Determination of Copper and Zinc Pollutants in Ludwigia prostrata Roxb Using Near-Infrared Reflectance Spectroscopy (NIRS)

Abstract

The feasibility of using near-infrared reflectance spectroscopy (NIRS) to determine the concentrations of copper (Cu) and zinc (Zn) in Ludwigia prostrata Roxb plants was investigated. Ludwigia prostrata Roxb plants were grown over a full growth cycle under controlled laboratory conditions in soils contaminated with heavy metals. The Cu and Zn concentrations in 72 L. prostrata Roxb samples were analyzed using fame atomic absorption spectrometry, and NIRS spectra were collected in the 1099–2500 nm range. Five mathematical treatments of the spectral data were compared prior to developing the calibration models (n = 48) using partial least squares regression methods. The two calibration models for Cu and Zn concentrations were evaluated according to the correlation coefficient of cross-validation (R_cv) and root mean squares error of cross-validation. The highest R_cv and the lowest RMSECV were obtained for Cu (0.9 and 7.24 mg kg⁻¹) and Zn (0.94 and 19.17 mg kg⁻¹), respectively. The results showed that near-infrared diffuse refectance spectroscopy can be used for the rapid determination of Cu and Zn in leaves of L. prostrata Roxb plants.

Keywords

Heavy metals Near-infrared reflectance spectroscopy NIRS Near-infrared diffuse reflectance spectroscopy NIDRS Partial least squares PLS

INTRODUCTION

Heavy metals are non-biodegradable and accumulate in nature, and metal emissions and their deposition over time can lead to anomalous enrichment and contamination of the surface environment. Heavy metal contamination in soils is a major environmental problem, one that poses significant risks to human health as well as to ecosystems.^1,2 Many scientific activities have been devoted to the determination of sources, types, and degree of heavy metal pollution in soils.^3,4 As a result of uncontrolled industrial activities, many hazardous chemical substances, including many heavy metals, have been released leading to the deterioration of the ecosystem.^5,6 Heavy metal contamination in the environment poses serious health risks because many heavy metals accumulate in living tissues throughout the food chain.^7,8

A number of studies have focused on the influence of heavy metals in contaminated soils of plants. The toxic levels of some heavy metals appear to be the result of environmental pollution.^9–11 Much concern has been expressed by those working in the environmental sciences, since Cu and Zn widely coexist with other heavy metals in polluted soils. Reduction of biomass production and nutritional quality has been observed in crops grown in soils contaminated with Cu and Zn,¹² and photosynthesis, transpiration, carbohydrate metabolism, and other metabolic activities have been found to be inhibited by Cu and Zn.^13,14 Ludwigia prostrata Roxb is a common weed, widely distributed in South Asia. The plant has proven very useful in the enrichment and tolerance of Cu and Zn contained in sewage polluted areas.¹⁵ In polluted soil near smelters, the total and extractable Cu and Zn of L. prostrata Roxb in soil samples measuring in the 100 m² range were 9.3 times and 34 times higher than those in pollution-free areas, respectively.¹⁶ The order of average concentrations of Cu and Zn in the parts of L. prostrata Roxb was roots > leaves > stems.¹⁷ The amounts of Cu or Zn in L. prostrata Roxb show significant correlation with their concentration of Cu or Zn in soil of the sewage irrigation area. For all these reasons, L. prostrata Roxb was selected as experimental material in this study.

Among spectroscopic techniques, near-infrared reflectance spectroscopy (NIRS) has been used as a rapid, inexpensive, and accurate method for analyzing petrochemicals, pharmaceuticals, and agricultural products.^18–20 The NIRS spectrum of an organic material can give a global signature of composition based basically on the assessment of the organic chemical structures containing O–H, N–H, and C–H bonds, which, with the application of chemometric techniques, can be used to elucidate particular compositional characteristics in the food matrix not easily detected by targeted chemical analysis.^21,22

