree biomass.