土壤有机质含量可见-近红外光谱反演模型校正集优选方法

土壤有机质含量可见-近红外光谱反演过程中校正集的构建策略对模型的预测精度有重要影响。以江汉平原洪湖地区水稻土为研究对象,采用Kennard-Stone(KS)法,Rank-KS(RKS)和Sample set Partitioning based on joint X-Y distance(SPXY)法,构建样本数占总校正集不同比例的子校正集,通过偏最小二乘回归,建立土壤有机质含量的可见—近红外光谱反演模型。结果表明:KS法无法提高模型预测精度,但可以在保证标准差与预测均方根误差比(ratio of performance to standard deviation,RPD)2.0的前提下减少30%的校正样本;基于SPXY法的模型,当子校正集样本比例为总校正集的50%时达到最佳的模型预测精度,RPD为2.557;RKS法能够在保证预测精度的情况下(RPD2.0),最多减少总校正集70%的样本,对应模型RPD为2.212。当校正集与验证集的有机质含量分布相近时,能够以较少的建模样本达到与总校正集相近甚至更高的模型预测精度,提升土壤有机质光谱反演模型的实用性。     

土壤有机质可见光-近红外光谱预测样本优化选择

土壤有机质可见光-近红外光谱预测中建模样本的优化选择对提高有机质模型估算精度具有重要作用。本文以湟水流域土壤有机质为例,采用基于土壤单一属性信息考虑的建模样本选择方法:浓度梯度法、Kennard-Stone(KS)方法,以及基于土壤多种信息考虑的建模样本选择方法:Rank-KS(RKS)法、土壤类型结合浓度梯度法以及土壤类型结合KS法。通过偏最小二乘回归建模,探索可见光–近红外光谱预测青海湟水流域有机质的最优样本集。结果表明:不同级别样本数的最佳建模样本选择方法不同,整体表现为基于土壤多种信息挑选的建模样本集的模型精度相比土壤单一信息均较高,特别是KS方法结合土壤类型后的建模样本集模型精度明显提高且在样本数较少时更为明显。土壤类型可以优化建模样本选择方法提高模型预测精度。在保证固定验证样本模型预测精度的情况下,土壤类型参与建模样本的选择可以有效减少建模样本数,进而降低了建模成本。     

基于灰度关联-岭回归的荒漠土壤有机质含量高光谱估算

为改善高光谱技术对荒漠土壤有机质的估测效果,该文采集了以色列Seder Boker地区的荒漠土壤,经预处理、理化分析后将土样分为砂质土和黏壤土2类,再通过光谱采集、处理得到6种光谱指标:反射率(reflectivity,REF)、倒数之对数变换(inverse-log reflectance,LR)、去包络线处理(continuum removal,CR)、标准正态变量变换(standard normal variable reflectance,SNV)、一阶微分变换(first order differential reflectance,FDR)和二阶微分变换(second order differential reflectance,SDR)。通过灰度关联(gray correlation,GC)法确定SNV、FDR、SDR为敏感光谱指标,采用偏最小二乘回归(partial least squares regression,PLSR)法和岭回归(ridge regression,RR)法,构建基于敏感光谱指标的土壤有机质高光谱反演模型,并对模型精度进行比较。结果表明:砂质土有机质含量的反演效果要优于黏壤土;基于SNV指标建立的模型决定系数R~2和相对分析误差RPD均为最高、均方根误差RMSE最低,所以SNV是土壤有机质的最佳光谱反演指标;对SNV-PLSR模型和SNV-RR模型综合比较得出,SNV-RR模型仅用全谱4%左右的波段建模,实现了更为理想的反演效果:其中,对砂质土有机质的预测能力极强(R_p~2为0.866,RMSE为0.610 g/kg、RPD为2.72),对黏壤土有机质的预测能力很好(Rp2为0.863,RMSE为0.898 g/kg、RPD为2.37)。荒漠土壤有机质GC-SNV-RR反演模型的建立为高光谱模型的优化、土壤有机质的快速测定提供了一种新的途径。     