Near-infrared reflectance spectroscopy has been proven an efficient, rapid, and nondestructive method for quantitatively analyzing complex samples that does not require sample preparation. It has extensive application potential in the environmental monitoring field.²³ Theoretically, inorganic compounds cannot produce a direct signal in the NIRS spectrum. Nevertheless, predicting heavy metals in herbaceous plants and agricultural products can still be possible when using NIRS because heavy metals interfere with the spectral features of organic compounds or hydrated inorganic molecules.^24,25 Using NIRS to detect heavy metals is generally dependent on the reactions occurring between those heavy metals and the organic or hydrated compounds that produce features of a metal organic complex in the NIRS region. Some reports show that NIRS was successfully used to determine the concentrations of calcium (Ca), potassium (K), and iron (Fe) in wine, and there are also a few reports of NIRS being used successfully to detect heavy metals in fresh or dried plant samples.²⁶ NIRS has also been used to measure S, Na, and B in Trifolium repens and Medicago sativa Linn²⁷ and to provide rapid quantitative analyses of Mo, Ca, Mg, Fe, and Mn elements in Chinese southwest tobacco (Nicotiana × sanderae).²⁸

Mineral analysis commonly involves using inductively coupled argon plasma, atomic absorption spectroscopy (AAS), and X-ray fluorescence spectroscopy where the sample preparation requirements can be time-consuming and in some cases are not environmentally friendly procedures. A study of accumulative characteristics of heavy metals in soil L. prostrata Roxb system has been reported²⁹ that concluded that the distribution of heavy metals in L. prostrata Roxb in sewage irrigation areas for Cu is root > leaf > stem and for Zn is stem > root > leaf. Zn is more mobile than Cu, which is likely to accumulate in the root of L. prostrata Roxb. The concentration of Cu in the L. prostrata Roxb stem has significant correlation with the concentration of Cu in the soil of sewage irrigation areas.

There are no studies using NIRS for the detection of Cu and Zn in L. prostrata Roxb plants in the literature. In this study, a method for quantitative analysis of Cu and Zn in L. prostrata Roxb was investigated. Partial least squares regression (PLS) was used for the calibration method due to the complexity of the spectra. The main purpose was to explore the potential and accuracy of the near-infrared diffuse reflectance spectroscopy (NIDRS) method for quantitative determination of Cu and Zn in L. prostrata Roxb. Dried and ground L. prostrata Roxb leaves were used directly for the NIDRS method to simplify the sample preparation. In addition, different pretreatments were compared to obtain the ideal chemometric models.

MATERIALS AND METHODS

Samples Preparation. Seventy-two soil samples of varying characteristics (texture, organic matter, N content, pH) were collected from various locations on several farms in Nanchang, Jiangxi, China. Soil samples were sieved with a 2 mm sieve and air-dried for 3 days. Fertilizers including KH₂PO₄ (0.41g) and NH₄NO₃ (0.41g) were added to the soil to provide 75 mg P kg⁻¹ of soil, 100 mg K kg⁻¹ of soil, and 100 mg N kg⁻¹ of soil as NH₄NO₃.³⁰ Then Cu (NO₃)₂ and Zn (NO₃)₂ solutions were added to the soils (about 30 kg). The soil samples were completely mixed, equilibrated for 10 days, and air-dried for several days. This was repeated several times to obtain contaminated soil samples with different levels of Cu and Zn (0, 125, 250, 500, 750, 1000 mg kg⁻¹) for growing L. prostrata Roxb.

The soil samples (5 kg) were placed in plastic pots (150 mm diameter × 250 mm height). Each treatment was replicated with four pots for experimental validation. The moisture level of the soil was kept to near field water content (40%) and equilibrated for two weeks. Three seedlings of L. prostrata Roxb (8 cm roots, 3 cm roots) were replicated in four pots and watered daily to 60% of the field water content.

After 180 days of growth, L. prostrata Roxb of each treatment were carefully removed from each pot at harvest and were washed thoroughly to remove adhered soil by a quick wash in deionized water. Samples (n = 72) were assigned for calibration (n = 48) and validation (n = 24) sets in the ratio of 2: 1 by the Kennard–Stone (KS) algorithm before being divided into roots and leaves.

