Estimation Of Daily/1-km Spatially Continuous Soil Moisture Content From The Combined Use Of Optical And Microwave Data Over The Tibetan Plateau Region

As one of the fundamental components of the Earth system,soil moisture is deeply involved in the global carbon,water and energy cycles,and occupies an important position in precipitation analysis,agricultural drought monitoring and flood risk assessment.However,due to radio frequency interference and limitations of the inversion algorithm,most of the soil moisture products cannot provide spatial and temporal continuity soil moisture data with high spatial resolution,and the observation data provided by optical remote sensing also has certain data gaps due to the cloud cover.As a result,the currently available satellite soil moisture observations cannot meet the actual needs of ecological,moisture and climate research in small areas.As the largest plateau in China and highest plateau in the world,the Tibet Plateau has a greater impact on the agricultural,ecological,and economic development of the surrounding areas due to the relatively special geological environment and climatic conditions.Aiming at the problems of low spatial resolution and data gaps of soil moisture products in the Tibet Plateau,this paper proposes two methods of disaggregation-first and predication-first based on the soil moisture data provided by ESA-CCI(European Space Agency-Climate Change Initiative)products and the remote sensing surface parameter data provided by MODIS(Moderate-resolution Imaging Spectroradiometer)products.Among them,the downscaling process mainly uses the DISPATCH(Disaggregation based on Physical And Theoretical scale Change)method to downscale the soil moisture from 25 kilometers to 1 kilometer;and the gaps filling process is mainly through the construction of a generalized regression neural network(GRNN)model to estimate the soil moisture data.Based on the above process,the soil moisture data of the kilometer-level continuous time and space of the Tibet Plateau was finally obtained.Comparing the results with the measured data,the results show:(1)The soil moisture obtained by the two estimation methods have very similar accuracy,and the Root Mean Square Errors(RMSE)are 0.074 m~3/m~3 and 0.075 m~3/m~3,respectively.But the method of predication-first cannot maintain a high-precision reduction in the simulation of a smaller value,and is relatively weak in stability;(2)Under different drought conditions,there is a certain difference in the accuracy of the two soil moisture estimation methods.Semi-arid areas and humid areas have higher accuracy.The average RMSE values are0.057 m~3/m~3and 0.048 m~3/m~3,respectively.The accuracy of arid areas is poor,the mean value of RMSE is 0.085 m~3/m~3.