几种卫星反演土壤湿度在中国地区的对比分析

鉴于土壤湿度是气象学、气候学研究领域中的重要环境因子和过程参数,目前对于各种卫星土壤湿度产品在中国地区的精度验证研究较少。该文在中国地区对ASCAT、SMOS、FY 3种全球范围卫星反演土壤湿度产品进行时空对比,评价三者一致性,并使用站点观测资料(土壤湿度、降水量)对其反演产品进行检验。结果表明,中国区域内ASCAT、SMOS、FY反演土壤湿度空间分布较为一致,但与ASCAT、SMOS相比,FY卫星产品的大值区域分布与另外2种有所不同;与站点土壤湿度数据相比,ASCAT卫星产品在各个地区表现一致,与地面数据相关性较好,SMOS和FY卫星产品在一些区域表现较差;与降水量数据相比,3种土壤湿度产品与降水量的变化趋势较为一致,FY卫星产品在一些地区的一些时间段与其他2种卫星产品数据表现出较大的偏差。     

土壤湿度再分析资料在江苏地区的适用性

针对欧洲中期天气预报中心的ERA-Interim再分析土壤湿度资料在江苏区域的适用性未知的问题,利用23年江苏省农业气象观测站的土壤湿度资料,对ERA-Interim再分析土壤湿度资料在江苏区域的适用性进行评估分析。结果表明:1)ERA-Interim土壤湿度的空间分布与农业气象观测站的资料有较好的一致性,2套数据整体上均呈现"南湿北干"的空间格局;ERA-Interim与农业气象观测站的资料在江苏大部分地区相关系数都达到0.75以上,ERA-Interim能较好地体现该地区的土壤湿度的时间变化特征。2)ERA-Interim季节性波动与观测数据较为一致,体现为春季、秋季较干的变化特征,但冬季明显低估了土壤湿度。3)就年际变化而言,ERA-Interim资料与观测资料在春、夏、秋、冬4个季节具有较好的一致性;冬季最高达0.88,夏季最低到0.75,因此能较好地反映土壤湿度的年际变化特征。     

AO辅助LSSVM的GNSS-IR土壤湿度反演方法

全球导航卫星系统多径干涉遥感技术(global navigation satellite system interferometric reflection, GNSS-IR)已成为目前研究的热点，用其测量的数据可以对土壤湿度值等进行估算。针对当前该方法存在土壤湿度反演精度较低的问题，文章以美国板块边界观测网络(PBO)中p043测站为研究对象，并对该测站的GNSS信噪比数据进行分析，提取L2频段反射信号的延迟相位作为输入，PBO H₂O的土壤湿度值作为输出，构建了基于AO-LSSVM土壤湿度反演模型，并将该模型与BP神经网络和PSO-LSSVM进行对比。实验结果表明，基于AO-LSSVM方法得到的PRN10卫星反演结果与土壤湿度真值之间的决定系数为0.920,均方根误差为0.021,平均绝对误差为0.017,相比BP神经网络和PSO-LSSVM更加贴近土壤湿度真值，证明了利用该方法能够有效提高土壤湿度反演的精度。     

AMSR-E积雪产品在内蒙地区的精度验证

使用地面积雪观测数据对2005年～2008年40°N～48°N、112°E～128°E区域的AMSR-E积雪产品进行了误差分析和精度验证,结果表明:2005年～2008年的AMSR-E积雪产品较好地反映了研究区域地面积雪信息的时间变化特征;AMSR-E积雪产品普遍地低估了地面积雪深度,相对而言,当地面积雪较薄时,AMSR-E可较好地反映积雪深度,当积雪较厚时,AMSR-E明显低估积雪深度;2005年～2006年、2006年～2007年以及2007年～2008年3个冬-春季时段AMSR-E和站点观测值的平均差值分别达7.38cm,6.87cm和22.07cm。     

