Short-term forecasting of halibut CPUE: Linear and non-linear univariate approaches

In the present paper, two univariate forecasting techniques were tested to evaluate the short-term CPUE capacity forecast for Pacific halibut, Hippoglossus stenolepis (Pleuronectidae). The first methodology, based on the Box–Jenkins approach (autoregressive integrated moving average models [ARIMA models]), assumes a linear relationship between the time series data. The second methodology, using artificial neural network models (ANNs), enables highly non-linear processes to be modelled. The best results from a seasonal ARIMA model indicated that one non-seasonal autoregressive term combined with a non-seasonal moving average term and a seasonal moving average term was suitable to explain a variance level of 32.6% in the validation phase, providing statistically acceptable but insufficiently satisfactory estimations. The configuration of the best ANN model (three autoregressive terms in the input layer and five neurons in the hidden layer) provided a significant improvement in the independent validation phase (91% of the variation explained), indicating a clear non-linear relationship between variables. Modelling of the abundance indices is a useful tool for understanding the dynamics of populations and may enable short-term quantitative recommendations for fisheries management to be made.