多源时序SAR数据土壤水分反演研究

发布时间：2018-02-09 00:26

本文关键词： 土壤水分合成孔径雷达多传感器多时相变化检测积分方程模型　出处：《中国矿业大学》2017年博士论文　论文类型：学位论文

【摘要】：土壤水分是地表过程中的关键变量,在全球能量平衡、水循环及气候变化研究中扮演重要的角色。土壤水分是农作物生长的基本条件,同时也是作物长势监测、估产及旱情监测的重要参数。因此,土壤水分时空动态估计和描述对水文、生态和农业具有重要的意义。合成孔径雷达(Synthetic Aperture Radar,SAR)具有全天时、全天候数据获取的能力,且雷达后向散射信息对土壤介电特性响应敏感,被广泛应用于土壤表面水分监测和反演研究。然而,雷达后向散射信息受到多个因素的共同作用,包括雷达系统参数、土壤介电常数、表面粗糙度和植被覆盖,因此实现土壤水分的精确反演需要有效去除土壤表面粗糙度和植被覆盖等因素的影响。该论文基于多时相Radarsat-2,Terra SAR-X和Sentinel-1A影像及相应的现场实测数据,对河北邯郸农田区域和宁夏盐池半干旱区域土壤表面水分进行反演研究。综合多极化、多波段和多时相SAR数据降低或去除表面粗糙度对雷达后向散射信息的影响,进而利用经验、半经验和理论模型进行土壤水分反演研究,结合现场实测数据对不同土壤水分反演方法进行评定分析。论文主要研究成果概括为以下几个方面:(1)全极化SAR数据土壤水分反演研究,提出了结合特征降维的多参数土壤水分反演方法。基于全极化SAR数据提取的雷达后向散射系数和极化参数,结合PCA特征降维方法和最小均方根误差准则,获取适用于不同实验区土壤水分反演的最优特征参数,进而用于估算土壤表面含水量。该方法充分利用了全极化数据散射信息丰富的特点,且有效去除了特征参数之间的信息冗余。(2)多波段SAR数据土壤水分反演研究。基于经验模型发展了结合C波段和X波段SAR数据的裸露农田土壤水分反演方法,利用不同波段SAR数据有效表征影响雷达后向散射信息的土壤表面参数,降低表面粗糙度的影响,结果表明多波段SAR数据的应用能够不同程度地提高土壤水分反演精度。针对植被覆盖对雷达后向散射信息的影响,发展了基于多波段SAR数据的结合水云模型和CIEM模型的土壤水分反演方法,在植被覆盖农田获得了可靠的土壤水分信息。(3)基于理论模型分析了不同传感器SAR数据在裸露地表土壤水分反演中的应用。针对表面粗糙度的影响,提出了基于IEM和CIEM模型的多源SAR数据土壤水分反演方法,有效克服了观测量少于未知参数的“病态”反演问题,在无需先验信息的条件下实现土壤水分的求解,在此基础上引入代价函数分布图评定反演结果的收敛性与唯一性。(4)基于IEM模拟数据分析了雷达后向散射系数差(d B)与土壤表面水分变化之间的关系,对于表面粗糙度恒定的裸露地表,土壤水分变化与雷达后向散射系数变化之间表现高度的相关性,验证了变化检测模型的合理性。在此基础上,利用河北邯郸农田实验区现场实测数据以及时序Radarsat-2和Sentinel-1A影像进行土壤水分变化检测分析,验证了在表面粗糙度变化较小的情况下,雷达后向散射系数差能够有效表征土壤水分的变化,进而可以利用时序SAR数据估算土壤水分变化信息。(5)融合时序Radarsat-2和Sentinel-1A数据的土壤水分反演研究。首先利用模拟数据以及时序SAR影像和实测数据评定验证Alpha模型的合理性,然后基于时序SAR数据利用Alpha比值模型去除表面粗糙度贡献,构建有效的观测方程组,结合土壤水分约束条件求解不同时相土壤含水量。针对观测方程少于未知参数的欠定方程组求解问题,提出了融合多源时序SAR数据的土壤水分反演方法,利用近同步获取的时序Radarsat-2和Sentinel-1A数据构建有效观测方程,将欠定方程组的求解转换为超定方程组求解,提高土壤水分反演的可靠性,结合现场实测数据验证了该方法的有效性。
[Abstract]:Soil moisture is a key variable in the process of the surface, in the global energy balance, play an important role in the research of water cycle and climate change. The soil moisture is the basic condition for the growth of crops, but also an important parameter in crop growth monitoring, evaluation and monitoring of drought. Therefore, dynamic soil moisture estimation and description of hydrology, has an important the significance of ecological agriculture. And the synthetic aperture radar (Synthetic Aperture, Radar, SAR) has all day long, all-weather data acquisition, and radar scattering information on response to sensitive soil dielectric properties, is widely used in surface soil moisture monitoring and inversion study. However, the radar backscatter information by the interaction of multiple factors, including radar system parameters, the soil dielectric constant, surface roughness and vegetation cover, thus realize the accurate inversion of soil moisture to effectively remove soil Effect of surface roughness and vegetation cover factors. This paper based on the multi temporal Radarsat-2, Terra SAR-X and Sentinel-1A images and the corresponding measured data, the inversion of Hebei farmland in Handan region and Ningxia Yanchi semi arid region of soil surface moisture. Multi polarization, multi band and multi temporal SAR data to reduce or remove the surface roughness the degree of influence to the radar scattering information, and then use the experience of soil moisture inversion of semi empirical and theoretical model, combined with the field data of different soil moisture inversion methods were evaluated and analyzed. The main research results are summarized as follows: (1) SAR data on soil moisture inversion method is proposed for full polarization, multi parameter soil moisture inversion combined with dimensionality reduction. Fully polarimetric SAR data extraction of radar