卷积神经网络提取风云影像土壤湿度

目的 时空分辨率较高的土壤湿度数据对于生产实践和科学研究具有重要意义。以国产的风云气象卫星为数据源,利用卷积神经网络自主学习输入变量间深层关联的优势,获取高质量土壤湿度数据,为科学研究和生产实践服务。方法 首先构建了一个土壤湿度提取卷积神经网络（soil moisture convolutional neural network,SMCNN）,SMCNN由温度子网络和土壤湿度子网络构成,每个子网络均包含特征提取器和编码器。特征提取器用于为每个像素生成一个特征向量,其中温度子网络的特征提取器由11个卷积层组成,湿度子网络的特征提取器由9个卷积层组成,卷积层均使用1×1的卷积核。编码器用于将提取到的特征拟合为目标变量。两个子网络均使用平均方差作为损失函数。使用随机梯度下降算法对模型进行训练,最后利用训练好的模型提取区域土壤湿度数据。结果 选择宁夏回族自治区为实验区,利用获取的2016-2019年风云3D影像和相应地面站点数据作为实验数据,选择线性回归模型、BP（back propagation）神经网络模型作为对比模型开展数据实验,选择均方根误差作为评价指标。实验结果表明,SMCNN的均方根误差为0.006 7,优于对比模型,SMCNN模型在从风云影像中提取土壤湿度方面具有优势。结论 本文利用卷积神经网络分别构建用于反演地表温度和土壤湿度的子网络,再组成一个完整的土壤湿度反演网络结构,从风云3D数据中获取数值精度、时空分辨率均较高的土壤湿度数据,满足了科学研究和生产实践对大范围高精度土壤湿度数据的需求。     

使用温度植被干旱指数法(TVDI)反演新疆土壤湿度

利用MODIS合成产品数据MOD11A2和MOD13A2获取的归一化植被指数(NDVI)和陆地表面温度(Ts)构建Ts-NDVI特征空间,依据该特征空间计算的温度植被干旱指数(TVDI)作为土壤湿度监测指标,反演了新疆8、9两个月份每16 d的土壤湿度。使用野外与卫星同步采样的土壤湿度数据进行验证,发现TVDI指标与实测土壤湿度数据显著相关,能够较好地反映表层土壤湿度,反映的新疆土壤湿度的空间分布与新疆的年降水量分布、年平均相对湿度分布很吻合;同时表明8、9两个月份期间新疆土壤湿度低的区域在不断扩大。     

使用温度植被干旱指数法(TVDI)反演新疆土壤湿度

利用MODIS合成产品数据MOD11A2和MOD13A2获取的归一化植被指数(NDVI)和陆地表面温度(Ts)构建Ts—NDVI特征空间，依据该特征空间计算的温度植被干旱指数(TVDI)作为土壤湿度监测指标，反演了新疆8、9两个月份每16d的土壤湿度。使用野外与卫星同步采样的土壤湿度数据进行验证，发现TVDI指标与实测土壤湿度数据显著相关，能够较好地反映表层土壤湿度，反映的新疆土壤湿度的空间分布与新疆的年降水量分布、年平均相对湿度分布很吻合；同时表明8、9两个月份期间新疆土壤湿度低的区域在不断扩大。     

基于SPOT-VGT可见光/短波红外波段数据对AMSR-E土壤湿度的降尺度研究

基于SPOT-VGT数据,由短波红外、红和蓝波段反射率计算了表征地表土壤湿度的可见光—短波红外干旱指数(VSDI),通过对1km空间分辨率的VSDI影像进行空间升尺度处理,采用多种函数建立了25km空间分辨率AMSR-E土壤湿度数据与VSDI指数的关系,发现二者关系最符合S型曲线模型,拟合残差在空间上呈现随机分布的特征。基于S曲线函数关系下的1km预测土壤湿度和残差值,对AMSR-E土壤湿度进行降尺度模拟,得到1km空间分辨率的土壤湿度。将原始AMSR-E土壤湿度和实测数据对降尺度结果分别比较验证后,表明基于该方法获得的土壤湿度模拟精度较高。     

