Differentiation of eight tea (Camellia sinensis) cultivars in China by elemental fingerprint of their leaves

Transcription

1 Research Article Received: 28 December 2008 Revised: 17 June 2009 Accepted: 20 June 2009 Published online in Wiley Interscience: 20 September 2009 ( DOI /jsfa.3716 Differentiation of eight tea (Camellia sinensis) cultivars in China by elemental fingerprint of their leaves Yingxu Chen, a Mingge Yu, a,b Jie Xu, a,c Xincai Chen a and Jiyan Shi a Abstract BACKGROUND: Tea is an infusion made from dried leaves of tea (Camellia sinensis) and can be a good dietary source of essential trace metals for humans. Therefore, it is necessary to consider variations in element content of tea leaves among tea cultivars. Thus, elemental fingerprint techniques, based on elemental contents(al,b,ca,cd,co,cr,cu,fe,k,mg,mn,na,p,pb,andzn) determined by inductively coupled plasma atomic emission spectrometry (ICP-AES) and multivariate statistical analysis, have been used to differentiate eight tea cultivars. RESULTS: The ranges of element concentrations in leaves of the eight cultivars were in good agreement with those obtained in previous studies and the level of most elements in tea leaves was significantly different among cultivars. The classifications of eight tea cultivars were 100% accurate in total by principal component analysis (PCA),hierarchicalcluster analysis (HCA),linear discriminant analysis (LDA), and back-propagation neural network (BPNN) analysis. CONCLUSION: Each cultivar presented a distinctive element fingerprint and the elements in tea leaves can be significant predictors in differentiating tea cultivars. c 2009 Society of Chemical Industry Keywords: tea; Camellia sinensis;cultivar; elemental fingerprint;pattern recognition;icp-aes 2350 INTRODUCTION Tea (Camellia sinensis), as a unique crop, a magical commodity or a popular beverage in the world, originated from China and has been cultivated by the Chinese people for at least 2000 and possibly up to 5000 years. 1,2 China is a leading tea producer, consumer and exporter in the world. The tea industry is of vital importance to the nation s economy and people s livelihood. 3,4 Tea is an infusion made from dried leaves and can be a good dietary source of essential trace metals for humans. Being a cross-pollinated plant, the tea plant is highly heterogeneous and there is much difference in element content between different cultivars. 5 8 The elemental composition of tea and tea leaves is the object of broad scientific studies from, for example, medical, toxicological and environmental points of view. Therefore, it is essential to consider variations in element content of tea leaves among tea cultivars. Elemental fingerprint techniques, based on elemental composition and multivariate statistical analysis, has primarily been used to determine the geographic origin of teas The most common procedures used for the classification of teas brands and determination of the geographic origin of different teas include several pattern recognition techniques, such as principal component analysis (PCA), linear discriminant analysis (LDA), cluster analysis (CA), soft independent modeling of class analogy (SIMCA), and artificial neural networks (ANNs) However, there are no previous studies on differentiation of tea cultivars by elemental fingerprint. In considering the variation of tea cultivars, the aim of this study is to differentiate eight tea cultivars by fingerprint of elements in leaves of the tea plant grown under the same conditions in Hangzhou, Zhejiang province, China. In the present work, the contentsofal,b,ca,cd,co,cr,cu,fe,k,mg,mn,na,p,pb,andzn in tea leaves have been determined by using inductively coupled plasma atomic emission spectrometry (ICP-AES). PCA, hierarchical cluster analysis (HCA), LDA, and back-propagation neural network (BPNN) analysis have been applied to accomplish the aim of this paper. Our study is expected to be valuable for future studies on classification of teas, the cultivation and breeding of the tea plants, and nutritional, medical