backscattering coefficient and polarization parameters based on the combination of PCA Feature reduction method and minimum mean square error criterion, the optimal parameters obtained for different soil moisture inversion of the experimentation area, and then used to estimate the soil surface moisture. The method makes full use of the characteristics of polarimetric scattering information rich data, and effectively removes the redundant information between the characteristic parameters of the soil. (2) SAR data study on multi band moisture inversion. Based on empirical model developed with bare farmland soil moisture inversion C band and X band SAR data, using effective characterization of different band SAR radar data to the soil surface scattering parameter information, reduce the effect of surface roughness, the results show that the application of multi band SAR data to a different extent to improve the accuracy of soil moisture inversion. The effects of vegetation coverage on the radar scattering information, combined with the development of cloud model and C multi band based on SAR data Soil moisture inversion method of IEM model, the vegetation coverage of farmland soil moisture obtained reliable information. (3) the theoretical model to analyze the data of different sensors in the SAR application of bare surface soil moisture retrieval. Based on the influence of surface roughness, the multi SAR data of soil moisture inversion method based on CIEM model and IEM and effectively overcome the measurement of unknown parameters less than the "sick" inversion problem, to solve the soil moisture without prior information, on the basis of introducing the convergence and uniqueness of the cost function evaluation of distribution results of inversion. (4) IEM simulation data analysis of the radar backscatter coefficient difference based on (D B) and the relationship between soil surface moisture changes, the surface roughness of exposed surface of constant degree, change of soil moisture and between radar backscattering coefficient changes showed high correlation, To verify the rationality of the change detection model. Based on the analysis of soil moisture change detection by using the measured data in site of Hebei Handan experimentation area of farmland and the temporal Radarsat-2 and Sentinel-1A images verified in surface roughness change is small, the radar backscatter coefficient difference can effectively characterize the change of soil moisture, and soil moisture estimation change information using time series SAR data. (5) study on soil moisture inversion sequence Radarsat-2 and Sentinel-1A data fusion. Firstly, using the simulation data and the rationality of time series SAR images and data evaluation to validate the Alpha model, and then the timing data using the SAR ratio of Alpha removal model of surface roughness based on contribution, construct the effective observation equations, and the solution soil moisture constraints and soil moisture. The observation equation is less than unknown parameters Underdetermined equations to solve problems, put forward soil moisture inversion method of fusion of multi-source SAR timing data, construct the effective observation equation using the near simultaneous collection of temporal Radarsat-2 and Sentinel-1A data, will owe conversion equations for solving overdetermined equations, improve the reliability of soil moisture inversion, verify the validity of the the method of combining the measured data.

【学位授予单位】：中国矿业大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：S152.7

【相似文献】