土壤水分是地-气能量交换的重要参数，对地表径流，蒸散发和地下深层排水有着重要的影响作用。土壤水分通过大气和地表之间的能量交换来调节地表能量循环。同时土壤水分也是地表碳循环的一个重要决定因子。因此，土壤水分的监测对理解全球水、能量和碳循环的组成和相互作用至关重要。随着遥感技术的发展，遥感反演土壤水分的相关研究和应用已经取得了一系列实质性的进展。可见光和热红外遥感数据的空间分辨率较高，但是数据的获取容易受天气状况的影响而缺失。微波遥感不受天气情况限制，且对土壤水分相对敏感成为了土壤水分反演的有效手段之一。但与热红外相比，目前被动微波遥感空间分辨率较低（25km~40km），仅适用于全球尺度研究，无法满足干旱监测、生态系统建模、水文气候模型分析（＜10km）等区域尺度上的实际需求。因此，如何提高土壤水分空间分辨率是本文研究的重点问题。本文围绕上述问题，针对复杂地表，特别是植被覆盖密集区对土壤水分反演过程中带来的误差和不确定性展开了研究，发展被动微波遥感土壤水分降尺度方法，以期提高获取高精度、高空间分辨率的土壤水分。研究内容主要包括： （1）被动微波遥感土壤水分产品在农林交错带地区的验证与分析：根河是中国东北部复杂地表异质性的代表地区，高精度的被动微波遥感土壤水分产品对后续土壤水分降尺度结果起着决定性的作用。文中选取了6种被动微波土壤水分产品，即AMSR2 四种土壤水分产品：JAXA, LPRM_C波段，LPRM_X波段，QDCA土壤水分产品以及SMOS L3土壤水分产品和SMAP L3土壤水分产品进行了验证。通过时间序列图、统计指标分析等日尺度和月尺度验证方法对6种土壤水分产品进行了比较。结果显示，SMAP土壤水分与实测数据具有较高的相关性，月均值验证中R为0.865，且其在升轨时间的误差值小于0.04m3m-3。其次，SMOS土壤水分比QDCA、LPRM、JAXA的土壤水分精度高，但其在时间序列的变化上噪音相对较大。QDCA和JAXA的土壤水分偏低，且偏差值的范围分别为：-0.053m3m-3~ -0.018m3m-3，-0.101 ~ -0.090 m3m-3。LPRM土壤水分偏高，且高估值范围为0.261~0.576 m3m-3。综合评估指标，与其他土壤水分相比，根河地区SMAP L3土壤水分与实测数据较为接近，其精度相对较好。 （2）针对热红外数据受天气影响造成的时空缺失问题，与被动微波遥感结合，发展了时空连续的高分辨率的土壤水分反演算法：针对光学/热红外数据中地表温度缺失状况，利用AMSR2多通道亮温数据和MODIS LST数据获得了时空连续的地表温度数据，填补了地表温度在时空上的部分缺失信息；针对光学/热红外数据中NDVI饱和点比LAI低的问题，使用GLASS LAI作为植被参数输入代替了NDVI。同时，引入了微波极化差指数，克服了光学数据的易饱和问题；采用时间序列上的逐日回归方法，获得降尺度因子，减小了季节性影响。针对复杂地表状况，本文分别分析了微波极化差比值MPDI与植被、土壤水分之间的关系，表明了MPDI能够同时表征土壤水分与植被信息。同时，在植被指数与地表温度空间特征分布分析中，LAI比NDVI在该研究区更能表征该区域的植被信息状况。基于参数敏感性分析改进了降尺度方法，获得了空间分辨率为1km的土壤水分数据。像元尺度验证结果表明了，降尺度结果与原有SMAP土壤水分相比，精度有一定提高，降尺度结果与实测数据的相关系数提高了0.027~0.059。站点验证表明，由于下午的SMAP L3土壤水分精度高于上午，导致了下午的降尺度精度比上午的降尺度精度高，在农田、草地、灌木、林地，下午时段的降尺度土壤水分数据和实测土壤水分的相关系数分别为：0.665、0.649、0.627、0.535；RMSE分别为0.029m3m-3、0.038 m3m-3、0.041 m3m-3、0.052 m3m-3。 （3）针对地表异质性，利用同极化和交叉极化后向散射系数构建了地表异质校正项，发展了主被动微波遥感高空间分辨率土壤水分算法：本文选择了水云模型和ω-τ模型分别建立了复杂地表综合模拟数据库。利用模拟数据库分析了不同地表下土壤发射率和后向散射系数之间的关系，后向散射系数、后向散射极化比值与土壤水分的关系。基于参数敏感性分析，结合Sentinel-1雷达数据构建了SMAP土壤水分降尺度方法，获得了1km空间分辨率的土壤水分。该方法考虑了地表粗糙度和植被生长对土壤水分反演的影响，引入了地表空间异质性校正项，减小了地表异质性的影响，具有一定的物理基础。验证结果表明，降尺度后的土壤水分数据在空间分布上显示了更多的土壤水分细节。站点验证显示，在农田、草地、灌木、林地，降尺度后的土壤水分数据和实测土壤水分的相关系数分别为：0.689、0.677、0.657、0.557；RMSE分别为0.043m3m-3、0.049 m3m-3、0.051 m3m-3、0.057 m3m-3。像元尺度验证表明，与SMAP土壤水分相比，降尺度后的土壤水分精度有所提高，即相关系数提高了0.028~0.081。 （4）通过两种降尺度结果的空间分布比较、实测站点验证误差统计比较、以及利用Sentinel-1土壤水分数据与SMAP L2_SP土壤水分数据对两种土壤水分降尺度结果进行了比较，评估了两种降尺度方法。结果表明，由于波段的穿透能力有限，低矮植被区土壤水分的反演精度高于林区。主被动结合方法的精度高于热红外与微波结合方法，其在低矮植被区的相关系数分别为0.686~0.756，0.647~0.696；在林区的相关系数分别为0.542~0.650，0.532~0.602。 综上所述，本文围绕复杂地表下土壤水分反演中面临的空间分辨率问题，基于主被动微波遥感、热红外遥感反演土壤水分的特点，发展了两种被动微波遥感土壤水分降尺度方法，提高了土壤水分空间分辨率，为农业、生态、水文等应用提供了更好的数据基础。

Soil moisture is a land state variable that controls several water-cycle fluxes, namely runoff, evapotranspiration, and deep drainage. In addition, soil moisture modulates the energy cycle through exchanges of energy between the atmosphere and the land surface. Soil moisture is also key determinants of the carbon exchange at the land surface. Thus, global measurements of soil moisture are vital to understanding the components and interactions between the global water, energy, and carbon cycles. With the development of remote sensing technology, the research and application of remote sensing for soil moisture retrieval have made a series of substantial progress. The visible/infrared remote sensing have a high spatial resolution, but it easily limited by weather conditions. Microwave remote sensing, with its sensitivity to soil moisture and ability to collect soil moisture information under all weather conditions, during the day or night, has become an effective method of monitoring regional and