and toxicological effects in relation to tea plants grown in China. EXPERIMENTAL Samples Eight different tea cultivars (Table 1) were chosen for the present study. They were planted at the tea plantation of the Zhejiang Correspondence to: Yingxu Chen, Department of Environmental Engineering, Zhejiang University, Hangzhou , China. yxchen@zju.edu.cn a Department of Environmental Engineering, Zhejiang University, Hangzhou , China b DepartmentofAgriculturalScience,CollegeofYingdongBiologicalEngineering, Shaoguan University, Shaoguan , China c CollegeofBiologyandChemicalEngineering,JiaxingUniversity,Jiaxing314001, China J Sci Food Agric 2009; 89: c 2009 Society of Chemical Industry

2 Elemental fingerprint of eight tea cultivars Table 1. Description of eight tea cultivars studied in the present work Code Name Taxon Suitability for tea type JF Jinfeng tea C. sinensis cv. a Black, green Jinfeng MZ Meizhan tea C. sinensis cv. Meizhan Oolong, green, black LJ Longjing 43 tea C. sinensis cv. Green Longjing 43 ZH Zhenghe dabai tea C. sinensis cv. Black, green Zhenghe-dabaicha YS Yingshuang tea C. sinensis cv. Black, green Yingshuang MX Maoxie tea C. sinensis cv. Maoxie Oolong, green, black JK Jiukengzhong tea C. sinensis cv. Green Jiukengzhong BY Biyun tea C. sinensis cv. Biyun Green a cv., cultivar. University Tea Research Institute, located in Hangzhou, Zhejiang Province, China. The basic soil characteristics are given in Table 2. At all times, cultivation and management were kept the same for each cultivar. In May of 2007, three replicate samples of a bud with three leaves were collected from each cultivar. All samples were thoroughly rinsed first with tap water and then with deionized water. During washing, each leaf was brushed gently with the fingers to remove dust on the leaf surface. After washing, the samples were dried in an oven at 80 C for 24h to a constant weight and ground for subsequent analyses. Sample analysis Briefly, tea samples weighing approximately 5 g were ashed at 500 C for 6 h, and the ash was dissolved with 2 ml of 6 mol L 1 HCl. The clear digest samples were then transferred quantitatively to 25 ml calibrated flasks with deionized water. The concentrations ofal,b,ca,cd,co,cr,cu,fe,k,mg,mn,na,p,pb,andzninthe dry ashing solution of tea leaves were determined using ICP- AES (IRIS/AP, Thermo Jarrell Ash Corp., Franklin, MA, USA). Blanks and a standard reference material of tea leaf (GB07605-GSV- 4, Institute of Geophysical and Geochemical Sciences, Chinese Academy of Geological Sciences, Langfang, Hebei, China) were analyzed concurrently for accuracy assurance. All analyses of tea leaf samples were replicated three times. Statistical analysis Data are expressed on a dry weight (DW) basis. Data of element concentrations in tea plants of eight cultivars were subjected to one-way analysis of variance (ANOVA) to evaluate the influence of cultivars for each element, performed using SPSS11.5 for Windows. Each tea sample was considered as an assembly of 15 variables constituting their chemical descriptors. A data matrix with objects as rows and the mentioned variables as columns was built. In this study, pattern recognition methods have been applied involving display methods such as PCA and CA, and supervised learning procedures for classification, e.g. LDA and BPNN. The statistical package SPSS11.5 was used for pattern recognition computations. Alternatively, the Matlab6.5 program was used for BPNN calculations. PCA was used to achieve a reduction of dimension and to observe a primary evaluation of between-class similarity. HCA is an unsupervised classification procedure that involves a measurement of the similarity between objects to be clustered. LDA, as a supervised technique, is a useful complement to PCA. It was applied to raw data by using stepwise modes to evaluate differences in eight tea cultivars. BPNN is the most widely used neural network system and it uses the gradient or steepest descent method to determine the weight of connective neurons. The key point is the error back-propagation technique. In