基于可见–近红外光谱的水稻土全磷反演研究

采用PLSR偏最小二乘法回归结合留一法交叉验证,利用长期定位试验田以及直湖港小流域面上的水稻土土壤样本建立最优模型,研究了不同光谱预处理方式对水稻土全磷可见-近红外高光谱反演精度的影响,探索水稻土全磷光谱反演的可行性;并结合简单相关系数法以及PLSR模型回归系数法分析了水稻土全磷光谱反演的重要波段。结果表明,光谱预处理方法对土壤全磷反演精度的影响不大;基于PLSR建立的水稻土全磷光谱反演模型的校正决定系数达0.85,交叉验证决定系数为0.70,RPD为1.8,有较好的模型精度;440~740 nm为土壤全磷光谱反演的重要波段。利用PLSR对水稻土全磷进行光谱预测是可行的。     

干旱区典型绿洲土壤有机质的反演及影响因素研究

为了大面积、高精度地反演土壤有机质含量,为农业可持续发展提供数据支撑。以新疆渭干河-库车河三角洲绿洲为研究区,采用波段平均法将实测高光谱窄波段拟合为Landsat 8 OLI遥感影像的宽波段,建立土壤有机质含量的估算模型,并将最优估算模型应用到经过波段校正的Landsat 8OLI遥感影像中。结果表明:(1)反射率进行倒数、对数、平方、一阶微分等数学变换后与有机质含量的相关性显著提高;(2)土壤有机质的高光谱估算模型拟合度较高,最优估算模型的决定系数R2为0.852,采用比值法对多光谱波段反射率进行校正,校正后的遥感影像反演结果得到了较大提高,检验样本的决定系数R2从0.711提升至0.849。从反演结果来看,将高光谱估算模型应用到经过订正的多光谱影像,土壤有机质反演模型的精度得到了大幅度提高,运用此方法可以实现高精度的土壤有机质区域化反演。(3)有机质的分布受土地利用类型、土壤颗粒组成、土壤质地的影响,其中土壤质地对有机质的空间分布影响最为显著。     

基于组合模型的庐山森林土壤有效铁光谱间接反演研究

铁是植物生长的重要微量营养元素之一,土壤有效铁含量对林地环境起着重要的影响,利用土壤光谱预测技术获取土壤有效铁含量信息具有重要意义。而要通过土壤光谱直接预测土壤有效铁含量是难以实现的,因此提出利用土壤有机质含量与有效铁含量之间的相关性,探讨间接估算土壤有效铁含量的可行性。以庐山森林土壤样本为研究对象,研究基于偏最小二乘回归(PLSR)和径向基函数(RBF)神经网络的组合模型预测土壤有机质含量的适用性,并且通过构建有机质含量与有效铁含量的二项式线性模型,对土壤有效铁含量进行间接反演,探讨不同权重下的最优组合模型。结果表明,组合模型的预测效果优于偏最小二乘回归和RBF神经网络单个模型,并且熵值组合为最优组合模型,其中,土壤有机质的反演模型验证的决定系数(R~2)为0.81,均方根误差(RMSE_p)为11.54 g kg~(-1),测定值标准差与标准预测误差的比值(RPD)为2.18;有效铁的间接反演模型R~2为0.70,RMSE_p为21.60 mg kg~(-1),RPD为1.77。通过土壤有机质构建土壤有效铁含量的光谱反演间接模型,在光谱反演模型中,组合模型能较大限度地利用各种预测样本信息,能有效减少单个预测模型中随机因素的影响,增强预测稳定性,提高模型的预测能力。因此,组合模型可对土壤有机质含量的光谱预测及土壤有效铁的间接预测发挥更好的作用。     

母质与土地利用类型对土壤光谱反演模型的影响

用可见光近红外(Vis-Near Infrared,VNIR)光谱建立的土壤反演模型可以快速高效测定土壤某些属性,但不考虑土壤自身特点的纯统计模型反演的精度会受到制约。本文研究了母质和土地利用类型对土壤光谱反演模型建立的影响。研究所用集合为SF(安徽宣城的林地样品)、SP(安徽宣城的水田样品)、DP(安徽定远的水田样品),结果显示:母质和土地利用类型的差异会显著影响异地模型的适应性,一个地区建立的反演模型不可随便用于母质和土地利用类型不同的其他地区;当异地模型不适用于反演时,可考虑采用精度稍低的全局模型进行预测。     

