Effect of data weighting on the mature male biomass estimate for Alaskan golden king crab

Abstract

Size-structured integrated population dynamics models are used to estimate the time-trajectories of mature male biomass (MMB) of Alaska crab stocks for stock status determination and harvest allocation. Lack of annual biomass surveys makes it difficult to assess the status and biomass of the Aleutian Islands golden king crab (Lithodes aequispinus). The assessment for this stock relies on commercial catch, size-composition, crab bycatch in groundfish (trawl and fish pot) fisheries, effort, catch-per-unit of effort, and tagging data to determine the biomass and other stock assessment parameters. The effect of data re-weighting (i.e., stage-2 weighting) methods on MMB estimates was investigated for this stock in relation to the sensitivity of the trends in MMB to the data re-weighting method. The McAllister and Ianelli, and Francis methods were used to re-weight the size-composition data and Punt’s method was applied to re-weight the tagging data. Model misspecification (e.g., natural mortality and growth) and the effect of omitting a potentially conflicting data source on estimates of MMB were also investigated. Re-weighting and model misspecification changed the magnitude of estimated values for MMB and their coefficients of variation, but not the MMB trends. The stage-2 weighting of tagging data led to slightly lower estimates of MMB. Under the robust multinomial likelihood for size-composition data, there was not much of a difference between the results of the McAllister and Ianelli method, which ignores correlations in residuals for size-compositions, and the Francis method, which explicitly accounts for these correlations. Specifically, both re-weighting methods led to similar trends, precision, and point estimates of MMB. The R0 profiles indicated that there was information for abundance estimation when all the data were considered under base or variable growth increment scenarios. The CPUE indices were more informative about absolute abundance than the size-composition data. Hence the issue of data weighting should continue to be explored using case studies.