Knowledge Resource Center for Ecological Environment in Arid Area
| 黑河流域机载L 波段微波辐射计数据反演土壤水分 及卫星产品精度评估研究 | |
| 其他题名 | Surface Soil Moisture Retrieval Using Airborne L-band Microwave Radiometer data and Evaluation of Satellite-Based Products in Heihe River Basin |
| 王增艳 | |
| 出版年 | 2017 |
| 学位类型 | 博士 |
| 导师 | 王建 |
| 学位授予单位 | 中国科学院大学 |
| 中文摘要 | 土壤水分是联系地球表层物质能量交换的重要纽带，在全球水循环、能量平衡及气候系统中意义重大，也是开展区域生态、水文、农业及旱情灾害等研究的重要参数指标。被动微波遥感因其直接、快速、灵敏度高的特点，在地表土壤水分监测方面具有独特优势，是目前星载、大范围、动态土壤水分监测的主要手段。当前随着微波传感器技术发展与大型航空及地面土壤水分观测试验的开展，被动微波遥感在正向微波辐射传输模型发展、土壤水分反演算法改进、长时间序列土壤水分产品制备等方面均取得了长足进展。但是，受卫星传感器观测空间分辨率限制，当前被动微波遥感土壤水分研究在反演算法中模型参数标定、产品的真实性检验方面仍存在以下问题：1. 已有的土壤水分反演算法大多以简化的微波辐射传输模型为基础，并对模型中若干关键变量，如地表温度、粗糙度、植被光学厚度等，建立了大量经验、半经验模型，应用中需要提前进行标定。但是，直接将基于航空或地面观测试验获得的参数标定结果应用于卫星粗分辨率的亮温观测数据时，会为卫星土壤水分反演结果带来较大误差。目前，对这部分来源误差的分析还缺乏足够的量化。2. 土壤水分是一个时空异质性十分强烈的地表参数，卫星被动微波遥感数据空间分辨率较低，而基于地面站点的实测数据在空间上代表范围较小。由于观测尺度不匹配，已有土壤水分产品在真实性验证方面仍然存在较大困难。3. 在我国，针对最新卫星（增强型微波辐射计2 （Advanced Microwave Scanning Radiometer 2 , AMSR2）、土壤水分海洋盐分卫星（Soil Moisture and Ocean Salinity mission, SMOS）和土壤水分主被动卫星（Soil Moisture Active and Passive mission, SMAP））遥感土壤水分产品的验证工作极少，并且集中在青藏高原地区，在其他区域特别是复杂地表下垫面条件下的验证还不够充分。 鉴于以上因素，本研究选择黑河流域为研究区，以量化和降低被动微波遥感土壤水分反演中参数不确定性，卫星遥感土壤水分产品的真实性验证和误差来源量化分析为研究目标开展了以下几个方面的研究：（1）微波辐射传输模型参数的敏感性量化分析采用一种基于方差的定量全局敏感性分析方法——扩展傅里叶振幅敏感性分析算法 (extended Fourier Amplitude Sensitivity Test, eFAST)，对L波段生物圈微波发射模型（L-band Microwave Emission of the Biosphere, L- MEB）中参数的敏感性进行量化分析，并对不同入射角度、地面粗糙程度及植被覆盖状况下的参数敏感性变化情况进行了分析，探讨了模型的主要敏感因子及其变化。对不同土壤水分条件下的待标定参数的敏感性进行对比，探讨参数标定结果的可靠性。首先，对不同极化、入射角度条件下L-MEB模型中参数敏感性结果的对比分析表明，地表土壤水分、土壤粗糙度因子、植被光学厚度和地表有效温度是L-MEB模型中的四个主要敏感因子，这一结果证明了当前SMOS卫星默认土壤水分反演3参数方法中三个参数的可反演性，特别强调了地表温度对模型输出结果的高敏感性，特别在地表较为粗糙或者地表植被覆盖较厚的情况下。其次，对L-MEB模型中待标定参数的敏感性分析结果显示，土壤粗糙度因子和植被光学厚度可以得到较好的标定结果，而植被结构因子、单次散射反照率和土壤粗糙度系数的敏感性较低，或者随土壤水分变化不稳定，表明这些参数在当前标定方法中可能无法获得令人满意的标定结果，未来可以发展考虑模型参数在不同入射角或者不同土壤水分条件下不同敏感性指数大小的新的参数标定方法，以提高这些参数标定结果的准确性。（2）联合航空及地面多源观测数据的被动微波遥感土壤水分反演算法发展与验证以生态-水文过程综合遥感观测联合(Heihe Watershed Allied Telemetry Experimental Research, HiWTER)试验提供的多源航空及地面观测数据为基础，首先，基于极化L波段多角度微波辐射计(Polarimetric L-band Multi-beam Radiometer，PLMR)的多角度亮温观测数据，发展了一种基于0°入射角亮度温度的单通道土壤水分反演方法，以减少反演过程中因入射角度造成的参数标定误差，获得了研究区6月30日、7月10日和8月2日共3景、高空间分辨率的土壤水分反演结果。其次，通过采用航空同步多尺度地面土壤水分观测数据，对土壤水分反演结果进行了验证。结果显示，反演结果在点尺度上验证的均方根误差（Root Mean Square Error, RMSE）在0.035-0.055m3/m3之间，田间尺度上验证的RMSE略高于点尺度，反演偏差（Bias）低于0.02m3/m3，村社尺度上，反演土壤水分在空间上的准确性较高，与村社距前次灌溉间隔日数间呈明显的负相关关系，相关系数（Correlation Coefficents, R）在0.3左右。时间上，三个反演结果分别对应于研究区玉米生长的前中后三个时期，航空被动微波遥感土壤水分反演精度逐渐下降。（3）卫星遥感土壤水分产品在黑河流域复杂下垫面条件下的精度验证与评估基于黑河流域多尺度水文气象观测网及无线传感器观测网络获得的长时间序列土壤温湿度观测数据和MODIS提供的逐月归一化植被指数（Normalized Difference Vegetation Index, NDVI）数据，对黑河流域复杂下垫面条件下，三种最新卫星传感器（AMSR2、SMOS和SMAP）6种微波遥感土壤水分产品（LPRM-C1、LPRM-C2、LPRM-X、JAXA、SMOS和SMAP）的进行了详细的精度验证和误差来源分析。首先，对卫星遥感土壤水分产品的精度验证表明：在下游荒漠地区，6种土壤水分产品均取得了较高的验证精度，其无偏均方根误差（unbiased RMSE, ubRMSE）均在0.04m3/m3以内，并且除LPRM-X产品外，其余产品均表现出一定的高估趋势。对于植被覆盖地表，SMAP、SMOS和JAXA产品均表现出低估趋势，上游高寒草地的低估程度均高于中游绿洲区。对于上游高寒草地，其平均低估偏差表现为SMAP（0.15 m3/m3） |
