A Machine-learning-based Model to Inverse Internal Solitary Wave Amplitude from Satellite Image

Internal solitary wave (ISW) has a large amplitude and spreads widely in the global oceans. Amplitude is an important parameter for ISW studies, while the accuracy of state-of-the-art models is limited. Here we show our study of ISW amplitude inversion using synchronous ISW observations from in-situ data and satellite images combined with machine learning techniques. The ISW amplitude can be inversed based on satellite extracted information. Four different machine learning techniques are applied, and results are compared. The results show that the XGBoost has the best performance with a root mean square of 22.08 m on an independent dataset. The proposed amplitude inversion model may shed light on the ISW energy studies.