基于近红外光谱分析的土壤全氮含量估测研究

应用近红外光谱分析技术对比研究基于土壤风干样本和鲜样来预测全氮含量的可行性。选取水稻土为研究对象,首先分析了不同水分土壤的光谱特征,显示随水分含量增加,吸光度升高,且鲜样的吸光度高于干样。通过比较不同预处理方法,对土壤干鲜样分别采用逐步多元回归(SMLR)和偏最小二乘法(PLSR)建立了相应的近红外模型。结果表明,利用近红外光谱均可预测干鲜土壤样本的全氮含量,特别是利用偏最小二乘法建立的标定模型,预测精度高,反演性较好,鲜样和干样外部验证决定系数分别达到0.89和0.91,相对误差仅为6.92%和5.92%,研究结果可以为田间土壤全氮含量的估测提供技术依据和参考。     

近红外光谱法在土壤有机质研究中的应用

近红外光谱技术(Near Infrared Reflectance Spectroscopy,NIRS)具有快速、低成本、无损等优点。目前利用NIRS获取土壤信息已成为国内外学者研究的重点,但是在我国利用NIRS对土壤成分进行定量分析才刚刚起步。本文简要介绍了近红外光谱分析的基本原理、模型的建立及评价,详细论述了该技术在预测土壤有机质及其组分方面的应用,并对NIRS在我国土壤有机质定量研究方面的应用前景进行了展望。     

Use of Near Infrared Reflectance Spectroscopy (NIRS) for Predicting Soil Fertility and Historical Management

This study tests the potential of near infrared reflectance spectroscopy (NIRS) for predicting soil fertility and management history from topsoil (0–10 cm deep) spectra. Soil fertility was assessed by measuring the growth of a test plant, and soil management history was determined through inquiries with farmers. Moreover, NIRS predictive value was compared with that of a group of topsoil parameters: total carbon and nitrogen, nitrate, potential respiration and denitrification, and microbial biomass. Modelling used partial and modified partial least square regressions to ensure comparisons between predictions by NIRS versus by soil parameters. Soil fertility and management history were well predicted by NIRS (Q² = 0.78 and R² = 0.89 both; Q² and R² are cross-validation and calibration coefficients of determination, respectively), as were the soil parameters (Q² = 0.79–0.92 and R² = 0.86–0.98). Soil fertility and management history were more accurately predicted by NIRS than by the set of soil parameters.     

近红外反射光谱分析在土壤肥料学的应用及发展方向

近红外反射光谱分析是一种"巨人"分析法,在土壤肥料及植物营养科学具有广阔的应用前景。对近红外反射光谱分析在土壤肥料及植物营养学科中的应用,存在问题,及其在土壤肥料及植物营养学科领域的发展方向进行了评述。     

Inability of Near Infrared Reflectance Spectroscopy (NIRS) to identify belowground earthworm casts in no-tillage soil

Several studies have emphasised the ability of Near Infrared Reflectance Spectroscopy (NIRS) to identify surface earthworm casts in the field. However, less is known about casts deposited within the soil, which usually represent the majority found in the field. This study tested the ability of NIRS to identify belowground casts in agricultural systems. Casts and surrounding soils were sampled at depths of 20–30 cm in a loamy soil under no tillage for 12 years. To distinguish different types of cast, sizes and orientations relative to the horizontal plane were measured. NIRS analyses and analyses of carbon and nitrogen content were also performed to compare casts to surrounding soils. Casts were classified into 4 size classes, with no preferential orientation. Cast carbon and nitrogen content were not influenced by their size and did not differ from surrounding soils. PCAs performed on the NIRS data did not allow casts to be differentiated from surrounding soils, regardless of size class. However, soil aggregates were clearly differentiated probably due to their spatial distribution in the soil. Although this study did not identify specific NIRS signatures for casts, it shows the utility of this method to investigate the origin of the soil consumed by earthworms. In our case, NIRS analyses suggest that the high bulk density of the soil (1.42 g cm⁻³) forced ingestion by endogeic earthworms, simply to move around, without preferential selection for organic matter. Consequently, their casts were deposited a few mm from where they had ingested soil with similar organic matter quality.     

