Modelling growth in tuna RFMO stock assessments: Current approaches and challenges

• Published in: Fisheries research Vol. 180; pp. 177 - 193
• Main Authors: Kolody, Dale S., Eveson, J.Paige, Hillary, Richard M.
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01-08-2016
• Subjects: age estimation; growth model; management strategy evaluation; Marine; RFMO; stock assessment; Thunnus; tuna; RFMO; growth model; tuna; age estimation; management strategy evaluation; stock assessment
• ISSN: 0165-7836; 1872-6763
• DOI: 10.1016/j.fishres.2015.06.016

•Growth modelling challenges and possible solutions are reviewed for key populations managed by tuna Regional Fisheries Management Organizations.•Depending on the population, growth curve estimates are potentially affected by age estimation problems and biased temporal, spatial, and sex sampling.•Estimation biases are likely to arise from overly-simplified statistical assumptions and size-based selectivity processes.•We recommend more effort to describe growth curve uncertainty, and management strategy evaluation to provide robust management and growth research prioritization.
We review the approaches used to model growth in recent tuna Regional Fisheries Management Organization (tRFMO) stock assessments, and the challenges encountered. The tRFMO fisheries span vast areas, with multinational fleets operating a diverse range of gear types, and are assessed with a range of modelling methods. Despite the high volume and/or value nature of many tuna and billfish fisheries, there remain substantial data input challenges, including biased size composition sampling, conflicting age estimates from hard parts (and inconclusive validation studies), and very high error rates in some large-scale tagging programmes. There is evidence for spatial and temporal variability in growth rates, but sampling is rarely adequate to quantify this variability, and it is not described in most tRFMO assessments. Sophisticated statistical methods have been developed to combine catch length frequency distributions, age-length data and tag growth increment observations into a single estimation framework (though the data are generally not sufficient to allow the variances to be objectively partitioned). Modelling individual growth variability with random effects has the potential to greatly reduce biases from tag growth increment analyses, but this is computationally prohibitive when the growth curve is estimated within the assessment model. In contrast, the effects of size-based selectivity may not be adequately described if growth is estimated outside of the assessment model. Different species are affected by these problems to different and largely unknown degrees. We discuss options for mitigating some of these problems, but doubt that entirely satisfactory solutions can be achieved in most cases. Accordingly, we recommend that i) greater emphasis should be placed on representing the plausible growth uncertainty in the assessments (i.e. using a model ensemble approach), and ii) management strategy evaluation should be used to develop harvest strategies that are robust to the growth (and other, potentially more urgent) uncertainties, and to prioritize research investment in the context of achieving management objectives.