꽁치

This describes the published Pacific Saury framework through WK19, with the research status reviewed October 14, 2026. The completion-ratio and FRA survey relationships form the prior arm; weekly integration forms the likelihood arm. Time-to-fill and SPAN describe time, rather than adding tonnes to the season estimate.

The base case assumes fishing ends October 31 as the fleet begins returning by the end of October. The model’s own ending dates remain a separate sensitivity. The registered smoothed-handover sensitivity is also separate: its material flag opens a review, never automatic adoption or replacement of the primary estimate.

The lower endpoint is a P20 model quantile. The upper endpoint is fleet capacity under the stated return assumption, with no quantile probability. The indicative band is not a calibrated two-sided 80% forecast interval. Each issue discloses the endpoint basis, one-sided misses, width, changing width and collapse condition. Historical allocation-ceiling results do not establish coverage for a capacity-ceiling issue.

WK19 reviews October 4–10 and gives references for October 11–17, the week already underway. A whole-week reference does not become a fresh forecast for only the remaining days at publication. The latest-catch and season-split references keep their own errors and matched baselines; the one-week publication bar remains unearned.

CONF-W reference ranges first become eligible in WK19; WK18 is excluded. The nominal 80% target uses retained historical log catch-rate errors scaled to tonnes. Actual weekly-tonne coverage is unestablished, and original live outcomes accrue prospectively. These ranges leave the points and existing season-split band unchanged; historical reconstructions never become original calls.

NOAA and ECMWF forecasts remain quoted as issued with cycle, valid time and native grid point. Observed wave exposure is weighted across hulls at the unchanged 4.0 m working line; missing days remain missing. Water maps, vessel positions, working assumptions and successful fishing outcomes are different evidence.

The latest work fits and checks detection by comparable Japanese research nets. Detection by a specified net, biological occupancy, Taiwan commercial encounter and catch in tonnes are separate targets. Positive fish samples alone supply no commercial failed-attempt or effort denominator.

The latest source-consistent comparison adds dated survey operations, distinguishes research gear, and compares thermal models with a thermal-plus-current arm. The added-current arm clears its registered retrospective research comparison; cruise-level calibration and transfer remain unresolved. This result cannot be compared directly with the earlier uncertainty-adjusted net study because the deciding populations differ.

The October 14 source review qualifies original surface and near-100 m temperature inputs for October 14–17. Temperature uses the October 13 cycle; currents retain the October 10 cycle with their age disclosed. Applying saved models at dated October 8 fleet positions reproduces independently without a new fit. This remaining-day support diagnostic predicts neither future vessels nor validated Taiwan fish encounters.

Publication still requires qualified dated fish observations, sampled non-detections and effort, supported future forcing and independent calibration for the stated event. Taiwan attempted-fishing outcomes remain absent and support for comparable individual detectors is sparse. Numerical reproduction and source identity do not establish forecast accuracy. No new Taiwan fish-presence percentage or fish-adjusted catch is adopted.

Provider-current paths at earlier fleet positions describe conditional water movement. They do not predict future vessels or establish fish migration. Fish-initialized scenarios that leave the retained grid or meet missing velocity nodes remain unsupported; missing support is not fish absence.

Earlier paired catch experiments did not show an improvement from the tested current adjustments. Those adjustments remain research candidates. The latest route work needs a continuous supported corridor, qualified missing-node and wet-mask treatment, and habitat and depth-aware movement before its deciding comparison.

A commercial catch bridge must validate the same encounter, effort and catch population before any weekly-tonne change. Season integration must carry shared weekly uncertainty through complete cumulative paths. Research construction, a successful download or green software checks do not establish predictive skill.