Soil moisture retrieval using multi-temporal Sentinel-1 SAR data in agricultural areas

Abstract: Soil moisture is a key variable that links the water and energy cycles. Its information is also essential for many applications, such as agricultural drought monitoring, crop status monitoring and crop yield prediction. Sentinel-1 of the European Space Agency (ESA) is composed of 2 satellites, Sentinel-1A and Sentinel-1B, which share the same orbital plane with a 180° orbital phasing difference. The Sentinel-1 mission can provide C-band synthetic aperture radar (SAR) data with a global revisit time of just 6 days and high spatial resolution of about tens of meters, thus showing a strong potential for global soil moisture monitoring at high/moderate spatial resolutions. The aim of this study was to investigate the capability of multi-temporal Sentinel-1 C-band SAR data with a short repeating cycle in soil moisture estimation over agricultural fields. In order to retrieve soil moisture, an algorithm based on the change detection technique was utilized. This algorithm (referred to as alpha approximation approach) relies on the assumptions that the contributions of vegetation and surface roughness to the radar backscattered signal are multiplicative. Therefore, the effects of vegetation and surface roughness on radar backscattering coefficients can be decoupled from the effects of soil moisture changes by rationing multi-temporal like-polarized (HH and VV) intensities between two close acquisition dates. The ratio is expected to track changes in soil moisture only since the changes of surface roughness, canopy structure and vegetation biomass take place at longer temporal scales than soil moisture changes. The alpha approximation approach was firstly evaluated by comparing with data sets simulated by a theoretical radiative transfer (RT) scattering model. It was found that the alpha approximation approach was overall in good agreement with the RT scattering model without introducing significant errors for bare surface and low vegetation area, which confirmed that the alpha approximation approach was a simple and effective way to reduce the influences of vegetation and surface roughness. Furthermore, under the assumption of alpha approximation, the ratio of 2 consecutive backscatter measurements could be approximately represented as the squared ratio of corresponding Bragg scattering coefficients. For Sentinel-1 SAR data with only one like-polarized channel (i.e. VV), N SAR acquisitions would result in N - 1 linear equations in N unknown Bragg scattering coefficients. To solve this underdetermined system of equations, a bounded linear least-squares optimization was applied. Once the unknown Bragg scattering coefficients were retrieved, the relative dielectric constant could be analytically derived with the soil moisture being estimated by the inversion of microwave dielectric model. The alpha approximation approach was then applied to 4 consecutive Sentinel-1 SAR images acquired over Huailai experiment field. Soil moisture maps were successfully obtained for each date. The results were validated using ground measurements on one acquisition date, with root mean squared error (RMSE) value of 0.06 cm3/cm3 and mean bias value of 0.01 cm3/cm3. The results demonstrated the overall good performance of the alpha approximation approach. These results imply that multi-temporal Sentinel-1 SAR data show great potential in achieving high resolution and accurate soil moisture retrievals over agricultural fields.