近红外光谱联合CARS-PLS-LDA的山茶油检测

为了寻找快速判别山茶油掺假的检测方法,本研究利用近红外光谱技术对掺杂大豆没油山茶油进行掺假检测研究.试验在350～1 800 nm波段范围内采集样本的透射光谱,利用CARS方法筛选重要的波长变量,应用偏最小二乘-线性判别分析(PLS-LDA)建立山茶油掺假的判别模型,并与未经变量优选的判别模型进行比较.结果表明,近红外光谱技术联合CARS-PLS-LDA方法可以有效判别纯山茶油和掺假山茶油,校正集、预测集及独立样本组样本的判别正确率、灵敏度及特异性均为100％.CARS-PLS-LDA判别模型性能优于未经变量优选的判别模型,表明CARS方法可以有效筛选重要波长变量,能简化判别模型及提高判别模型的稳定性和判别精度.本研究可为山茶油掺假快速检测提供理论依据.     

Comparison of Near Infrared Reflectance Spectroscopy with Combustion and Chemical Methods for Soil Carbon Measurements in Agricultural Soils

As interest in soil organic carbon (SOC) dynamics increases, so do needs for rapid, accurate, and inexpensive methods for quantifying SOC. Objectives were to i) evaluate near infrared reflectance (NIR) spectroscopy potential to determine SOC and soil organic matter (SOM) in soils from across Tennessee, USA; and ii) evaluate potential upper limits of SOC from forest, pasture, no-tillage, and conventional tilled sites. Samples were analyzed via dry-combustion (SOC), Walkley–Black chemical SOM, and NIR. In addition, the sample particle size was classified to give five surface roughness levels to determine effects of particle size on NIR. Partial least squares regression was used to develop a model for predicting SOC as measured by NIR by comparing against SOM and SOC. Both NIR and SOM correlated well (R² > 0.9) with SOC (combustion). NIR is therefore considered a sufficiently accurate method for quantifying SOC in soils of Tennessee, with pasture and forested systems having the greatest accumulations.Abbreviations SOC, soil organic carbon; NIR, Near Infrared Reflectance Spectroscopy; MTREC, Middle Tennessee Research and Education Center; RECM, Research and Education Center at Milan; PREC, Plateau Research and Education Center; PLS, Partial least squares.     

Measurement of Blend Concentrations of Conventional and Waxy Hard Wheats Using NIR Spectroscopy

Breeding development of waxy (amylose‐free) hard wheat lines adapted to the North American climate has been underway for more than a decade, with releases of competitive varieties imminent. Because of required identity preservation and a possible premium value placed on waxy lots, a rapid and accurate method is desired to identify and quantify the mixing of conventional wheat with waxy wheat, a condition that might occur at harvest or any point downstream. Our previous work demonstrated that lines pure with waxy starch can be identified from nonwaxy lines by use of near‐infrared (NIR) spectroscopy applied either on a whole kernel or ground meal basis. However, mixture quantification by NIR techniques has not been examined until now. Using hard winter wheat grown in two seasons (2011 and 2012) and at two locations (Nebraska and Arizona), a series of mixtures ranging in proportion (conventional/waxy) percentage by weight, from 0:100 to 100:0, were formed from nine pairs of waxy and nonwaxy varieties or lines, with year and location being consistent within a pair. Twenty‐nine mixtures (0, 1, 2, 3, 4, 5, 10, 15, …, 85, 90, 95, 96, 97, 98, 99, and 100%) were formed for each pair. Partial least squares regression models were developed by using eight of the nine pairs, with model validation accomplished by using the pair excluded. This procedure was repeated for each pair. The results indicate that, regardless of sample format or spectral pretreatment, the optimal models typically produce coefficients of determination in excess of 0.98, with standard errors of 4–7%, thus demonstrating the feasibility of the use of the NIR technique to predict the mixture level to within 10% by weight.     

