Generation of High-Resolution Surface Soil Moisture over Mountain Areas by Spatially Downscaling Remote Sensing Products Based on Land Surface Temperature-Vegetation Index Feature Space

Spatial downscaling has been a key solution to get high-resolution surface soil moisture (SSM), which has attracted wide attention in remote sensing society. However, the impact from topographic reliefs, complexifying SSM spatial heterogeneity, has been rarely considered in previous downscaling studies. Here, we propose a novel approach for SSM downscaling based on the physical connection between the land surface temperature (LST)-vegetation index triangle feature space and SSM, where a self-adaptive calibration method was applied to determine the estimation coefficients via a sliding window with the topographic effect of LST alleviated in advance. The proposed method was evaluated at a typical mountain region in central USA from 2015 June 1 to September 30 via the 25-km original European Space Agency Climate Change Initiative SSM product and Moderate Resolution Imaging Spectroradiometer/Terra LST and normalized difference vegetation index products. Through the direct validation with the in situ soil moisture measurements from the Snow Telemetry network, the downscaled results show better performance than other previous methods, with the average value of the correlation coefficient, root-mean-square error, and unbiased root-mean-square error derived at the site level of 0.47, 0.103 m3/m3, and 0.056 m3/m3, respectively. Meanwhile, the good downscaling effect can be reflected by the downscaling performance evaluation index. Furthermore, an intercomparison with the Soil Moisture Active Passive-HydroBlocks SSM product also reveals the consistent spatial distribution and strong correlation of the downscaled results. Overall, these results confirm the potential application of the proposed method in generating seamless high-resolution SSM over mountain areas, which will contribute to related mountain studies.