Determination of Cu and Zn Concentrations. The roots and leaves of L. prostrata Roxb were oven-dried at 70 °C until their weights were constant. All of the samples (72 roots and leaves) were ground with agate mortars, sieved to pass through a 100 mesh sieve, and kept in sealed storage in polyethylene bags. Cu and Zn concentrations in each sample were determined (digested with conc. HNO₃, and conc. HClO₄ = 5:1) with flame AAS (Analyst 300, Perkin Elmer, Germany).³¹

Near-Infrared Spectroscopy Data Collection and Analysis. Spectra from 1099 nm to 2500 nm were collected in diffuse reflectance mode with an NIR spectrometer (TENSOR 37, Bruker Technical and Service Co., Germany) coupled with an integrating sphere diffuse reflectance accessory. Powder from each plant tissue sample was loaded into a quartz bottle. The collected spectra were digitalized at about 1 cm⁻¹ intervals, so each spectrum comprised 4148 data points. Before the measurement of samples, a reference spectrum was obtained with a gold-filled material provided within the spectrometer. To increase the signal-to-noise ratio, both reference and sample spectra were the average of 32 scans. Three parallel measurements were taken, and the average of the spectra was used for PLS model development. In addition, the samples were scanned at 20 °C in random order to decrease the systematic errors of the instrument.

Statistical Analysis and Interpretation. The Cu and Zn concentration in leaves, roots, and soil were analyzed with OriginPro 8.5 (OriginLab, USA) within each treatment. The Cu and Zn accumulation concentrations were expressed by means ± standard deviation (SD) of 12 replicates within each treatment. Student's t-test was applied to determine the significance of results between the different samples. The statistical significance was set at the P < 0.05 confidence level.

Calibration models between heavy metals and NIRS spectra were developed using PLS regressions with leave-one-out cross-validation (LOOCV). In order to obtain the optimized model, transform baseline, first derivative calculation, multiplicative scatter correction (MSC), and standard normal variate (SNV) were used to develop Cu and Zn prediction models, separately. The data were analyzed by using The Unscrambler 10.0 (CAMO AS, Norway) and OriginPro 8.5 statistical package software. Model performance was estimated and judged by comparing the predicted and measured value data in the validation set.

The performance of the PLS calibration model was evaluated in terms of the correlation coefficients (R_cv)of cross-validation and root mean square error of cross-validation (RMSECV), which was determined by LOOCV.^32,33 The factor number in the PLS model was determined by using the LOOCV with an F test. RMSECV and R_cv are defined as follows:

R_{CV} = \sqrt{1 - [\frac{\sum_{i = 1}^{n} {(y - {\hat{y}}_{i})}^{2}}{\sum_{i = 1}^{n} {(y_{i} - y_{m})}^{2}}]} \times 100 %

(1)

RMSECV = \sqrt{\frac{1}{n} \sum_{i = 1}^{n} {(y_{i} - {\hat{y}}_{i})}^{2}}

(2)

where y_i is the reference value of the ith sample, ŷ_i is the predicted value of the ith sample, y_m is the average of the referenced value, and n is the number of samples.

The residual predictive deviation (RPD) was recorded, which is the ratio of the standard deviation of the reference data to the standard error of prediction of the model.³⁴ The RPD values were classified as follows: RPD < 1.0 indicates a very poor model or predictions and their use is not recommended; RPD between 1.0 and 2.5 indicates a poor model or predictions where only high and low values are distinguishable; RPD between 2.5 and 5.0 is considered fair and recommended for screening purposes; and RPD > 5.0 indicates a good quantitative model or prediction.³⁵

RESULTS AND DISCUSSION

Effects of Cu- and Zn-Contaminated Soils on Plants. During the growth cycle, the effects of Cu and Zn on the external appearance of the leaves of L. prostrata Roxb were very obvious. For plants growing in the soils with relatively low heavy metal-level treatments (125 and 250 mg kg⁻¹), the leaves of the plant were bigger, greener, and denser compared to those growing in higher metal-contaminated soils. The leaves of the plants growing in the soils with a medium level of heavy metals (500 mg kg⁻¹) had chlorosis, elongation, and curling leaf edges. Those growing in soils with high levels of heavy metals (750–1000 mg kg⁻¹) had necrotic leaf tips. The changes of L. prostrata Roxb leaves observed in this study are common responses of plants to stress, such as metal pollution. Similar symptoms were also observed by Audet and Charest³⁶ for tobacco plants subjected to Cu and Zn pollution.

