Diversity and Subjectivity of Model Uncertainty Specification in Global Tuna Stock Assessment

ABSTRACT As a major source of scientific uncertainty in stock assessments and fisheries management, model uncertainty in tuna stock assessments is primarily examined using sensitivity analysis and ensemble modelling. However, uncertainty specifications (i.e., factors and levels included in uncertainty analyses) for tuna assessments vary within and among tuna Regional Fisheries Management Organisations (tRFMOs). We reviewed the uncertainty specification in 133 stock assessments conducted over the past 20 years by five tRFMOs, covering seven major commercial tuna species. Our analysis revealed four key findings: although a wide range of uncertainty specifications has been applied, only a few factors are commonly considered across assessments; patterns in uncertainty specification are more closely aligned with management organisation than with species; the emphasis on different factors has shifted over time (e.g., increasing attention to growth and decreasing focus on data weighting); and a change to the modelling platform is a key factor leading to changing uncertainty specifications between successive assessments. Our findings demonstrated the diverse range of uncertainty specifications in tuna stock assessments, and revealed the important role of subjective factors in making these selections. We propose some best practices for specifying uncertainty to enhance consistency and reducing subjectivity across tuna stock assessments.