User menu
This working paper briefly describes key developments on the IOTC yellowfin operating model since the 2018 WPTT and WPM. Key points include:
• The yellowfin OM was updated in relation to the 2018 stock assessment. This was not a specific request from the WPTT/WPM, but was undertaken in recognition that a number of potentially important assumptions had changed (including data revisions), and that this might prove useful in the context of the 2019 yellowfin assessment review process.
• A fractional factorial design (OMgridY19.1) was used to set up a 144 model grid that can, in principle, quantify the main effects of 11 factors (two 3-level and nine 2-level factors). The full factorial grid would have required 4608 models. We attempted to repeat the convergence 3 times per configuration, however this could not be achieved for 13 cases (>20 failures each). Some (possibly all) convergence failures appeared to be associated with biomass too low to allow the observed catches to be removed from the population. Many of the results were most sensitive to the CL weighting assumptions, often resulting in bimodal distributions.
• A second fractional factorial grid (OMgridY19.2) was run with an intermediate CL weighting assumption (the assessment post-fit ESS raised to the power of 0.75, and capped at 100). 121 of 144 models converged (twice), and forms the basis of the results presented here.
• The alternative growth curve of Dortel et al. (2015) was added as a new reference case uncertainty dimension, as parameterized in the Fu et al (2018) assessment. This was not considered as a robustness test, because we could not identify any strong reason to expect it to be less reliable than the main assessment assumption (one possible criticism arises from the process used for filtering the tag data, which was not clear, and could bias the estimates)
• Tag mixing time assumptions were changed from 3 quarters to 4 and 8 quarters. Note that these mixing rates are only relevant to 25% of scenarios each, as the other 50% of scenarios have a negligible tag weighting.
• An additional regional scaling factor was admitted to the reference case OM (the most extreme with respect to the biomass ratio in the NW region to the other regions, from among the two alternatives considered as worthy of consideration in the recent assessment).
• Dome-shaped (double normal) selectivity was admitted as an alternate assumption to the usual logistical longline selectivity in gridB19.2, however it was omitted from projections pending further investigation.
• The new set of YFT conditioning results is better behaved than previous iterations, with key inferences generally showing a unimodal distribution with few outliers, and a central tendency slightly more pessimistic than the recent assessment. Post hoc model sampling on the basis of stock status characteristics (as applied in the two previous OM development iterations) appears unnecessary this time. We do not know the extent to which this new stability is attributable to: i) the revised data and assumptions in the 2018 assessment (possibly including the phasing issue in which it appears that the minimizer can be trapped fitting to seasonal CPUE patterns rather than annual trends), ii) the requirement for repeated convergence and/or iii) confounding of interactions in the fractional factorial design.
• MP evaluations are presented as a subset of gridOMref19.2, in which only the logistic selectivity was retained. While this modified grid is balanced (i.e. all factor levels
represented an equal number of times), we are not certain if there are unintended consequences in the fractional design.
• Example MP results are presented for the three tuning objectives defined in the 2018 TCMP. The results suggest that the rebuilding objectives (rebuild to BMSY by 2024, 2029 and 2034) are attainable, but only if TAC change constraints are relaxed from 15%, and catches in the short-medium term are considerable lower than recent catches.
• A subset of the requested robustness tests was completed, showing qualitatively unsurprising results (e.g. adverse model assumptions increase the risk of not meeting management objectives).
Summary points are presented for discussion and/or endorsement from the IOTC MSE task force with respect to requirements for i) the 2019 TCMP (primarily reference case requirements) and the 2019 WPTT/WPM (further development of robustness scenarios and candidate MPs). Decisions for the next iteration of the YFT MSE are documented in the 2019 MSE Task Force report.
