基于多波段多极化SAR数据的草原地表土壤水分反演方法研究

发布时间：2018-05-29 16:11

本文选题：土壤水分反演 + 合成孔径雷达　；参考：《电子科技大学》2017年博士论文

【摘要】：土壤水分在全球和区域水文和气象过程中发挥着重要作用,被认为是地球科学研究中不可或缺的状态变量。特别是在自然环境恶劣的干旱、半干旱和高寒草原区域,土壤水分被认为是影响草原物候的最重要因素。遥感技术的快速发展使得多时空尺度的土壤水分变化监测成为可能。由于对土壤水分的高度敏感性及全天时全天候的观测能力,使得微波遥感在裸露地表和植被覆盖区域土壤水分反演中应用越来越广泛。然而地表粗糙度和植被的散射贡献会降低微波信号对土壤水分的敏感性,从而增加土壤水分反演的复杂度和难度。因此同时消除地表粗糙度和植被对土壤水分反演的影响成为构建草原地表土壤水分反演方法的关键技术。本文以青海省乌图美仁草原、四川省若尔盖草原和青海省青海湖流域为研究区域,以合成孔径雷达(synthetic aperture radar,SAR)数据、光学遥感数据和地面实测数据为重要数据源,通过耦合不同的土壤介电常数模型、地表散射模型和植被散射模型,构建适合草原地表的土壤水分反演方法。本论文主要工作概括如下:(1)采用实测地表粗糙度参数初始化地表散射模型,发展了基于高级积分方程模型(advanced integral equation model,AIEM)和比值方法的草原地表土壤水分反演方法。该方法将实测地表粗糙度参数作为先验知识用于模拟裸露地表同极化后向散射系数与土壤水分之间的经验关系,同时实现对比值方程中未知系数的求解。构建的土壤水分反演方法中,AIEM模型用于模拟裸露地表的后向散射系数,比值方程用于消除植被的散射贡献,其中四个不同的植被参数包括叶面积指数(leaf area index,LAI)、植被含水量(vegetation water content,VWC)、归一化植被指数(normalized difference vegetation index,NDVI)和增强植被指数(enhanced vegetation index,EVI)分别用于植被散射贡献的参数化。通过在乌图美仁草原和青海湖流域的实验结果表明,该方法可以用于草原地表的土壤水分反演。同时研究发现,LAI同其它植被参数相比更适合于乌图美仁草原植被的参数化,而LAI、NDVI和EVI都可以用于表征青海湖流域的植被散射特征。(2)充分利用SAR数据的多极化信息消除对实测地表粗糙度参数的依赖,建立土壤介电常数与观测到的同极化后向散射系数间的定量函数关系,从而实现对草原地表的土壤水分反演。该方法解决在实测地表粗糙度参数缺失的情况下如何充分利用多极化SAR数据实现对草原地表的土壤水分反演。首先通过化简Dubois模型建立土壤介电常数与裸露地表同极化后向散射系数间的函数关系,然后分别采用比值方法和水云模型(water cloud model,WCM)实现对植被散射贡献的分离,其中植被的散射机制分别采用LAI、VWC、NDVI和EVI四个植被参数进行表征。通过实验结果表明,该方法有效地解决了地表粗糙度参数和植被对土壤水分反演的影响。从植被参数化对土壤水分反演结果影响发现,在乌图美仁草原LAI参数化植被的效果较好,而在若尔盖草原EVI效果较好。该方法不依赖于任何的地表粗糙度参数,极大地提高了土壤水分反演方法的适用性。(3)基于AIEM模型、比值方法和有效粗糙度参数构建了适合草原地表的土壤水分反演方法。该方法主要是针对在考虑地表粗糙度参数的同时且不依赖于实测地表粗糙度参数的情况下实现对草原地表的土壤水分反演。AIEM模型用于模拟裸露地表后向散射系数,其中给定的粗糙度参数作为模型的输入参数。比值方法用于将植被的散射贡献从总后向散射系数中进行分离,其中四个植被参数LAI、VWC、NDVI和EVI分别用于表征植被的散射贡献。通过实验结果发现,该方法可以用于草原地表土壤水分反演并且算法精度明显提高。针对比值方程中植被的参数化问题,研究结果表明LAI适合表征乌图美仁草原的植被散射,EVI适合描述若尔盖草原的植被生长状况,而LAI、NDVI和EVI均可以表征青海湖流域的植被散射贡献。该方法考虑了地表粗糙度参数对后向散射系数的贡献,但同时有效粗糙度参数的加入消除了对地面实测数据的依赖,提高了该方法的普适性。(4)基于全极化Radarsat-2数据提取的极化特征参数以及多元线性回归方程,探索极化特征参数用于估算草原地表土壤水分的可行性。该方法考虑如何在不消除地表粗糙度参数和植被散射贡献的前提下直接采用极化特征参数实现对草原地表的土壤水分估算。本文考虑的极化特征参数包括Cloude分解参数极化熵、散射角和反熵,三个特征值参数,特征值组合参数包括单次反射特征值相对差异度、双次散射特征值相对差异度和雷达植被指数,以及Freeman分解参数表面散射分量、二次散射分量和体散射分量。通过乌图美仁草原和若尔盖草原的实验结果发现,极化特征参数可以辅助草原地表的土壤水分反演。针对草原地表土壤水分反演存在的最大问题是如何同时消除地表粗糙度参数和植被对土壤水分反演的影响。本文提出同时耦合土壤介电常数模型、地表散射模型和植被散射模型,构建适合草原地表的土壤水分反演方法。这些理论和方法的突破将为草原区域的土壤水分反演提供新的理论与方法支持。
[Abstract]:Soil moisture plays an important role in the global and regional hydrological and meteorological processes. It is considered to be an indispensable state variable in the research of earth science. Especially in the arid, semi-arid and alpine steppe regions of the natural environment, soil moisture is considered as the most important factor affecting the grassland phenology. The rapid development of remote sensing technology makes it possible. It is possible to monitor soil moisture changes in a space-time scale. Due to the high sensitivity to soil moisture and the all-weather observational ability all day, microwave remote sensing is becoming more and more widely used in the inversion of soil moisture in the exposed and vegetation cover areas. However, the surface roughness and the contribution of vegetation scattering will reduce the microwave signal. The sensitivity of soil moisture can increase the complexity and difficulty of soil moisture inversion, so eliminating the surface roughness and the influence of vegetation on soil moisture inversion is the key technology to build the soil moisture inversion method of the grassland. This paper takes the Qinghai province of uuumeen