Satellite-Based Retrieval of the Vertical Structure of Internal Solitary Waves in Continuous Stratification

The vertical structure of internal solitary waves (ISWs), defined by the amplitudes at different water depths, is critical for understanding and predicting how the waves affect ocean mixing and sediment transport. Spaceborne synthetic aperture radar (SAR) data have been used to retrieve the vertical structures of ISWs in many parts of the global oceans. However, the commonly used theories for determining the vertical structure from SAR images are often based on a two-layer ocean model, deviating from the continuous stratification found in the real ocean environment. In this study, we retrieved the vertical structures of ISWs in continuous stratification from SAR images through a machine learning approach. We first proposed an improved parameterization scheme for ocean stratification incorporating a component of deep-layer stratification. With this scheme, we simulated the vertical structures of ISWs and their corresponding radar surface signatures. These physical simulations lead to a key input, the modulation depth, excluded in existing methods or models. Then, the relationship between radar signals and vertical structures of ISWs was found through a backpropagation neural network. We validated the relationship by two contemporaneous experiments conducted in the Andaman Sea and the Strait of Gibraltar. This study aims to provide a fresh perspective on retrieving the vertical structure of ISWs in the real ocean environment by replacing the ideal two-layer ocean model and demonstrates the applicable potential of satellites in retrieving three-dimensional structures.