CMP results and development history

SPRFMO

South Pacific Regional Fisheries Management Organisation
Jack Mackerel Candidate Management Procedures
Current jmMSE Evidence and SCW17 Development History

This page summarizes the available results; it does not record an adoption decision. The current candidate interpretation comes from jmMSE, while the SCW17 papers describe earlier development. The published candidate results use the completed common 500-draw reference and nine-OM robustness evaluation.

1 Current candidate results from jmMSE

The current result set contains eight CMPs from two empirical rule families: HS+20 (MP29), HS-20 (MP43), HSsym (MP45), HS-30 (MP47), PR+20 (MP32), PR-20 (MP44), PRsym (MP46), and PR-30 (MP48). MP43 and MP44 are the focal cases; MP29 and MP32 are sensitivity cases retaining the earlier +20 annual-limit specification. This designation describes their role in the current analysis, not a formal recommendation or adoption decision.

The eight-CMP comparison provides the following results:

  1. All CMPs use the same five observed abundance indices and transparent empirical HCR structures, while the variants isolate alternative annual TAC-change constraints.
  2. The completed analysis contains 500 posterior draws for the reference OM and nine robustness OMs, with spawning biomass, fishing pressure, catch, IACC, VB/VB[2025], and VB/VB[MSY] outputs.
  3. The OMs simulate index uncertainty, temporal autocorrelation, and cross-correlation rather than treating the five indices as independent confirmations of stock change.
  4. Reference and robustness results show trade-offs among stock condition, catch, stability, and responsiveness; no CMP dominates every measure and OM.
  5. The scorecard is a relative trade-off tool; its ordering depends on metric selection, direction, scaling, and weighting and is not an acceptability or focal-status decision.

The SC14 MSE report presents the current synthesis and the Slick explorer provides the saved comparison file. Both use the completed 500-draw outputs; archived 100-draw results remain provenance and must not be mixed into that performance comparison.

Preliminary banking-and-borrowing results have also been supplied for HS+20 and PR+20. The transferred archive requires a corrected, matched rerun because its saved transaction fields contain no applied banking or borrowing events. See the banking-and-borrowing section of the SC14 report for the reported results, eligibility reconstruction, and implementation caveat.

2 Paper-08 subset screen

SCW17/Paper-08 documented a repeatable subset test of an empirical index procedure using Chile_CPUE and Offshore_CPUE, cpues.ind with a mean combination, and buffer.hcr. A structured sample of 12 parameter combinations varied the limit and lower and upper buffer values.

The paper reported that the subset spanned a wide range of Kobe-green performance and that higher green probability could coincide with lower catch over the tuning window. It explicitly cautioned that this subset was not the intended full parameter screen. That qualification is retained here.

Evidence retained from this screen includes the lesson that performance must be interpreted across biological and fishery outcomes rather than from a single tuning measure. Evidence still needed before relying on any historical screen includes:

  • saved run results and the table of tested parameter combinations;
  • confirmation against the merged jmMSE implementation;
  • evaluation across the agreed reference and robustness OMs;
  • the complete agreed performance-metric set;
  • checks for model convergence, failed runs, and repeatability; and
  • a recorded working-group interpretation of trade-offs.

3 Paper-09 hockeystick test set

SCW17/Paper-09 converted workshop proposals into a structured set of index-based hockeystick tests. It established several useful implementation controls:

  • use exact jmMSE index names;
  • define all candidates in one table rather than changing each rule by hand;
  • stop when required indices are unavailable rather than substituting silently;
  • distinguish tests based on a mean index from tests using loess smoothing; and
  • retain unavailable tests as unavailable until their methods exist.

Those rows remain historical test proposals. They no longer define the primary CMP registry, which is based on the current jmMSE code and retains legacy MP identifiers alongside the current HS/PR working labels.

4 Current interpretation

SCW17 narrowed dozens of exploratory MPs to simple hockey-stick and power-ramp families and reasonable variants for continued evaluation. Subsequent work expanded robustness testing, corrected implementations, added vulnerable biomass metrics, documented the observation-error correlation structure, and developed reproducible trade-off and scorecard views.

The machine-readable development history links the SCW17 run18/run20/run22 subset to the later MP29/MP32 family refinement and the MP43–MP48 annual-limit variants. The workshop runs are an antecedent development set rather than one-to-one aliases for the current CMPs. The expanded eight-CMP comparison was added after the 24 July meeting and remains subject to Task Team and Scientific Committee review.

The current results support carrying a reduced subset forward for Scientific Committee and Commission consideration. It does not by itself select an adopted CMP. Final advice requires a frozen and reviewed metric set, sensitivity to any weighting scheme, corrected banking-and-borrowing tests, and a formal exceptional-circumstances protocol.

The external experts’ positive overall assessment is retained together with their caveats: candidate design would have benefited from additional stakeholder input, and earlier agreement on the assessment and reference points would have kept the technical discussion more focused. The authors’ record of engagement and their response to those caveats are presented separately in the SC14 synthesis.

5 Evidence promotion checklist

Check Required before shortlist Required before adoption
Exact CMP specification Yes Yes
Reproducible reference-OM runs Yes Yes
Agreed robustness set Planned or complete Complete
Complete 500-sample comparison Desirable Yes
Full performance metrics and definitions Yes Yes
Failure and warning audit Yes Yes
Frozen worked example Desirable Yes
Independent implementation review Desirable Yes
Exceptional-circumstances protocol Planned Yes
Formal decision record Yes Yes