高光谱估算土壤有机质含量的波长变量筛选方法

土壤高光谱数据量大、波段维数高,存在光谱信息无效、冗余和重叠现象,导致基于全波段构建的土壤有机质含量反演模型不稳定、精度难以提升。因此,探寻筛选关键波长变量的方法,通过滤除干扰、冗余、共线信息,提高模型预测性能,是目前土壤高光谱研究的热点之一。该文对江汉平原公安县的土壤样本进行室内理化分析、光谱测量与处理等工作获取了实证数据,采用无信息变量消除法(uninformative variables elimination,UVE)剔除无效变量,利用竞争性自适应重加权算法(competitive adaptive reweighted sampling,CARS)滤除冗余变量,运用连续投影算法(successive projections algorithm,SPA)消除共线变量,并尝试将不同类型的筛选方法进行耦合筛选关键波长变量,应用偏最小二乘回归(partial least squares regression,PLSR)分别建立土壤有机质含量估算模型,对比各种变量筛选方法的优缺点,最终,构建筛选土壤高光谱数据关键变量的方法体系。研究结果表明,除SPA方法的模型精度低于全波段外,其他6种变量筛选方法的建模效果均优于全波段;在3种单个变量筛选方法中,CARS方法优于UVE、SPA变量筛选方法,能有效地筛选出重要波长变量,其预测集相对分析误差RPD值为2.84;综合比较各种变量筛选方法,发现CARS-SPA方法从全波段2 001个波长中筛选出37个特征波长建立的土壤有机质含量的PLSR模型效果最好,其模型预测集的决定系数R2和相对分析误差RPD值分别为0.92、3.60,所选波段仅为全波段的1.85%。CARS-SPA-PLSR模型简单、预测效果好,可作为该区域土壤有机质含量估测的重要方法,对今后土壤近地传感器设备的开发具有一定的指导作用。     

基于可见光近红外光谱的南疆荒漠土壤有机质反演研究

有机质含量是土壤肥力评价的重要依据,快速监测南疆大面积分布的荒漠土壤有机质含量,可为后备耕地资源的合理开发与利用提供重要数字依据。通过野外采样与室内光谱及有机质含量测定,获取了240个荒漠土样的有机质含量与可见光近红外光谱数据。在此基础上,分析了不同有机质含量土样的光谱特征、有机质与光谱数据的相关性,并采用多种数据处理方式构建了反演模型。研究结果表明,不同有机质含量的土样在反射率和曲线形态上均存在一定的差异,土壤有机质在400~842 nm波段与反射率具有较高的相关性,相关系数最大值-0.32位于588 nm波段,反射率经连续统去除处理后相关性得到了明显提升,达极显著相关水平波段数量明显增加,尤其在797~1330 nm、1852~1872 nm、2155~2338 nm波段的改善效果显著,连续统去除的最大相关系数0.55位于86 4 nm波段。不同建模方法的精度对比结果表明偏最小二乘法(PLSR)具有最高的建模精度,在7种不同数据处理方式的PLSR模型中,以SG平滑+峰值归一化+正交信号校正(SGS+MAN+OSC)处理模式下的建模集的决定系数(R~2)最高、均方根误差(RMSE)最低,分别为0.81和0.98;该模型预测集的R~2、RMSE、残余预测偏差(RPD)分别为0.76、0.99和2.01,表明该模型具有较好的预测能力,可推荐为南疆荒漠土壤有机质的光谱定量反演模型。     

原状水稻土土壤湿度高光谱特征及反演模型

室内高光谱遥感定量反演土壤湿度的相关研究,可以为利用星载或机载遥感数据源进行大尺度土壤水分监测提供理论基础;原状土土样相较于当前广泛使用的研磨土土样,能更好地保持土壤的原始光谱特性和持水性能。为了得到更可靠的土壤湿度光谱特征与更具有适用性的反演模型,利用ASD Field Spec 3光谱仪采集了江苏省金坛市竹林镇的62个原状水稻土土样12套不同湿度梯度的可见光-近红外反射光谱,通过反射率、相对反射率以及水分吸收峰特征参数探究原状湿土光谱特征;在此基础上,选择对土壤湿度变化响应明显的变量,采用偏最小二乘回归反演土壤湿度,并对比选出最佳反演模型。结果表明:(1)随着风干过程中土壤湿度的降低:土壤反射率整体先下降后上升,可见光波段在土壤湿度较低时存在误差;1450 nm和1950 nm吸收峰的位置向短波方向偏移,吸收峰深度、宽度和面积不断减小,2 200 nm处吸收峰右肩高度不断降低。(2)土壤水分吸收峰特征参数对土壤湿度响应明显;反射率和相对反射率与土壤湿度相关性较高的波段均出现在>1300 nm以后。(3)反射率(R)反演原状土湿度精度较低,R2val=0.752,RPD=1.81,经过数学形式和相对反射率(Rr)变换后,[lg(1/Rr)]'建模中R2val=0.820,RPD=1.99;基于水分吸收峰特征参数建立的模型R2val=0.800,RMSEP=62.70 g kg-1,RPD=2.29,最佳主成分数为3,是反演原状土土壤湿度的最佳建模变量。     