| 英文摘要 | Surface soil moisture (SM) is a crucial variable linking the material and energy exchange over land surface. It plays an important role in the global water cycle, energy balance and climate system, and makes a key indicator in the regional researches on ecology, hydrology, agricluture and drought disaster . The passive microwave remote sensing that can directly and rapidly monitoring SM with high sensitivity hasbecome the main tool in satellite SM monitoring. Currently, with the development of microwave remote sensor techonogy and the conductions of large ground- and airborne-based SM observation experiments, great progresses have been witnessed in the improvements of microwave radiative transfer models and SM retriveal algorithms, as well as the development of long-term satellite-based SM products. However, due to the limitation of course spatial resolution in passive microwave observations, problems still exist in the model parameters calibrations and the validitions of satellite-based SM products:I. Most of the SM retrieval algorithms are based on the simplified microwave radiative transfer model. Key parameters in the model, like land surface temperature (Ts), roughness (HR), vegetation optical depth (tNAD), which need to be calibrated before the retrievals, are parameterized empirically- or semi-empirically. Errors are introduced to the course-resolution satellite observation when the calibration results based on the high-resolution ground or airborne observations are used directly. However, these errors have not attracted sufficient attentions and effectively been estimated. II. SM is a variable with strong heterogeneity. Ground obaervations can only represent a very limited scale of land surface, but the spatial resolutions of satellite observations are always on the scale of tens of kilometers. This mismatching in spatial scale is also facing great challenges in the validation of satellite-based SM products. III. The validation of newest SMAP have not been fully conducted in China, while most validations other satellite-based soil moisture products, e.g., AMSR2, SMOS are mainly conducted in Tibetan Plateau. Thus, more valitiaons are indisapensible in other regions and more landcover categories.In this thesis, Heihe River Basin (HRB) in the arid region of northwestern China is selected as the study area. The main objective of this thesis is to quantify and compare the sensitivities of input parameters in the microwave radiative transfer model, and to reduce the uncertainties of input parameters in the existing retrieval algorithm, as well as to estimate the accuracy of the newest satellite-based SM products. Main contents are as follows:(1) Quantification and comparison of the sensitivities of parameters in a microwave radiative transfer modelA global sensitivity analysis method of extended Fourier Amplitude Sensitivity Test (eFAST) is adopted to conduct the quantitive sensitivity analysis of parameters in the L-band Microwave Emission of the Biosphere (L-MEB). The sensitivities are compared under different incidence angles and land surface conditions of roughness and vegetation.First, surface SM, HR, tNAD, and effective land surface temperature (Teff) are the four most sensitive parameters in the L-MEB model, demonstrating their possibility to be retrieved in the multiparameter retrieval approaches. Then, the highly total sensitivity