Deep Learning-driven Forecasting of Daily Sea Surface Temperatures in the South China Sea

Sea Surface Temperature (SST) plays an important role in influencing oceanic and climatic changes. Accurate prediction of SST variations is crucial for marine ecological environments, and meteorology. Traditional SST prediction methods primarily rely on numerical models, which are often computationally expensive. In this study, we introduce a rapid intelligent forecasting model for the South China Sea (SCS) SST based on a deep learning model (3D U-Net), utilizing SST, Sea Surface Height Anomalies (SSHA), and Sea Surface Wind (SSW) as input variables. The results demonstrate that the 3D U-Net model exhibits higher accuracy across all forecasting periods than the Convolutional Long Short-Term Memory (ConvLSTM) model. The 3D U-Net model exhibits smaller prediction errors in different regions of the SCS and demonstrates robust performance during the monsoon season. Additionally, the model achieved an accuracy of over 80% in predicting marine heatwave (MHW) events in the SCS in 2021. In summary, the 3D U-Net model, integrating multi-source sea surface data, can rapidly and accurately predict SST in the SCS, offering a new method for forecasting MHW events.