【作者】 陈文谦； 李杰； 洪学宝； 王峰； 杨东凯；

【Author】 CHEN Wenqian;LI Jie;HONG Xuebao;WANG Feng;YANG Dongkai;Electronics and Information Engineering, Beihang University;

【机构】 北京航空航天大学电子信息工程学院；

【摘要】 利用GNSS反射信号进行陆地表面土壤湿度反演经历了数十年发展,已经成为一种成熟的遥感技术。根据接收机安装的方式,可分为地基、机载和星载三种方式。为发挥不同方式的优势,提升土壤湿度的反演精度,提出一种基于地基和星载GNSS反射信号土壤湿度遥感的天地一体化反演算法。在地基情况下,通过希尔伯特变换计算利用低阶多项式拟合分离的GNSS干涉信号中的直射和反射信号的幅值比,并利用半经验模型进行土壤湿度的反演。在星载情况下,利用美国航空航天局的旋风全球导航卫星系统(cyclone global navigation satellite system,CYGNSS)提供的散射系数等数据,土壤湿度主被动项目(soil moisture active passive,SMAP)发布的土壤湿度产品作为建模数据以及同比数据,提出了一种利用列文伯格-马夸尔特(Levenberg-Marquardt,LM)算法建立神经网络模型的方法,可大范围面积地反演土壤湿度数据。实验结果表明,在地基情况下,同比和反演的土壤湿度相关系数为0.89,RMSE为0.05 cm3/cm3;在星载情况下,利用2018年1月的CYGNSS和SMAP数据建立模型之后,通过与SMAP数据提供的土壤湿度数据进行对比,均方根误差最小可达0.068 cm3/cm3,相关系数为0.81。所提出的天地一体化GNSS反射信号土壤湿度反演系统不仅可提供大范围的土壤湿度数据,还可以提供单点土壤湿度精确数据。更多还原

【Abstract】 The use of GNSS reflection signals for land surface soil moisture inversion has undergone decades of development and has become a mature remote sensing technology. According to the installation method of the receiver, it can be divided into three types: ground-based, airborne, and satellite-based. To leverage the advantages of different methods and improve the accuracy of soil moisture inversion, a fusion of ground-and satellitebased inversion algorithm based on ground and satellite GNSS reflection signals for soil moisture remote sensing is proposed. In the case of ground-based, the amplitude ratio of direct and reflected signals in GNSS interference signals separated by low order polynomial fitting is calculated through Hilbert transform, and then soil moisture inversion is carried out based on a semi empirical model. In the case of spaceborne, using data such as scattering coefficients provided by NASA’s cyclone global navigation satellite system(CYGNSS), and soil moisture products released by the soil moisture active passive(SMAP) project as modeling data and year-on-year data, a method is proposed to establish a neural network model using the Levenberg-Marquardt(LM) algorithm, which can retrieve soil moisture data on a large scale. The experimental results show that in the case of ground-based, the correlation coefficient of soil moisture for year-on-year and inversion is 0.89, and the RMSE is 0.05 cm~3/cm~3; In the case of satellite-based, after establishing the model using CYGNSS and SMAP data in January 2018, by comparing with the soil moisture data provided by SMAP data, the minimum root-mean-square deviation can reach 0.068 cm~3/cm~3, and the correlation coefficient is 0.81. The proposed integrated soil moisture inversion model for heaven and earth can not only provide large-scale soil moisture data,but also provide accurate soil moisture data at a single point.更多还原