Estimation of subsurface salinity and analysis of Changjiang diluted water volume in the East China Sea

In the East China Sea (ECS), the sea surface salinity (SSS) changes as the Changjiang Diluted Water (CDW) propagates toward the Korean Peninsula via the ocean current and winds every summer annually. Although the vertical stratifications resulting from the CDW volume changes are important, it has not been analyzed yet. Therefore, in this study, we aimed to estimate the salinity at a depth of 10 m (S10m) using convolutional neural network (CNN) model based on multi-satellite measurements and analyze CDW volume variations. The main CDW mass in the ECS reaches approximately 10 m in depth; thus, the CNN model was developed using sea surface physical factors as input and in situ S10m obtained from the National Institute of Fisheries Science (NIFS) as ground truth data from 2015 to 2021. The CNN tests result showed a determination coefficient (R2) of 0.81, root mean square error (RMSE) of 0.63 psu, and relative RMSE (RRMSE) of 2.00%. Unlike the sea surface distribution, the spatial distribution of S10m showed that the CDW was predominantly present in the center of the ECS. From SHapley Additive exPlanations (SHAP) analysis, SSS exhibited a strong positive relationship with S10m, and the sea level anomaly showed a strong negative relationship. After calculating the volume of the CDW from the surface to a depth of 10 m, the maximum (3.01x1012 m3) and minimum volumes (1.31x1012 m3) were represented in 2016 and 2018, respectively. Finally, the warming effect induced by the CDW volume changes was analyzed in two different years: 2016 and 2018. Specifically, in 2016, the sea surface temperature increased by more than 4.79 degrees C in the Ieodo location, while in 2018, it increased by 2.19 degrees C. Thus, our findings can obtain information about the volume variation of the CDW and its effect on the ECS in summer.