卫星微波遥感结合可见光遥感估算黄河源区土壤湿度研究

利用欧洲环境卫星(ENVISAT)搭载的高级合成孔径雷达ASAR(Advanced Synthesis Aperture Radar)交叉极化模式(APP)2009年8月9日和10月6日的数据对青藏高原东北部玛曲地区土壤湿度进行了估算。对于裸土区域采用表层微波后向散射几何光学模型GOM(Geometry Optics Model),对于植被覆盖度大的区域利用“水-云”模型处理植被层对后向散射系数的影响,取得了较好的结果:遥感估算的土壤湿度值和地面实测值之间的均方根误差RMSE<0.05,决定系数R²>0.82,表明该方法适合反演玛曲地区的土壤水分。从遥感估算的总体结果可以看出:山谷和陡峭山坡的反演结果相对较差,而在相对平坦的地区反演结果较好,估算的土壤湿度值在0.20~0.50 m³/m³之间。     

微波遥感土壤湿度研究进展

发展实用的微波遥感土壤湿度卫星反演算法,以提供区域尺度上的土壤湿度信息,对水文学、气象学以及农业科学研究与应用至关重要。简要分析了主动微波遥感土壤湿度的研究进展情况,重点对被动微波遥感土壤湿度的原理、算法发展及研究趋势进行了详细论述。主动微波具有卫星数据空间分辨率高的特点,随着携载主动微波器的一系列卫星的发射,主动微波遥感土壤湿度将受到重视。被动微波遥感研究历史长、反演算法特别是卫星遥感算比较成熟,是今后区域尺度乃至全球尺度监测土壤湿度的重要手段。     

An evaluation of AMSR-E derived soil moisture over Australia

This paper assesses remotely sensed near-surface soil moisture over Australia, derived from the passive microwave Advanced Microwave Scanning Radiometer - Earth Observing System (AMSR-E) instrument. Soil moisture fields generated by the AMSR-E soil moisture retrieval algorithm developed at the Vrije Universiteit Amsterdam (VUA) in collaboration with NASA have been used in this study, following a preliminary investigation of several other retrieval algorithms. The VUA-NASA AMSR-E near-surface soil moisture product has been compared to in-situ soil moisture data from 12 locations in the Murrumbidgee and Goulburn Monitoring Networks, both in southeast Australia. Temporally, the AMSR-E soil moisture has a strong association to ground-based soil moisture data, with typical correlations of greater than 0.8 and typical RMSD less than 0.03 vol/vol (for a normalized and filtered AMSR-E timeseries). Continental-scale spatial patterns in the VUA-NASA AMSR-E soil moisture have also been visually examined by comparison to spatial rainfall data. The AMSR-E soil moisture has a strong correspondence to precipitation data across Australia: in the short term, maps of the daily soil moisture anomaly show a clear response to precipitation events, and in the longer term, maps of the annual average soil moisture show the expected strong correspondence to annual average precipitation.     

利用归一化微波差异指数和表面散射模型反演裸露地表土壤水分

在给定土壤质地和粗糙度状况条件下,用AIEM模型模拟AMSR-E的6.925GHz、10.65GHz和18.7GHz频率下不同含水量时土壤表面发射率和土壤温度的关系,分析表明V极化的发射率受土壤温度的影响很小,其变化主要由土壤水分的变化引起。通过计算不同频率组合V极化通道的归一化微波差异指数,并模拟与土壤水分的关系,然后利用这一关系对塔克拉玛干沙漠中部某地的土壤水分进行反演。结果发现用18.7GHz和10.65GHz V极化通道组合的反演值与AMSR-E Level 3土壤水分产品的吻合程度最好。在此基础上分别用3种常见的半经验表面散射模型:Q/H模型、Hp模型和Qp模型,通过计算上述通道组合的NMDI来反演研究区的土壤水分,结果表明利用3种半经验模型得到的反演值之间差异非常小,并且与用AIEM模型计算NMDI时的反演结果吻合较好。     

Validation of SMOS Soil Moisture Products on Different Vegetation Types in Qilian Mountain

Observation data of 34 in-situ stations located in seven main vegetation types were used to evaluate the performance of SMOS soil moisture products in Qilian Mountain,Northwest China.SMOS data were processed to correspond to the observation data,and three indices：R、Biasand RMSE were calculated at both annual and seasonal scales for each observation station.Results show that SMOS products were credible in the study area,but underestimated soil moisture in Qilian Mountain,and failed to achieve the intended accuracy target of 0.04 m3/m3.SMOS performed better in estimating vegetation emission than soil emission,leading to its better performance in areas with higher vegetation coverage.Similarly,SMOS performed better in the humid condition than the arid condition,and also better in areas with smaller soil moisture variability than those with large soil moisture variability.At seasonal scale,SMOS products fitted the observations better in the summer and autumn than the spring.     

