I made six sensitive runs of CASAL models (LENGTH/AGE_IUU_MODEL) for stock status assessments of *Dissostichus eleginoides *in research block C in Division 58.4.4 b following the recommendation during the latest WG-SAM meeting. The six models with three kinds of assumptions on IUU catch were arranged referring to WG-SAM-14/18 as follows: 1) catch at length/age models on the assumption of no IUU fishing occurrence (Len_1/AgeSing_1 models), 2) catch at length/age models with estimation of IUU catch within the model using disease mortality functionality (Len_2/AgeSing_2 models), 3) catch at length/age models on the assumption that 22 % of IUU catch to the total observed across Division 58.4.4 occurred in the block (Len_3/AgeSing_3 models). Single age-length key (ALK) was used for each age structured model. Selectivity and associated index for disease mortality were estimated for Len_2/AgeSing_2 models but the selectivity for IUU fishery was fixed for Len_3/AgeSing_3 models. The other common conditions among six models were set as follows: parameters for ALK (k, Linf and t0) were fixed; year class strength (YCS) was estimated; double-normal selectivity for legal fishery was estimated; mean-based weighting method was used for proportion-at-length/age and tag-recapture; uniform prior was used for initial biomass (*B*_{0}) estimate.

The MPD estimates of the initial and current spawning biomass and current vulnerable biomass were highest for models on the assumption of 22% of total IUU catch and lowest for models with assumption of no IUU catch for both size- and age structured models. The age structured models were higher in each kind of biomass than size structured models for each assumption of IUU catch condition.

Every model showed more convergence in MCMC posterior trace for *B*_{0} compared to my previous versions. MCMC estimates for median biomasses became larger than the MPD ones for each model, but the difference between them was considerably smaller compared to my previous versions.

Further sensitivity runs derived from Len_1 model (LEN_1_DER_MODEL) were examined using eight options changing conditions, i.e. estimate/fix for given parameters, uniform-log prior for *B*_{0}, and multinomial-based weighting for proportion-at-length and/or tag-recapture. From the results, the smaller differences between MPD estimates of median *B*_{0} and MCMC ones and more convergence in MCMC posterior trace for *B*_{0} for the current LENGTH/AGE_IUU_MODEL runs compared to my previous versions seem to be due to the use of the mean-based weighting instead of multinomial-based weighting used in my previous versions.

This report does not yet address the investigation on likely fixed selection patterns for estimation of IUU catches for Len/AgeSing_2 models (para. 2.24 (iii) of WG-SAM-2014 report). I need advice about the methods during WG-FSA meeting.