A closed-loop simulation framework and indicator approach for evaluating impacts of retrospective patterns in stock assessments

Abstract

Abstract A retrospective pattern within a stock assessment occurs when historical estimates systematically increase or decrease as data are removed and has been cited as a cause of persistent overfishing. For two case studies, Gulf of Maine cod and New England pollock, we demonstrated how closed-loop simulation can be used to evaluate the impacts of retrospective patterns with respect to management objectives. Operating models (OM) representing alternative states of nature were developed and various management procedures (MP) that account for retrospective patterns in the fitted assessment models were applied. From the cod example, downward adjustment of the catch advice based on Mohn's rho was more beneficial over model averaging (MA) to meeting biomass objectives from the cod example and avoiding stock crashes. For pollock, downward adjustment and MA were robust to meeting biomass objectives at the cost of foregone catch. The ability to discriminate OMs, using indicators generated from the simulated projections, varied by MP and time for cod, but was poorer overall for pollock. This framework could be used to identify if retrospective issues generate poor management outcomes and, in some cases, alleviate pressure to identify the single most credible state of nature.