Report of the Benchmark Workshop on Sandeel (WKSand)

The Benchmark Workshop on Sandeel Stocks (WKSand) 2016 met in Bergen, Norway, from 31 October to 4 November 2016, following data assessment and web conference meetings earlier in the year. Its remit was to review the definition of sandeel stocks and appropriate stock assessment methods for sandeels in the North Sea.
27 people participated including stakeholders. An External Expert and an External Chair from outside the ICES community took part in the process. See participants list in Annex 1. The outputs of the workshop are this report, and the new stock annexes to be used for the assessment of the stocks in the ICES advisory process.
The main outcomes of the workshop were:
Agreement that larval drift in relation to hydrography, otolith chemistry, and inde-pendent dynamics justify definition of 7 separate stocks of sandeels in the North Sea. There was agreement that boundaries between these 7 stocks should be adjusted to reflect new information on hydro-dynamics, differences in management and in be-haviour of the fishing fleet between areas. The main changes were primarily advocated by fishermen and their representatives, and it was agreed to split area 3, with the EU part of area 3 merged with area 2, rationalise the boundary between areas 5 and 7, and 4 and 7, and make small adjustments to the boundary of area 1, making the areas better fit the underpinning science while also being appropriate for practical concerns of the fishing industry.  
It was agreed that the SMS model should be used for stocks in areas SA1, SA2, SA3 and SA4 but that data were inadequate for an analytical assessment for areas 5, 6 or 7. However, the External Experts identified a problem with the SMS model that needs to be addressed: the model is not designed to provide reliable estimates of variance, and the variance estimate derived from the model is a critical component of the escapement strategy TAC setting process. Much variance in the real world is not in-cluded in the SMS model (an obvious example being uncertainty in sandeel consumption by predators, and hence the variance around the estimate of M) and so variance is underestimated by the model, which results in the TAC being set at a higher level than should be the case to achieve the desired metric of less than a 0.05 risk of SSB falling below the reference point threshold.  
It was agreed that the SESAM model seems to be valuable as an exploratory tool. It is possible that SESAM may provide a way to estimate variance with a more compre-hensive coverage of input parameter variability and uncertainty.
In view of evidence derived from analysis of kittiwake breeding success, it was agreed that assessments in SA1 and SA4 should at least take note of total stock biomass thresholds recommended by recent published reviews to avoid depletion of the stock below levels likely to have adverse effects on dependent predators.
A number of recommendations and further research needs were listed for consideration.