GA-PLS方法提取土壤水盐光谱特征的精度分析

光谱定量遥感已成为土壤盐渍化大尺度调查的有效手段之一,但黄河三角洲地区盐渍化土壤的光谱响应特征尚未明确。以黄河三角洲野外测定土壤体积含水率、电导率为例,应用遗传偏最小二乘法(GA-PLS)在小样本集条件下提取盐渍土壤的水分-盐分的光谱响应特征,利用蒙特卡罗方法随机模拟结果表明:在不同土壤水盐含量条件下,GA-PLS方法所提取的光谱特征具有鲁棒性,含水率模型稳定在23个波段变量,即响应特征为365~425,500~515,720~740,755~765与955~965 nm;土壤电导率模型的特征集数目为20个波段变量,特征为370~385,405~425\,500~535,650~660,755~760与1 030~1 050 nm。实验在不同预处理模型下,GA-PLS算法所建立水盐光谱模型较PLSR模型均显示出更高的精度。其中,包络线预处理方法与GA-PLS算法相结合效果最优,其水分光谱模型测试集拟合精度(R²),预测残差平方和(PRSS)与残差预测方差(RPD)分别为0.88,9.36与15.80;土壤光谱模型测试集精度R²,PRSS与RPD分别为0.71,15.68与13.76。     

基于Sentinel-1与FY-3C数据反演植被覆盖地表土壤水分

基于新一代的Sentinel-1SAR数据与FY-3C的MWRI数据,研究植被覆盖地表土壤湿度反演方法。为消除植被对土壤湿度反演影响,首先利用FY-3C/MWRI的微波极化差异指数MPDI,建立植被含水量反演模型;然后,结合植被含水量反演模型和水—云模型,发展一种主被动微波联合反演植被覆盖地表土壤含水量模型;最后,在江淮地区开展反演试验,利用观测的土壤湿度数据进行反演结果的精度验证。结果表明:(1)对于植被覆盖地表土壤湿度反演,由FY3C/MWRI提取的MPDI对于去除植被影响效果较好;(2)相比于VH极化哨兵1号卫星数据,VV极化数据更适用于土壤含水量的反演,能够得到更高的土壤湿度反演精度;(3)哨兵1号卫星数据能够获得较高精度的土壤含水量反演结果,试验反演的土壤湿度值与实测值相关系数为0.561 2,均方根误差为0.044cm~3/cm~3。     

基于多源遥感数据协同反演森林地表土壤水分研究

土壤水分在土壤监测中是一项重要的指标,对于农业生产、生态环境以及水资源管理有着重要的影响。随着遥感建模与反演理论的不断成熟,其逐渐成为分析土壤指标的重要技术与手段。因此,利用光学影像与雷达影像数据,以大兴安岭地区漠河市为研究区域,分别建立以Landsat 8为数据源的土壤水分反演模型和由Landsat 8影像数据与GF-3卫星数据协同反演的土壤水分反演模型,将反演结果与实际测得数据进行对比验证,并评价所建立的反演模型。结果表明：①对研究区地温进行反演,利用地表温度（Ts）与归一化差异湿度指数NDMI构建Ts-NDMI特征空间,结合实测数据可以发现Ts-NDMI特征空间土壤水分反演模型的反演结果与实测土壤含水量为负相关性;②协同GF-3卫星数据和Landsat 8遥感影像数据所建立的土壤水分反演模型能得到质量较高的反演结果,且在高植被覆盖度地区,利用该协同反演模型得到的反演结果比利用单一光学数据源所建模型得到的反演结果精度高,为今后高植被覆盖度地区土壤湿度的研究提供了新途径。     

干旱/半干旱区MODIS地表温度反演与验证研究

劈窗算法是目前热红外遥感反演地表温度最常用的方法,根据Coll提出的劈窗算法建立基于MODIS适用干旱/半干旱区地表温度反演算法,并用同期的LP DAAC发布的MODIS地表温度产品和相应的53个气象站点的实际观测数据进行验证。通过分析,模型的反演精度与MODIS地表温度产品的反演精度相当,与气象观测数据相一致,反演精度较好,能够较精确地反演干旱/半干旱地区地表温度的时空变化特征。     

