Surface soil moisture relationship model construction based on random forest method

Aiming to solve the limitation of coarse spatial resolution of passive microwave soil moisture product, a soil moisture relationship model based on random forest method was constructed with land surface temperature (LST), normalized difference vegetation index (NDVI), and surface albedo (ALB) from MODIS products and surface soil moisture (SSM) from AMSR-E soil moisture product in the study area at the east edge of the Tibetan Plateau. The results show better performance of the proposed compared with the commonly used purely-empirical method, with the R² values above 0.88 and the RMSE values lower than 0.05m³/m³, respectively. It suggested that the proposed soil moisture relationship is able to capture the spatio-temporal variation of surface moisture well. There should be great potential to improve the downscaling soil moisture accuracy when the model is used in the passive microwave soil moisture downscaling scheme.