Portable Probes to Measure Electrical Conductivity and Soil Quality in the Field

Abstract

Soil electrical conductivity (EC) is a useful indicator in managing agricultural systems, but tools for convenient and inexpensive measurements in the field are generally lacking. Handheld conductivity probes were designed to evaluate in‐field naturally occurring and human‐induced total soluble electrolyte levels in soil and water. The probes were used to survey and monitor EC in the field and to assess soil and water quality as related to environmental stability and sustainable food production. A pencil‐sized 16‐cm probe (PP) was connected to a handheld Hanna (DiST WP 4) conductivity meter, resulting in an economical, compact, and easy to use device. The tool provided accurate and precise results compared with laboratory instrumentation under standardized conditions of soil water content and temperature. Soil samples, varying widely in texture and organic matter content, and having ECs ranging from 0.13 to 2.32 dS m^?1 were used for comparison. Mean values and coefficients of variation were similar for the PP and the commercial laboratory EC meter with values determined with the two instruments being strongly correlated (r²=0.96–0.99). The handheld and PP probes effectively replaced expensive and cumbersome laboratory and field instruments used to measure EC in water and soil samples. The probe measurements were useful alternatives to conventional methods as they enabled accurate and precise measurement of EC, were a manageable size for field use, and were reliable and economic. The utility of EC as an indicator of soil health, plant‐available N, and environmental quality is also presented.     

Assessment of the species composition of forest floor horizons in mixed spruce-beech stands by Near Infrared Reflectance Spectroscopy (NIRS)

How the mixture of tree species modifies short-term decomposition has been well documented using litterbag studies. However, how litter of different tree species interact in the long-term is obscured by our inability to visually recognize the species identity of residual decomposition products in the two most decomposed layers of the forest floor (i.e. the Oe and Oa layers respectively). To overcome this problem, we used Near Infrared Reflectance Spectroscopy (NIRS) to determine indirectly the species composition of forest floor layers. For this purpose, controlled mixtures of increasing complexity comprising beech and spruce foliage materials at various stages of decomposition from sites differing in soil acid-base status were created. In addition to the controlled mixtures, natural mixtures of litterfall from mixed stands were used to develop prediction models. Following a calibration/validation procedure, the best regression models to predict the actual species proportion from spectral properties were selected for each tree species based on the highest coefficient of determination (R²) and the lowest root mean square error of prediction (RMSEP). For the validation, the R² (predictions versus true proportions) were 0.95 and 0.94 for both beech and spruce components in mixtures of materials at all stages of decomposition from the gradient of sites. The R² decreased only marginally by 0.04 when models were tested on independent samples of similar composition. The best models were used to predict the beech-spruce proportion in Oe and Oa layers of unknown composition. They provided in most cases plausible predictions when compared to the composition of the canopy above the sampling points. Thus, tedious and potentially erroneous hand sorting of forest floor layers may be replaced by the use of NIRS models to determine species composition, even at late stages of decomposition.     

近红外光谱结合小波变换预处理在蜂蜜产地快速检测中的应用

应用近红外光谱结合化学计量学方法对蜂蜜产地进行了判别分析。kennard-Stone法划分训练集和预测集。光谱用一阶导数加自归一化预处理后,再用小波变换（WT）进行压缩和滤噪。结合滤波后光谱信息,分别用径向基神经网络（RBFNN）和偏最小二乘-线性判别分析（PLS-LDA）建立了苹果蜜产地和油菜蜜产地判别模型。对不同小波基和分解尺度进行了详细讨论。对苹果蜜,WT-RBFNN模型和WT-PLS-LDA模型都是小波基为db1、分解尺度为2时的预测精度最好,都为96.2%。对油菜蜜：WT-RBFNN模型在小波基为db4和分解尺度为1时,预测精度最好;WT-PLS-LDA模型在小波基为db9、分解尺度也为1时,预测精度最好,为90.5%;预测精度WT-PLS-LDA模型优于WT-RBFNN模型。研究表明：WT结合线性的PLS-LDA建模比WT结合非线性的RBFNN建模更适于蜂蜜产地鉴别;近红外光谱结合WT-PLS-LDA可实现对蜂蜜产地的快速无损检测,为蜂蜜产地鉴别提供了一种新方法。     

烤烟烟叶钾含量的近红外光谱法快速测定

随机选取烤烟建模集样品(150个)和检验集样品(35个),利用傅里叶变换近红外光谱仪测定烤烟样品的近红外光谱,并用常规化学分析法测定烤烟样品的含钾量。采用偏最小二乘法(PLS)把测得的烤烟样品的光谱值与烤烟钾含量的数值拟合建立定标模型,经分析得出:预测模型分析烤烟钾含量的决定系数(R2)为0.909,预测标准差(RMSEP)为0.119%。近红外法测定结果与常规化学分析方法的结果具有较好的相关性,能够应用于烤烟钾含量的快速诊断。     