The Cu and Zn concentrations in plant leaves increased with an increased amount of Cu and Zn in soils. The Cu concentration in the plant leaves significantly increased from 55.17 ± 11.44 in the control treatment to 67.42 ± 12.15 mg kg⁻¹ in the 250 Cu mg kg⁻¹ soil, 77.93 ± 11.27 in 500 Cu mg kg⁻¹ soils, and 88.97 ± 14.82 mg kg⁻¹ in 750 Cu mg kg⁻¹ soils. The Cu concentration at 107.95 ± 12.61 mg kg⁻¹ was obtained in 1000 Cu mg kg⁻¹ soils. The same trends were also observed in Zn pollution experiments. Along with an increase of Zn in soils, Zn concentration in leaves significantly increased from 161.51 ± 12.65 in the control treatment to 262.26 ± 10.68 mg kg⁻¹ in the 1000 Cu mg kg⁻¹ soil. However, the Zn concentration in leaves of L. prostrata Roxb was about 2.5-fold higher than the Cu concentration, although the plants grew in the soils with the same amount of Cu and Zn. The addition of Zn in soils significantly triggered Zn accumulation in L. prostrata Roxb leaves.

Heavy Metals Concentration Statistics. The range, mean, and SD for the concentration of Cu and Zn in L. prostrata Roxb leaves measured in both calibration and validation sets are shown in Table I. The concentration ranges of Cu and Zn in the calibration set used to build the PLS models were 45.73–120.56 and 47.94–296.91 mg kg⁻¹, respectively. In the validation set, the concentration ranges of the Cu and Zn were 46.73–119.53 and 48.94–296.71 mg kg⁻¹, respectively. Heavy metals in farm soils have a wide concentration range due to different plant species, soil characteristics, and farm management practices. The results of the samples are consistent with the values usually found for this type of sample. Therefore, the variability in elemental composition in both calibration and validation sets was considered suitable for developing NIRS calibrations for Cu and Zn concentration.

TABLE I.

Statistics of the concentrations of Cu and Zn (mg kg⁻¹) of samples in calibration and validation sets.

	Calibration				Validation

Analyte	n^a	Range	Mean	SD^b	n	Range	Mean	SD
Cu (mg kg⁻¹)	48	45.73–120.56	69.67	19.04	24	46.73–119.53	68.22	19.55
Zn (mg kg⁻¹)	48	47.94–296.91	205.29	57.44	24	48.94–296.71	203.28	57.72

n = sample numbers.

SD = standard deviation.

Relationship Between Spectral Properties and Heavy Metals Concentration. Different metal concentrations in L. prostrata Roxb leaves did not cause the obvious change of spectrum with different metal concentrations. From Fig. 1, it can be seen that there is no obvious change in the spectra after the uptake of the ions by plant. Cu and Zn accumulation had an effect on growth, lipid peroxidation, anti-oxidative enzymes, leaf chlorophyll, root activity, and protein content, due to Cu and Zn being taken up into the plant root system through nutrient metabolic pathways. It is not a simple linear relationship between the spectra of powder samples and concentrations of the heavy metal ions. It is difficult to use the ordinary calibration method for quantitative analysis. Therefore, multivariate calibration was used in this study.

Fig. 1.

NIRS of L. prostrata Roxb samples.

Quantitative Model and Partial Least Squares Loadings. Table II shows the results of PLS calibration models with different spectral pretreatments including first derivative calculation, MSC, and SNV for Cu and Zn concentration in leaves. Both the background removal (transform baseline) and the background correction (SNV and MSC) techniques can improve the performance of the PLS models of Zn. As noted in Table II, only SNV can improve the PLS models of Cu. The optimized pretreatment was SNV because the RMSECV values of Cu (6.82 mg kg″¹) and Zn (18.51 mg kg″¹) were all relatively low, and the optimized PLS factor numbers were found to be 16 and 14 for Cu and Zn, respectively.

TABLE II.

PLS models and the calculation results of results of leave-one-out cross-validation and the prediction set.

Analyte	Preprocessing	PLS factors	Slope	R_cv	RMSECV^a	RPD^b
Cu (mg kg⁻¹)	Raw spectra	18	0.84	0.93	6.94	2.7
	Transform baseline	17	0.82	0.91	7.92	2.4
	First derivative	9	0.65	0.81	11.14	1.7
	MSC^c	16	0.83	0.93	7.07	2.7
	SNV^d	16	0.84	0.93	6.82	2.8
Zn (mg kg⁻¹)	Raw spectra	17	0.87	0.89	20.22	2.8
	Transform baseline	16	0.89	0.93	18.77	3.1
	First derivative	9	0.79	0.92	22.84	2.5
	MSC^c	13	0.92	0.95	18.87	3.0
	SNV^d	14	0.90	0.94	18.51	3.1

RMSECV = root mean square error of cross-validation.

