Downscaling and validation of 1km ESA Soil Water Index for soil moisture monitoring on agricultural land in temperate climate conditions

Near-surface soil moisture is an important parameter for estimating the water balance of ecosystems, especially for the exchange of water between atmosphere and soil. In the past decades, global and continental products, e.g. the Soil Water Index (SWI) of ESA, have been developed for large-scale monitoring from passive and active microwave data. ESA SWI was developed for Europe based on the surface soil moisture product derived from Sentinel-1 C-band SAR data (1km resolution) and Metop ASCAT surface soil moisture (12.5km). However, for practical, small-scale applications on the local level, there is currently little data available. This contribution elaborates on the validation of the SWI for temperate agricultural regions at the example of the agrometeorological network of the JECAM test site DEMMIN in Northeast Germany. For this purpose, two resampling methods, a bilinear interpolation and a statistical downscaling approach using Random Forest were tested on SWI data from 2019. The statistical downscaling approach integrates Sentinel satellite data and the Topographical Wetness Index based on a 10m resolved elevation model. Both resampling methods showed similar results. Over time, the SWI significantly overestimates the in situ data before and after the crop growing season. A higher agreement is observed in the summer months. For 19 of the 29 investigated agrometeorological stations a statistically positive correlation with R>0.5 was found. The remaining stations showed little to no correlation, most likely due influences of various crops types on the remote sensing data. The results suggest a temporally limited applicability of the SWI 1km data for local assessments of soil moisture.