Optimizing harvest control rules in the presence of natural variability and parameter uncertainty. pp 124-145 In ‘Providing scientific advice to implement the precautionary approach under the MagnusonStevens Fishery Conservation and Management

Abstract.- Classical one-parameter harvest policies (such as those based on maintaining a constant optimal catch, constant optimal fishing mortality rate, or constant optimal escapement) and full optimal control solutions (such as those generated through stochas-tic dynamic programming) represent two ends of a spectrum of possible harvest control rules. The classical one-parameter policies have little flexibility and may be substantially sub-optimal, but are easy to describe. True optimal control policies, on the other hand, are completely flexible and fully optimal, but they can be inaccessible. As a compromise between the classical one-parameter policies and a full optimal control solution, several authors have suggested that fisheries be managed by specifying the functional form of a control rule a priori and then choosing values for one or more of the parameters so as to maximize a management objective function. The purpose of this paper is to gain an increased understanding of how such harvest control rules can be used to address the problem of optimal fishery management. This is undertaken in three stages, proceeding in order of increasing complexity. In Stage 1, the analysis assumes that population dynamics are completely deterministic and that the values of all biological parameters are known with certainty. Here, the focus of optimization is on maximum sustainable yield (MSY). In Stage 2, the analysis is generalized to the case in which natural (stochastic) variability is present but the values of all biological parameters are still known with certainty. Here, the focus of optimization is on a stochastic analogue of MSY, with attention paid also to tradeoffs between long-term average yield and the level of variability around that average. In Stage 3, the analysis is further generalized to the case in which natural variability is present and the values of biological parameters are uncertain. Here, the focus of optimization is o