Learning and adaptation in the management of waterfowl harvests

Abstract

A formal framework for the adaptive management of waterfowl harvests was adopted by the U.S. Fish and Wildlife Service in 1995. The process admits competing models of waterfowl population dynamics and harvest impacts, and relies on model averaging to compute optimal strategies for regulating harvest. Model weights, reflecting the relative ability of the alternative models to predict changes in population size, are used in the model averaging and are updated each year based on a comparison of model predictions and observations of population size. Since its inception the adaptive harvest program has focused principally on mallards ( Anas platyrhynchos), which constitute a large portion of the U.S. waterfowl harvest. Four competing models, derived from a combination of two survival and two reproductive hypotheses, were originally assigned equal weights. In the last year of available information (2007), model weights favored the weakly density-dependent reproductive hypothesis over the strongly density-dependent one, and the additive mortality hypothesis over the compensatory one. The change in model weights led to a more conservative harvesting policy than what was in effect in the early years of the program. Adaptive harvest management has been successful in many ways, but nonetheless has exposed the difficulties in defining management objectives, in predicting and regulating harvests, and in coping with the tradeoffs inherent in managing multiple waterfowl stocks exposed to a common harvest. The key challenge now facing managers is whether adaptive harvest management as an institution can be sufficiently adaptive, and whether the knowledge and experience gained from the process can be reflected in higher-level policy decisions.