反射光谱估算滨海土壤黏粒含量

探明反射光谱估算土壤黏粒含量的成因是实现黏粒含量测定、提高估算精度的基础。该文以江苏省滨海平原的150个土壤样品为研究对象,将测得的原始光谱数据进行平滑、一阶导数、连续统去除和倒数等数据变换,采用逐步多元线性回归(stepwise multiple linear regression,SMLR)和偏最小二乘回归(partial least squares regression,PLSR)方法估算黏粒含量,并在此基础上分析建模的影响波段,探讨反射光谱估算土壤黏粒含量的成因。结果表明,连续统去除光谱数据的SMLR分析估算精度最高,建模集和验证集决定系数分别为0.941和0.750。360~900、1 800~2 490 nm是黏粒含量的重要建模影响波段,该建模影响波段主要包括铁离子(410 nm附近)、土壤有机质(500~800 nm)、层状硅酸盐中的结晶水(1 900 nm附近)、绿泥石和蛭石等黏土矿物(2 325 nm)的吸收特征波段;PLSR分析表明,1 400 nm附近回归系数出现的双峰特征源于高岭石的双峰吸收。黏粒中的黏土矿物、黏粒对铁离子的吸附特性以及黏粒与有机质的高度相关性是实现反射光谱估算滨海土壤黏粒含量的原因。     

Improved analysis and modelling of soil diffuse reflectance spectra using wavelets

Diffuse reflectance spectroscopy using visible (vis), near‐infrared (NIR) and mid‐infrared (mid‐IR) energy can be a powerful tool to assess and monitor soil quality and function. Mathematical pre‐processing techniques and multivariate calibrations are commonly used to develop spectroscopic models to predict soil properties. These models contain many predictor variables that are collinear and redundant by nature. Partial least squares regression (PLSR) is often used for their analysis. Wavelets can be used to smooth signals and to reduce large data sets to parsimonious representations for more efficient data storage, computation and transmission. Our aim was to investigate their potential for the analyses of soil diffuse reflectance spectra. Specifically we wished to: (i) show how wavelets can be used to represent the multi‐scale nature of soil diffuse reflectance spectra, (ii) produce parsimonious representations of the spectra using selected wavelet coefficients and (iii) improve the regression analysis for prediction of soil organic carbon (SOC) and clay content. We decomposed soil vis‐NIR and mid‐IR spectra using the discrete wavelet transform (DWT) using a Daubechies’s wavelet with two vanishing moments. A multiresolution analysis (MRA) revealed their multi‐scale nature. The MRA identified local features in the spectra that contain information on soil composition. We illustrated a technique for the selection of wavelet coefficients, which were used to produce parsimonious multivariate calibrations for SOC and clay content. Both vis‐NIR and mid‐IR data were reduced to less than 7% of their original size. The selected coefficients were also back‐transformed. Multivariate calibrations were performed by PLSR, multiple linear regression (MLR) and MLR with quadratic polynomials (MLR‐QP) using the spectra, all wavelet coefficients, the selected coefficients and their back transformations. Calibrations by MLR‐QP using the selected wavelet coefficients produced the best predictions of SOC and clay content. MLR‐QP accounted for any nonlinearity in the data. Transforming soil spectra into the wavelet domain and producing a smaller representation of the data improved the efficiency of the calibrations. The models were computed with reduced, parsimonious data sets using simpler regressions.     

Accounting for the effects of water and the environment on proximally sensed vis–NIR soil spectra and their calibrations

