Data-driven integrated assessment of global wild-caught seafood exported to Hong Kong by 2030 in different representative concentration and shared socioeconomic pathways

Abstract

Wild-caught seafood is an important commodity traded globally. As climate change and socioeconomic development is affecting global marine capture fisheries, the impact on regional supply remains unexplored, especially for areas like Hong Kong relying on global trading to meet high seafood consumption. However, it is challenging to assess the global marine capture fisheries production using complex process-based models. In this study, a data-driven integrated assessment approach was developed to evaluate the change of global seafood supply from wild catch. With the catch data available from 1990 to 2014, machine learning models were trained and tested including environmental, socioeconomic, geographic, and technological features to estimate the catch by ocean grid cells for individual species. Nine popular seafood categories in Hong Kong were studied, which include 68 species in total. Important input features for estimating the catch were compared across species and the impacts of these input features were interpreted using partial dependence plots. The global marine wild catch of the 68 species by countries and the export to Hong Kong were projected by 2030 in RCP2.6-SSP1, RCP4.5-SSP2, RCP7.0-SSP3, and RCP8.5-SSP5. Performances of machine learning models demonstrate the reliability of data-driven methods to estimate the catch by ocean grid cells. The importance of geographic features rank top for the estimate while that of climate change and socioeconomic development varies significantly across species. The projection reflects a drop of squid exported to Hong Kong due to the reduction of squid supply from China's mainland during 2015–2019. The export of wild-caught seafood of the nine categories to Hong Kong will have a slight decline by about 16% from the 2020 level by 2030. The projection also suggests no significant differences among the four climatic-socioeconomically interrelated scenarios regarding the export to Hong Kong before 2030. Top producers include China's mainland, United States, and Japan. However, China's mainland and Japan will suffer from the decline. The data-driven integrated assessment approach can be improved to provide more insights into the long-term change and sustainable management.