RTM-Based Downscaling of Medium Resolution Soil Moisture Using Sentinel-1 Data Over Agricultural Fields

High temporal soil moisture at field scale resolution (10 m–100 m) is important for smart farming decisions. Although, medium and coarse resolution (1 km–50 km) soil moisture information is operationally available on a large scale, high resolution (field scale) datasets are not. This study propose a data assimilation approach to downscale medium resolution (1 km × 1 km) soil moisture information–of intense agriculturally cultivated areas–to field scale. For achieving high transferability of the proposed method, the used input data (Sentinel-1 VV backscatter, Sentinel-2 derived vegetation water content, literature values) can be provided systematically from global operational satellites. Microwave and optical data are used together as input data of a radiative transfer model to derive soil moisture information with high temporal and spatial resolution. The retrieval approach shows a mean ubRMSE for soil moisture estimates of all test fields (Munich-North-Isar test site, Bavaria, Germany) with 0.045 m³/m³ and 0.037 m³/m³ for 2017 and 2018. Furthermore, the retrieved soil moisture estimates cover a broad range of values from 0.05 m³/m³ to 0.4 m³/m³. In addition, the temporal evolution of the soil moisture patterns are in line with precipitation events. Moreover, the drying behavior is matched as well. The proposed method showed that for the test area, high resolution soil moisture time series can be provided by only using remote sensing derived input data. In this way, this study is another step towards providing high spatio-temporal soil moisture information for precision farming purposes.