Visible–near infrared (vis–NIR) spectroscopy can be used to estimate soil properties effectively using spectroscopic calibrations derived from data contained in spectroscopic databases. However, these calibrations cannot be used with proximally sensed (field) spectra because the spectra in these databases are recorded in the laboratory and are different to field spectra. Environmental factors, such as the amount of water in the soil, ambient light, temperature and the condition of the soil surface, cause the differences. Here, we investigated the use of direct standardization (DS) to remove those environmental factors from field spectra. We selected 104 sensing (sampling) sites from nine paddy fields in Zhejiang province, China. At each site, vis–NIR spectra were recorded with a portable spectrometer. The soils were also sampled to record their spectra under laboratory conditions and to measure their soil organic matter (SOM) content. The resulting data were divided into training and validation sets. A subset of the corresponding field and laboratory spectra in the training set (the transfer set) was used to derive the DS transfer matrix, which characterizes the differences between the field and laboratory spectra. Using DS, we transferred the field spectra of the validation samples so that they acquired the characteristics of spectra that were measured in the laboratory. A partial least squares regression (PLSR) of SOM on the laboratory spectra of the training set was then used to predict both the original field spectra and the DS‐transferred field spectra. The assessment statistics of the predictions were improved from R² = 0.25 and RPD = 0.35 to R² = 0.69 and RPD = 1.61. We also performed independent predictions of SOM on the DS‐transferred field spectra with a PLSR derived using the Chinese soil spectroscopic database (CSSD), which was developed in the laboratory. The R² and RPD values of these predictions were 0.70 and 1.79, respectively. Predictions of SOM with the DS‐transferred field spectra were more accurate than those treated with external parameter orthogonalisation (EPO), and more accurate than predictions made by spiking. Our results show that DS can effectively account for the effects of water and environmental factors on field spectra and improve predictions of SOM. DS is conceptually straightforward and allows the use of calibrations made with laboratory‐measured spectra to predict soil properties from proximally sensed (field) spectra, without needing to recalibrate the models.     

Exploring the predictability of soil texture and organic matter content with a commercial integrated soil profiling tool

In soil mapping, combining information from conceptually different proximal soil sensors can increase the accuracy of prediction and robustness of the model when compared with using individual sensors. In this study the predictability of soil texture (clay, silt and sand fractions) and soil organic matter (SOM) content was tested with a commercial integrated soil profiling tool that included sensors for measuring apparent electrical conductivity (EC_a), reflectance in the visible and near‐infrared (vis‐NIR) parts of the electromagnetic spectrum and insertion force (IF). The measurements were made at 20 locations on each of two Swedish farms. At every location, sensor measurements were made at 1.5‐cm intervals from the soil surface to a depth of 0.8 m. Soil samples were collected close to the sensor measurement points and analysed for texture and SOM content. Farm‐specific calibrations were developed for texture and SOM with each sensor separately and with combinations of all three sensors. The calibrations were made using both partial least squares regression (PLSR) and simple linear regression. The results for the two farms were quite consistent in terms of rank in prediction performance between the individual sensors and the sensor combinations. The vis‐NIR spectrometer was the best individual sensor for predicting the soil properties tested on both farms, with root mean square error of cross‐validation (RMSECV) of 0.3–0.5% for SOM, about 6% for clay and silt and 10–11% for sand. The inclusion of IF reduced the RMSECV for predictions of SOM content by about 10%. For soil texture, including EC_a reduced the RMSECV on average for all particle size fractions by 5–10%. However, the small improvements obtained by combining sensors do not provide strong support for combining vis‐NIR sensor measurements with measurements of EC_a and or IF.     

Multivariate calibration of hyperspectral γ-ray energy spectra for proximal soil sensing

The development of proximal soil sensors to collect fine‐scale soil information for environmental monitoring, modelling and precision agriculture is vital. Conventional soil sampling and laboratory analyses are time‐consuming and expensive. In this paper we look at the possibility of calibrating hyperspectral γ‐ray energy spectra to predict various surface and subsurface soil properties. The spectra were collected with a proximal, on‐the‐go γ‐ray spectrometer. We surveyed two geographically and physiographically different fields in New South Wales, Australia, and collected hyperspectral γ‐ray data consisting of 256 energy bands at more than 20 000 sites in each field. Bootstrap aggregation with partial least squares regression (or bagging‐PLSR) was used to calibrate the γ‐ray spectra of each field for predictions of selected soil properties. However, significant amounts of pre‐processing were necessary to expose the correlations between the γ‐ray spectra and the soil data. We first filtered the spectra spatially using local kriging, then further de‐noised, normalized and detrended them. The resulting bagging‐PLSR models of each field were tested using leave‐one‐out cross‐validation. Bagging‐PLSR provided robust predictions of clay, coarse sand and Fe contents in the 0–15 cm soil layer and pH and coarse sand contents in the 15–50 cm soil layer. Furthermore, bagging‐PLSR provided us with a measure of the uncertainty of predictions. This study is apparently the first to use a multivariate calibration technique with on‐the‐go proximal γ‐ray spectrometry. Proximally sensed γ‐ray spectrometry proved to be a useful tool for predicting soil properties in different soil landscapes.     

