This is a living to-do list for the Chatham sablefish stock assessment, which is transitioning from a mark-recapture / YPR analysis to an integrated statistical catch-at-age model in 2020. It has been updated for the 2020 assessment season and should be updated before and after each assessment. Any new or remaining items should be reported in the Discussion section of the stock assessment.
Within each section, items are labeled and defined as follows:
HIGH PRIORITY: tasks that must be completed for the statistical catch-at-age model to be considered for management.
SHORT TERM: tasks that should be completed within 1-2 years of implementation. They are critical components of a well-developed statistical catch-at-age model.
LONG TERM: tasks that should be completed within 2–4 years of implementation. While they are not critical to the implementation of the model, they will improve model-based inference, understanding of stock dynamics, and data quality
Data
[x] Work with groundfish staff to develop rationale for the choice of fishery-dependent data sources to include in the model (these indices do not appear to follow population trends well) and whether fishery selectivity should be fixed or estimated. This relates to the challenge of accounting for unobserved fishery releases and estimating fishery selectivity and fitting to landed catch compositions. HIGH PRIORITY
I've opted to leave the fishery and survey CPUE within the model for now to maintain consistency with the Feds and make full use of the data we have. I plan on leaving selectivity fixed at the Federal values until we can revisit the issue of changing retention probability over time. I will run a simple sensitivity analysis on selectivity to evaluate the impact of this assumption.
[ ] Federal assessment uses survey weight-at-age to fit catch biomass. Due to differences between survey and fishery weight-at-age, fishery weight-at-age may be more appropriate. If we choose to use this instead, we need to provide justification in write up. HIGH PRIORITY
Because discarding is permitted in the state fishery, there are large differences in survey and fishery weight-at-age, especially at younger ages. Consequently, I opted to use fishery weight-at-age to fit to landed catch biomass, and survey weight-at-age to estimate exploitable biomass, spawning biomass, and other quantities of interest in the model. This has been documented as a difference between the state and federal models. #44
[ ] Work with groundfish biologists to obtain the raw mark-recapture data from 1997-2002. 2003 data were partially lost, 2004 used PIT tags. LONG TERM
[ ] Develop Chatham-specific ageing error matrices. This should include an examination of the choice of plus group (the model currently doesn't fit the plus group well) and potentially use a rectangular instead of square ageing error matrix that may better account for older aged fish in the model (contact P. Hulson for more information). LONG TERM
[ ] Develop Chatham-specific age-length keys. LONG TERM
[ ] Improve methods for accounting for fishery releases, including conducting research to better understand discarding behavior and how it has changed over time. LONG TERM
[ ] Assess alternative sources of data, especially historical biological and catch data (Carlile et al. 2002). LONG TERM
[ ] Develop methods to account for additional sources of mortality, including sport, subsistence and personal use harvest; estimated bycatch mortality in the halibut fishery; and estimated deadloss, which includes mortality from sand fleas, sharks, and whales. LONG TERM
[ ] Review indices of abundance. The fishery and survey CPUE have little contrast and may not be useful indices of abundance. This may include standardizing CPUE through generalized linear or addition modeling to account for variables to affect CPUE. It may also include developing algorithms to identify trip and set targets and allocating total trip landings to set effort. LONG TERM
Model
[x] Complete the development and estimation of management reference points, Bonus: estimate reference ABC recommendation retrospectively. HIGH PRIORITY
[x] Analyze the impact on biomass and max ABC estimates if we had a biannual or triennial instead of an annual mark-recapture experiment HIGH PRIORITY
[x] Improve weighting methods and tune model to composition data.
HIGH PRIORITY
I implemented the McAllister-Ianelli (1997) using the harmonic mean across years to tune age and length comp data. See tmb/tune_comps.r script for analysis.
[ ] Conduct retrospective analysis to determine model performance over time. SHORT TERM
[x] Implement Bayesian analysis to evaluate posterior densities of estimated and derived quantities of interest. SHORT TERM
I used the tmbstan package and developed figures to show 95% CI for abundance indices and derived quantities of interest, including spawning stock biomass and management reference points. This will likely be an on-going process to improve this method.
[ ] Evaluate estimation and assumptions about recruitment variability. Note that moving to estimation of recruitment variability may impact data weighting. SHORT TERM
[ ] Develop framework to conduct projections to evaluate stock status and assess risk to future spawning stock. SHORT TERM
[ ] Develop priors for catchability and other relevant parameters. LONG TERM
[ ] Evaluate alternative harvest policies and biological reference points. LONG TERM
Notes to improve assessment text and figures
[ ] Section in text for selectivity sensitivity analysis (fixed at Federal values).
[ ] Section in text for maturity sensitivity analysis (fixed at NSEI-estimated values from a length-based analysis that's converted to age using vonB curve).
[ ] Section in text for natural mortality sensitivity analysis (fixed at 0.1).
[ ] Are we plotting standardized or raw residuals?
This is a living to-do list for the Chatham sablefish stock assessment, which is transitioning from a mark-recapture / YPR analysis to an integrated statistical catch-at-age model in 2020. It has been updated for the 2020 assessment season and should be updated before and after each assessment. Any new or remaining items should be reported in the Discussion section of the stock assessment.
Within each section, items are labeled and defined as follows: HIGH PRIORITY: tasks that must be completed for the statistical catch-at-age model to be considered for management. SHORT TERM: tasks that should be completed within 1-2 years of implementation. They are critical components of a well-developed statistical catch-at-age model. LONG TERM: tasks that should be completed within 2–4 years of implementation. While they are not critical to the implementation of the model, they will improve model-based inference, understanding of stock dynamics, and data quality
Data
Model
Notes to improve assessment text and figures