Forecasts of recruitment in South African anchovy from SARP field data using a simple deterministic expert system

Abstract

A rule-based deterministic model is used to forecast recruitment of anchovy Engraulis capensis in the southern Benguela from field measurements of biological and environmental variables collected during two consecutive years. Field data were obtained on a monthly basis during the anchovy spawning season as part of the South African Sardine and Anchovy Recruitment Programme (SARP). A total of six indicator variables was considered for the analysis: distance offshore of the 16°C isotherm; southerly wind stress; anchovy egg abundance; incidence of oocyte atresia in adult females; an index of fish starvation; and oil : meal ratios obtained from commercial data (which were used as an index of fish condition). In the model, data for each variable were assessed against a threshold value. “Active” variables were assumed to contribute towards below-average recruitment by being either above or below their threshold values, as defiNed by the rules. The model was not based on the SARP field data, which were used to test the model. Predicted results from the expert system compared favourably with final estimates of recruitment strength for both 1994 and 1995 (years of below average recruitment), indicating that field measurements of biological and environmental variables may be used in a structured manner to obtain forecasts of anchovy recruitment to the fishery.