The seas surrounding the Korean Peninsula exhibit large spatiotemporal variability in the marine carbonate system due to the combined effects of monsoons, freshwater input, and ocean currents. In particular, the Ieodo Ocean Research Station, located on the continental shelf margin of the East China Sea where Changjiang River discharge and the Kuroshio Current interact, is a key site for investigating ocean acidification mechanisms. In this study, we estimated high-resolution carbonate system variables, including pCO₂ and total alkalinity (TA), in the East China Sea using a Physics-Informed Neural Operator (PINO) model and analyzed their spatiotemporal variability and acidification trends.
The model was designed with an FNO-ConvLSTM-based spatiotemporal encoder to learn spatial patterns and short-term variability in marine environmental variables, followed by a PINO framework that incorporates carbonate system equilibrium. The PINO model was trained using both data loss and equation loss. The equation loss included CO₂ solubility, carbonic acid dissociation constants K₁ and K₂, and a carbonate-equilibrium-based pCO₂ formulation to maintain chemical consistency between pCO₂ and TA. pH was not directly trained; instead, model-estimated pCO₂ and TA were used as inputs to PyCO2SYS to calculate pH, dissolved inorganic carbon (DIC), aragonite saturation state (Ωarag), and Revelle factor.
Input data included sea surface temperature, sea surface salinity, chlorophyll-a, 10 m wind fields, distance-to-coast data, atmospheric CO₂, and CDOM. Observed pCO₂ and TA from the National Institute of Fisheries Science and the Ieodo Ocean Research Station were used as target data. Based on the estimated carbonate system dataset, EEMD analysis showed that residual pH in the Ieodo region decreased at a rate of -0.0544 ± 0.0002 decade⁻¹, approximately twice as fast as the Pacific average, while surface seawater pCO₂ increased by +5.60 ± 2.66 μatm yr⁻¹.
The pCO₂ anomaly analysis showed that summer warming generally increased pCO₂ through a positive thermal effect. However, in the Ieodo region, a strong negative non-thermal effect offset the thermal effect, resulting in weak or negative pCO₂ anomalies. This was accompanied by decreased DIC and increased Ωarag and pH, suggesting that nutrient supply from the mixing of Changjiang diluted water and offshore waters promoted phytoplankton blooms and biological carbon removal. These results indicate that summer carbonate system variability in the Ieodo region is controlled by the combined effects of freshwater input, water mass mixing, and biological pumping.