Investigation of the sensitivity of the Ross Sea toothfish assessment to withholding subsets of the available data

Antarctic and Patagonian toothfish stocks in the Antarctic have been assessed within CCAMLR areas using integrated statistical catch-at-age stock assessment models. These models were initially run annually, but have been run biennially since 2007. This document investigates the impact on model outputs of withholding from a stock assessment between one and eight years of age frequency and/or tag recapture data. Examining the impact of considering a subset of the available data simulates both the effect of not updating an assessment every year, and also the evolution of stock assessments with time as more years’ data becomes available.

The 2009 stock assessment of Antarctic toothfish in the Ross Sea region was used as the working model. The 2009 stock assessment of Antarctic toothfish in Subarea 88.2E was used as a sensitivity to examine the impact of withholding data from an assessment based on a smaller and shorter data set. In both instances the ‘base case’ model was used, i.e., the models that were used to set the precautionary yields for each fishery in 2009. The change in the biomass estimated by the models arising from withholding data was found to be highest in the earlier years of stock assessments, when each extra year of withheld data represented a large contribution to the model. In later years, the change in estimated biomass arising from withholding data was much smaller, with values generally within the 95% percentile of the MCMC run from the 2009 assessment. For the 2009 assessment removing up to four years of the early tag data had little effect on the estimate of the initial biomass, or the range of the estimates.

Running updated assessments with catch data, or with catch data and age frequencies, provided no further information on the biomass trajectories in the model considered. Tag recapture data provided the most information to the model, although the model results were not very sensitive to changes in years of tag release and recapture data available to the model. For example even when the number of tag recaptures in the most recent year was doubled or halved, the corresponding estimate of current biomass changed by less than 10%; furthermore, these estimates remained within the 95% percentile envelope estimated by the MCMC posteriors.