We examined spatio-temporal variability of krill body length and number of bycatch fish as variables of interest based on the scientific observer data for the 2010/2011 fishing season. Both krill length and number of bycatch fish were analyzed by using a hierarchical Bayesian model composed of multistage cluster units (i.e., month, sub-area, fishing gear, flag state, vessel, cruise, and haul) incorporated in a state-space model that separates biological process error from fishery process error and observation error. The parameters of the model were estimated by the Markov chain Monte Carlo (MCMC) method in WinBUGS with statistical software R. Although the posterior distribution adequately converged for the krill length model, some parameters did not converge well in the bycatch fish model. The interaction between month and sub-areahas large effects on krill length. Krill length varies among cruises, but there is no clear difference among cruises within a vessel. The uncertainty in parameter estimation is large in the sub-area effect and the interaction effect between month and sub-area. For the bycatch fish model, there is no obvious influence in biological process and fishery processes except for cruise effect on the number of bycatch fish. Some cruises on which fishing gear TMB was used showed large number of bycatch fish with large its variance, which suggests the necessity of reviewing the procedure of data collection with considering the difference of fishing gear.
Sarah Mackey (CCAMLR Secretariat)