Development and evaluation of an accounting model for estimating deer population sizes

We developed a deterministic population model to simulate dynamics of the white-tailed deer (Odocoileus virginianus) herd in the Mille Lacs Wildlife Management Area, Minnesota. The model was driven by three primary parameters representing reproduction, harvest, and natural mortality. Reproductive rates were age-specific, while harvest and natural mortality rates were age- and sex-specific. Natural mortality rates were estimated using survival data collected from 81 radio-collared female deer during 1996–2000. Minimal variation was observed in annual summer survival rates (range = 96–100%), but annual winter survival rates varied (range = 76–96%) and were related to winter severity indices (P = 0.02). Deer herd dynamics were simulated from 1990 to 2001 and population estimates derived from aerial surveys conducted in 1997 and 2001 were used to validate simulated model estimates. Sensitivity analyses were performed to identify important model parameters and quantify error in model estimates caused by inaccurate input data used during deterministic simulations. Model estimates were most sensitive to biased summer survival, winter survival, and productivity rates of adult females. High variance in model estimates was caused by annual variability in adult female winter survival, adult female productivity, and fawn female summer survival. We recommend collecting data for sensitive model parameters to improve model accuracy. A stochastic model was used to simulate herd sizes that could have occurred in the real population and these population estimates were compared to simulated estimates derived from the deterministic model to assess model performance. This analysis indicated that population estimates produced by the deterministic model would be reasonably accurate for up to 4 years. However, accuracy in modeled estimates declined overtime which suggested that accounting models should not be used exclusively as long-term population monitoring devices for deer management. Additionally, we developed an individual-based population model to examine the effect demographic stochasticity had on error in model estimates for small deer herds. The relative variation in model output decreased in a curvilinear fashion as the number of adult females in the initial population increased. Our findings indicate that while deterministic accounting models are useful to guide research activities and assess potential consequences of proposed management strategies, they are not designed to monitor deer abundance over an extended period of time.