NOAA-SPC Assessment Model Workshop on SWPO Striped Marlin

Summary of Outcomes and Recommendations

Nicholas Ducharme-Barth, Claudio Castillo-Jordán, Michelle Sculley, Robert Ahrens, and Felipe Carvalho

January 20-24, 2025

Workshop context

  • 2024 SWPO striped marlin stock assessment presented at SC20
  • SC20 identified several technical issues requiring attention:
    • Poor fits to both size composition and abundance index data
    • Conflicts between different data types
    • Inconsistencies in fits to length/weight from same fisheries
    • Concerns about selectivity (and interactions with the new growth curve)
    • Inability to estimate model initial conditions
  • United States offered to host joint modeling meeting with SPC
  • Workshop held January 20-24, 2025 in Honolulu, Hawaii

Workshop participants

SPC:

  • Claudio Castillo-Jordán

NOAA PIFSC:

  • Rob Ahrens
  • Felipe Carvalho
  • Nicholas Ducharme-Barth
  • Michelle Sculley

Pre-workshop status

  • 2024 assessment used MULTIFAN-CL (MFCL) framework
  • Cubic splines defined selectivity as function of age
    • Limited by 10 annual age classes
    • Constraints on selectivity curve shape
  • SPC developed quarterly age model (mfcl-1979-20p3)
    • Improved selectivity resolution
    • Created mismatch with striped marlin reproductive biology

Key challenge

Quarterly vs Annual Population Dynamics

  • Striped marlin have restricted spawning season (late austral spring/early summer)
  • MFCL quarterly model introduced year-round recruitment
  • Need for modeling approach that could:
    • Provide finer resolution for selectivity
    • Maintain annual population dynamics aligned with biology

Workshop approach

  • Shift from MULTIFAN-CL to Stock Synthesis
  • Investigate key issues
    • Poor fits to size and index data
    • Data conflicts
    • Model initial conditions
    • Growth curve

Advantages of Stock Synthesis:

  • Can model selectivity as function of length
  • Maintains annual population dynamics

Model development process

Stepwise development:

  1. Base Stock Synthesis model (1979 start): 01-mls-base-1979
  1. Refined selectivity & data treatment: 03-chg-selex-1979
  1. Reverted to 1952 start year: 04-start-1952
  1. Excluded problematic size data: 06-exclude-more-comp
  1. Added catch uncertainty: 07-catch-uncertainty
  1. Refined growth modeling: 12-CAAL-old-growth-SD

Key changes

Initial conditions:

  • Returned to 1952 start year with unfished state
  • Better addressed population initialization concerns
  • Consistency with diagnostic-2019

Key changes

Selectivity:

  • Length-based selectivity provided greater resolution
  • Double-normal functional form with estimated parameters
  • Further refinements needed (beyond range of observations)

Key changes

Size data exclusion:

  • Excluded problematic size composition data:
    • Mixed-fleet data with temporal patchiness
    • Fleets with small sample sizes
    • Data showing internal conflicts
    • Data with quality concerns
  • Generally improved fit to remaining data components

Key changes

Growth:

  • Internal estimation using conditional age-at-length data
  • Addressed some limitations from externally derived growth curve
  • Improved overall model fit
  • Growth slower but still struggles fitting smallest observations

Key changes

Catch uncertainty:

  • Implemented catch uncertainty for early high-catch periods
  • Bifurcated Japanese LL fleet in sub-region 2
  • Reduced sensitivity to potentially unreliable historical catch

Additional investigation

  • Investigated source of misfit to the index seen in 2000 - 2002
  • Ran additional models to remove the influence of:
    • the index highpoint in 1998 (16-CAAL-rm-spike)
    • Fleet 6 (Australian LL; sub-region 2) weight frequency data (17-CAAL-noAU-ASPM)
  • Conflict is driven by the Fleet 6 weight data which suggests an increasing trend

Remaining challenges

  • Spatiotemporal heterogeneity in fleet coverage
    • No single fleet samples all areas consistently
    • Different spatial coverage during different time periods
  • Potential non-representativeness in mixed-fleet size data
  • Limitations in age data from opportunistic sampling
    • Longline will select for fastest growing younger individuals
  • Data weighting remains challenging

Assessment recommendations

  • Continue development in Stock Synthesis framework
  • Begin model from 1952 assuming unfished initial conditions
  • Develop multiple alternative abundance indices
  • Exclude problematic size composition data
  • Consider ensemble modeling approaches
  • Explore two-sex model development

Prioritized tasks

  1. Develop alternative relative abundance indices
    • JP/TW/AU (all region data)
    • JP only (sub-region 2 data)
    • JP/AU (sub-regions 2 & 3)
    • JP only (all region data)
    • AU (CSIRO index)
  2. Finalize updated size composition data

Summary

  • Successfully transitioned assessment model to Stock Synthesis
  • Addressed some key technical concerns raised by SC20
  • Improved fit to data while maintaining biological realism
  • Identified clear path forward for 2025 assessment

Thank you

Questions or Comments?


Repository website: https://n-ducharmebarth-noaa.github.io/2025-swpo-mls-meeting/