global soil moisture. However, passive microwave remote sensing have a low spatial resolution (25km~40km), which is applicable to global scale research and cannot meet the actual needs of regional standards（＜10km）such as drought monitoring, ecosystem modeling, and hydro-climatic model analysis. It is difficult to meet application requirements on the regional scale. Therefore, the focus of this paper is on how to improve the spatial resolution of soil moisture. This paper develops two soil moisture downscaling methods based on the analysis of the uncertainty and errors of soil moisture retrieval with dense vegatetion in complex surface conditions. The contents of this paper mainly include the following parts: (1) Evaluation and analysis of passive microwave remote sensing soil moisture product in the argo-purlieu ecotone. Genhe area provides representative coverage of the complex land surface hydrometeorological conditions in northeastern China. High-precision soil moisture products determines the subsequent downscaling soil moisture results. In this part, we evaluated the Soil Moisture and Ocean Salinity (SMOS) L3 product, the Soil Moisture Active Passive (SMAP) L3 product, and four soil moisture products derived from the Advanced Microwave Scanning Radiometer 2 (AMSR2), i.e., the Dual Channel Algorithm based on the Qp model (QDCA) product, the Japan Aerospace Exploration Agency (JAXA) L3 product, and the Land Parameter Retrieval Model (LPRM) C band and X band products in the Genhe area of China. The results indicated that SMAP soil moisture has a high correlation with the in situ data, R is 0.865, and RMSE is less than 0.04m3m-3 at the ascending orbit. Secondly, SMOS soil moisture has higher accuracy than QDCA, LPRM and JAXA, but its have a big noise in the time series. The soil moisture of QDCA and JAXA is underestimate the in situ data, and the range of the Bias range are -0.053m3m-3~ -0.018m3m-3, -0.101 ~ -0.090 m3m-3. The LPRM soil moisture is overestimate the in situ data and the Bias range is 0.261~0.576 m3m-3. Compared with other soil moisture, the SMAP L3 soil moisture is closest to the ground measurements and have a low RMSE and bias. (2) Aiming at the space-time missing problem caused by the influence of weather on thermal infrared data, we combined with passive microwave remote sensing, and developed a high-resolution soil moisture retrieval algorithm. In order to fill in MODIS land surface temperature (LST) missing data, we adopt multiple regression approach based on AMSR2 multi-channel brightness temperature and MODIS/Terra and Aqua LST data to obtain space-time continuum LST data Because of the saturation point of optical data NDVI is lower than LAI, we use the GLASS LAI instead of NDVI. At the same time, we introduces the microwave polarization difference index (MPDI) to overcome the problem of saturation of optical data. Moreover, we use the daily regression method on time series to reduce the impact of the seasonal. This paper analyzes the relationship between the MPDI and vegetation index，the MPDI and soil moisture, It found that MPDI could represent the informaion of soil moisture and vegetation. In addition, we analyzes the spatial distribution of vegetation index and LST in the complex land surface, found that LAI is more suitable than NDVI for represent the vegetation information in this area. Based on the parameter sensitivity analysis, this part modified the downscaling method. The results show