Development of Improved Algorithms for Downscaling SMAP-Derived Soil Moisture Using Visible/Inferred satellite observations

In this research, a new downscaling method was proposed to improve the spatial resolution of Soil Moisture Active Passive (SMAP) soil moisture estimates with the use of higher resolution from the visible/infrared satellite (MODIS). The model relies on a physical linear relationship between the soil moisture content and shortwave infrared transformed reflectance (STR). Finally, the physical downscaling model was compared against the traditional empirical triangle method to evaluate the performance of using STR in place of land surface temperature (LST). In this study, MODIS and SMAP satellites observations from 2015 to 2018 were used and the evaluation of the downscaled soil moisture was undertaken at the COSMOS sites in Arizona. Results show that while both downscaling algorithms have decreased Root Mean Square Error (RMSE) values, the performance of physical algorithm generally is better than empirical. However, the spatial correlation (R) values decreased using downscaling algorithms, but the RMSE values improved from 0.032 of SMAP soil moisture to 0.3 and 0.031 for the physical and empirical downscaling algorithms, respectively. Also the values of bias improved from 0.011 of SMAP soil moisture to 0.0016 and 0.0076 for the physical and empirical downscaling algorithms, respectively. In conclusion, the result show the proposed physical downscaling algorithm nicely improves the limited spatial variability of SMAP soil moisture and replacing LST with STR can yield a new insight on the downscaling issues.