User menu
This working paper describes developments on the Indian Ocean Tuna Commission (IOTC) yellowfin (YFT) Management Procedure Evaluation work, since the 2020 Working Party on Tropical Tunas (WPTT) and Working Party on Methods (WPM). Despite a number of recent investigations, we have not identified a satisfactory Operating Model (OM) upon which we would be confident in providing Management Procedure (MP) advice to the 2021 TCMP. Any attempt to present YFT MP evaluation results at this time might create a misleading perspective that could be counter-productive to the longer term MP adoption process and efforts to provide stock assessment advice. The status of the IOTC YFT stock assessment and management advice provided by the WPTT are in a state of flux, subject to a substantial collaborative review and development process, and with a new assessment scheduled for 2021. Given the close association between the assessment and the OM conditioning, we expect that there should be new perspectives arising from the 2021 assessment, with which to guide the YFT OM revision. The insights provided here are hopefully useful to the assessment as well. Our key concerns about the OM at this time include:
• The MP evaluations conducted here suggest that very large and immediate catch reductions would be required to hit the BMSY rebuilding objective by 2034. However, there are good reasons to believe that the OMs requested by WPTT (2020) are too pessimistic.
• The widening gap between the data used to condition the OMs (2017 end year) and data collected subsequently, provides a very informative hindcast prediction diagnostic, similar to those being encouraged in recent IOTC assessments. When projecting forward from 2017, a large portion of the proposed YFT OM models struggled to remove the catch that was actually reported in 2019. It seems inescapable that any model is too pessimistic (or the space-time distribution of fish and fisheries is not sufficiently realistic), if these observed catches cannot be attained. Among those models that could attain the 2019 catch, the majority failed to remove the remaining bridging catches (assumed equal to 2019) for the 2020-2021 period, i.e. prior to the first MP quota setting.
• In previous development iterations, we have recognized that sometimes, not all fleets are able to remove the TAC recommended by the MP. It was expected that this might cause a disconnect between TACs and catches and how change constraints are applied, that could merit further consideration in the MP definitions. But this was expected to be a minor problem, and it was implicitly assumed that the estimated OM model parameters would still provide an adequate description of the fishery dynamics whether or not this was occurring. However, recent investigation suggests that if the stock is now, or soon to be, near the pessimistic state that the current OM suggests, then how the failure to remove recent catches is modelled could potentially have non-trivial consequences for MP evaluation results and MP selection. i.e. the space-time structure introduces refugia that may prevent some fleets from extracting their allocated quota, and it is not clear that the models adequately represent these refugia and how the fisheries would respond under these conditions.
• The OM has essentially been unrestrained with respect to how effort is allowed to increase to extract the TACs (and bridging catches before the MP is implemented). Due to the way that the Baranov equations are implemented in the C++ projection sub-routine (but not the modified Pope’s approximation originally implemented in R), there is an “effort ceiling” parameter that is set relative to the recent Fs estimated during conditioning. When this value is set to some reasonable number (e.g. if we speculate that the effective fishing mortality could at most double in the period from 2017-2019), the number of minimally plausible realizations (i.e. those that can remove the 2019 catch) is substantially reduced. The role of effort changes may require further consideration in all of the MP evaluations (or at least added as a standard set of robustness tests), to more realistically speculate about how fisheries will respond when quotas cannot be easily reached (i.e. would effort increase, would fleets move or shift targeting?)
• When the MP evaluation results are subset after running, retaining only those realizations that were actually able to extract >95% of the observed 2019 catch, it is notable that:
o The assumed CPUE CV of 30% in the conditioning was over-represented (97% of realizations vs 3% for the CPUE CV of 10%). The CPUE catchability trend of 0% per year was over-represented relative to 1% per year. This suggests the CPUE series are not very compatible with the recent observed catches in the context of the current model structure.
o The high M option was retained at a much higher rate than the low M option (despite the fact that the low M option seemed to be more compatible with the tagging studies in the Indian Ocean, and emerging inferences from direct ageing studies in the Atlantic).
o The down-weighted tag λ option was over-represented relative to the full weighting.
It seems clear that a better mechanism for combining OM characteristics and weighting on the basis of some sort of plausibility diagnostics may be required. However, it is not clear that the OM is in the right structural space that would enable this to be a sufficient solution. The 2 area OMs demonstrated most of the same problems as the 4 area OMs, though to a lesser extent.
A number of additional investigations are described in the main text, but it is not clear how useful the specific inferences are, given the big picture problems with the OM. A full suite of MP evaluation results has not been provided at this time, but it would still be feasible to produce these results before the TCMP 2021, if the MSE Task Force considered this to be useful. We recommend that the OM developers should continue to engage with the YFT assessment team to update everything in 2021, consider broader interpretations of the baseline assumptions to try to resolve the retrospective problem.
