【作者】 杨虎；

【导师】 郭华东； 施建成；

【作者基本信息】 中国科学院研究生院（遥感应用研究所） ， 地图学与地理信息系统， 2003， 博士

【摘要】 在许多水文和气候模型应用中，地表土壤水分含量(～5cm深度)的时空分布信息十分重要，如降雨制图，干旱模式监测，植被需水分析等。表面土壤水还可作为土壤水模型的输入参数，用以预测植被根系的土壤水分含量。(Hymer et al．，2000)。在干旱半干旱地区，监测地表土壤水分的时空变化特性对理解土壤—植被相互作用过程，提高土壤和植被的有效利用率尤为必要。目前，在要求精度范围内获取大范围地表土壤水分时空分布信息仍是一个迫切需要解决的问题。传统的测量方法，如重量法测量和时域反射计都是基于点的测量方法，需要实地操作和烦杂的后处理过程，而且无法在要求的时间精度范围内得到地表土壤水分的空间分布信息。 研究证明，星载合成孔径雷达(SAR)得到的地表后向散射系数与地表介电常数有直接相关关系，从而能够在水文模型要求的精度范围内有效提取地表土壤水分信息。但由于电磁波与地表相互作用的复杂性，雷达后向散射系数除受地表介电常数(土壤水分)影响外，还受到地表粗糙度、土壤类型、植被覆盖以及雷达入射角、频率、极化等多种因素的影响。特别是在植被覆盖地表，对其下土壤水分的监测更带有极大的困难性。因此，利用雷达后向散射系数反演土壤水分必须首先充分理解微波与地表的相互作用过程。此外，目前星载合成孔径雷达如ERS-1／2、Radarsat等均为单一频率、单一极化的雷达，无法从得到的单参数雷达后向散射系数中直接提取地表土壤水分信息。 本研究中，首先利用基于微波辐射传输方程的微波植被模型和积分方程(IEM)模型模拟了各种地表土壤水分含量情况下，植被覆盖、地表粗糙度(包括地表均方根高度和相关长度)、雷达入射角对C波段(频率4.7Ghz)水平极化(HH)雷达后向散射系数的影响，在此基础上，建立模型消除了植被覆盖、地表粗糙度、及雷达入射角对雷达后向散射的影响，利用多时相50m分辨率Radarsat ScanSAR雷达后向散射系数图像反演得到了地表土壤水分变化模式信息。通过与实测地表土壤水分含量对比，反演结果均方根误差(RMSE)为0.44。本论文研究工作取得的创新性研究成果主要有以下几方面： 1)．利用最新发展的电磁波散射模型研究了不同植被覆盖地表雷达波对地 表土壤水分的敏感性，建立了半经验植被雷达后向散射模型;2).研究发现在农作物等矮小植被覆盖地表，植被层直接后向散射与植被 类型相关，且在植被生长期，雷达后向散射系数对植被含水量的敏感 性要高于对植被高度变化的敏感性;3).解决了单参数雷达地表土壤水分反演问题中，雷达入射角和地表粗糙 度的影响这一难点问题;4).利用土壤介电模型校正了不同土壤类型对反演地表土壤体积含水量的 影响;5).在以上成果基础上，建立了完整的单参数雷达地表土壤水分变化探测 反演算法，经地表验证，模型反演地表土壤水分变化值的精度为RMSE =0 .44;6).将建立的反演模型应用于多时相50m分辨率Radarsat ScanSAR雷达 图像，得到了相应分辨率地表土壤变化值。通过对比分析，发现模型 反演得到的地表土壤水分变化信息与相应的地表降雨、植被、土壤类 型具有一定的相关性。更多还原

【Abstract】 Monitoring soil moisture dynamics is very important for understanding soil-vegetation interactions in both space and time. The effects of the surface roughness and vegetation covers are well-understood problems in estimating soil moisture with SAR imagery. Retrieving soil moisture information with radar measurements could be is achievable by using the multi-frequency and/or multi-polarization measurements to separate the vegetation and surface roughness effects. The currently available satellites, however, are single polarization, single frequency sensors such as ERS-1/2, Radarsat, and JERS-1. There is a need to develop a technique to estimate soil moisture information from these available data sources at both regional and local scales.In this study, we demonstrate a technique using the multi-temporal C band HH polarized Radarsat SCANSAR data to estimate the relative soil moisture change. The experiment data from SGP97 covered a whole range of vegetation growing season and different type agriculture fields. This technique is mainly involved two steps:1) Vegetation effects correction: We used NDVI (Normalized Difference Vegetation Index) derived from TM and AVHRR measurements for spatial andtemporal variations of vegetation covers at different scales. Using a simple radiative transfer model for vegetation volume scattering and the Integral Equation Model (IBM) for surface scattering with the field in situ measurements as the input, we compared the simulated and SAR measured backscattering coefficients in different agricultural fields. We, then, parameterized a semi-empirical model for the different land surface cover types. This semi-empirical model was applied to minimize the effects of the vegetation volume scattering and extinction in radar measurements.2) Radar incidence angle and surface roughness correction: To make radar incidence correction and eliminate the surface roughness effects, a wide range of surface parameters (soil moisture, surface RMS height, correlation length, incidence angle) was input to the IBM model to simulate the effect of surface roughness and radar incidence angle on the sensitivity of soil moisture to the radar backscattering coefficient. A simple model was established to simulate the effects of incidence angle and surface roughness.3) Establishment of soil moisture change inversion model: According to a modified IBM model simulation results, the bare surface backscattering coefficients can be expressed as a funtion of the dielectric component for a given surface roughness when the surface slope greater than 2.0, which is valid for most nature surface:in above equation, R0 is the surface reflectivity at normal incidence. A( 9 ,sr) is a function of surface roughness and Radar incident angle, and B is only influenced by incident angle. IBM simulation results show that in our analysis incident angle range from 20?to 40? the parameters is almost kept constant, its value is from 1.59-1.61. for parameter A, there is a close relationship exist between A( 9 ,sr) in two different Radar incident angle that can be expressed as:with considering the effects of soil texture, we get the final expression of the inversion model:where mv(t1) , mv(t2) is volumetric soil moisture content in two different temp, c,dis soil type related parameters, and v(t1), S(t2) is coresponding bare soil radarbackscattering coefficients.Inversion results show that for the C band HH polarized Radarsat SCANSAR data with a range of incidence angle from 20 to 40 , the soil moisture change value can be derived with an acceptable accuracy using the above model. The temporal and spatial soil moisture change patterns are associated with rainfall and vegetation cover, as well as the soil hydraulic characteristics.更多还原