RPD = ratio performance deviation.

MSC = multiplicative scattering correction.

SNV = vector normalization.

A qualitative interpretation of the RPD was provided by the authors of this paper in which NIRS calibration models with RPD values greater than 2.5 were considered fair and recommended for screening purposes. In the study, the RPDs obtained for the calibration models were higher than 2.5, so the models would be considered fair and could be used as screening tools to diagnose pollution levels of Cu and Zn in L. prostrata Roxb.

Cu and Zn can be combined with diverse organic complexes such as chelates, which represent the main source of variance for Cu and Zn treatments. Using NIRS we are able to track the Cu and Zn concentration by response. Figure 2 shows PLS loadings for Cu and Zn measured in L. prostrata Roxb leaves using NIDRS. The observed valleys and peaks can be attributed to overtone bands related to molecular groups in the matrix. Loadings in the NIRS region show the wavenumber at 7140 cm⁻¹, between 6250 and 5882 cm⁻¹ (around 6010 and 5950) related with O–H and C–H tones (e.g., water, sugars, and ethanol), around 4440 cm⁻¹ related with C–H and –CH2 combinations (e.g., cellulose and carbohydrates), between 6540 and 6250 cm⁻¹ (around 6365) related with N–H overtones(e.g., secondary amide and proteins), and between 4762 and 4545 cm⁻¹ (around 4651) related with C–H combinations and overtones (e.g., benzene and aromatic series).^37,38 Additionally, the other loadings were difficult to assign to any particular bond or chemical structure.

Fig. 2.

Partial least squares (PLS) loadings with SNV for Cu (red line) and Zn (black line) measured in L. prostrata Roxb.

Validation of the Method. Optimized PLS models for Cu and Zn were established using PLS regression with SNV pretreatment. In order to test the feasibility of the model, the validation set was further done with 24 validation samples. The scatter diagrams of near-infrared predicted values versus reference values for Cu and Zn are shown in Fig. 3 and Fig. 4, respectively. The correlation coefficient (r) and SD were found to be 0.93 and 7.24 mg kg⁻¹ for the Cu model and 0.94 and 19.17 mg kg⁻¹ for the Zn model, respectively, corresponding to the relatively satisfactory linear relationship in the scatter diagrams. The corresponding RPD was 2.7 and 3.0 for Cu and Zn. Therefore, these two heavy metals had better correlation with the spectra of L. prostrata Roxb samples, and the quantitative models could be accepted.

Fig. 3.

Near-infrared predicted values versus reference values for Cu in L. prostrata Roxb (n = 24).

Fig. 4.

Near-infrared predicted values versus reference values for Zn in L. prostrata Roxb (n = 24).

CONCLUSION

Some relationships exist between NIRS spectra and the concentration of heavy metals in plants. The study has shown that NIRS can be used to determine the concentrations of Cu and Zn in L. prostrata Roxb plants. Five mathematical treatments of the spectral data were compared prior to developing the calibration models using PLS regression methods. The best calibration models for Cu and Zn concentrations were obtained for Cu of R_cv = 0.9 and RMSECV = 7.24 mg kg⁻¹, Zn of R_cv = 0.94 and RMSECV = 19.17 mg kg⁻¹. The study will offer useful information for determining Cu and Zn concentrations in plants.

Footnotes

ACKNOWLEDGMENTS

The authors gratefully acknowledge the financial support provided by National Science and Technology Support Program (31160250, 61178036), Graduate Students Innovation Foundation (YC 2013-S157), Center of Photoelectric Detection Technology Engineering of Jiangxi Province (2012-155).

References

Nahmani

Lavelle

. “Effects of Heavy Metal Pollution on Soil Macrofauna in a Grassland of Northern France”. Eur. J. Soil Biol. 2002. 38(3–4): 297–300.

Škrbić

Durišić-Mladenović

“Chemometric Interpretation of Heavy Metal Patterns in Soils Worldwide”. Chemosphere. 2010. 80(11): 1360–1369.

Lee

C.S.

X.D.

Shi

W.Z.H.

Cheung

S.C.

Thornton

. “Metal Contamination in Urban, Suburban, and Country Park Soils of Hong Kong: A Study Based on GIS and Multivariate Statistics”. Sci. Total Environ. 2006. 356(1–3): 45–61.

Franco-Uría

López-Mateo

Roca

Fernández-Marcos

M.L.

