0 10 20 30 40 50 60 Harvest rate resource users, should several models continue to make similar predictions. For another, passive management makes it difficult to discriminate between good management and good luck. High harvests could accrue by chance during a series of good years, despite application of a wrong model.

As an example of the potential utility of active adaptive management, let us consider waterfowl harvesting, specifically mallard ducks, in more detail. Harvest quotas for a variety of ducks are determined in part using a sophisticated system of stratified aerial surveys criss-crossing the extensive area of breeding habitat on the North American prairies (Nichols et al. 1995). Density levels and pond availability are used to predict stochastic variability in duck recruitment rates, and these recruitment rates are interpreted as a harvestable surplus (Anderson 1975). Much of the stochastic variability in demographic parameters stems from variation in rainfall on the prairies. Wet weather generates large numbers of small ponds and pothole lakes on the prairies, which in turn generates increased success in offspring recruitment. Banding records, obtained from the recovery of identification bands (in the hunting season), allow an estimation of mortality rates. This information on offspring recruitment and mortality is then used in quantitative population models to predict safe harvesting levels year-to-year. The remarkable consistency in duck numbers over time attests to the robustness of this program (Nichols et al. 1995).

There are indications, nonetheless, that the harvest allocation for some waterfowl species could be considerably improved. A key uncertainty in the harvest evaluation procedure is whether mortality is compensatory or not (Anderson 1975; Williams et al. 1996). In this context, perfect compensation means that increased duck mortality due to harvest has no effect on overall duck mortality, at least over some range of harvest rates, because survival in the wild adjusts perfectly to losses imposed by man (Fig. 15.7). The alternative hypothesis is that there is no compensation, hunting mortality is in addition to natural mortality, and so total mortality is linearly related to harvest rates (Fig. 15.7). Current data are inadequate to discriminate between these two hypotheses, yet they have critical implications with respect to both the risk of over-harvesting, particularly in poor years, and the optimal harvest policy (Anderson 1975). Simulation models have been used to show that by far the most efficient way to decide which of these alternative models is correct is through active adaptive management (Nichols et al. 1995; Williams et al. 1996). Indeed, this may be the only realistic way to reduce the uncertainty in biological processes, at least within our lifetime.

Such an active adaptive management procedure has been implemented, despite the inherent difficulty in coordinating agencies and resource users in a variety of jurisdictions (Nichols et al. 1995). If this kind of coordinated model evaluation can be conducted for waterfowl, it can be used even more readily for less mobile species. The key may be that there has been a long-standing tradition in waterfowl management to apply biomathematical models to the production of recruitment and harvest management. Such models have been applied rarely to wildlife species, for which harvesting policy is often developed in a more haphazard fashion. The adaptive approach demonstrates a more productive option.

15.6 Summary Statistical hypothesis testing is not always the best way to make informed decisions about causal factors associated with wildlife population dynamics, because of preoccupation with rejection of null hypotheses rather than evaluation of the merits of a suite of more plausible models. We outline an alternative approach to inference that is based on information theory, allowing one to decide which model or suite of models offers the best explanation for existing patterns of data. Such an approach complements the practical need to make the best management decisions possible on the basis of incomplete scientific information. A cornerstone of all model evaluation procedures is some measure of goodness-of-fit between models and data. Such model evaluation is an essential component of adaptive management regimes, where alternative explanatory models are vigorously pursued using historical data or experimental perturbation. We show how adaptive management can be used to improve management of harvesting in migratory waterfowl populations in North America.

Was this article helpful?

0 0

Post a comment