Update on IOTC Bigeye Tuna MSE Operating Model Development October 2019

IOTC bigeye (BET) Management Strategy Evaluation (MSE) development requests since the 2018 WPTT and WPM were mostly addressed for the IOTC MSE Task Force meeting in Mar 2019 and are documented in a separate information paper (Kolody and Jumppanen 2019a). This paper highlights key changes in the BET reference set OM requested by the IOTC 2019 MSE Task Force meeting and outlines issues to be addressed to progress the bigeye OMs to the next iteration. Issues related to selecting OM ensembles that are relevant to both bigeye and yellowfin are documented in the yellowfin companion paper (Kolody and Jumppanen 2019g). A stand-alone document (attachment 1) summarizes the current state of the bigeye reference set OM as used for MP evaluation in Kolody and Jumppanen (2019c).Key points include:
• The reference set OM is stochastically sampled from 94 Stock Synthesis model specifications, retained from a fractional factorial grid of 144 models with uncertainty in 8 dimensions. Models were rejected based on repeated (usually 5-10) failed attempts to reach numerical convergence from jittered initial parameter values, or a non-trivial Stock Synthesis catch penalty term. We interpret the catch penalty term to mean that the model would require implausible levels of fishing effort to remove some component of the observed catch (in at least one quarter-age-region strata). This could result from unrealistically pessimistic overall dynamics or a problematic space/time distribution of fish (most likely the latter for bigeye).
• The reference set OM is similar to previous iterations and is generally optimistic about current stock status and future stock status at current catch levels. This is consistent with the most recent assessment (Langley 2016), upon which the OM is based. The bigeye reference set OM will need to be compared with the 2019 stock assessment, to ensure that the inferences are still compatible.
• Most of the contrasting OM assumptions appear to have a non-negligible effect, usually in a predictable manner (e.g. lower steepness, M, and increasing CPUE catchability trend tend to be associated with more pessimistic stock status). Tags tend to be associated with more pessimistic outcomes. The least influential assumption appears to be the longline selectivity assumption (logistic or “double-normal”), presumably because the double-normal option tends to estimate only a weak domed shape.
• We have employed and presented a series of model diagnostics that should be able to identify many model interactions and outlier behaviour in a large model grid. However, we recognize that the approach is somewhat qualitative, and there might be other features of interest that are over-looked in these aggregate summary statistics. We continue to welcome suggestions for improving the process of evaluating and selecting models (retention/rejection or non-binary weighting).
• Resources have been identified to support the ongoing technical and scientific requirements for bigeye MSE until at least Dec 2020.