. “Source Identification of Heavy Metals in Pasture Land by Multivariate Analysis in NW Spain”. J. Hazard. Mater. 2009. 165(1–3): 1008–1015.

Morselli

Brusori

Passarini

Bernardi

Francaviglia

Gataleta

Marchionni

Aromolo

Benedetti

Olivieri

. “Heavy Metals Monitoring at a Mediterranean Natural Ecosystem of Central Italy. Trends in Different Environmental Matrixes”. Environ. Int. 2004. 30(2): 173–181.

Frey

Stemmer

Widmer

Luster

Sperisen

. “Microbial Activity and Community Structure of a Soil After Heavy Metal Contamination in a Model Forest Ecosystem”. Soil Biol. Biochem. 2006. 38(7): 1745–1756.

Angelova

Ivanova

Delibaltova

Ivanov

. “Bio-Accumulation and Distribution of Heavy Metals in Fibre Crops (Flax, Cotton and Hemp)”. Ind. Crops Prod. 2004. 19(3): 197–205.

Bai

L.Y.

Zeng

X.B.

L.F.

Pen

S.H.

. “Effects of Land Use on Heavy Metal Accumulation in Soils and Sources Analysis”. Agric. Sci. China. 2010. 9(11): 1650–1658.

Smith

S.R.

. “A Critical Review of the Bioavailability and Impacts of Heavy Metals in Municipal Solid Waste Composts Compared to Sewage Sludge”. Environ. Int. 2009. 35(1): 142–156.

10.

Park

B.-Y.

Lee

J.-K.

H.-M.

Kim

Y.-H.

. “Effects of Heavy Metal Contamination from an Abandoned Mine on Nematode Community Structure as an Indicator of Soil Ecosystem Health”. Appl. Soil Ecol. 2011. 51: 17–24.

11.

Gunawardena

Egodawatta

Ayoko

G.A.

Goonetilleke

. “Role of Traffic in Atmospheric Accumulation of Heavy Metals and Polycyclic Aromatic Hydrocarbons”. Atmos. Environ. 2012. 54: 502–510.

12.

Liu

H.Y.

Probst

Liao

B.H.

. “Metal Contamination of Soils and Crops Affected by the Chenzhou Lead/Zinc Mine Spill (Hunan, China)”. Sci. Total Environ. 2005. 339(1–3): 153–166.

13.

Aibibu

Liu

Y.G.

Zeng

G.M.

Wang

Chen

B.B.

Song

H.X.

. “Cadmium Accumulation in Vetiveria zizanioides and Its Effects on Growth, Physiological and Biochemical Characters”. Bioresour. Technol. 2010. 101(16): 6297–6303.

14.

Ayeni

Ndakidemi

Snyman

Odendaal

. “Chemical, Biological and Physiological Indicators of Metal Pollution in Wetlands”. Sci. Res. Essays. 2010. 5(15): 1938–1949.

15.

Öncel

Keles

Üstün

A.S.

. “Interactive Effects of Temperature and Heavy Metal Stress on the Growth and Some Biochemical Compounds in Wheat Seedlings”. Environ. Pollut. 2000. 107(3): 315–320.

16.

Chen

Shen

. “Leaching and Uptake of Heavy Metals by Ten Different Species of Plants During an EDTA-Assisted Phytoextraction Process”. Chemosphere. 2004. 57(3): 187–196.

17.

Wang

Zhang

Wang

. “Effects of Cu and Zn Stress on Physiological Indications of Ludwigia prostrate Roxb”. Nanjing Linxueyuan Xuebao. 2009. 33(4): 43–47.

18.

Candolfi

Maesschalck

Massart

Hailey

Harrington

. “Identification of Pharmaceutical Excipients Using NIR Spectroscopy and SIMCA”. J. Pharm. Biomed. Anal. 1999. 19(6): 923–935.

19.

Balabin

Safieva

. “Gasoline Classification by Source and Type Based on Near Infrared (NIR) Spectroscopy Data”. Fuel. 2008. 87(7): 1096–1101.

20.

Sun

Cai

Hao

Ouyang

Liu

. “Visible-NIR Spectroscopy and Least Square Support Vector Machines Regression for Determination of Vitamin C of Mandarin Fruit”. Sens. Lett. 2012. 10(1–2): 506–510.

21.

Osborne

B.G.

