During the first week of June, colleagues from the Alaska and Northwest Fisheries Science Centers, the University of Washington, and the Chilean fisheries agency INPESCA (Instituto de Investigación Pesquera) convened an informal technical workshop on methods for developing management strategies evaluations (MSEs).
The activities centered on a variety of case studies including Bering Sea yellowfin sole and Chilean hake. Additional overviews on methods used for Pacific sardine and Gulf of Alaska pollock were also presented. Approaches for modifying current models within an MSE context were reviewed and developed. This included converting "estimation models" into "operating models" that can be tested over a broad range of alternative hypotheses and sources of estimation uncertainty. In particular, the workshop participants developed code to generically call operating models which will then provide output that can be processed and evaluated with estimation models (and by extension catch specification/control rules).
A side benefit of this exercise was honing skills in simulating data from known population dynamics parameters to test assessment model estimation performance.
Activities focused on two main topics: 1) ways to condition an operating model that reflects (to the extent practical) uncertainty in the resource dynamics with the ability to generate new data (given a future catch level) and 2) testing the robustness of control rules (quota specification processes) given new data sequences arising from the operating model. Ideally, operating models should reflect the uncertainty of processes and measurement errors for the stock in question.
One method is to "grid" over different structural models (and use the point estimates obtained). This was contrasted with a separate technique in which operating model simulations are derived from Markov chain Monte Carlo (MCMC) estimates of the posterior probability densities.
The group also developed and reviewed a hybrid approach which used MCMC-derived posterior distributions for different model configurations (fixed values of stock recruitment steepness). With the generous assistance and expertise from NWFSC scientists, methods for simulating data using the stock synthesis program (using the R package r4ss) were developed.
For further details and technical documentation, contact James Ianelli (Jim.Ianelli@noaa.gov).
