Mitigating the impact of dense vegetation on the Sentinel-1 surface soil moisture over Europe

Abstract

The current generation of Synthetic Aperture Radars (SAR) has a high potential to retrieve surface soil moisture (SSM) at a kilometer-scale resolution. Research has shown that a change detection approach applied to the backscatter from the Sentinel-1 mission was able to yield a consistent kilometer-scale SSM product over Europe. This product is operational and available on the Copernicus Global Land Service (CGLS) website (https://land.copernicus.eu/global/). A known problem of the CGLS algorithm is its reduced performance over areas with dense vegetation. The combined influence of vegetation water content and geometry on the backscatter signal results in a lower sensitivity to SSM. This effect is especially observed over woody vegetation such as broadleaved or coniferous forests. In addition, a wet bias is detected in the CGLS SSM data during the growing season over land cover with seasonal variation of vegetation.

This study utilizes the native resolution of Sentinel-1 in its interferometric wide swath mode (20x22m), resampled to a 20m pixel spacing, to assess three dense vegetation masks over Europe. The masks are derived from forest/tree cover maps based on Sentinel-1, Sentinel-2, or a combination of both. At 20m, the backscatter pixels are selectively filtered to discard the ones flagged as non-soil moisture sensitive. The masked backscatter at 20m sampling is then resampled to a kilometer scale and used as input for the CGLS change detection model algorithm. The resulting SSM product is compared to in-situ stations from the International Soil Moisture Network (ISMN) and with modeled soil moisture from ERA5-Land. The results sug gest that masking dense vegetation consistently improves the SSM signal over regions containing both forested areas, and croplands or grasslands.

This study highlights the potential of masking non-soil moisture sensitive pixels at the native resolution of the Sentinel-1 backscatter. The results demonstrate the ability of high-resolution forest masking to mitigate the effect of dense vegetation on the CGLS SSM product.