grassland, Ruoergai grassland in Sichuan province and the Qinghai Lake of Qinghai province. The basin is a study area, with synthetic aperture radar (synthetic aperture radar, SAR) data, optical remote sensing data and ground measured data as important data sources. By coupling different soil dielectric constant model, surface scattering model and vegetation scattering model, the soil moisture inversion method suitable for grass ground surface is constructed. The main work of this paper is the main work of this paper. The following are summarized as follows: (1) the soil surface soil moisture inversion method based on the advanced integral equation model (Advanced integral equation model, AIEM) and the ratio method is developed by using the measured surface roughness parameters. This method uses the measured surface roughness parameters as a priori knowledge to simulate the bare surface homopolar. The empirical relationship between the back scattering coefficient and soil moisture, and the solution of the unknown coefficient in the ratio equation. In the soil moisture inversion method, the AIEM model is used to simulate the backscatter coefficient of the bare surface, and the ratio equation is used to eliminate the scattering contribution of the vegetation, of which four different vegetation parameters include the leaf area. Leaf area index (LAI), vegetation water content (vegetation water content, VWC), normalized vegetation index (normalized difference vegetation index) and enhanced vegetation index for parameterization of the contribution of vegetation scatter respectively. Through the experimental results in the uuumeen grassland and the Qinghai Lake Basin This method can be used to invert soil moisture on the ground surface of grassland. At the same time, it is found that LAI is more suitable to parameterize the vegetation of uuumeen grassland compared with other vegetation parameters, and LAI, NDVI and EVI can be used to characterize the characteristics of vegetation scattering in the Qinghai Lake basin. (2) fully utilize the multi polarization information of SAR data to eliminate the measured surface roughness. Based on the dependence of roughness parameters, the quantitative function relationship between the soil dielectric constant and the observed polarization backscattering coefficient is established to realize the inversion of soil moisture on the ground surface of the grassland. This method solves how to make full use of the multipolar SAR data to realize the soil moisture on the grassland surface under the absence of the measured surface roughness parameters. First, the function relationship between the soil dielectric constant and the scattering coefficient of the bare surface is established by the simplified Dubois model. Then the ratio method and the water cloud model (water cloud model, WCM) are used to separate the contribution of the vegetation scattering, in which the vegetation scattering mechanism uses four vegetation parameters, LAI, VWC, NDVI and EVI, respectively. The experimental results show that the method can effectively solve the effect of surface roughness parameters and vegetation on soil moisture inversion. The effect of parameterized vegetation on LAI vegetation in uuumumeen grassland is better, and the effect of EVI in Ruoergai grassland is better. The applicability of the soil moisture inversion method is greatly improved by any surface roughness