Progress in Investigating Differences Amongst SCAA and ASAP Assessment (including Reference Point) Estimates for Gulf of Maine Cod

Various differences between SCAA and ASAP assessments of the Gulf of Maine cod, and their implications, are investigated further. Amongst the results are that neither all nor very few elements of the starting numbers-at-age vector for the assessment should be estimated, but rather an intermediate number which depends on the data available for years close to the starting year selected. This in turn leads to a demonstration that assessments commencing in 1970 incorporate reliable information about recruitment levels for the preceding 5 years, which therefore ought also to be taken into account in fitting stock-recruitment relationships and estimating MSY related reference points. Stock-recruitment relationships incorporating a downturn in recruitment at higher biomass levels are favoured in terms of the AIC statistical model selection criterion for analyses incorporating those earlier years, and suggest lower values for BMSY as well as that the current status of the stock is not overfished. Investigations of the adjusted lognormal and multinomial distributions assumed for fitting proportions-at-age data for the SCAA and ASAP approaches respectively show that neither is appropriate, with each resulting in overweighting of data for younger compared to older ages, and hence to results of lesser precision than need be the case. Further consideration of the domed vs flat survey selectivity issue depends on appropriate modelling of the proportions-at age distributions, and therefore needs to await further work on that topic. Estimation of additional variances in fitting to abundance indices is clearly justified on statistical grounds, while use of numbers rather than mass in fitting to abundance indices decreases the precision of estimates of current (2010) spawning biomass. For most of the SCAA model variants considered, this 2010 biomass is estimated in the 15 – 17 thousand ton range, though under the multinomial surrogate distribution for proportions-at-age data this drops to 14 thousand, while increasing to 20 thousand if numbers rather than mass are used in fitting to abundance indices (though reasons are offered for preferring the mass-based approach).