Lifetime Citizen Portal Access — AI Briefings, Alerts & Unlimited Follows
Supplemental review of widow rockfish assessment constrained by federal staff absence; panel flags sensitivity to data-weighting and age inputs
Loading...
Summary
The Pacific Fishery Management Councils groundfish subcommittee met for a supplemental review of the widow rockfish assessment and found that changes to data-weighting, fleet structure and age-composition inputs in the proposed base model produce materially different estimates for natural mortality and catch recommendations. The panel warned that the absence of federal stock-assessment staff constrained the depth of review.
The Pacific Fishery Management Councils groundfish subcommittee convened a supplemental review of the widow rockfish stock assessment to evaluate a proposed updated base model and the responses to 14 review requests. Chair Cheryl Barnes opened the session by noting that the meeting was recorded and that the "PAM" website contains the assessment reports, model files and previous assessments for review.
The panels charge was to determine whether the update represents the best scientific information available and is appropriate for use in conservation and management, and to identify any additional research needs. Barnes told the group that the proposed model incorporates several notable changes since the August draft: a simplified fleet structure, revised historical catch reconstructions for Washington, a change in composition data weighting (McAllister nd—[Ianelli] -> Francis weighting with lambda adjustments), a length-based fecundity relationship, and a revised recruitment-bias adjustment start year.
Why it matters: those technical choices matter because they change key outputs managers use to set catch limits. The stock assessment teams sensitivity runs show that some combinations of those choices produce higher estimates of natural mortality (M) and materially different projections. Barnes summarized the STAT results: yield increased about 3% relative to the August version in the proposed base model, and the 2027 ACL rose roughly 13%, largely driven by revised estimates of recent recruitment and by changes in the data-weighting and age-composition inputs.
Panel discussion focused on two interrelated technical issues. First, several reviewers noted that switching weight methods and tuning the lambda parameter had one of the largest single effects on model outcomes. As one reviewer put it, the Francis weighting plus lambda set to 1 is the current standard and tends to reduce the influence of composition data relative to abundance indices; applying the older McAllister-Ianelli approach or ad hoc lambda values can move results significantly. Second, the panel repeatedly returned to conflicts between age compositions from the West Coast bottom-trawl survey and those from the midwater trawl fishery. Several reviewers said removing or down-weighting the bottom-trawl survey age data produced a higher estimate of M and a much larger calculated OFL in sensitivity tests, demonstrating that a relatively small sample of older ages in some years can have outsized effects on projections.
Panel members and industry advisers emphasized geographic and fishery-sampling context: fishing effort has been concentrated in parts of the coast (notably central Oregon and the Columbia River region), and some fleets (especially the midwater trawl) tend to target narrow areas and times. Advisers warned that such targeting can make fishery age and length samples non-representative of the coastwide population and therefore alter how the model interprets recruitment and mortality signals.
The panel repeatedly flagged process limits: federal STAT members who produced the assessment were absent because of a government shutdown, and reviewers stressed that proceeding without those scientists limits the depth of the exchange and the ability to run new model experiments on demand. Several reviewers recommended explicitly documenting any analysis that could not be completed because lead STAT staff were unavailable.
The panel did not take any formal votes. Members agreed to continue deliberations Thursday with attention to (1) clearly documenting the limits imposed by the STAT absence, (2) sharpening the reports explanation of which modeling changes drive the largest differences in outcomes, and (3) identifying a small set of high-value additional runs (if STAT becomes available) to test whether large changes in OFL and ABC/ACL estimates are robust to plausible alternative assumptions.
The meeting adjourned with public comment from industry representatives who thanked the panel for engaging with their concerns about spatial sampling and the interpretation of age-structure signals.
