Application of cluster analysis to characterize fishing-ground formation of the large purse seine fishery
This study classified environmental regimes at inferred fishing locations of the Korean large purse seine fishery by integrating automatic identification system-based fishing activity data from Global Fishing Watch with public marine environmental data. Temperature and salinity from GOFS3.1 and gridded NIFS serial oceanographic observations, together with MODIS-Aqua chlorophyll a, were matched to fishing locations from 2021 to 2023. Correlation, principal component, and variance inflation factor analyses identified temperature and salinity at 30 m depth and chlorophyll a as representative variables. Seasonal Gaussian mixture models identified five clusters for each environmental dataset, which were integrated into four meta-groups using Ward’s hierarchical clustering. The meta-groups represented low-temperature/high-chlorophyll a, stable, high-temperature/low-salinity, and high-temperature/high-salinity conditions. Spatial similarity between datasets varied among meta-groups, with Pearson correlations of 0.438-0.643 and Top 10% Jaccard indices of 0.303-0.524. Major distribution areas were broadly consistent around southeastern Korea, the Korea Strait, and Jeju Island. These results support the complementary use of public marine environmental data for classifying environmental conditions at fishing locations. Since species-specific catch data were unavailable, the meta-groups should be interpreted as environmental regimes rather than species-specific fishing grounds.