Status of Stocks & Multispecies Assessment Program
Multispecies Modeling Workshop
In August REFM scientists hosted a 4-day multispecies modeling workshop at the AFSC. The goal of the workshop was to review and develop technical aspects of fitting statistical age-structured multispecies models. Three of these types of models are being developed for the eastern Bering Sea (EBS), Aleutian Islands, and Gulf of Alaska (GOA). The Aleutian Islands study is being developed by a University of Washington graduate student whereas the GOA work is being done by a student at the University of Alaska. The EBS study extends from earlier studies published by REFM scientists. Workshop participants from South Korea and from New England (Grand Banks) also presented approaches to multispecies modeling applications from their regions.
The common linkage over all regions and models was that conventional age-structured models were at the core of the approach. That is, configuration options allow normal stock assessment data (e.g., observer size and age composition estimates, survey abundance indices) to be tuned independently among species included in the model. This provides a realistic diagnostic check for assumptions about individual stocks since they can be compared with standard single-species stock assessments. Also, this allows for the gradual introduction of trophic interactions through the use of bioenergetics and stomach content data.
For the Aleutian Islands, the statistical model was the most highly developed since it implemented the ability to evaluate a variety of functional feeding responses. In particular, the parameterization for these functional responses was done in a manner that allowed for robust estimation methods. Also, consumption rates as predicted from the model were used to compute ration, and the ration was subsequently tuned to information on the available bioenergetics for each species. Data availability and the complexity of the boundary current Aleutian Islands system were seen as problems that may require additional assumptions for parameter estimation purposes.
For the EBS, the data are more extensive with a somewhat simpler environmental setting. The development of this model was closer to traditional multispecies virtual population analysis (VPA) methods where age-specific prey “suitabilities” were taken as values based on observed stomach content data. This model implements an algorithm that stabilizes the populations (which are generally conditioned on each other’s abundances) and allows for Bayesian evaluation of model uncertainties. Model developments underway include a careful evaluation of the impact of arrowtooth flounder abundances in the EBS and the ability to use size-based stomach content data rather than approximated age. Also, there appears to be the potential to link growth data with temperatures and evaluate interannual variability in ration, particularly for EBS pollock.
