Estimating stock depletion level from patterns of catch history

Abstract

AbstractThe degree to which a stock is depleted is one of the most important quantities in fisheries management because it is used to quantify the success of management and to inform management responses. However, stock depletion is extremely difficult to estimate, particularly with limited data. Using the RAM Legacy database, we developed a boosted regression tree (BRT) model to correlate depletion with a range of predictors calculated from catch data, making the model usable for many fisheries worldwide. The most important predictors were found to be catch trends obtained from linear regressions of scaled catch on time, including regression coefficients for the whole catch time series, the subseries before and after the maximum catch, and in recent years. Eight predictors explain about 80% of variation in depletion. There is a correlation of .5 between measured levels of depletion and the predictions of the BRT model. Predictions are less biased when the stock is fished down below half of the carrying capacity. The BRT model outperforms comparable existing catch‐based depletion estimators and could be used to provide priors for depletion for data‐poor stock assessment methods, or used more directly to provide estimates of the probability that depletion is below a given threshold value.