Visible and Near-Infrared Reflectance Spectroscopy for Assessment of Soil Properties in the Caucasus Mountains,Azerbaijan

ABSTRACT

This study aimed to predict soil properties using visible–near infrared (VIS-NIR) spectroscopy combined with partial least square regression (PLSR) modeling. Special emphasis was given to evaluating effect of pre-processing methods on prediction accuracy and important wavelengths. A total of 114 samples were collected and involved in chemical and spectral analyzes. PLSR model of each soil property was calibrated for all pre-processing methods using all samples, and leave-one-out cross-validation was used to make comparisons between them. Then, PLSR model of each best pre-processing method was calibrated using a 75% of all samples and correspondingly validated with the remaining a 25%. Model accuracy was evaluated based on coef?cient of determination (R²), root mean-squared errors (RMSE), and residual prediction deviations (RPD). The high correlation coefficients were found between the tested soil properties and reflectance spectra. The pre-processing methods considerably improved prediction accuracy and filtering methods outperformed linearization methods, and the latter outperformed normalization methods. The performance of cross-validation, calibration and independent validation was similar. An excellent prediction (RPD>2.5) model was obtained for soil organic carbon (SOC) and calcium-carbonate (CaCO₃), good quantitative (2.0< RPD<2.5) prediction for sand, silt, and clay, fair prediction (1.4< RPD<1.8) for pH, and poor prediction (1.0< RPD<1.4) for hygroscopic water content (WC). Important wavelengths varied depending on soil property, but some wavelengths were common. This study can be a precursor to building a pioneering soil spectral database, calibrating satellite data, and hyperspectral image mapping of soils as well as digital soil mapping, environmental, and erosion modeling in the Caucasus Mountains.     

Rapid determination of lime requirement by mid‐infrared spectroscopy: A promising approach for precision agriculture

Mid‐infrared spectroscopy (MIRS) has proven to be a cost‐effective, high throughput measurement technique for soil analysis. After multivariate calibration mid‐infrared spectra can be used to predict various soil properties, some of which are related to lime requirement (LR). The objective of this study was to test the performance of MIRS for recommending variable rate liming on typical Central European soils in view of precision agriculture applications. In Germany, LR of arable topsoils is commonly derived from the parameters organic matter content (SOM), clay content, and soil pH (CaCl₂) as recommended by the Association of German Agricultural Analytical and Research Institutes (VDLUFA). We analysed a total of 458 samples from six locations across Germany, which all revealed large within‐field soil heterogeneity. Calcareous topsoils were observed at some positions of three locations (79 samples). To exclude such samples from LR determination, peak height at 2513 cm^?1 of the MIR spectrum was used for identification. Spectra‐based identification was accurate for carbonate contents > 0.5%. Subsequent LR derivation (LR_SPP) from MIRS‐PLSR predictions of SOM, clay, and pH (CaCl₂) for non‐calcareous soil samples using the VDLUFA look‐up tables was successful for all locations (R² = 0.54–0.82; RMSE = 857–1414 kg CaO ha^?1). Alternatively, we tested direct LR prediction (LR_DP) by MIRS‐PLSR and also achieved satisfactory performance (R² = 0.52–0.77; RMSE = 811–1420 kg CaO ha^?1; RPD = 1.44–2.08). Further improvement was achieved by refining the VDLUFA tables towards a stepless algorithm. It can be concluded that MIRS provides a promising approach for precise LR estimation on heterogeneous arable fields. Large sample numbers can be processed with low effort which is an essential prerequisite for variable rate liming in precision agriculture.     

