Research On The Theory And Method Of Soil Moisture Retrieval Using A Spaceborne GNSS-R Technique

As a new remote sensing retrieval technique,GNSS-R has been applied in ocean surface wind speed inversion,ocean elevation measurement,sea ice detection,soil moisture monitoring,and other remote sensing fields.With the successful launch of the CYGNSS mission,spaceborne GNSS-R technique began to enter a period of rapid development.Soil moisture plays an important role in the global water cycle and energy exchange,and is one of the important indicators for predicting all kinds of extreme weather events.Therefore,how to use the spaceborne GNSS-R technique to detect global soil moisture is the focus of practical application.In this paper,the sampling area of CYGNSS was taken as the research area between 38°north and south latitudes,and the theory and method of Soil Moisture inversion by spaceborne GNSS-R technique was researched and explored by combining with CYGNSS data and SMAP data.The following contents were mainly carried out.（1）Research on the basic theory of GNSS-R on board,and the theoretical basis of spaceborne GNSS-R inversion of soil water was given,including GNSS signal reflection characteristics,geometric relations,as well as the study of the flashing area and the two-dimensional time-delay doppler correlation power map.On this basis,the spaceborne GNSS reflection signal power model was analyzed.（2）To carry out the research on spaceborne GNSS-R soil moisture inversion method,the basic process of soil moisture inversion process in this paper is firstly introduced,and the characteristic parameters related to the construction of spaceborne GNSS-R soil moisture inversion model by random forest algorithm are analyzed.The machine learning algorithm model was constructed based on CYGNSS data and soil moisture of SMAP products from March 2017 to March 2019.（3）Based on the analysis of the final inversion results and the verification of the ground data,the optimal characteristic input was determined as the surface reflectance at the peak power point,the reflection Angle,the front slope,the delay difference between the peak power reflection point and the specular reflection point,and the auxiliary characteristic parameters,vegetation optical thickness,surface roughness and soil texture.The correlation coefficient between soil moisture and SMAP products of CYGNSS retrieved from March 2018 to March2019 was 0.842,and the root-mean-square error was 0.056 cm³cm^-3.Based on different land cover types analysis model inversion ability difference,comparing results showed that the best inversion effect under the bare land,correlation coefficient of 0.811,the root mean square error was only 0.039 cm³cm^-3.At the same time,the influence of seasonal vegetation change on the inversion results is analyzed.Finally,the soil moisture predicted by CYGNSS was verified on the ground using measured data from the ground station of SCAN.The correlation coefficient between CYGNSS soil moisture and the soil moisture directly observed by the selected ground station is between 0.58 and 0.64,and the maximum root mean square error is 0.054 cm³cm^-3.