2019 STAR Panel: Weight-at-age

[chantelwetzel-noaa]

Future assessments should consider the use of empirical weights-at-age. One approach would be to begin with the WCGBT data and use it to fit (outside model) a smoother outside the model (see Jim Ianelli for code that uses a random effects model to estimate weights using empirical data by cohort and year). This would avoids confounding weight at age with growth estimates (e.g., Lee et al. 2019, Whitten et al. 2013), should allow the model to run faster, and should enable the model to better accommodate variability in growth (in both age and over time). Another advantage is that somatic body masses are based on actual measurements instead of model estimates which predicts mean length and then converts length to mass via a fixed set of length-weight parameters.

This request is linked to issue #2.

[chantelwetzel-noaa]

I have started reviewing the data we have that could be used to inform a weight-at-age matrix. The data available to inform this are:

• AFSC Slope (N = 7226), WCGBT (N = 25,664), Triennial (N = 2,211), and PacFIN data from Oregon only (N = 8,206). The data across these sources are sparse before 1995 at the earliest. We may want to omit the data from PacFIN since they are limited until 2017 and are only collected in Oregon, not coastwide. There are two data file for the fishery and survey in the data-raw/weight_at_age folder (not on github).

The follow tasks will need to be accomplished to determine if developing a weight-at-age matrix is a viable option for sablefish:

• Identify the start of when data are sufficient to estimate sex-specific weight-at-age.
• Determine the maximum age to estimate age-specific weight-at-age by sex and where the extrapolation across older ages should begin. The 2023 assessment estimated that weight-at-age stabilizes around age 30 for females and age 25 for males.
• Modify existing code to estimate sex-specific weight-at-age for sablefish.
• Determine how to extrapolate estimates for ages where sufficient data may not be available for a specific year and sex.
• Evaluate the annual weight-at-age estimates for plausibility (e.g., are there clear patterns or do the annual estimates jump around).