基于相似光谱匹配预测土壤有机质和阳离子交换量

土壤可见光-近红外波段光谱(350~2 500 nm)包含了大量的土壤属性信息,相同类型的土壤具有相似的光谱曲线特征,但相似光谱曲线是否具有相似的属性含量?探讨此问题可为土壤光谱库的应用提供依据,从而最终服务于快速获取土壤信息技术体系的构建。该研究以安徽宣城为研究区,根据母质、地形特征和土地利用等信息,采集91个典型土壤剖面,共含400个土壤发生层样品,测定了有机质(soil organic matter,SOM)和阳离子交换量(cation exchange capacity,CEC)含量,同时采用VARIAN公司的Cary 5000分光光度计测定了土壤光谱,并将光谱数据变换为反射率(R)、反射率一阶导数(FDR)和吸收度(Log(1/R))3种形式。该文采用光谱角(spectral angle mapper,SAM)、偏最小二乘回归(partial least square regression,PLSR)和SAM-PLSR(spectral angle mapper-partial least square regression,SAM-PLSR)3种方法预测土壤SOM和CEC。SAM方法是通过对测试集104个光谱曲线与参考集的296个光谱曲线进行相似性计算,并以此实现土壤SOM和CEC含量的预测。SAM-PLSR方法以SAM算法下的匹配结果作为建模样本建立PLSR模型和进行预测分析。结果表明,具有相似光谱曲线的土壤具有相似的SOM和CEC含量,SAM算法下相似光谱匹配可直接预测SOM(R2=0.78,RPD=2.17)和CEC(R2=0.82,RPD=2.41)。PLSR方法可很好地预测SOM(R2=0.87,RPD=2.77)和CEC(R2=0.87,RPD=2.59);相较之下,SAM-PLSR方法不仅可以更加准确预测SOM(R2=0.89,RPD=3.00)和CEC(R2=0.91,RPD=3.06),而且大大减少了建模样本的数量。该研究使可见光-近红外光谱可更加高效地用于土壤属性分析,并为土壤光谱数据库的建设及应用提供技术参考。     

基于近红外光谱和支持向量机的土壤参数预测

应用支持向量机算法对实时土壤光谱数据进行处理,获得了土壤全氮和有机质的回归模型并研究了模型随参数变化的规律。从中国农业大学试验田采集了150个土样,用光谱仪获取了原始土壤样本的近红外光谱,用实验室分析法获取了各样本的全氮和有机质含量。以近红外光谱数据为自变量对2个土壤参数进行了回归建模并评价了算法各参数对模型的影响。研究表明土壤参数适合于全谱支持向量回归。对于土壤全氮,基于小波降噪NIR光谱的SVM回归模型的标定R2为0.9224,验证R2为0.3667;对于土壤有机质,基于原始NIR光谱的SVM回归模型     

Prediction of soil organic matter using a spatially constrained local partial least squares regression and the Chinese vis–NIR spectral library

We need to determine the best use of soil vis–NIR spectral libraries that are being developed at regional, national and global scales to predict soil properties from new spectral readings. To reduce the complexity of a calibration dataset derived from the Chinese vis–NIR soil spectral library (CSSL), we tested a local regression method that combined geographical sub‐setting with a local partial least squares regression (local‐PLSR) that uses a limited number of similar vis–NIR spectra (k‐nearest neighbours). The central idea of the local regression, and of other local statistical approaches, is to derive a local prediction model by identifying samples in the calibration dataset that are similar, in spectral variable space, to the samples used for prediction. Here, to derive our local regressions we used Euclidean distance in spectral space between the calibration dataset and prediction samples, and we also used soil geographical zoning to account for similarities in soil‐forming conditions. We tested this approach with the CSSL, which comprised 2732 soil samples collected from 20 provinces in the People's Republic of China to predict soil organic matter (SOM). Results showed that the prediction accuracy of our spatially constrained local‐PLSR method (R² = 0.74, RPIQ = 2.6) was better than that from local‐PLSR (R² = 0.69, RPIQ = 2.3) and PLSR alone (R² = 0.50, RPIQ = 1.5). The coupling of a local‐PLSR regression with soil geographical zoning can improve the accuracy of local SOM predictions using large, complex soil spectral libraries. The approach might be embedded into vis–NIR sensors for laboratory analysis or field estimation.     

