A Machine Learning-Based Approach for Surface Soil Moisture Estimations with Google Earth Engine

Due to its relation the Earth’s climate and weather phenomena like drought, flooding, or landslides, knowledge of soil moisture content is valuable many scientific professional users. Remote-sensing offers unique possibility for continuous measurements this variable. Especially agriculture, there a strong demand high spatial resolution mapping. However, operationally available products exist with medium coarse only (?1 km). This study introduces machine learning (ML)—based approach (50 m) mapping based on integration Landsat-8 optical thermal images, Copernicus Sentinel-1 C-Band SAR modelled data, executable in Google Earth Engine. The novelty lies applying an entirely data-driven ML concept global estimation surface content. Globally distributed situ data from International Soil Moisture Network acted as input model training. Based independent validation dataset, resulting overall accuracy, terms Root-Mean-Squared-Error R², was 0.04 m3·m?3 0.81, respectively. Beyond retrieval itself, article framework collecting training stand-alone Python package Engine API facilitates execution collection which cloud-based. For retrieval, it eliminates requirement download preprocess any datasets.