An assessment of beaked redfish (S. mentella and S. fasciatus) in NAFO division 3M based on revised 2005-2008 catches (is a retrospective biased assessment necessarily useless in terms of scientific advice?)

The 3M redfish assessment is focused on the beaked redfish, regarded as a management unit composed of two populations from two very similar species: the Flemish Cap S. mentella and S. fasciatus. The reason for this approach is the historical dominance of this group in the 3M redfish commercial catch until 2005. However a new golden redfish fishery (S. marinus) started on September 2005 on shallower depths of the Flemish Cap bank above 300m. This new reality implied a revision of catch estimates, in order to split recent redfish commercial catch from the major fleets on Div. 3M into golden (S. marinus) and beaked (S. mentella and S. fasciatus) redfish catches. An Extended Survivor Analysis (Shepherd, 1999) was performed using the previous XSA framework and the 1989-2008 EU survey abundance at age matrix as the tuning input file. The sequence of negative-positive regions on the log q’s residuals that mark the 1989-2000 interval lead the authors to consider an alternate XSA framework: the 1989-2008 EU survey series was sliced into “three consecutive surveys” (EU survey I, 1989-1993; EU survey II, 1994-2000; and EU survey III, 2001-2008) corresponding to the three regions of the log q’s matrix of residuals. This option resulted on improved catchability diagnostics but with a worst XSA performance in terms of retrospective bias, correlation with the tuning survey and convergence. Sensitivity analysis shown that the model runs better with long survey series based mean log q’s, despite the well defined residual patterns inherent to this option. Very high fishing mortalities until 1996 forced a rapid and steep decline of the 3M beaked redfish stock. With lower fishing mortalities since then the stock decline was halted. But the weak 1991-1997 year classes kept the stock size at a low level till 2002, basically sustained by the survival and growth of the existing cohorts. The cohort’s size at age 4 increased continuously from 1998 onwards, reaching an historical high with the 2002 year class. The 2000-2004 sequence of above average year classes pumped abundance and 4+ biomass to high levels at the beginning of 2008. The reproductive potential of the stock has been well above average over the past ten years and female spawning stock biomass (SSB) is now increasing at a faster rate. But in 2008 female SSB was still far away from a target of 40,000 tons, beyond which the two consecutive above average recruitments from the beginning of the series occurred. Retrospective analysis show more severe retrospective patterns from 2008 to 2009 (however, and regardless the downward revisions of their trajectories, exploitable biomass and SSB on the final year of most recent assessments have very close estimates). Assuming that the retrospective bias observed between the two last assessments can be repeated between the present and next year assessment, cohort dependent retrospective log linear functions were adjusted to observed log XSA2009/2008 ratios, for abundance and fishing mortality. These functions provide the initial population and fishing mortality at age of a projection with two sets of age/cohort specific bias correctors that can be used to minimize in the short term likely discrepancies on SSB and yield trajectories due to retrospective patterns between consecutive age based assessments.