SMAP土壤水分产品在淮河流域的适用性评估

选取淮河流域为研究区域,利用2016年6月至2019年5月流域内的313个土壤水分观测站0~10 cm土壤体积含水量数据,使用多种指标分析SMAP卫星（Soil Moisture Active Passive）9 km分辨率土壤水分产品（L2_SM_P_E）精度的空间和时间（年、月、日尺度）特征,并讨论植被、土壤、地形等对精度影响。结果表明：①整体来看,L2_SM_P_E在淮河流域达不到0.04 m³/m³的预期精度,存在湿区高估、干区低估的现象,但可以较好地反映流域土壤水分的空间分布特征,也能较为准确地指示高湿区和低湿区。②L2_SM_P_E的精度存在明显的区域差异和季节差异。冬季精度明显优于其他季节,流域大部分地区的无偏均方根误差（ubRMSE）均接近预期精度,且在流域北部的部分地区、伏牛山区和大别山区达到了预期精度。在春秋季,流域北部和大别山区的精度较高。夏季L2_SM_P_E的可用性较差。③L2_SM_P_E和降水有较好的一致性,对降水的响应比土壤水分观测值敏感。在降水过程中和降水结束后,L2_SM_P_E的误差以随机误差为主;当土壤相对干燥,则以系统性负偏差为主。④L2_SM_P_E的精度与采样点的土壤类型关系并不密切,山地地区的精度要优于其他地区。     

Intercomparison of microwave remote-sensing soil moisture data sets based on distributed eco-hydrological model simulation and in situ measurements over the North China Plain

Intercomparisons of microwave-based soil moisture products from active ASCAT (Advanced Scatterometer) and passive AMSR-E (Advanced Microwave Scanning Radiometer for the Earth Observing System) is conducted based on surface soil moisture (SSM) simulations from the eco-hydrological model, Vegetation Interface Processes (VIP), after it is carefully validated with in situ measurements over the North China Plain. Correlations with VIP SSM simulation are generally satisfactory with average values of 0.71 for ASCAT and 0.47 for AMSR-E during 2007–2009. ASCAT and AMSR-E present unbiased errors of 0.044 and 0.053 m³ m^?3 on average, with respect to model simulation. The empirical orthogonal functions (EOF) analysis results illustrate that AMSR-E provides more consistent SSM spatial structure with VIP than ASCAT; while ASCAT is more capable of capturing SSM temporal dynamics. This is supported by the facts that ASCAT has more consistent expansion coefficients corresponding to primary EOF mode with VIP (R = 0.825, p < 0.1). However, comparison based on SSM anomaly demonstrates that AMSR-E and ASCAT have similar skill in capturing SSM short-term variability. Temporal analysis of SSM anomaly time series shows that AMSR-E provides best performance in autumn, while ASCAT provides lower anomaly bias during highly-vegetated summer with vegetation optical depth of 0.61. Moreover, ASCAT retrieval accuracy is less influenced by vegetation cover, as it is in relatively better agreement with VIP simulation in forest than in other land-use types and exhibits smaller interannual fluctuation than AMSR-E. Identification of the error characteristics of these two microwave soil moisture data sets will be helpful for correctly interpreting the data products and also facilitate optimal specification of the error matrix in data assimilation at a regional scale.     

A comparison study on empirical microwave soil moisture downscaling methods based on the integration of microwave-optical/IR data on the Tibetan Plateau

