Estimating spatio-temporal distribution of fish and gear selectivity functions from pooled scientific survey and commercial fishing data

• Published in: Fisheries research 2021-11, Vol.243, p.106054, Article 106054
• Main Authors: Martin Gonzalez, Guillermo, Wiff, Rodrigo, Marshall, C. Tara, Cornulier, Thomas
• Format: Article
• Language: English
• Subjects: Cod; GAM; Selectivity; Spatio-temporal; Surveys
• ISSN: 0165-7836; 1872-6763
• DOI: 10.1016/j.fishres.2021.106054

Model-based prediction of fish distribution at fine resolutions in space and time has the potential to inform area-based and dynamic forms of management, such as permanent marine protected areas or real-time temporary closures. A major limitation to the spatial and temporal mapping resolution that is achievable is the amount of high quality, standardised data that can be utilized for fitting statistical models. To achieve an adequate spatio-temporal resolution from sparse data, one option is pooling information from several sources, such as scientific surveys and fisheries data. Because surveys and fisheries data usually use different sampling methods, pooling information from different sources requires cross-calibration of catch rates values across multiple gears. However, the individual gear efficiency and selectivity curves (the ratio between catch and availability at a given length) for all fishing gears and species are typically unknown. Using cod (Gadus morhua) in the northern North Sea as a case study, we developed a new formulation of spatio-temporal generalised additive models (GAM) of relative abundance of fish, combining catch data from multiple sources. Differences in gear efficiency and selectivity were internally calibrated within the model by the estimation of the local spatio-temporal variation in abundance. We show that pooling data sources enables the prediction of multi-annual and seasonal spatial variation in cod relative abundance-at-size, at spatio-temporal resolutions that are relevant for informing fishing strategies, e.g., reducing bycatch in real-time, or management objectives, e.g., real-time closed areas. We also show that GAM models fit to catch and effort data can reveal the relative efficiency and selectivity of different survey and commercial gears. The selectivity curve estimates that emerged as a by-product of our analysis are consistent with expert knowledge of the performance of the gears employed for cod. Our analytical approach can therefore serve two useful purposes: to estimate spatio-temporal variation in relative abundance of fish and to estimate relative gear efficiency and selectivity.