that compared with the SMAP soil moisture, the accuracy of the downscaling results is improved, and the R between the downscaling results and the in situ data is improved by 0.027~0.059. Meanwhile, the accuracy of downscaling in the afternoon is higher than that in the morning because of the accuracy of SMAP L3 soil moisture in the afternoon is higher than that in the morning. The R between downscaling soil moisture data and the in situ soil moisture in the afternoon with the cropland, grassland, shrubs and forest are 0.665, 0.649, 0.627, 0.535; RMSE are 0.029m3m-3, 0.038 m3m-3, 0.041 m3m-3, 0.052 m3m-3, respectively. (3) Aiming at the surface heterogeneity, we use the VH and VV polarization-backscattering coefficient to constructe the term of surface heterogeneity correction, and develop the high spatial resolution of soil moisture retrieval algorithm. This paper use the water cloud model and the model to simulate the surface backscattering coefficient and surface emissivity at the different surface conditions, and construct a simulation database. Based on the simulation database, we analyzes the relationship between surface emissivity and backscattering coefficient, the relationship between soil moisture and backscattering coefficient, the relationship between backscattering polarization ratio and soil moisture, respectively. With the results of parameter analysis, we proposed a downscaling method. This method use the Sentinel-1 radar data to downscale SMAP soil moisture products, and obtained 1km soil moisture data in Genhe area. It considers the influence of surface roughness and vegetation growth, and add the term of spatial heterogeneity correction to reduce the influence of surface heterogeneity. It has a certain physical basis. The results show that the downscaling soil moisture can present more details in the spatial distribution. The R between the downscaling soil moisture and in situ soil moisture in cropland, grassland, shrub and forest are 0.689, 0.677, 0.657, and 0.557, respectively; RMSE are 0.043m3m-3, 0.049m3m-3, 0.051 m3m-3, 0.057 m3m-3, respectively. Compared with SMAP L3 soil moisture, the accuracy of downscaling soil moisturei is improved, and R is increased by 0.028~0.081. (4) This paper use the spatial distribution of soil moisture (36km, 1km), in situ data, Sentinel-1 soil moisture data and SMAP L2_SP soil moisture data to evaluate the results of the two downscaling methods. The results show that the accuracy of soil moisture in low vegetation is higher than the forest areas because of the penetration of band's limited by land type. Moreover, the accuracy of the combination of active-passive method is higher than the combination of thermal infrared and microwave method. The R in the low vegetation area with the two menthods are 0.686~0.756，0.647~0.696, respectively. The R in the forest area are 0.542~0.650，0.532~0.602, respectively. In a word, this paper developed two methods for downscaling soil moisture based on the advantages of soil moisture retrieval with passive microwave remote sensing, visible/infrared remote sensing and active microwave remote sensing. These two methods are mainly to solve the problem of spatial resolution in retrieval soil moisture with the passive microwave remote sensing. The results show that these two methods achieve high spatial resolution (1km) soil moisture and provide a better data foundation for agricultural, ecological, hydrological and other applications.