Recruitment prediction with genetic algorithms with application to the Pacific Herring fishery

Recruitment prediction is a key element for management decisions in many fisheries. Nevertheless, accuracy of predictions is very low. Generally, recruitment is related to spawning biomass and other factors, but with relative low success (i.e. low correlations). We developed a new approach using genetic algorithms (GA) as a tool to produce a formula to predict very high, high, medium, low, and very low levels of recruitment in the Pacific Herring (Clupea pallassi) fishery stock of the west coast of Vancouver Island, British Columbia, Canada. With spawning biomass, sea surface temperature, salinity, and pacific hake biomass data from 1948 to 1989, the corresponding prediction formula is searched through a population of 100 individuals and 100,000 generations with 5% mutation rate. The formula produced 61.9% correct predictions within the time series available. Spawning biomass seems an insignificant factor to establish a recruitment level. Recruitment seems to be marked rather by the environment.