A stochastic simulation model, designed to test the effects of alternative management schemes on refuging waterfowl populations, was constructed from data on fall-migrating snow geese (Chen c. caerulescens) at DeSoto National Wildlife Refuge, Iowa. Components of the model include goose population level, food density and distribution, food-searching flight characteristics, feeding rates, activity and energy budgets, migration rates, and the effects of weather, hunting pressure, and land management practices on the system. Independent data were collected to test model validity. The model validly describes movement from the refuge core to distant feeding areas and provides insight into factors affecting emigration. Refuge population level was not sensitive to shifts ((+OR-)20%) in the input values of 25 selected parameters, but hunting mortality and the daily distances moved to feeding areas were sensitive to several combinations of parameter perturbations. Model outcome was most sensitive to changes in digestive efficiency, mean food density, and the proportion of refuge fields in which food was available. In other model experiments, increased hunting pressure caused significant (P < 0.05) increases in hunting mortality and a reduction in the refuge population. The direct affect of hunting was relatively less important in reducing waterfowl population size than the associated disturbance of feeding geese by hunters. Disturbance reduced energy gains and subsequently hastened emigration. Geese in the DeSoto Refuge vicinity feed almost exclusively on waste corn. Simulated hunting mortality increased significantly (P < 0.0001) when waste-corn density was moderately reduced because increased goose movement resulted in more hunting opportunities. But at extremely low levels of food density, the use of the refuge by waterfowl declined and hunting mortality was reduced. By providing super-abundant food on the refuge, hunting mortality was reduced and waterfowl use increased. An artificial reduction in size of the feeding arena, from a radius of 75 miles to 5 miles, did not affect waterfowl population levels because food supplies were adequate within 5 miles of the refuge core. The model can be used to simulate the effects of other management scenarios, and is a valuable tool for identifying and meeting management objectives.