A Preliminary AI Framework for the Predictability of Heat Flux Around South China Sea

In the past decades, ocean heat accumulation accounts for more than 90% of the excess energy in the Earth system, moderating the global warming. However, it is still of great challenging to understand how the acceleration in global ocean heat accumulation functions by mode. In this paper, we make an initial attempt to develop an AI framework via Transformer, to explore the predictability for ocean heat flux around South China Sea (SCS), with regards to cross-quantilogram in essential ocean variables. It has been demonstrated that our proposed scheme could perform as a preliminary means to predict ocean heat flux, with RMSE, MAE, MAE/STD respectively 3.68E-02W/m², 2.64E-02W/m², 46.18% for a 14-day prediction duration, and the predictability horizon could even reach up to 42-day with a reasonable performance.