SAR Soil Moisture Retrieval Using Physically Based AI Model

Soil moisture is a key factor in monitoring the water cycle on land surfaces. Generally, in-situ soil moisture measurements have been used to observe soil moisture content and it is adopted as the accurate soil moisture observation. However, in-situ soil moisture sensing has limitations for capturing spatial soil moisture distribution, then the remote sensing technique was suggested. In this research, Synthetic Aperture Radar (SAR) which is set on the satellite data is used to retrieve high-resolution soil moisture on various land surface conditions. The physical-based SAR soil moisture retrieval has been noticed as the traditional method for considering vegetation at the sensing spot. However, it is suitable for sparsely vegetated areas and difficult to get reliable results with vegetation. In this study, the Artificial Intelligence (AI) method is suggested which has advantages for accurate soil moisture retrieval with plenty of land surface information. To avoid the generation of an empirical AI model for specific sites, a physical-based AI model is used. Finally, the evaluation of only the AI model, physical model, and physically based AI model is conducted to derive the most robust method for complex terrain, the Korean peninsula.