Report of the Workshop on management strategy evaluation for the Pandalus in Subdivision 3.a.20 and Division 4.a East fishery (WKPandMSE)

To follow up a request from Norway to ICES regarding a new management plan for the northern shrimp stock in Skagerrak and the Norwegian Deep, the Workshop on management strategy evaluation for the Pandalus in Subdivision 3.a.20 and Division 4.a East (WKPANDMSE), chaired by Guldborg Søvik (Norway), met on 23–25 August 2016 in Bergen, Norway. The report was finalized by correspondence.
The MSE was conducted using a modified version of the HCS harvest rules simula-tion program. The version HCS16L has a stock model structured by age groups, but with life history and selection modelled depending on length-at-age. All evaluations were made relative to a base case simulation reflecting current stock status and dynamics, as well as current assessment accuracy. Discarding level was assumed to occur as presently. Thirty years were simulated (1000 iterations) and mean values calculated for the last 11 years of the simulation period. Two different recruitment functions were used in the simulations, one representing the recent (since 2008) peri-od of predominantly poor recruitment, and one parameterized over the longer time-series with high recruitment values (1988–2007). This allowed modelling the current situation that management must deal with, as well as confirming results against the historical fishery and giving guidance as to how the fishery might be expected to change if recruitment returns to its previous higher level. 

For the “historic recruitment” scenario (1988–2007), an F of 0.56 was found (corre-sponding to long-term average yield of 13 200 tonnes), while for the “low recruitment” scenario (2008–2014), an F of 0.32 was found (corresponding to long-term average yield of 6300 tonnes). These FMSY are the highest values consistent with the precautionary principle of having less than a 5% chance of dropping below B_lim, implying that point estimates are given (not ranges). The yields are removals from the sea. Discards must be deducted to obtain landings estimates. Historical discards (2009–2014) are believed to be running at around 10% of biomass, thus predicted landings can be expected to be approximately 10% lower than the yields presented here. The values obtained for the historic recruitment scenario roughly tally with the historical development of the fishery (range of 1988–2007 landings: 10 200–16 000 tonnes, and F: 0.30–0.74). The low recruitment scenario represents the optimum that can be achieved at current recruitment levels. Until, and unless, recruitment returns to the previous higher levels it seems unlikely that the stock would be able to sustain harvest levels similar to those obtained prior to 2008. The analysis also allowed for an exploration of what would be likely to happen if the F_MSY from one recruitment sce-nario was applied to the other scenario. Applying the F_MSY = 0.56 to the low recruit-ment scenario (i.e. recruitment is lower than anticipated) would give a risk level of 28.9% and expected long-term average yield would rise from 6300 tonnes to 7100 tonnes. Conversely, applying the F_MSY = 0.32 to the historic recruitment scenario (i.e. recruitment is higher than anticipated) reduces the risk to 0.1% and the long-term projected yield to 10 800 tonnes (down from 13 200 tonnes). There is thus a modest loss of yield (2400 tonnes) from underestimating recruitment, but a high risk (28.9%) of going below B_lim if one were overly optimistic about future recruitment. Including a 10% interannual quota flexibility (’banking and borrowing’) in the harvest control rule had an insignificant effect on the performance of the rule. 

The structure of the HCS tool, which does not include an explicit assessment model, did not allow simulating an in-year adjustment of the TAC (update assessment). Yield-per-recruit analyses were, however, carried out. Ignoring discards, they showed that, at any realistic F level, the overall yield is higher if the shrimp are al-lowed to become two year olds before being fished, and hence an increase in catches of the one year olds will decrease the total yield from each year class. Thus, no rationale for increasing the TAC based on the size of the incoming year class was found. However, to improve the accuracy of the assessment it was recommended that in years where the stock is projected to be near or below Bpa at the end of the TAC year, an in-year assessment should be performed once the survey data have been collected.
The group was not able to investigate the effect of different discarding levels on the sensitivity of the results. Increasing discards had almost no effect on predicted yields. Simulating reduced discards compared to historical patterns did not work with the software used in our study. Most likely, discarding was not modelled correctly. For each year, the model applies a specified (average) fraction of the total catch in numbers as discards within a certain size range. This is an extremely poor match for the real discard pattern. One would expect to see a higher number of discards of a large incoming year class than of a small one. The model seems to do the opposite. By imposing a fixed number of discards based on the overall catch you would get a high fraction of a small year class discarded and a low fraction of a large year class discarded.