Predicting soil properties from the Australian soil visible–near infrared spectroscopic database

There are reflectance spectra in the visible and near infrared wavelengths from some 20 000 archived samples of soil in Australia. Their particular forms depend on absorbances at specific wavelengths characteristic of components in the soil such as water, iron oxides, clay minerals and carbon compounds, and so one might expect to be able to predict soil properties from the spectra. We tested a tree‐based technique for the prediction of 24 soil properties. A tree is first constructed by the definition of rules that separate the data into fairly homogeneous groups for any given property on both the absorptions at specified wavelengths and other, categoric, variables. Then within each group the property is predicted from the absorptions at those wavelengths by ordinary least‐squares regression. The spectroscopic predictions of the soil properties were compared with actual values in a subset of sample data separated from the whole data for validation. The criteria of success that we used were the root mean squared error (RMSE) to measure the inaccuracy of our predictions, the mean error (ME) to measure their bias and the standard deviation of the error (SDE) to measure their imprecision. We also used the ratio of performance to deviation (RPD), which is the ratio of the standard deviation of the observed values to the RMSE of the predictions; the larger it is the better does the technique perform. We found good predictions (RPD>2) for clay and total sand content, for total organic carbon and total nitrogen, pH, cation exchange capacity, and exchangeable calcium, magnesium and sodium. Several other properties were moderately well predicted (1.5 ≤ RPD < 2); they included air‐dry water content, volumetric water content at field capacity and wilting point, bulk density, the contents of silt, fine sand and coarse sand, total and exchangeable potassium, total phosphorus and extractable iron. Properties that were poorly predicted (RPD < 1.5) include the carbon‐to‐nitrogen ratio, available phosphorus and exchangeable acidity. We conclude that even though the predictions are less accurate than direct measurements, the spectra are cheap yet valuable sources of information for predicting values of individual soil properties when large numbers of analyses are needed, for example, for soil mapping.     

滩涂土壤有机质含量的反射光谱估算

Rapid determination of soil organic matter (SOM) using regression models based on soil reflectance spectral data serves an important function in precision agriculture. “deviation of arch”(DOA)-based regression and partial least squares regression (PLSR) are two popular modeling approaches to predict SOM. However, few studies have explored the accuracy of the DOA-based regression and PLSR models. Therefore, the DOA-based regression and PLSR were applied to the visible near-infrared (VNIR) spectra to estimate SOM content in the case of various dataset divisions. A two-fold cross-validation scheme was adopted and repeated 10 000 times for rigorous evaluation of the DOA-based models in comparison with the widely used PLSR model. Soil samples were collected for SOM analysis in the coastal area of northern Jiangsu Province, China. The results indicated that both modelling methods provided reasonable estimates of SOM, with PLSR outperforming DOA-based regression in general. However, the performance of PLSR for the validation dataset decreased more noticeably. Among the four DOA-based models, the linear model of the DOA provided the best estimation of SOM and a cutoff of SOM content (19.76 g kg^-1), and the performance for calibration and validation datasets was consistent. As the SOM content exceeded 19.76 g kg^-1, SOM became more effective in masking the spectral features of other soil properties to a certain extent. This work confirmed that reflectance spectroscopy combined with PLSR could serve as a non-destructive and cost-efficient way for rapid determination of SOM when hyperspectral data were available. The DOA-based model, which requires only 3 bands in the visible spectra, also provided SOM estimation with acceptable accuracy.     

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。     

含水率对土壤有机质含量高光谱估算的影响

土壤含水率对有机质(soil organic matter,SOM)含量高光谱估算精度有很大的影响。为了探讨SOM高光谱估算中土壤含水率的影响,该文对烘干土、风干土和质量含水率为5%~40%(按5%递增)的土壤样本进行了室内高光谱测量,对光谱数据进行了反射率、反射率一阶导数和反射率倒数对数3种光谱数据变换,运用偏最小二乘回归法(partial least squares regression,PLSR)建立了相应的SOM估算模型。结果表明,风干土的SOM高光谱估算精度较好;当含水率水平小于25%时,SOM估算模型精度受含水率的影响较大,光谱数据进行反射率倒数对数变换后的模型精度最高;当含水率水平大于等于25%时,水分对土壤光谱反射率的影响要大于SOM,不适宜利用土壤光谱数据进行SOM含量高光谱估算。该研究可为大田环境不同含水率情况下光谱估算SOM提供参考。     

