Assessment of density-dependent responses in white-tailed deer population management (Odocoileus virginianus)

White-tailed deer (Odocoileus virginianus), the most widely pursued game species in the United States, contribute millions of man-days of recreation and billions of dollars of income to society. Conversely, deer herds can cause substantial problems in terms of agricultural damage, vehicular damage, ecological health, and even human health. For decades, managers and biologists have relied on relatively inexpensive and easily collected data from hunter-harvested deer to provide information for making harvest management decisions. Using such data, I sought to improve the utility of this approach by developing quantitative models that assessed the density-dependent responses of physical parameters and recruitment.; To conduct this analysis I obtained long-term data sets for 12 populations in the Southeast that spanned several physiographic provinces and a wide range of densities. Reconstructions for two population segments, total adult and total herd density, were related to 11 commonly used physical parameters. Recruitment models using a simple quadratic expression related fawn density to adult female and total adult densities. Physical parameter and recruitment models both included densities lagged one and two-years in addition to contemporary data.; Buck yearling weight (BYWT), yearling antler measures, notably number of points and spike rate, and doe yearling weight (DYWT) all proved to be useful predictors of density (r2 = 0.16–0.89), with BYWT being the strongest and most consistent. Time lags and population segment proved to be very influential to these relationships with the total adult density lagged two years being the preferred context. Fawn weights (buck and doe) and lactation rate were seldom significantly related to density. Linear rather than curvilinear models were most appropriate across the range of densities examined.; A three-year running average on physical parameters and density estimates improved the predictive power of the models with very little loss in precision. Such an average may have helped minimize the influence of a number of other factors such as mast, weather, and disease.; Recruitment models indicated density-dependent dynamics were operative for eight of the nine populations. While the two-year lag was also the preferred context for these models, the one-year lag was nearly as strong indicating that recruitment responded to changes in density more quickly than physical condition. I used second, third, and fifth order terms in the quadratic recruitment model. The later two appeared to fit the data well indicating that a right-shifted stock-recruitment curve was an appropriate model.; Relative densities based on the third-order recruitment models were related to physical parameters (BYWT, DYWT, and antler measures) allowing comparisons among populations. Slopes were similar with no apparent pattern along a presumed habitat quality gradient suggesting that food resources, the range of relative densities and local genetic variation may be more influential. The small variation in slopes (all between 2 and 4% change in relative density for each one pound change in weight) among populations suggests that relative density models could be widely used.; Validations for both the physical parameter and recruitment models using independent data showed that predicted and observed densities were highly correlated (r = 0.45–0.96).; Some populations did not show significant density-dependent responses for recruitment, physical condition, or both. This underscores the fact that some habitats may be too poor for such a process to be operative or detectable. Managers should be cautious applying these results on exceptionally poor habitats. The efficacy of time lags should also serve to caution managers not to look for responses in herds too quickly.; The populations examined in this study provided long-duration data that undoubtedly captured a great deal of stochasticity...