Soil moisture plays a vital role in land surface energy and the water cycle. Microwave remote sensing is widely used because of the physically based relationship between the land surface emission observed and soil moisture. However, the application of retrieved soil moisture data is restricted by its coarse spatial resolution. To overcome this weakness, downscaling methods should be developed to disaggregate coarse resolution microwave soil moisture data to fine resolution. The traditional method is the microwave-optical/IR synergistic approach, in which land surface temperature, vegetation index, and surface albedo are key parameters. Five purely empirical methods based on the triangle feature are selected in this study. To evaluate their performance on downscaling microwave soil moisture, these methods are applied to the Zoige Plateau in China using the Advanced Microwave Scanning Radiometer on the Earth Observing System (AMSR-E) Land Parameter Retrieval Model (LPRM) soil moisture product and Moderate Resolution Imaging Spectroradiometer (MODIS) optical/IR products. The coarse-resolution AMSR-E LPRM soil moisture data are disaggregated into the high resolution of the MODIS product, and the surface soil moisture measurements of the Maqu soil moisture observation network located in the plateau are used to validate the downscaling results. Results show that (1) the relationship models used in these methods can generally capture the variation in soil moisture, with R² around 0.6, but have a relatively high uncertainty under conditions of high soil moisture; (2) the methods can provide high-resolution soil moisture distribution, but the downscaled soil moisture presents a low level correlation with field measurements at different spatial and temporal scales. This comparative study provides insight into the performance of popular purely empirical downscaling methods on enhancing the spatial resolution of soil moisture on the Tibetan Plateau. Although synergistic methods can improve the spatial resolution of AMSR-E soil moisture data, additional studies are needed to exclude the uncertainty from AMSR-E soil moisture estimation, the low sensitivity of the relationship model under high soil moisture, and the spatial representativeness difference between coarse pixels and point measurement.     

Soil moisture estimation through ASCAT and AMSR-E sensors: An intercomparison and validation study across Europe

Global soil moisture products retrieved from various remote sensing sensors are becoming readily available with a nearly daily temporal resolution. Active and passive microwave sensors are generally considered as the best technologies for retrieving soil moisture from space. The Advanced Microwave Scanning Radiometer for the Earth observing system (AMSR-E) on-board the Aqua satellite and the Advanced SCATterometer (ASCAT) on-board the MetOp (Meteorological Operational) satellite are among the sensors most widely used for soil moisture retrieval in the last years. However, due to differences in the spatial resolution, observation depths and measurement uncertainties, validation of satellite data with in situ observations and/or modelled data is not straightforward. In this study, a comprehensive assessment of the reliability of soil moisture estimations from the ASCAT and AMSR-E sensors is carried out by using observed and modelled soil moisture data over 17 sites located in 4 countries across Europe (Italy, Spain, France and Luxembourg). As regards satellite data, products generated by implementing three different algorithms with AMSR-E data are considered: (i) the Land Parameter Retrieval Model, LPRM, (ii) the standard NASA (National Aeronautics and Space Administration) algorithm, and (iii) the Polarization Ratio Index, PRI. For ASCAT the Vienna University of Technology, TUWIEN, change detection algorithm is employed. An exponential filter is applied to approach root-zone soil moisture. Moreover, two different scaling strategies, based respectively on linear regression correction and Cumulative Density Function (CDF) matching, are employed to remove systematic differences between satellite and site-specific soil moisture data. Results are shown in terms of both relative soil moisture values (i.e., between 0 and 1) and anomalies from the climatological expectation.Among the three soil moisture products derived from AMSR-E sensor data, for most sites the highest correlation with observed and modelled data is found using the LPRM algorithm. Considering relative soil moisture values for an ~ 5 cm soil layer, the TUWIEN ASCAT product outperforms AMSR-E over all sites in France and central Italy while similar results are obtained in all other regions. Specifically, the average correlation coefficient with observed (modelled) data equals to 0.71 (0.74) and 0.62 (0.72) for ASCAT and AMSR-E-LPRM, respectively. Correlation values increase up to 0.81 (0.81) and 0.69 (0.77) for the two satellite products when exponential filtering and CDF matching approaches are applied. On the other hand, considering the anomalies, correlation values decrease but, more significantly, in this case ASCAT outperforms all the other products for all sites except the Spanish ones. Overall, the reliability of all the satellite soil moisture products was found to decrease with increasing vegetation density and to be in good accordance with previous studies. The results provide an overview of the ASCAT and AMSR-E reliability and robustness over different regions in Europe, thereby highlighting advantages and shortcomings for the effective use of these data sets for operational applications such as flood forecasting and numerical weather prediction.     