苏北沿海滩涂地区土壤有机质含量的高光谱预测

基于反射高光谱快速、无损的检测优势,以苏北沿海滩涂地区不同成陆年代土壤作为光谱信息源,应用偏最小二乘回归（PLSR）方法,研究了原始反射光谱（REF）、微分光谱（FDR）、反射率倒数的对数（lg(1/R)）和波段深度（BD）对不同成陆年代土壤有机质含量的预测精度。结果表明,不同成陆年代土壤有机质含量预测的最佳光谱指标存在差异。REF是构建总体样本有机质含量PLSR预测模型的最佳光谱指标,均方根误差（RMSE）和相关系数（r）分别为2.7231和0.8701;FDR是预测成陆千年土壤样本有机质含量的最佳光谱指标,RMSE和r分别为2.0110和0.9436;BD所构建的成陆百年土壤有机质含量的PLSR预测模型为最优,RMSE和r分别为2.7051和0.8770。相关分析表明,可见光波段、以1 400 nm为中心及1 900～2 450 nm的红外波段是估算土壤有机质含量的最佳波段。     

Improving the prediction performance of a large tropical vis‐NIR spectroscopic soil library from Brazil by clustering into smaller subsets or use of data mining calibration techniques

Effective agricultural planning requires basic soil information. In recent decades visible near‐infrared diffuse reflectance spectroscopy (vis‐NIR) has been shown to be a viable alternative for rapidly analysing soil properties. We studied 7172 samples of seven different soil types collected from several regions of Brazil and varying in organic matter (OM) (0.2–10.3%) and clay content (0.2–99.0%). The aim was to explore the possibility of enhancing the performance of vis‐NIR data in predicting organic matter and clay content in this library by dividing it into smaller sub‐libraries on the basis of their vis‐NIR spectra. We used partial least square regression (PLSR) models on the sub‐libraries and compared the results with PLSR and two non‐linear calibration techniques, boosted regression trees (BT) and support vector machines (SVM) applied to the whole library. The whole library calibrations for clay performed well (ME (modelling efficiency) > 0.82; RMSE (root mean squared error) < 10.9%), reflecting the influence of the direct spectral responses of this property in the vis‐NIR range. Calibrations for OM were reasonably good, especially in view of the very small variation in this property (ME > 0.60; RMSE < 0.55%). The best results were, however, found when dividing the large library into smaller subsets by using variation in the mean‐normalized or first derivative spectra. This divided the global data set into clusters that were more uniform in mineralogy, regardless of geographical origin, and improved predictive performance. The best clustering method improved the RMSE in the validation to 8.6% clay and 0.47% OM, which corresponds to a 21% and 15% reduction, respectively, as compared with whole library PLSR. For the whole library, SVM performed almost equally well, reducing RMSE to 8.9% clay and 0.48% OM.     

Spectroscopy-Based Soil Organic Matter Estimation in Brown Forest Soil Areas of the Shandong Peninsula, China

Soil organic matter (SOM) is important for plant growth and production. Conventional analyses of SOM are expensive and time consuming. Hyperspectral remote sensing is an alternative approach for SOM estimation. In this study, the diffuse reflectance spectra of soil samples from Qixia City, the Shandong Peninsula, China, were measured with an ASD FieldSpec 3 portable object spectrometer (Analytical Spectral Devices Inc., Boulder, USA). Raw spectral reflectance data were transformed using four methods:nine points weighted moving average (NWMA), NWMA with first derivative (NWMA + FD), NWMA with standard normal variate (NWMA + SNV), and NWMA with min-max standardization (NWMA + MS). These data were analyzed and correlated with SOM content. The evaluation model was established using support vector machine regression (SVM) with sensitive wavelengths. The results showed that NWMA + FD was the best of the four pretreatment methods. The sensitive wavelengths based on NWMA + FD were 917, 991, 1 007, 1 996, and 2 267 nm. The SVM model established with the above-mentioned five sensitive wavelengths was significant (R²=0.875, root mean square error (RMSE)=0.107 g kg^-1 for calibration set; R²=0.853, RMSE=0.097 g kg^-1 for validation set). The results indicate that hyperspectral remote sensing can quickly and accurately predict SOM content in the brown forest soil areas of the Shandong Peninsula. This is a novel approach for rapid monitoring and accurate diagnosis of brown forest soil nutrients.     