index (TSI) of Teff in the analyses emphasizes the importance of high-precision Ts data in the surface SM retrievals, especially for rougher or densly vegetated surface conditions. Finally, the analysis indicates that TSI values of HR and tNAD are high and TSI values of vegetation structure single scattering albedo (w), and soil roughness coefficient (NR) are relative low at incidence angles near nadir,. This suggests that calibration experiments performed at small incidence angles may be appropriate for some but not all of the model parameters, characterizing the effect of soil surface roughness and vegetation on the terrestrial brightness temperature. As a consequence, new calibration procedures that account for the different sensitivities of these model parameters at larger incidence angles may need to be developed in the future.(2) Improvement and validition of single channel SM retrieval algorithm based on airborne PLMR data.Firstly, through the combinations of multi-sources of remote sensing and ground observations in Heihe Watershed Allied Telemetry Experimental Research (HiWTER) in the middle reaches of HRB, a single channle SM retrieval method based on the Polarimetric L-band Multi-beam Radiometer (PLMR) data at the incidence angle of 0° is developed, and the surface SM are retrieved at the resolution of ~700m on three dates of June 30th, July 10th and Auguest 2th, 2012.Secondly, the validations of the retrieval are conducted on three different scales. Results show that the accuracy of the SM retrievals on the field scale is slightly higher than that on the point scale, with the Root Mean Square Error (RMSE) varying between 0.035-0.055m3/m3 on point scale and the biases lower than 0.02m3/m3 on field scale. The SM retrievals and the villiage-scale irrigation data (days to the last irrigation), are negatively correlated with each other in space, with the R value of about 0.3.(3) Evaluation and quantitative error analysis of the newest satellite-based SM products in HRBA comprehensive evaluation of six passive microwave remotely sensed SM products (AMSR2/LPRM-C1, AMSR2/LPRM-C2, AMSR2/LPRM-X, AMSR2/JAXA-X, SMOS and SMAP) based on three newest satellite is performed using a multi-scaled and long-term ground-based surface soilmoisture observationconsisting of automatic weather stations and the wireless sensor network in HRB. Quantive error analyses are also conducted with the help of long-term soil temperature and MODIS Normalized Difference Vegetation Index (NDVI ) observations.Validation results show that: (1) For the desert in the downstream of HRB, high accuracies are observed for all the six SM products with the unbiased RMSE (ubRMSE) of about 0.04 m3/m3, and all products except for LPRM-X show a trend of overestimation against ground observations. (2) For the vegetated regions, SMAP, SMOS and JAXA products underestimate SM with respect to ground observations. The underestimation values are generally higher for grassland in the upstream region with a dryer Bias of about 0.15 m3/m3 for SMAP, 0.18 m3/m3 for SMOS and 0.26 m3/m3 for JAXA product, and the dryer Bias are lower of about 0.01 m3/m3 for SMAP, 0.05 m3/m3 for SMOS and 