The merit of estimating high-resolution soil moisture using combined optical, thermal and microwave data

Tremendous progress has been made in estimating soil moisture from satellite remote sensing data. Several global-scale coarse-resolution products have also been generated and released for various applications in the Earth system. However, high-resolution soil moisture estimation is still in its infancy. Currently, two main methods are used for this purpose: downscaling approaches and direct retrieval from microwave and optical/thermal data. Several studies have attempted to comprehensively evaluate the performance of these approaches and have found that each method has its own strengths and weaknesses, with no single method outperforming the others. In this study, we aim to investigate the advantages of integrating optical, thermal, and microwave data to estimate soil moisture by leveraging an intensive soil moisture network and triple collocation method. Firstly, we determined the best-performing coarse-resolution microwave soil moisture product through the triple collocation approach. Secondly, we generated 1-km soil moisture using a downscaling approach based on land surface temperature and vegetation index, utilizing the best-performing SMAP L3 descending product. Thirdly, we evaluated the high-resolution downscaled soil moisture, Sentinel 1 soil moisture, and SMAP/Sentinel 1 combined soil moisture products using soil moisture measurements from the REMEDHUS station network, ETOPO1 elevation, CHIRPS precipitation, and the ESA CCI land cover map. Finally, we investigated and demonstrated the advantages of merging these products through point-scale evaluation and large-scale spatial pattern comparison.