基于高光谱的复垦农田土壤有机质含量估测

为了快速准确估算矿区复垦土地土壤有机质含量,以永城矿区复垦农田为例,在土样有机质含量测定和高光谱数据测量的基础上,对土壤高光谱数据进行多种预处理并与有机质实测含量进行相关性分析,利用相关系数进行P=0.01水平显著检验,确定敏感波段,建立一元线性回归、多元逐步回归和偏最小二乘回归等多种有机质含量与高光谱估测模型。结果表明:经过数学变换的土壤光谱反射率与土壤有机质含量相关性显著提高,复垦区土壤光谱经过多元散射校正和一元微分处理并利用偏最小二乘回归模型建模预测效果最好。当前较少有研究对矿区复垦农田土壤有机质进行高光谱估测,本研究成果可为有效利用高光谱遥感技术,快速、有效地测定复垦农田土壤有机质含量提供技术支撑。     

基于高光谱的北京铁矿区土壤重金属镍元素含量反演研究

运用高光谱数据对北京典型铁矿区土壤重金属镍含量进行建模反演,探索高光谱遥感技术在土壤重金属污染快速监测上应用的可行性。使用便携式地物光谱仪采集研究区土壤样本光谱反射率数据,光谱反射率数据经多种数学变换后,经逐步回归方法筛选最佳特征波段,利用多元线性回归（SLR）和偏最小二乘回归（PLSR）方法建立模型以光谱反射数据对土壤重金属镍元素含量进行反演。基于光谱二阶微分的多元线性回归模型（SD-MLR）的稳定性和精度最高（R2 = 0.842,RMSE = 4.474）,能够良好地预测研究区土壤镍元素含量。光谱数据数学变换能够有效提高其与土壤镍元素含量间的相关性。不同的光谱变换形式建立模型的预测能力和精度有如下关系,光谱二阶微分 > 光谱倒数对数一阶微分 > 光谱一阶微分 > 光谱倒数对数 > 光谱连续统去除 > 原始光谱。采用光谱二阶微分建立多元线性回归模型为研究区土壤镍元素含量反演的最佳模型,可为土壤重金属污染快速监测提供技术参考。     

淮北平原土壤高光谱特征及有机质含量预测

以安徽省淮北平原的蒙城县为研究区,采集131个表层土壤(0～20 cm)样品。采用Cary 5000分光光度计测定土壤光谱反射率,分析该地区典型土壤类型的光谱特征,利用偏最小二乘回归方法建立土壤有机质光谱预测模型。首先比较不同光谱变换对土壤有机质含量光谱预测建模的影响;其次根据光谱相似性对土壤样品进行分类,比较不同土壤类型和不同光谱分类的有机质光谱预测精度。结果表明:①不同土壤有机质含量和不同土壤类型光谱曲线在整体波段范围内趋势基本一致;有机质含量与光谱反射率呈显著负相关;有机质含量越低,曲线特征差异明显,可能是受其他因素的影响;②土壤光谱反射率经倒数的对数处理后,有机质光谱建模的决定系数和相对分析误差均有所提高,均方根误差降低,模型预测效果较优;③按照光谱相似性分类后建立的有机质光谱预测模型,比按土壤类型建立的光谱预测模型精度明显提高。     

Use of different chemometric approaches for an estimation of soil properties at field scale with near infrared spectroscopy