0.07 m3/m3 for SMAP product. Although with higher biases for the grassland, the correlation coefficents (R) are generally higher (higher than 0.4) than that for the croplands(lower than 0.3), which may be explained by the strong heterogeneity in the middle stream of HRB. (3) All the three sets of LPRM products show overestimation of SM. The overestimations are always higher in C-band product with the wetter bias of about 0.33 m3/m3 for grassland and 0.1 m3/m3 for cropland, and wetter biases are about 0.18 m3/m3 for grassland and 0.08 m3/m3 for cropland. R values decrease as the frequencies increase. For the grassland, the R value is about 0.36 and R value decreases from about 0.5 for C1 band product to below 0.3 for the X-band product.Error analyses show that: (1) LPRM, SMOS and SMAP products show high accuracy in the input parameter of surface soil temperature, with the accuracy of about -1.08K, -2.99K and -3.69K for the nighttime (1:30am for LPRM, 6:00am for SMOS and SMAP) products, respectively. The R values are always high with R values always higher than 0.75. However, the daytime LPRM product shows a high soil temperature bias of about 6.73K, resulting the worse evaluation results in daytime LPRM SM validation. The constant input soil temperature of 293K in JAXA algorithm may be the main reason resulting in the bad estimation of the its products. For SMOS product, however, the unsatisfactory accuracy may be attributed to the higher Radio Frequency Interference (RFI) in the daytime observations. (2) The validation of vegetation optical depth using MODIS NDVI data shows higher R of about 0.5 for LPRM algorithm, indicating the good simulation of vegetation effectsin the algorithm. However, the R are always low (lower than 0.15) in SMOS algorithm, which may be caused by the initial value setting and iterative method used in the retrieval algorithm.Overall, this study aims to reduce the uncertainties of parameters in the forward models, to improve the SM retrival algorithm, and to conduct accuracy estimation and error analyses of satellite-based SM products in the HRB. Quantitative sensitivity analyses of parameters in a microwave radiative transfer model are conducted, and single channel algorithm is improved and validated, as well as the accuracy estimation and error analyses researches are conducted. Results and conculsions in this study can provide method and idea for algorithm improment, products application and validation of passive microwave remote sensing of SM in the future. |
| 中文关键词 | 被动微波遥感 ; 土壤水分 ; L-MEB模型 ; 敏感性分析 ; 验证 |
| 英文关键词 | Passive Microwave Remote Senging Soil Moisture L-MEB Sensitivity Analysis Validation |
| 语种 | 中文 |
| 国家 | 中国 |
| 来源学科分类 | 地图学与地理信息系统 |
| 来源机构 | 中国科学院西北生态环境资源研究院 |
| 资源类型 | 学位论文 |
| 条目标识符 | http://119.78.100.177/qdio/handle/2XILL650/287958 |
| 推荐引用方式 GB/T 7714 | 王增艳. 黑河流域机载L 波段微波辐射计数据反演土壤水分 及卫星产品精度评估研究[D]. 中国科学院大学,2017. |
| 条目包含的文件 | 条目无相关文件。 | |||||
| 个性服务 |
| 推荐该条目 |
| 保存到收藏夹 |
| 导出为Endnote文件 |
| 谷歌学术 |
| 谷歌学术中相似的文章 |
| [王增艳]的文章 |
| 百度学术 |
| 百度学术中相似的文章 |
| [王增艳]的文章 |
| 必应学术 |
| 必应学术中相似的文章 |
| [王增艳]的文章 |
| 相关权益政策 |
| 暂无数据 |
| 收藏/分享 |
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。
