In order to predict the distributions of nine breeding birds in the state of Georgia, I built models consisting of four hierarchical levels built on nested mapping units of decreasing area. The models were fit to the hierarchical logistic regression model using MCMC through the program WinBugs. Results of three fold cross validation showed an average overall correct classification rate of 72%. I tested the models in two management scenarios, choosing between three possible reserves (coarse) and recommending an optimal size for wildlife openings (fine). I evaluated model performance by their ability to differentiate between alternatives despite model uncertainty. We tested the impact of using AIC or DIC and of estimation based on model averaging. The models were able to choose a clear alternative in the coarse scenario, but not the fine one. Using AIC versus DIC or model averaging had no impact on the ranking of alternatives.