Role of Parameter Uncertainty in Assessing Harvest Strategies

Abstract

Research methods aimed at comparing harvest strategies often implicitly assume that the form of the underlying population dynamics is known but that there is uncertainty in the abundance forecasts that are used to implement the harvest strategy. In reality, however, there is uncertainty in both the model structure and estimated parameters. In this paper, I illustrate the effect of parameter uncertainty on the performance of five different harvest strategies using a simulation model for Chinook salmon Oncorhynchus tshawytscha. The assumed management goal is to determine the harvest control parameters that maximize harvest subject to a minimum extinction criterion. In general, the different harvest strategies performed similarly in terms of the expected mean harvest when optimized using the true simulation model (i.e., when the population dynamics were assumed to be known). However, parameter uncertainty affected the various strategies differently. Harvest strategies with optimal control parameters (e.g., thresholds or exploitation rates) that fell near the region of the parameter space with a high risk of extinction were more likely to result in failed management objectives. The simulation results indicate that simple buffers (e.g., a systematic lowering of exploitation rates from predetermined targets) are likely to substantially decrease the risk of choosing harvest control parameters with a higher than desired level of extinction risk.