Satellite ocean data can inform precision fishing

Abstract

Commercial fishing is important both economically and socially around the world. However, there is a need for fishing to become more efficient to maximise the protein caught to emissions released ratio. Predicting fish abundance and location using ocean predictors may enable greater fishing efficiency and precision. The recent release of catch data by the Norwegian Directorate of Fisheries stimulated the Nordic AI Meet to initiate a competition to apply artificial intelligence to increase commercial fishing efficiency. Satellite ocean model predictions and data were used to predict gross catches of ten commercial fish species landed in Norwegian ports, seven days into the future. Random forest regression ensembles generated monotonic linear predictions of actual catch. Predicting the exact location of maximum landings in the future week was elusive, however the predicted distribution maps provide skippers with an additional tool to supplement their knowledge and experience and enable more precise and efficient fishing.