Stock assessment of yellowfin tuna in the Western and Central Pacific Ocean: 2023. A. Magnusson, J. Day, T. Teears, J. Hampton, N. Davies, C. Castillo Jordán, T. Peatman, R. Scott, J. Scutt Phillips, S. McKechnie, F. Scott, N. Yao, R. Natadra, G. Pilling, P. Williams, P. Hamer

Executive Summary

This paper describes the 2023 stock assessment of yellowfin tuna (Thunnus albacares) in the western and  central  Pacific  Ocean.  An  additional  three  years  of  data  were  available  since  the  previous assessment in 2020,  and the model extends through to the end of 2021.  The assessment moved to  a  new  5  region  spatial  structure  with  improved  convergence  properties  and  which  achieved  a positive definite Hessian solution, which was a requirement for future assessments from SC18.  This change  was  made  during  the  stepwise  model  development  process.  The  5  region  model,  given its  superior  convergence  properties  to  the  original  9  region  model,  is  used  as  the  basis  for  the structural uncertainty grid and stock status conclusions. Additional new developments to the stock assessment, many of which have emanated from the independent peer review of the 2020 yellowfin assessment, include:

• Conversion from a catch-errors to a catch-conditioned approach, and the inclusion of a likelihood component for the CPUE from the index fisheries.
• Change from using VAST to sdmTMB to standardise the input CPUE series and the inclusion of additional covariates in the CPUE model.
• Different CPUE variances used for the CPUE associated with each index fishery, applying anew approach to estimate these variances.
• Internal  estimation  of  natural  mortality  and  application  of  the  Lorenzen  form  of  natural mortality at age.
• Additional procedures implemented for achieving more reliable model convergence, including jittering and checking positive definite Hessian status for all grid models.
• Integration  of  parameter  estimation  uncertainty  with  model-based  uncertainty  across  the model grid for the key management reference points.
• Additional size composition filtering.
• Modifications to selectivity estimation settings, changes to fisheries with non-decreasing selectivity.
• Adoption of revised tagger effect modelling framework, reverting to assumptions similar to those used in 2017.
• Changes  to  size  data  weighting  and  downweighting  the  conditional  age-at-length  data  for internal growth estimation.