Scientia Marina, Vol 67, No S1 (2003)

Forecasting recruitment and stock biomass of Northeast Arctic cod using neural networks

Geir Huse
Institute of Marine Research, Norway

Geir Ottersen
Division of Zoology, Department of Biology, University of Oslo, Norway


We apply an artificial neural network (ANN) to predict recruitment and biomass development of Northeast Arctic cod. The ANN is trained using a genetic algorithm with input time series such as spawning stock biomass of cod, herring and capelin biomass, and temperature. Forecasts were made by training the ANN on parts of the time series (training set), and then using a trained ANN to predict cod recruitment or biomass in years outside of the training set. In general the predictions corresponded well to observations. The correlation (r2) between observed and predicted stock recruitment at age 3 was 0.74, based on a model with temperature, spawning stock biomass, and capelin biomass. The correlation between observed and predicted stock biomass was 0.89, 0.72 and 0.57 for one, two and three year predictions respectively. The best model for the one year predictions was based on input information on cod biomass, temperature, and cod landings. These results illustrate the strong forecasting ability of ANN models. In the light of our findings we discuss the potential benefit of applying ANN models as a forecasting technology in fisheries assessment.


cod; forecasting; neural networks; genetic algorithm; recruitment; abundance; temperature; growth; population dynamics

Full Text:


Copyright (c) 2003 Consejo Superior de Investigaciones Científicas (CSIC)

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Contact us

Technical support