基于近红外漫反射光谱的云芝提取物多糖含量快速检测研究

为探索快速无损测定云芝提取物中多糖含量的方法,通过采集粉末状云芝提取物近红外光谱,经预处理和波段选择,结合间隔偏最小二乘法(iPLS)和反向区间偏最小二乘法(Bi-PLS),建立并优化云芝提取物多糖含量检测模型。结果表明,光谱区间为9 365.92~8 918.76 cm~(-1)和5 341.48~4 894.32cm~(-1),二阶导数(SD)预处理后,建立的反向区间偏最小二乘法模型更优,其校正决定系数(R_(cal))、校正均方根差(RMSECV)、验证决定系数(R_(val))和验证均方根差(RMSEP)可分别达到0.9089、0.00781、0.9879和0.00292。该模型可以更有效地优选建模所需波段,降低模型复杂度,降低多糖含量的检测成本,提高检测效率,实现云芝提取物多糖含量的快速、无损检测。     

基于极限学习机利用可见-近红外光谱数据判别土壤类型的方法研究

为了实现土壤类型的快速无损识别,提出了一种利用可见-近红外光谱、基于极限学习机的土壤类型鉴别方法。首先,获取4种不同类型土壤的320个样本波长在325～1075 nm范围内的可见-近红外光谱数据;其次,用主成分分析的数学方法对数据进行降维处理,最终提取了三个主成分来代表原光谱数据;再次,将320个样本的数据随机分为测试集和预测集两个部分,建立极限学习机模型,利用该模型对土壤类型进行识别。实验结果表明,将极限学习机应用于土壤类型的识别精度可达100%,其训练速度和泛化性优于BP神经网络和支持向量机,能够快速、准确、无损鉴别土壤类型,使用方便,具有推广价值。     

近红外漫反射光谱技术快速无损识别灵芝和云芝提取物研究

为了满足食用菌提取物实际生产监管需要,本研究采用近红外漫反射光谱技术对来自不同地区的灵芝和云芝提取物样品进行定性识别研究。在800~2 750nm波段范围,采集灵芝和云芝提取物的漫反射光谱,应用主成分聚类分析和偏最小二乘判别法分别建立识别模型,用146个样品进行建模和48个外部样品集进行验证。结果表明:采用主成分聚类判别分析法,灵芝和云芝提取物的判别界线清晰,正确率达到88.54%;采用偏最小二乘判别法,建立的鉴别分类模型能较好地对灵芝和云芝提取物进行鉴别,校正集和预测集样品的识别正确率均为100%。因此,近红外结合主成分聚类分析和偏最小二乘判别法识别灵芝和云芝提取物是可行的,同时研究结果为灵芝和云芝提取物的快速识别提供了理论依据和使用方法。     

不同傅里叶变换红外光谱法研究土壤光谱特征的比较与应用

为了将傅里叶变换红外光谱技术更好地应用于土壤中官能团的定量分析,通过比较3种不同傅里叶变换技术下土壤特征吸收峰的差异,来选取最佳的光谱技术应用于土壤的相关研究。采用透射(T-FTIR)、衰减全反射(ATR-FTIR)和漫反射(DR-FTIR)3种光谱技术分别对有机物(苯甲酸、硬脂酸)和土壤(辽东栎)进行了分析。并将辽东栎和草地土壤样品按不同比例混合,使用T-FTIR和DR-FTIR对混合土样进行光谱测定,用于定量分析研究。结果表明:(1)有机物样品在3种光谱技术中均出现特征吸收峰,苯甲酸在1 600～1 400 cm^-1出现苯环C=C骨架特征峰,硬脂酸在2 900～2 800 cm^-1显现甲基的特征峰。有机物中羧酸-COOH内羰基-C=O在1 720～1 680 cm^-1出现伸缩振动吸收峰,羟基-OH分别在1 430～1 410 cm^-1,940～930 cm^-1附近出现面内和面外弯曲振动吸收峰。有机物在T-FTIR技术中需要用溴化钾(KBr)对其稀释。(2)辽东栎土壤样品在T-FTIR和DR-FTIR技术测试中发现,其土壤谱图中有较多有效特征吸收峰,土壤样品在T-FTIR技术中也需要用KBr进行稀释并使样品均匀的分布在锭片中; 而ATR-FTIR技术测试中仅出现个别有效特征吸收峰,不利于对土壤谱图鉴别与进一步分析。(3)辽东栎和草地土壤样品按不同比例混合的测试结果表明:T-FTIR和DR-FTIR技术测试中质量分数与峰面积比呈正相关,线性拟合分别为R²=0.70和R²=0.88。土壤在3种不同红外技术中,DR-FTIR光谱具有较好的土壤特征吸收峰,对土壤样品可以不用KBr稀释。测试步骤简单易操作,可用于土壤样品定量分析研究。     

