Hook selectivity models assessment for black spot seabream. Classic and heuristic approaches.

Abstract

Size selectivity of the deep water longline used in the black spot seabream (Pagellus bogaraveo) fishery in the Strait of Gibraltar was studied with data of four sizes of hooks. Logistic (classic) and Artificial Neural Networks (heuristic) selectivity models were fitted for two experimental fishing trials. Logistic selectivity model was adequate for only one of the two periods analysed and the inferior results obtained with the classical approach were significantly improved by ANNs. These results indicate that in the event that the classic models do not fit well, perhaps due to poor quality of the data (such as a smaller sample size or highly overlapped distributions), the simpler ANNs models, with capacity to combine linear relationships and highly non-linear, are most appropriate to establish the functional relation between variables.