三种秸秆有机肥改良土壤次生盐渍化的效果及生物效应

通过盆栽实验的方法,研究了三种秸秆有机肥对次生盐渍化土壤的EC值、硝酸根离子含量和油菜生物量的影响。结果表明,M1有机肥显著增加油菜的生物量,但在一定程度上增加土壤盐分含量;M2有机肥的效果与M1有机肥效果趋势相似,但增加土壤中的盐分含量的程度要小的多;而施用秸秆肥料M3,在增加油菜生物量的同时显著降低土壤盐分含量,减轻了土壤的次生盐渍化程度。结果表明,三种秸秆有机肥均能提高油菜对盐渍化的耐受能力,增加其生物量。采用M3秸秆有机肥改良土壤盐渍化是一条有效途径,但其最佳施用量与土壤盐渍化程度有关,非盐渍化土壤施用过多M3肥料将降低生物量。     

基于Landsat ETM+遥感数据的组分温度反演方法研究

地表温度在干旱监测和模拟地表热通量中有重要作用。在干旱半干旱地区,双源能量平衡模型(TSEB)通常用于计算地表的热通量。以黑河中游典型灌区为研究区域,选取4个时相的Landsat-7 ETM+遥感影像,通过植被指数与TSEB模型结合的方法反演土壤表面温度和植被冠层温度,并重点讨论土壤表面温度和植被冠层温度的分解算法。结果表明:土壤表面温度和植被冠层温度具有较好的时空一致性;土壤表面温度与植被冠层温度的反演精度通过地表净辐射与地表热通量得到了间接验证。地表净辐射与地表热通量的计算值与观测值相关性好,相关系数大于0.92。地表净辐射与地表热通量的线性回归分析表明拟合精度高。通过地表温度分解的方法获得的土壤表面温度和植被冠层温度,对监测典型区域的干旱和模拟地表热通量是可行的。     

基于MODIS影像的中巴经济走廊荒漠化程度时空动态监测研究

荒漠化是全球最为严重的生态环境问题之一,中巴经济走廊荒漠化问题尤为严重,干旱和大面积的荒漠是其主要的生态环境约束因素。以MODIS数据为基础,提取关键的地表特征参量,定量化研究荒漠化程度与地表特征参量间的关系与规律;构建了基于地表反照率-植被特征空间、决策树的遥感监测模型,并以2015年数据为例,分析了中巴经济走廊荒漠化程度,结果表明：Albedo-MSAVI、Albedo-NDVI和决策树C5.0共3种方法的总体精度分别为88.33%、85.83%和89.2%,Kappa系数分别为0.836 3、0.802 3和0.847 1,分析认为决策树方法最适宜反演中巴经济走廊荒漠化程度。最后基于决策树方法计算了2000~2015年中巴经济走廊荒漠化程度分布数据,分析结果表明在中巴经济走廊极度和重度荒漠化土地占整个区域的50%~60%,中度和轻度荒漠化土地占20%左右,非荒漠化土地和冰雪水体占20%左右。由于2000年左右,巴基斯坦经历了50 a来最严重的旱灾,2000年的重度和极度荒漠化达到总体面积的61.8%,从2005~2015年极度荒漠化土地有所减少,转化为重度荒漠化土地,有部分轻度荒漠化土地转化为非荒漠化土地。总体来说极度荒漠化程度呈下降趋势。     

滨海盐渍土可见近红外高光谱特征

土壤光谱是遥感监测的物理基础,盐渍化土壤光谱特征研究对于盐渍化土地的监测有着重要的意义。以黄河三角洲地区的滨海盐渍土为研究对象,通过野外土壤样品采集和高光谱测量,研究在去除水分以及剔除土壤质地影响后,不同盐渍化程度的滨海盐渍土在350~1 100 nm区间的高光谱反射和吸收特征,并且试图构建光谱预测模型。结果表明: 平滑后的光谱曲线能更准确有效地描述光谱的反射特征及吸收峰。不同盐化程度的土壤光谱曲线形态一致,但反射率大小差异较大。连续统去除后,490 nm处轻度盐化土吸收最小,在760~920 nm区间重度盐化土的吸收更强烈。原始光谱不能预测土壤盐渍化信息,但是二阶微分变换能够提高波段敏感性,建立的光谱预测模型能够基本满足预测要求。     