parameters. (3) based on the AIEM model, the ratio method and the effective roughness parameter, the soil moisture inversion method suitable for the grassland surface is constructed. This method is mainly aimed at considering the surface roughness parameters and does not depend on the measured surface roughness. In the case of parameters, the.AIEM model of soil moisture inversion on the ground surface is used to simulate the backscatter coefficient of the bare surface, in which the given roughness parameter is used as the input parameter of the model. The ratio method is used to separate the contribution of the vegetation from the total backscatter coefficient, of which four vegetation parameters are LAI, VWC, NDVI and EVI. It is not used to characterize the scattering contribution of vegetation. Through the experimental results, it is found that this method can be used to inverse the soil moisture in the grassland and improve the precision of the algorithm. According to the parameterization of vegetation in the ratio equation, the results show that LAI is suitable for the characterization of the vegetation scattering of uuus meaden, and EVI is suitable for describing the vegetation of Ruoergai grassland. Long condition, and LAI, NDVI and EVI can all represent the contribution of the vegetation scattering in the Qinghai Lake basin. This method considers the contribution of the surface roughness parameters to the backscatter coefficient, but the addition of the effective roughness parameters eliminates the dependence on the ground measured data and improves the universality of the method. (4) based on the full polarization Radarsat-2 data extraction The polarization characteristic parameters and multiple linear regression equations are used to explore the feasibility of the polarization characteristic parameters used to estimate the soil moisture in the grassland. This method considers how to estimate the soil moisture content directly on the grassland surface without eliminating the surface roughness parameters and the contribution of vegetation scattering. The polarization characteristic parameters include the polarization entropy of the Cloude decomposition parameter, the scattering angle and the anti entropy, three eigenvalues, and the combination parameters of the eigenvalues include the relative difference degree of the eigenvalue of the single reflection, the relative difference of the double scattering eigenvalue and the radar vegetation index, the surface scattering component of the Freeman decomposition parameter, the scattering component and the body scattering component. Through the experimental results of the urumumeen grassland and Ruoergai grassland, it is found that the polarization characteristic parameters can assist the inversion of soil moisture in the grassland surface. The biggest problem for soil moisture inversion in the grassland is how to eliminate the influence of surface roughness parameters and vegetation on soil moisture inversion at the same time. The dielectric constant model, the surface scattering model and the vegetation scattering model are used to construct the soil moisture inversion method suitable for the grassland surface. The breakthrough of these theories and methods will provide a new theory and method support for the soil moisture inversion in the grassland area.
【学位授予单位】：电子科技大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：S812.2;TN957.52

【参考文献】