To investigate spatio-temporal variability of krill body length and number of bycatch fish, observer data sets on Japanese krill fishery from 1995 to 2008 were analyzed by using a hierarchical Bayesian model. The model was composed of multistage cluster units (i.e., years, subarea, vessels, and haul) based on a state-space model, which can separate biological process error in the population dynamics from observation error caused by the fishery activity and individual observation. In both krill length and bycatch fish number, the parameters estimated by MCMC showed the variation among years, subareas, and vessels. The potent interaction effect between year and subarea suggests large spatiotemporal variability of krill population structure and a difficulty in predicting krill population dynamics. Variances of krill length and bycatch fish number by sampling stages were calculated by the multistage sampling formula with the variance terms derived from the Bayesian model. For both krill length and bycatch fish number, haul coverage ranging 0–50% showed marked effects on CV, although vessel coverage hardly changed CV. The results of this study suggest that scientific data collectionby commercial fishery is an important source of information for the management of krill resources and Antarctic ecosystem, while enlarging haul coverage > 50% have smaller effects in improving data accuracy and may not balance the rising cost of observer program.