link to AFSC home page

link to AFSC home page link to NMFS home page link to NOAA home page

Auke Bay Laboratory (ABL)

AFSC Quarterly
Research Reports
April-June 2006
Contents
Feature
ABL Reports
FMA Reports
HEPR Reports
NMML Reports
RACE Reports
REFM Reports
Items
Quarterly Index
Quarterly Home

Groundfish Assessment Program

Rockfish Modeling Workshop at Auke Bay Laboratory

A Rockfish Modeling Workshop was held on 23-25 May 2006 at the Auke Bay Laboratory (ABL) to advance assessment analyses of rockfish stocks in Alaska. The workshop objectives were to review the modeling history of rockfish, evaluate where improvements are needed, and identify key assumptions and sources of uncertainty in current rockfish assessment models. The potential for incorporating ecosystem components in rockfish stock assessments was also discussed along with approaches for communicating model results for the annual Stock Assessment and Fishery Evaluation (SAFE) reports. Participants included managers and scientists from the ABL, the Resource Ecology and Fisheries Management (REFM) Division, the National Marine Fiseries Service (NMFS) Alaska Regional Office, and the Alaska Department of Fish and Game (ADF&G).

Key life history features among rockfish species and stocks were compared. While age-at-maturity data for Pacific ocean perch (POP) appear to be adequate for assessment purposes, data on many other rockfish species were lacking. Natural mortality assumptions and estimates for Alaskan rockfish species were generally consistent with growth and longevity patterns. Estimates of area-swept survey “catchability” for bottom trawling varied considerably among species. Simulation analyses presented at the workshop revealed some evidence that patchy rockfish populations which tend to cluster by age or size may result in biased catchability estimates. The modeling impact of expanding within stratum area-swept estimates to cover untrawlable regions needs to be researched further. Presentations also showed that the impact of estimating recruitment variability terms is problematic. The workshop recommended that fixed values for dispersion terms for recruitment estimates be used.

Evaluations on the rockfish models note that they are fundamentally Bayesian models, in that prior distributions are assumed and for measures of uncertainty, posterior distributions are traditionally presented. The workshop concluded that the current SAFE report could be improved by better documentation on how prior distributions were developed. The group also suggested that the shape and distribution of priors be put into the SAFE documents. For posterior distribution analyses, the Markov Chain Monte Carlo (MCMC) integration is typically used. The group discussed using a basic set of chain diagnostics to check for convergence (i.e., that the posterior distribution is adequately represented). The workshop group recommended establishing a standardized approach for MCMC presentation and developing a common set of libraries, perhaps available via an Intranet site.

The group examined different SAFE reports and developed a standard list of tables and graphs. Regarding the “Ecosystem Considerations” section of the SAFE report, the workshop noted that developing models to evaluate environmental covariates may be most useful. For improving rockfish stock assessments, it may be more important to include environmental covariates affecting transport and recruitment rather than predation effects, because rockfish mortality does not appear to be dominated by predation.

Comparative reviews of POP population and fishery trends revealed a number of similarities between the Gulf of Alaska and Bering Sea/Aleutian Islands rockfish stocks. Overfishing during the early 1960s led to large declines, but since the late 1970s, POP abundances have steadily increased. Age-structured model results across other species of rockfish were also compared (e.g., northern rockfish, dusky rockfish, and rougheye rockfish) during this portion of the workshop.

A workshop summary and set of recommendations for future rockfish age-structured models was developed. In the short term, participants suggested: 1) adding tables in SAFE reports that clearly document management activities, 2) carefully evaluating different data sets for quality, 3) developing a system to evaluate model configurations where hypotheses about model assumptions can be easily performed, 4) evaluating the impact of different data sets on model results, 5) describing prior assumptions clearly and including associated posterior distributions, 6) standardizing computer code among rockfish stocks, particularly for generating standardized output and evaluations, 7) comparing results between areas and models to understand where assumptions and differences may exist, and 8) revising maturity estimates for the northern rockfish model prior to the next assessment cycle.

For the longer term, suggestions were to 1) collect more maturity samples for a number of species, 2) continue research into viability of larvae produced by older rockfish females, 3) explore the utility of environmental covariates in rockfish models, 4) consider time-varying parameters, 5) where data exist, investigate more spatially explicit model configurations, and 6) evaluate the use of alternative likelihood specifications (e.g., robust forms).

By Kalei Shotwell (ABL) and Jim Ianelli (REFM)
 

 

next >>>


            Home | FOIA | Privacy | USA.gov | Accessibility      doc logo