The application of proximal visible and near-infrared spectroscopy to estimate soil organic matter on the Triffa Plain of Morocco

Soil organic matter (SOM) is a fundamental soil constituent. The estimation of this parameter in the laboratory using the classical method is complex time-consuming and requires the use of chemical reagents. The objectives of this study were to assess the accuracy of two laboratory measurement setups of the VIS-NIR spectroscopy in estimating SOM content and determine the important spectral bands in the SOM estimation model. A total of 115 soil samples were collected from the non-root zone (0–20 cm) of soil in the study area of the Triffa Plain and then analysed for SOM in the laboratory by the Walkley–Black method. The reflectance spectra of soil samples were measured by two protocols, Contact Probe (CP) and Pistol Grip (PG)) of the ASD spectroradiometer (350–2500 nm) in the laboratory. Partial least squares regression (PLSR) was used to develop the prediction models. The results of coefficient of determination (R²) and the root mean square error (RMSE) showed that the pistol grip offers reasonable accuracy with an R² = 0.93 and RMSE = 0.13 compared to the contact probe protocol with an R² = 0.85 and RMSE = 0.19. The near-Infrared range were more accurate than those in the visible range for predicting SOM using the both setups (CP and PG). The significant wavelengths contributing to the prediction of SOM for (PG) setup were at: 424, 597, 1432, 1484, 1830,1920, 2200, 2357 and 2430 nm, while were at 433, 587, 1380, 1431, 1929, 2200 and 2345 nm for (CP) setup.     

基于条件拉丁超立方抽样的县域耕地土壤有机质空间插值合理样本密度的确定

  目的  采样密度与耕地土壤有机质（SOM）的空间预测精度密切相关,为提高耕地SOM空间预测精度,需要确定合理的采样密度。  方法  以湖南省岳阳县为例,用R语言设计了条件拉丁超立方体抽样（cLHS）方案,从7399个（采样密度为14.82个 km?2）耕地土壤样本中独立抽取不同采样密度的8个训练集（采样密度分别为10.01、7.41、3.70、1.85、0.93、0.46、0.23、0.12个 km?2）,为了兼顾样本特征空间与地理空间,地形部位、坡度、成土母质、土壤类型、乡镇和经纬度等信息被添加到了cLHS中。结合普通克里格方法,分析和探讨了不同采样密度的耕地SOM空间预测效果。  结果  不同采样密度训练集SOM均值高于湖南省平均水平,具有中等程度变异,描述性统计结果差异不大,各训练集对总体均具有较强的代表性;半方差函数模型均为指数模型,具有较好的半方差结构（结构性比例:87.5% ~ 94.5%）,空间相关性较强,变程与拟合优度呈现出正相关关系(相关系数r = 0.96),与结构性比例则表现为负相关关系(相关系数r = ?0.79);在采样密度为3.70个 km?2时,探测到的SOM变异结构中结构性组分最完整,精度最佳。当采样密度达到1.85个 km?2以上时可较稳健地揭示其空间结构特征,继续增加采样密度并不能大幅提升预测精度。  结论  考虑预测精度要求和工作成本,与研究区自然地理条件相似的地区将耕地土壤采样密度控制在1.85个 km?2以上可获得预期的效果。     

基于局部加权回归的土壤全氮含量可见-近红外光谱反演

全氮是土壤肥力的重要指标,对作物产量具有决定性作用,采用土壤可见-近红外(Vis-NIR)光谱预测技术及时获取土壤全氮含量信息具有重要意义。采用来自5省的450个土壤样本来验证局部加权回归方法(LWR)结合Vis-NIR光谱技术预测大面积土壤全氮含量的适用性。结果表明,LWR模型的预测效果优于偏最小二乘回归(PLSR)、人工神经网络(ANN)和支持向量机(SVM),选取主成分数为5,相似样本为40时,模型验证的决定系数(RP2)为0.83,均方根误差(RMSEP)为0.25 g kg-1,测定值标准偏差与标准预测误差的比值(RPD)达到2.41。LWR从建模集中选取与验证样本相似的土样作为局部建模样本,降低了差别大的样本对模型的干扰,从而提高了模型的预测能力。因此,LWR建模方法通过大范围、大样本土壤光谱数据进行大尺度区域的全氮等土壤属性预测时能够发挥更好的作用。