Spatial disaggregation of ASCAT soil moisture under all sky condition using support vector machine

With recent advances in downscaling methodologies, soil moisture (SM) estimation using microwave remote sensing has become feasible for local application. However, disaggregation of SM under all sky conditions remains challenging. This study suggests a new downscaling approach under all sky conditions based on support vector regression (SVR) using microwave and optical/infrared data and geolocation information. Optically derived estimates of land surface temperature and normalized difference vegetation index from MODerate Resolution Imaging Spectroradiometer land and atmosphere products were utilized to obtain a continuous spatio-temporal input datasets to disaggregate SM observation from Advanced SCATterometer in South Korea during 2015 growing season. SVR model was compared to synergistic downscaling approach (SDA), which is based on physical relationship between SM and hydrometeorological factors. Evaluation against in situ observations showed that the SVR model under all sky conditions (R: 0.57 to 0.81, ubRMSE: 0.0292 m³ m^?3 to 0.0398 m³ m^?3) outperformed coarse ASCAT SM (R: 0.55 to 0.77, ubRMSE: 0.0300 m³ m^?3 to 0.0408 m³ m^?3) and SDA model (mean R: 0.56 to 0.78, ubRMSE: 0.0324 m³ m^?3 to 0.0436 m³ m^?3) in terms of statistical results as well as sensitivity with precipitation. This study suggests that the spatial downscaling technique based on remote sensing has the potential to derive high resolution SM regardless of weather conditions without relying on data from other sources. It offers an insight for analyzing hydrological, climate, and agricultural conditions at regional to local scale.