Rapid Analysis of Apple Leaf Nitrogen using Near Infrared Spectroscopy and Multiple Linear Regression

A quick method was developed for diagnosis of nitrogen (N) in apple trees based on multiple linear regressions to establish the relationship between near-infrared reflectance spectra (NIRS) and the N contents of fresh and dry tissue. Spectral pretreatment methods such as derivatives, smoothing, and normalization were used. The derivatives appeared to be the most effective. The best calibration for fresh leaf gave 0.842 for the correlation coefficient of validation (R_v), 1.119 g kg^?1 for the root mean square error of prediction (RMSEP), and 8.311 for the ratio of the range in reference data from the validation samples to the root mean square error of prediction (RER). The best calibration for dried ground samples was obtained with R_v = 0.952, RMSEP = 0.633 g kg^?1, the ratio performance deviation (RPD) = 3.27, and RER = 13.728. The results showed that calibrations of dry-apple leaf are robust enough for an accurate prediction of N.     

基于近红外光谱技术的灵芝提取物掺假定量快速无损检测研究

为了探索快速测定灵芝提取物掺假(淀粉)含量的方法,采用近红外光谱扫描掺杂0%、5%、10%、20%、40%、60%、80%淀粉的灵芝提取物,对其光谱进行预处理和波段选择,并结合偏最小二乘法(PLS)建立灵芝提取物掺假定量快速无损检测方法。结果表明,使用多元散射校正(MSC)预处理方法,波数范围8 000~7 500、6 000~5 500和5 000~4 000cm~(-1),主因子数为8时,建立的偏最小二乘法模型的校正决定系数(R_(cal))、校正均方根差(RMSECV)、验证决定系数(R_(val))和验证均方根差(RMSEP)分别为0.9962、0.0249、0.9960和0.0241。运用该模型对验证集样品进行预测并统计分析,可知预测值与实际掺假值之间无显著差异。本研究为灵芝功能食品市场检测提供了方法基础。     

应用伽马射线和可见近红外光谱测定土壤容重

现有的土壤容重测定方法存在诸多不足,不能满足快速发展的精准农业、生态环境模拟、土壤碳储量估算等对大量、准确容重数据的需求。鉴于此,有研究提出γ射线衰减与可见-近红外并用测定土壤容重的方法,并成功地将该方法应用于土壤碳储量估算中,得到了较好的应用结果。为了检验该方法在我国南方丘陵区土壤容重测定上的准确性和适用性,本研究采集了广西南宁丘陵区的土壤样品,使用该方法计算出土壤容重,并与传统环刀采样烘干称重法的容重结果进行比较。结果表明,两者测定的土壤容重具有较高的回归决定系数,R~2可达0.92,且两者间的均方根误差较低,仅占土壤容重平均值的4.48%。因此,本研究认为,γ射线衰减与可见-近红外并用测定土壤容重在我国南方丘陵区有较好的准确性和适用性。     

基于可见/近红外反射光谱的梨树叶片钾含量的快速测定研究

利用可见/近红外反射光谱定量分析技术对梨树鲜叶钾素含量进行快速测定研究。对150个梨树叶片样本进行光谱扫描,其中120个做建模集,30个做验证集。通过对样品的可见/近红外光谱进行多种预处理,并建立钾素预测模型,探讨了可见/近红外光谱数据预处理对预测精度的影响。结果表明,通过原始光谱与S-G(3)平滑相结合的预处理方法,用17个主成分建立的偏最小二乘法模型最好,其交叉验证集和预测集模型的决定系数(R2)分别为0.722 7和0.679 1,交叉验证均方根误差(RMSECV)为1.171,预测的平均相对误差为6.81%,能高效、快速地预测梨树叶片钾素含量,为梨树钾素快速测定提供了新的手段。