Estimates of population parameters from different configurations of an integrated model under development for Antarctic krill are compared to estimates from CMIX and TrawlCI using the same net trawl data. The different configurations of the integrated model were varied based on whether acoustic data or net densities were used as model inputs for the biomass data, on how the biomass data were weighted, by whether natural mortality was estimated by the model or pre-assigned a value of 0.8, and by whether a narrow or wide version of the age-length transition matrix was used. The biomass data calculated for most combinations of samples grouped by year, area, and leg was somewhat lower when it was extrapolated from net densities than when biomass was calculated from acoustics. This was reflected in lower model estimates of total population abundance for most years and areas when the input data were based on net biomass densities in either TrawlCI or the integrated model than when acoustic measures of biomass were supplied as data to the integrated model. All estimates from the integrated model but not TrawlCI included selectivity parameters for the effects of differential availability to the surveys of different ages in the population. Thus the integrated model estimated two population sizes, the "vulnerable" population based on the relative availabilities of individuals of different ages to the surveys, and the total population including the individuals not available to the surveys. The "vulnerable" population sizes in the integrated models were usually more similar to TrawlCI estimates of population abundance than to the integrated estimates of total population abundance, particularly when acoustic measures of biomass were used to inform the integrated model. Time relationships in the length compositions were more evident in the integrated model estimates than in those from CMIX. Modifications to improve the estimation of selectivity parameters when multiple sources of biomass survey data are available are continuing.
Sarah Mackey (CCAMLR Secretariat)