土壤水盐介电模型对比与分析

土壤介电常数是微波遥感反演土壤水分和盐分的基础,是微波遥感研究的主要参数之一,选用模拟精度较高的土壤水盐介电模型对提高土壤水分和盐分反演精度具有重要意义。目前,土壤介电模型无法定量描述盐分对土壤介电常数的影响。采用Dobson模型及考虑盐分影响的Dobson-S模型、GRMDM模型、HQR模型和WYR模型分别模拟了土壤温度为25℃时不同土壤质地、含水量和含盐量土样在L、C、X波段的复介电常数实部与虚部,并将模拟结果与微波网络分析仪测量值进行对比分析,得到以下结论:(1)Dobson模型和GRMDM模型可较好地实现非盐渍土介电常数实部的模拟,而低频波段虚部的模拟值小于测量值;(2)Dobson-S模型对盐渍土介电常数实部的模拟精度较高,在L、C、X 3个波段相关系数(R)均为0.97,均方根误差(RMSE)小于2.10;但对于盐渍土介电常数虚部,在土壤体积含水量不同的情况下,Dobson-S模型、HQR模型和WYR模型的模拟精度不同。研究结果对选取适当的土壤介电模型反演土壤水分和盐分具有一定的参考价值。     

滨海盐渍土可见近红外高光谱特征

土壤光谱是遥感监测的物理基础,盐渍化土壤光谱特征研究对于盐渍化土地的监测有着重要的意义。以黄河三角洲地区的滨海盐渍土为研究对象,通过野外土壤样品采集和高光谱测量,研究在去除水分以及剔除土壤质地影响后,不同盐渍化程度的滨海盐渍土在350～1 100 nm区间的高光谱反射和吸收特征,并且试图构建光谱预测模型。结果表明:平滑后的光谱曲线能更准确有效地描述光谱的反射特征及吸收峰。不同盐化程度的土壤光谱曲线形态一致,但反射率大小差异较大。连续统去除后,490 nm处轻度盐化土吸收最小,在760～920 nm区间重度盐化土的吸收更强烈。原始光谱不能预测土壤盐渍化信息,但是二阶微分变换能够提高波段敏感性,建立的光谱预测模型能够基本满足预测要求。     

Remote Sensing Extraction of Soil Salinity in Yellow River Delta Kenli County based on Feature Space

Soil salinization is an important challenge to achieve sustainable use of land resources. The appropriate method for remote sensing quantitative inversion in the coastal Yellow River Delta region of China can provide technical reference for regional salinization monitoring and prevention. Utilizing Landsat 8 OLI image and field measured data, we extracted key surface characteristic parameters, quantitatively discussed the law and relationship between soil salinity and surface biophysical parameters and established a soil salinity inversion model. The results show that the inversion precisions of Albedo-MSAVI, SI-Albedo and SI-NDVI feature space are 83.4%, 88.8% and 80.6% respectively. The analysis shows the SI-Albedo model is suitable for the inversion of salinization level in Binhai areas. For Albedo-MSAVI and SI-NDVI models, they have certain reference significance for salinization information extraction in inland arid and semi-arid areas. Based on the inversion of the SI-Albedo feature space with the highest accuracy, the level of salinization in Kenli County is generally high-low-high trends from the east to the west, which is consistent with the formation mechanism of salt accumulation in this area.     

吉林西部盐碱地特性微波双频差分无损探测

针对吉林西部盐碱地特性(内地苏打盐类型),应用车载双频段被动微波遥感系统,对不同盐碱状态的4个裸盐碱区进行双极化多角度微波辐射无损探测。基于多角度双频率双极化的观测数据优势,选择逼近式迭代算法来反演其介电常数虚部,Dobson模型反演介电常数实部。在此基础上,应用双频差分法研究了该区域盐碱地介电特性与含水量、含盐量的关系。双频差分结果表明:实部双频差分与盐碱土含水量呈线性关系,相关系数为0.9996;虚部双频差分与含盐量呈线性关系,相关系数为0.9977。这为应用被动微波遥感定量反演盐碱地特性(含水量、含盐量)奠定了理论基础。     

Generating soil electrical conductivity maps at regional level by integrating measurements on the ground and remote sensing data

In arid and semi‐arid areas, salinization of soil and water resources is one of the major threats to irrigated agriculture. For management purposes, quantifying both the extent and distribution of salinization is important, but accurate data with sufficient spatial resolution are often not available. Commonly used techniques such as soil sampling and geophysical methods are time‐consuming and yield only point data. A method is described in which multispectral remote sensing images can be used to regionalize point data measured on the field. Field data consist of measurements of electrical conductivity and are obtained by the combination of geophysical methods and the analysis of field soil samples. Uncalibrated salinity maps were calculated with spectral correlation mapping using image‐based reference spectra of saline areas. As an alternative indicator for soil salinity, the NDVI was used. The method was verified in the Yanqi Basin, northwestern China. Correlations between field data and the uncalibrated salinity maps were found over non‐irrigated sites for all images. Good correlations (R ² up to 0.85) resulted for images collected during the winter months. The high correlation coefficients allow the uncalibrated salinity maps to be scaled to electrical conductivity maps.     

