Soil Moisture Inversion Of ENVISAT/ASAR Dual-polarized Data

Soil moisture is an important factor in controlling water cycle and energy exchange between terrestrial and atmospheric systems.It reflects the interaction between the surface and the atmosphere to a certain extent,has an impact on the precipitation and groundwater reserves.Soil moisture is also the basic condition for the growth and development of crops and vegetation,and plays an important role in estimating crop yields and monitoring drought conditions.Traditional soil moisture monitoring has higher accuracy in single-point measurement,but it cannot obtain soil moisture in large area,quickly and effectively.Synthetic Aperture Radar(SAR),as an advanced remote sensing technology,can solve these problems,and it has the characteristics of all-day,all-weather monitoring,and also has certain penetration into the soil.The use of SAR data to retrieve soil moisture has a solid physical theoretical basis,providing new methods and ways for large-area,real-time monitoring of soil moisture.In this paper,the relationship between backscattering coefficient and soil water content and surface roughness is simulated by AIEM model,and their general expressions are obtained in turn.Since the backscattering coefficient is affected by the surface parameters at the same time,it is necessary to combine them to further derive the empirical model of soil moisture inversion.The BP neural network model was also used to invert soil moisture.Based on the ENVISAT/ASAR dual-polarization data,combined with the measured data of Linze grassland and Arou grassland in the Heihe River Basin,the two inversion methods were verified and compared.The main research results include the following aspects:(1)Considering the influence of root mean square height and correlation length on backscattering coefficient,a new roughness parameter Zs is proposed,and then the relationship between Zs and VV and VH polarization combinations is derived.It has the advantages of simplified soil moisture inversion model and no need to measure roughness data to better invert soil moisture.(2)Based on the AIEM model,the relationship between the backscattering coefficient and the surface parameters is simulated,and the soil moisture inversion model is finally obtained.The model is extended to the incident angle of 20°~70°.Then,using the measured data to verify the inversion model,and the soil moisture distribution map is obtained.Compared with two other different empirical models,the verification results show that the empirical model inversion is better.(3)When inverting soil moisture based on BP neural network,first analyze the importance of three VV and VH polarization combinations on soil moisture.It is found that the inversion effect of(VV-VH)/(VH/VV)added to the neural network is relatively best.The reason for this analysis is that this polarization combination indirectly considers the effect of roughness.Then use the established BP neural network to invert the soil moisture of Linze grassland and Arou grassland.(4)Comparing the empirical model with the BP neural network model,the empirical model is better than the BP neural network in terms of running time(t)and root mean square error(RMSE).However,in terms of fitting effect,the BP neural network has a better coefficient of determination(R~2)and can effectively control the influence of noise in the image.