Estimating high-resolution soil moisture by combining data from a sparse network of soil moisture sensors and remotely sensed MODIS LST information

Abstract

The present work demonstrates a methodology for acquiring high-resolution soil moisture information, namely at 1 km at a daily time step, utilizing data from a sparse network of soil moisture sensors and remotely sensed Land Surface Temperature (LST). Building on previous research and highlighting the strong correlation between surface soil moisture and LST, as a result of the thermal inertia, we first evaluated the correlation between Moderate Resolution Imaging Spectroradiometer (MODIS) LST and ground-based soil moisture information from soil moisture sensors installed in a pilot area in Northeastern Greece. Second, a regression formula was developed for three out of six soil moisture sensors, keeping the three remaining monitoring stations serving as a validation set. Furthermore, regression coefficients were interpolated at 1 km and the regression equations were applied for the entire study area, thus acquiring soil moisture information at a spatial resolution of 1 km at the daily time step. The verification process indicated a reasonable accuracy, with a mean absolute error (MAE) of,0.02 m3/m3. The results were considerably better than using a simple interpolation or downscaled daily 1-km SMAP soil moisture.