Remote sensing of soil salinity: potentials and constraints

Soil salinity caused by natural or human-induced processes is a major environmental hazard. The global extent of primary salt-affected soils is about 955 M ha, while secondary salinization affects some 77 M ha, with 58% of these in irrigated areas. Nearly 20% of all irrigated land is salt-affected, and this proportion tends to increase in spite of considerable efforts dedicated to land reclamation. This requires careful monitoring of the soil salinity status and variation to curb degradation trends, and secure sustainable land use and management. Multitemporal optical and microwave remote sensing can significantly contribute to detecting temporal changes of salt-related surface features. Airborne geophysics and ground-based electromagnetic induction meters, combined with ground data, have shown potential for mapping depth of salinity occurrence. This paper reviews various sensors (e.g. aerial photographs, satellite- and airborne multispectral sensors, microwave sensors, video imagery, airborne geophysics, hyperspectral sensors, and electromagnetic induction meters) and approaches used for remote identification and mapping of salt-affected areas. Constraints on the use of remote sensing data for mapping salt-affected areas are shown related to the spectral behaviour of salt types, spatial distribution of salts on the terrain surface, temporal changes on salinity, interference of vegetation, and spectral confusions with other terrain surfaces.As raw remote sensing data need substantial transformation for proper feature recognition and mapping, techniques such as spectral unmixing, maximum likelihood classification, fuzzy classification, band ratioing, principal components analysis, and correlation equations are discussed. Lastly, the paper presents modelling of temporal and spatial changes of salinity using combined approaches that incorporate different data fusion and data integration techniques.     

基于Landsat影像和不规则梯形方法遥感反演延安城市森林表层土壤水分

反演模型对土壤水分评估结果有重要影响,基于此,以黄土沟壑区城市森林表层土壤为研究对象,以3期Landsat影像和实地土壤水分传感器测定数据为数据源,分别通过像元在二维空间（LST-NDVI与STR-NDVI,LST为地表温度,NDVI为归一化植被指数,STR为短波红外转换反射系数）中的散点图及其拟合的干燥边界与湿润边界,获取TOTRAM（热学—光学不规则梯形模型）和OPTRAM（光学不规则梯形模型）的参数,然后在像素水平上（30 m×30 m）反演出延安城市森林表层土壤水分（W）,验证两模型的精度,并比较两模型估算结果的差异及线性边界与非线性边界对反演结果的影响。结果发现：①除OPTRAM 模型在Landsat 7和Landsat 8上干湿边界呈现非线性外,像素在LST-NDVI空间和STR—NDVI空间中的干湿边界均呈线性,且包络成不规则梯形形状;②与实地测定数据相比,TOTRAM与OPTRAM两模型的平均误差（ME）分别为0.009和0.0455,表明两模型估算结果均偏高,但OPTRAM模型的均方根误差（RMSE）较TOTRAM模型更接近0。OPTRAM模型估算的W值均匀地分布在1∶1参考线两侧,且位于参考线上的点数多于TOTRAM模型,表明OPTRAM准确度高于TOTRAM模型,且非线性边界的反演精度高于线性边界;③与TOTRAM模型相比,OPTRAM模型估算出的W空间分异规律与土地利用/覆被类型具有较高的相关性,且OPTRAM模型对植被覆盖度极低的区域敏感。因此,在后续研究中,应在OPTRAM模型中探讨干湿边界复杂性与模型准确性改善之间的关系,同时考虑周围环境、降雨量、森林干扰和NDVI饱和等因素对两模型估算准确性的影响。     

Comparing Sentinel-2 MSI and Landsat 8 OLI in soil salinity detection: a case study of agricultural lands in coastal North Carolina

