Using a genetic algorithm to optimize a data-limited catch rule

Abstract

Many data-limited fish stocks worldwide require management advice. Simple empirical management procedures have been used to manage data-limited fisheries but do not necessarily ensure compliance with maximum sustainable yield objectives and precautionary principles. Genetic algorithms are efficient optimization procedures for which the objectives are formalized as a fitness function. This optimization can be included when testing management procedures in a management strategy evaluation. This study explored the application of a genetic algorithm to an empirical catch rule and found that this approach could substantially improve the performance of the catch rule. The optimized parameterization and the magnitude of the improvement were dependent on the specific stock, stock status, and definition of the fitness function. The genetic algorithm proved to be an efficient and automated method for tuning the catch rule and removed the need for manual intervention during the optimization process. Therefore, we conclude that the approach could also be applied to other management procedures, case-specific tuning, and even data-rich stocks. Finally, we recommend the phasing out of the current generic ICES "2 over 3"advice rule in favour of case-specific catch rules of the form tested here, although we caution that neither works well for fast-growing stocks.