Long-Term Variability of Phytoplankton Size Classes in the Littoral Seas of Korea Using Deep Neural Networks and Satellite Data

Understanding the dynamics of phytoplankton size classes (PSCs), highly sensitive to environmental conditions in marine ecosystems, is crucial for comprehending variations in primary production and biogeochemical processes. Over the past decades, the littoral seas of Korea have undergone significant environmental shifts, yet long-term studies on PSC distribution remain limited. Employing a regionally developed deep neural network model and 20 years (2003–2022) of satellite ocean color data, we assessed spatiotemporal variability in dominant PSCs in the Yellow Sea (YS), South Sea of Korea (SS), and East/Japan Sea (EJS). Micro-size phytoplankton dominated turbid nearshore waters of the YS and western SS year-round, while nano-size phytoplankton were seasonally prevalent in the central YS and EJS. Pico-size phytoplankton exhibited strong summer dominance under warm, stratified, nutrient-depleted conditions, showing a sustained long-term expansion across all regions, particularly in the southwestern EJS. This expansion was closely linked to rising sea surface temperatures and changes in nutrient stoichiometry. The increasing dominance of smaller phytoplankton may reduce primary production, alter food web structure, and ultimately diminish fishery productivity. These findings provide new insight into climate-driven ecological shifts in marginal seas and underscore the need for integrated long-term monitoring to anticipate future ecosystem responses in a rapidly warming ocean.