A soil moisture assimilation scheme based on the microwave Land Emissivity Model and the Community Land Model

Applications of microwave remote-sensing data in land data assimilation are a topic of current interest and importance due to their high temporal and spatial resolution and availability. However, there have been few studies on land surface sub-grid scale heterogeneity and calculating microwave wetland surface emissivity when directly assimilating gridded Advanced Microwave Scanning Radiometer-Earth Observing System (AMSR-E) satellite brightness temperature (BT) data to estimate soil moisture. How to assimilate gridded AMSR-E BT data for land surface model (LSM) grid cells including various land cover types, especially wetland, is worthy of careful study. The ensemble Kalman filter (EnKF) method is able to resolve the non-linearity and discontinuity in forecast and observation operators, and is widely used in land data assimilation. In this study, considering the influences of land surface sub-grid scale heterogeneity, a satellite data simulation scheme based on the National Center for Atmosphere Research (NCAR) Community Land Model version 2.0 (CLM2.0), microwave Land Emissivity Model (LandEM), Shuffled Complex Evolution (SCE-UA) algorithm and AMSR-E BT observation data is presented to simulate AMSR-E BT data and calibrate microwave wetland surface emissivity; then, a soil moisture data assimilation scheme is developed to directly assimilate the gridded AMSR-E BT data, which consists of the CLM2.0, LandEM and EnKF. The experimental results indicate that the calibrated microwave wetland surface emissivities possess excellent transportability, and that the assimilation scheme is practical and can significantly improve soil moisture estimation accuracy. This study provides a promising solution to improve soil moisture estimation accuracy through directly assimilating gridded AMSR-E BT data for various land cover types such as bare soil, vegetation, snow, lake and wetland.     

基于CIMISS的土壤水分质控应用研究

气象资料业务系统MDOS(meteorological data operation system)的质控方法使用通用阈值去质控土壤水分数据,会漏检和误检出部分疑误数据,且疑误结果多以“未通过降水关系检查”为主,降低值班人员的数据审核效率。为了提高土壤水分质控效率,结合本地土壤特性,分析海南土壤水分历史数据,总结适合本地的阈值范围参数,提出基于CIMISS(China integrated meteorological information service system)的MQCSM(multiple quality control method of soil moisture)算法。该算法引入时变检查、持续性检查等检查方法,多重质控原始土壤水分数据,能快速、准确地质控出疑误数据,并分类展示疑误结果于web监控页面工值班人员筛查。业务试用结果表明,对比现有质控方法,该算法能准确、有效地筛查出疑误土壤水分疑误数据,且质控监视平台实时展示疑误结果,提高了值班人员对疑误数据的审核效率。     

Evaluation of AMSR-E soil moisture results using the in-situ data over the Little River Experimental Watershed, Georgia

An operational global soil moisture data product is currently generated from the observations of the Advanced Microwave Scanning Radiometer (AMSR-E) aboard NASA's Aqua satellite using the retrieval procedure described in Njoku and Chan [Njoku, E.G. and Chan, S.K., 2006. Vegetation and surface roughness effects on AMSR-E land observations, remote sensing environment, 100(2), 190-199]. We have generated another soil moisture dataset from the same AMSR-E observed brightness temperature data using the Land Surface Microwave Emission Model (LSMEM) adopting a different estimation method. This paper focuses on a comparison study of soil moisture estimates from the above two methods. The soil moisture data from current AMSR-E product and LSMEM are compared with the in-situ measured soil moisture datasets over the Little River Experimental Watershed (LREW), Georgia, USA for the year 2003. The comparison study was carried out separately for the AMSR-E daytime and night time overpasses. The LSMEM method performed better than the current operational AMSR-E retrieval algorithm in this study. The differences between the AMSR-E and LSMEM results are mostly due to differences in various simplifications and assumptions made for variables in the radiative transfer equations and the soil and vegetation based physical models and the accuracy of the input surface temperature datasets for the LSMEM forward model approach. This study confirms that remote sensing data have the potential to provide useful hydrologic information, but the accuracy of the geophysical parameters could vary depending on the estimation methods. It cannot be concluded from this study whether the soil moisture estimation by the LSMEM approach will perform better in other geographic, climatic or topographic conditions. Nevertheless, this study sheds light on the effects of different approaches for the estimation of geophysical parameters, which may be useful for current and future satellite missions.