Exploring the Potential to Improve Fish Stock Assessments through the Use of Non-Parametric Time Series Modeling

Traditional stock assessment models incorporate varying levels of mechanistic detail in describing fish population dynamics. As a complementary approach, non-parametric time series models offer flexibility in describing non-linear stock dynamics without requiring assumptions about underlying mechanistic relationships. Thus, non-parametric models provide another perspective for thinking about the assumptions and uncertainties associated with traditional stock assessment models. Non-parametric models have proven particularly useful for making short-term forecasts of highly non-linear dynamical systems, but few applications of these models have been made to fisheries management. Potential uses of non-parametric time series modeling include: informing selection of relative biomass indices to be included in stock assessments, forecasting of short-term recruitment trends, and identifying key environmental drivers of population dynamics. We describe an application of non-parametric forecasting models to help inform an existing stock assessment process for king mackerel (Scomberomorus cavalla) fisheries within the US South Atlantic and Gulf of Mexico. We discuss ways in which this approach can potentially be incorporated into the stock assessment process, both in the preparation of data and in the analysis of population dynamics.