Study Of Retrieving Surface Soil Moisture Based On Sentinel-1 SAR Data

This paper selected the farming area of Baoji City as the research area,made full use of active microwave remote sensing in soil moisture monitoring,Sentinel-1 synthetic aperture Radar images were used as the main data sources,based on a brief description of the theory and methods of retrieving soil moisture with SAR data,the water cloud model for low vegetation cover surface was used to simulate the surface microwave scattering characteristics of the study area.After correcting the influence of vegetation scattering,the soil backscattering coefficients of VV and VH polarized were obtained.Finally,the soil moisture prediction model was constructed by three-layers RBF neural network to achieve the retrieval of relative water content of soil.The main conclusions of the study were as follows:(1)The water cloud model can be used to simulate the surface microwave scattering characteristics of the study area.The extracted soil backscattering coefficient had a good correlation with the measured soil moisture data.And compared with VH polarization,the correlation between soil backscattering coefficient and measured soil moisture was higher under VV polarization.(2)In the area covered by vegetation,the interference of the vegetation on microwave backscattering signal can't be ignored.For VV polarization and VH polarization,the average contribution of the vegetation to Radar backscattering coefficient was 1.68 dB and 3.36 dB,respectively.Estimated the water content of vegetation using optical remote sensing data,took it as an input parameter of water cloud model to calculate the backscattering coefficient of soil,which could effectively reduce the influence of vegetation scattering and improve the sensitivity of soil backscattering coefficient to soil moisture.(3)The relative water content of soil predicted by the RBF neural network model ranged from46.08%to 87.54%,with an average of 64.01%.Variation coefficient of soil moisture belonged to moderate spatial variability.The soil moisture level was mainly mild drought,accounting for68.21%of the area.Correlation analysis between relative water content of soil and environmental factors showed that soil types such as black loess,bauxite and new soil had higher average relative water content,mainly distributed in plains and valleys in the south of the study area.loess,red clay and purple soil had lower average relative water content,mainly distributed in the loess hilly area of the north.With elevation and slope increasing,the distribution advantages of drought-free and mild drought were generally reduced,while the distribution advantages of moderate drought and severe drought were generally increased.There was a positive correlation between relative water content of soil and vegetation coverage(R~2=0.386).(4)The root mean square error of the relative water content of soil and the measured soil moisture data was 5.65%,R~2=0.649,and it had good consistency in spatial distribution with0.01°soil moisture products.It was indicated that based on Sentinel-1 SAR data,the relative water content of soil retrieved by water cloud model and RBF neural network had high precision.It is feasible to use the above data and methods to carry out large-scale dynamic monitoring of farmland soil moisture.