Evaluating the utility of the Gulf Stream Index for predicting recruitment of Southern New England‐Mid Atlantic yellowtail flounder

The justification for incorporating environmental effects into fisheries stock assessment models has been investigated and debated for a long time. Recently, a state‐space age‐structured assessment model which includes the stochastic change in the environmental covariate over time and its effect on recruitment was developed for Southern New England‐Mid Atlantic yellowtail flounder (Limanda ferruginea). In this paper, we first investigated the correlations of environmental covariates with Southern New England‐Mid Atlantic yellowtail flounder recruitment deviations. The covariate that was most strongly correlated with the recruitment deviations was then incorporated into the state‐space model and alternative effects on the stock‐recruit relationship were estimated and compared. For the model that performed best as measured by Akaike information criterion, we also compared the estimates and predictions of various population attributes and biological reference points with those from an otherwise identical model without the environmental covariate in the stock‐recruit function. We found that the estimates of population parameters are similar for the two models but the predictions differed substantially. To evaluate which model provided more reliable predictions, we quantitatively compared the prediction skill of the two models by generating two series of retrospective predictions. Comparison of the retrospective prediction pattern suggested that from an average point of view, the environmentally explicit model can provide more accurate near‐term recruitment predictions especially the one year ahead recruitment prediction. However, the accuracy of the near‐term recruitment prediction from the environmentally explicit model was largely determined by the accuracy of the corresponding environment prediction the model provides.