. “Theory of Near Infrared Spectrophotometry”. In: Osborne

B.G.

Fearn

Hindle

P.H.

, editors. Practical NIR Spectroscopy with Applications in Food and Beverage Analysis. Essex, UK: Longman Scientific and Technical, 1993. 2nd ed. Pp. 227.

22.

W.Z.

. “The Basic Principle of Near Infrared Absorption Spectrum”. In: Lu

W.Z.

, editor. Modern Near Infrared Spectroscopy Analytical Technology. Beijing, China: China Petrochemical Press, 2007. 2nd ed.

23.

Teixeira dos Santos

C.A.

Lopo

Páscoa

R.N.

Lopes

J.A.

. “A Review on the Applications of Portable Near-Infrared Spectrometers in the Agro-Food Industry”. Appl. Spectrosc. 2013. 67(11): 1215–1233.

24.

Clark

Mayland

Lamb

. “Mineral Analysis of Forages with Near Infrared Reflectance Spectroscopy”. Agron. J. 1987. 79(3): 485–490.

25.

Cozzolino

Cynkar

Shah

Smith

. “Quantitative Analysis of Minerals and Electric Conductivity of Red Grape Homogenates by Near Infrared Reflectance Spectroscopy”. Comput. Electron. Agr. 2011. 77(1): 81–85.

26.

Cozzolino

Kwiatkowski

M.J.

Dambergs

R.G.

Cynkar

W.U.

Janik

L.J.

Skouroumounis

Gishen

. “Analysis of Elements in Wine Using Near Infrared Spectroscopy and Partial Least Squares Regression”. Talanta. 2008. 74(4): 711–716.

27.

Cozzolino

Moron

. “Exploring the Use of Near Infrared Reflectance Spectroscopy (NIRS) to Predict Trace Minerals in Legumes”. Anim. Feed Sci. Technol. 2004. 111(1–4): 161–173.

28.

Duan

Huang

Zheng

Xiong

Min

. “Determination of 27 Chemical Constituents in Chinese Southwest Tobacco by FT-NIR Spectroscopy”. Ind. Crops Prod. 2012. 40: 21–26.

29.

Aibibu

Liu

Y.G.

Song

H.X.

Chen

B.B.

. “Effects of Zn and Cu on Physiological and Biochemical Processes and Their Accumulation Characteristics of Vetiver”. J. Agro-Environ. Sci. 2010. 29(1): 54–59.

30.

Chen

Shen

. “The Use of Vetiver Grass (Vetiveria zizanioides) in the Phytoremediation of Soils Contaminated with Heavy Metals”. Appl. Geochem. 2004. 19(10): 1553–1565.

31.

Yang

Shu

Lan

Wong

. “Growth and Metal Accumulation in Vetiver and Two Sesbania Species on Lead/Zinc Mine Tailings”. Chemosphere. 2003. 52(9): 1593–1600.

32.

Haaland

Thomas

. “Partial Least-Squares Methods for Spectral Analyses. 1. Relation to Other Quantitative Calibration Methods and the Extraction of Qualitative Information”. Anal. Chem. 1988. 60(11): 1193–1202.

33.

Abdi

. “Partial Least Squares Regression and Projection on Latent Structure Regression (PLS Regression)”. Wiley Interdise. Rev.: Comput. Stat. 2010. 2(1): 97–106.

34.

Næs

Isaksson

Fearn

Davies

. A User Friendly Guide to Multivariate Calibration and Classification. Chichester, UK: NIR Publications, 2002.

35.

Martens

. “T. Næs. “Multivariate Calibration by Data Compression”. In: Williams

Norris

, editors. Near-Infrared Technology in the Agricultural and Food Industries. St. Paul, MN: American Association of Cereal Chemists, 2001. 2nd ed.

36.

Audet

Charest

. “Effects of AM Colonization on Wild Tobacco Plants Grown in Zinc-Contaminated Soil”. Mycorrhiza. 2006. 16(4): 277–283.

37.

Suszcynsky

Shann

. “Phototoxicity and Accumulation of Mercury in Tobacco Subjected to Different Exposure Routes”. Environ. Toxicol. Chem. 1995. 14(1): 61–67.

38.

Workman

Jr. Weyer

. Practical Guide to Interpretive Near Infrared Spectroscopy. Boca Raton, FL: CRC Press, 2008. Pp. 235–247.