Data weighting for tagging data in integrated size-structured models

• Published in: Fisheries research Vol. 192; pp. 94 - 102
• Main Authors: Punt, André E., Deng, Roy A., Siddeek, M.S.M., Buckworth, Rik. C., Vanek, Vicki
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.08.2017
• Subjects: Data weighting; Size-structured stock assessment methods; Tagging data; Tagging data; Size-structured stock assessment methods; Data weighting
• ISSN: 0165-7836; 1872-6763
• DOI: 10.1016/j.fishres.2015.12.010

•Assessment results can be sensitive to the assumptions for data weighting.•Tagging data may be over dispersed relative to the commonly assumed Bernoulli likelihood.•‘Francis weighting’ can be extended to apply to tagging data. Increasingly, stock assessments for hard-to-age species such as crabs, prawns, rock lobsters, and abalone are being based on integrated size-structured population dynamics models that are fit to a variety of data sources. These data sources include tagging data to inform growth. Diagnostic statistics and plots have been developed to explore how well integrated population models fit the data types typically used for assessment purposes (index data, size- and age-compositions, and conditional age-at-length data). However, such statistics and plots are not available for tagging data, when these data are used to estimate growth. This paper outlines two diagnostic statistics that can be used to evaluate fits to tagging data, and develops a method based on ‘Francis weighting’ for weighting tagging data in integrated models. For illustration, the methods are applied to Aleutian Islands golden king crab (Lithodes aequispinus) in Alaska, and tiger prawns (Penaeus semisulcatus and P. esculentus) in Australia's Northern Prawn Fishery. Some degree of growth model mis-specification was revealed for P. semisulcatus, and there were conflicts in the data for the tiger prawns. The standard errors for the estimates of mature male biomass for golden king crab were larger when the tagging data were downweighted based on the proposed weighting method. This serves to emphasise that assessments and their interpretations can be impacted by how tagging data are weighted.