This study presents the first comparison of Landsat 8 Operational Land Imager (OLI) and Sentinel-2 Multispectral Instrument (MSI) in identifying soil salinity using soil physiochemical, spectral, statistical, and image analysis techniques. By the end of the century, intermediate sea level rise scenarios project approximately 1.3 meters of sea level rise along the coast of the southeastern United States. One of the most vulnerable areas is Hyde County, North Carolina, where 1140 km² of agricultural lands are being salinized, endangering 4,200 people and $40 million USD of property. To determine the best multispectral sensor to map the extent of salinization, this study compared the feasibility of OLI and MSI to estimate electrical conductivity (EC). The EC of field samples were correlated with handheld spectrometer spectra resampled into multispectral sensor bands. Using an iterative ordinary least squares regression, it was found that EC was sensitive to OLI bands 2 (452 nm – 512 nm) and 4 (636 nm – 673 nm) and MSI bands 2 (457.5 nm – 522.5 nm) and 4 (650 nm – 680 nm). Respectively, the R²_Adj and Root Mean Square Error (RMSE) of 0.04–0.54 and 1.15 for OLI, and 0.05–0.67 and 1.17 for MSI, suggests that the two sensors have similar salinity modelling skill. The extracted saline soils make up approximately 1,703 hectares for OLI and 118 hectares for MSI, indicating overestimation from the OLI image due to its coarser spatial resolution. Additionally, field samples indicate that nearby vegetated land is saline, indicating an underestimation of total impacted land. As sea levels rise, accurately monitoring soil salinization will be critical to protecting coastal agricultural lands. MSI’s spatial and temporal resolution makes it superior to OLI for salinity tracking though they have roughly equivalent spectral resolutions. This study demonstrates that visible spectral bands are sensitive to soil salinity with the Blue and Red spectral ranges producing the highest model accuracy; however, the low accuracies for both sensors indicate the need of narrowband sensors. The HyspIRI to be launched in the early 2020s by NASA may provide ideal data source in soil salinity studies.     

Robust speech recognition based on joint model and feature spaceoptimization of hidden Markov models

The hidden Markov model (HMM) inversion algorithm, based on either the gradient search or the Baum-Welch reestimation of input speech features, is proposed and applied to the robust speech recognition tasks under general types of mismatch conditions. This algorithm stems from the gradient-based inversion algorithm of an artificial neural network (ANN) by viewing an HMM as a special type of ANN. Given input speech features s, the forward training of an HMM finds the model parameters lambda subject to an optimization criterion. On the other hand, the inversion of an HMM finds speech features, s, subject to an optimization criterion with given model parameters lambda. The gradient-based HMM inversion and the Baum-Welch HMM inversion algorithms can be successfully integrated with the model space optimization techniques, such as the robust MINIMAX technique, to compensate the mismatch in the joint model and feature space. The joint space mismatch compensation technique achieves better performance than the single space, i.e. either the model space or the feature space alone, mismatch compensation techniques. It is also demonstrated that approximately 10-dB signal-to-noise ratio (SNR) gain is obtained in the low SNR environments when the joint model and feature space mismatch compensation technique is used.     

Remotely Sensed Retrieving the Surface Soil Moisture of Yan’an Urban Forest based on Landsat Image and Trapezoid Method

It is crucial for soil moisture assessment to know the prediction accuracy of inversion model. Urban forest surface soil in a gully-loess region (Yan’an), was taken as the research object, and the three scenes of Landsat satellite remotely sensed imagery in different periods and soil moisture sensor in situ measurement data were used as the data source. The parameters of TOTRAM (Thermal-Optical TRApezoid Model) and OPTRAM (OPtical TRApezoid Model) were obtained through the scatter diagram of pixels in two-dimensional spaces (LST-NDVI and STR-NDVI, LSTis land surface temperature,NDVIis normalized vegetation index, and STR is shortwave infrared conversion reflection coefficient) and their fitting dry edge and wet edge, respectively. Then, the w values (soil moisture in percentage) of Yan’an urban forest were retrieved at the pixel level (30 m by 30 m), the accuracy of the two models was verified, the differences between the estimated results of the two models, and the influence of linear and nonlinear edge on the inversion results were compared. The results indicate that: (1) Except that the dry edge and wet edge of OPTRAM models on Landsat 7 and Landsat 8 were non-linear, the other dry and wet edges of pixels in LST-NDVI space and STR-NDVI space are almost linear and enveloped into a trapezoid shape. (2) Compared with the field measurement data, the mean error (ME) of TOTRAM and OPTRAM were 0.009 and 0.045 5, respectively, which indicating that the estimation results of both models were relatively high, but the root mean square error (RMSE) of the OPTRAM model was closer to zero than the TOTRAM model. The value of w estimated by the OPTRAM model is evenly distributed on both sides of the 1∶1 reference line, and the number of points on the reference line is more than that of the TOTRAM model in scatterplots, indicating that the accuracy of OPTRAM is higher than that of the TOTRAM model, moreover, the inversion precision of nonlinear edge is higher than that of linear edge. Thus, in further research, the relationship between the complexity of the dry edge and wet edge and the model’s accuracy improvement should be discussed in the OPTRAM model, and the influences of surrounding environment, rainfall, forest disturbance and NDVI saturation on the estimation accuracy of the two models need to be considered.