Visible and near infrared spectroscopy (vis‐NIRS) may be useful for an estimation of soil properties in arable fields, but the quality of results are often variable depending on the applied chemometric approach. Partial least squares regression (PLSR) may be replaced by approaches which employ supervised learning methods or variable selection procedures in order to increase the proportion of informative wavelengths used in the estimation procedure, to reduce the noise of the spectra and to find the best fitting solution. Objectives were (1) to compare the usefulness of PLSR with either PLSR combined with a genetic algorithm (GA‐PLSR) or support vector machine regression (SVMR) for an estimation of soil organic carbon (SOC), total nitrogen (N), pH, cation exchange capacity (CEC) and soil texture for surface soils (0–5 cm, n = 144) of an arable field in Bangalore (India) and (2) to test and optimize different calibration strategies for GA‐PLSR for an improved estimation of soil properties. PLSR was useful for an estimation of SOC, N, sand and clay. In the cross‐validation (n = 96), accuracies of estimated soil properties generally decreased in the order GA‐PLSR > SVMR > PLSR. However, the order of estimation accuracies for the random validation sample (n = 48) changed to SVMR > GA‐PLSR > PLSR for SOC, N, pH, and CEC, whereas for clay the order changed to SVMR > PLSR > GA‐PLSR. A sequential procedure, which used the most frequently selected wavelengths of the GA‐PLSR runs, proved to be useful for an improved estimation of SOC and N. Overall, SVMR especially improved estimations of SOC and clay, whereas GA‐PLSR was particularly useful for SOC and N and it was the only approach which successfully estimated CEC in cross‐validation and validation.     

基于差异化光谱指数的盐渍土水分含量预测——以滨海盐土为例

以滨海盐土为研究对象,通过添加不同浓度的盐溶液并模拟蒸发过程,获取不同含水、含盐量的土壤样品,并测定土壤光谱和土壤含水量,分别运用光谱指数法和偏最小二乘回归法(PLSR)对土壤含水量进行预测。结果表明：由2027 nm和1878 nm构建的土壤水分差异化光谱指数(NDMI2027,1878)是预测土壤水分的最优指数,且适用于任何等级的盐渍化土壤,其建模集和验证集的预测结果均优于PLSR方法,验证集R2达0.99,RMSE仅为21.84 g/kg,可比较准确地预测盐渍化土壤的含水量。     

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.     

辽宁省黄土状母质发育土壤有机质含量高光谱预测模型的构建

  目的  建立辽宁省黄土状母质发育土壤有机质含量的高光谱预测模型,以便快速获取土壤样品的有机质含量。  方法  对省域内黄土状母质发育土壤进行了样品采集,获取样品有机质含量和高光谱数据;选择原始光谱及其一阶微分、二阶微分、倒数对数、倒数对数一阶微分、倒数对数二阶微分6种光谱变换数据作为自变量,与土壤有机质含量进行相关分析,选取特征波段,分别建立多元逐步线性回归（SMLR）、偏最小二乘回归（PLSR）和主成分回归（PCR）3种土壤有机质高光谱线性预测模型,并进行了支持向量机（SVM）方法的非线性模型拟合。  结果  土壤有机质含量与其光谱反射率呈负相关关系,对光谱进行不同的数学变换,可以提高土壤有机质含量与光谱反射率的相关性,其中一阶微分和二阶微分的提升效果最佳;相同光谱数据在不同模型中建模精度存在显著差异,以原始光谱反射率一阶微分为自变量的PLSR模型精度最高,建模集和验证集的决定系数（R2）分别为0.958和0.976;3种线性方法建立的最佳预测模型的检验精度为:PLSR > SMLR > PCR。  结论  PLSR模型是辽宁省黄土状母质发育土壤有机质含量的最佳高光谱预测模型,且基于特征波段的建模效果优于全波段;SVM非线性模型的预测精度较低。     

土壤含水量对反射光谱法预测红壤土壤有机质的影响研究

研究土壤含水量对有机质预测的影响,可为野外红壤有机质快速测定提供理论依据。本文在实验室条件下测量了不同含水量红壤的可见光-近红外光谱反射率,运用偏最小二乘回归(PLSR)建立不同含水量的土壤有机质预测模型。结果显示,随土壤含水量的增加,有机质与一阶微分光谱的相关性先增加后下降,含水量为100～150 g/kg时相关系数最大。分380～2 400、380～1 300、1 300～2 400 nm三个波段建立不同含水量的有机质预测模型,模型预测精度均随土壤含水量增加而呈现先增加后下降的趋势。利用1 300～2 400 nm建立有机质预测模型可以有效避开氧化铁影响,建立的模型预测精度最高。本研究认为,当土壤含水量小于200 g/kg时,可以利用在室内控制条件下测定的土壤反射率,建立1 300～2 400 nm波段的PLSR模型,进行红壤土壤有机质含量预测。