the learning process of BPNN, the interconnection weights are adjusted from back layers to front layers to minimize output error. The merit of BPNN is that it can approach any nonlinear continuous function after being trained and has the fastest convergence ability. In this research BPNN was used to classify eight tea cultivars. RESULTS AND DISCUSSION Element content The concentrations of 15 elements in leaves of eight tea cultivars are shown in Table 3. In general, it can be seen that the Table 2. Soil ph, organic matter content, and concentrations of elements (mg kg 1 ) at the Zhejiang University Tea Plantation, presented as means (standard error of mean) Sample no. ph Organic matter (%) Al B Ca Cd Co Cr (0.02) 1.24 (0.03) (842.49) (0.09) (1.32) 0.92 (0.02) (0.08) (0.29) (0.01) 1.25 (0.01) (843.8) (0.34) (20.69) 0.93 (0.03) (0.04) (4.82) (0.01) 1.22 (0.01) (721.71) (0.34) (10.47) 0.92 (0.01) (0.2) 46.4 (2.15) (0.01) 1.2 (0.01) ( ) (0.43) (12.33) 0.88 (0) (0.32) (1.08) Total 5.98 (0.05) 1.23 (0.01) (386.71) (0.19) (8.28) 0.91 (0.01) (0.12) (1.18) Sample no. Cu Fe K Mg Mn Na P Pb Zn (0.22) (31.37) (41.43) (11.63) (3.18) 3062 (4.41) (7.9) (0.12) (0.02) (0.25) (9.67) (79.93) (5.7) (2.14) (12.98) (2.43) (0.49) (0.65) (0.07) (41.14) (53.37) (12.01) (3.96) (30.49) (41.45) (0.24) 67.7 (0.24) (0.14) (33.44) (12.21) (4.43) (2.68) (50.09) (14.03) (0.78) (2.49) Total (0.24) (15.92) (34.96) (8.12) (2.47) (17.7) (17.14) 26.9 (0.53) 67.3 (1.32) 2351 J Sci Food Agric 2009; 89: c 2009 Society of Chemical Industry

3 Y Chen et al ranges of element concentrations in leaves of the eight cultivars were in good agreement with the levels obtained in previous studies. 7 9,11,14,16,17 It was interesting to note that the level of the majority of elements in tea leaves was significant different among cultivars, using ANOVA analysis. Among the minerals and essential trace elements, Ca, Na, K, Mg, and Mn are present in tea leaves at g kg 1 levels, while Cr, Fe, Co, Ni, Cu, and Zn are present at mg kg 1 levels. For K, the most abundant metal among the macroelements, it was found that BY accumulated the highest concentration, and ZH contained the lowest concentration, ranging extensively from to mg kg 1. Ranges of Al concentration could be arranged according to their levels in the eight cultivars in the following order: MX > MZ > ZH > LJ > JK > YS > BY > JF, and varied from to mg kg 1. Mn, another important element in tea, was reported with a range of mg kg 1 in the tea leaves. 8 In the present paper, level of Mn was observed within the range mg kg 1, which was normal comparing with former results. The highest and lowest concentrations for Mn were in MZ and ZH, respectively. Among the tested toxic heavy metals, the most abundant metal in the tea leaves was Pb, which is of most concern by tea drinkers, ranging between 2.28 and 5.61 mg kg 1 among the eight cultivars and much lower than the range of mg kg 1 ever reported. 5,8 MX had the maximum concentration of Pb, which is slightly above the level permitted (5 mg kg 1 ) by the Chinese Ministry of Health. The elemental content of teas depends on several factors, including soil characteristics (e.g. mineral composition, ph, cation exchange capacity, and particle size), environmental conditions (e.g. temperature, rainfall, humidity, and air quality), and tea plant (e.g. cultivar, biological status). 5 8 In the present study, the eight cultivars were grown on the same site, and subjected to the same conditions of soil and environment (Table 2). Therefore, the differences of element content in tea leaves among cultivars were mainly attributed to cultivar variations. Tea cultivars have large variations in element concentrations largely because of their heterogeneity due to cross-pollination. 2 Ruan and Wong 7 have also reported that the concentrations of F and Al in tea plants were significantly different among the four cultivars. In the further view, tea cultivar variation might embody the different uptake abilities due to factors such as root soil interaction influencing soil characteristics and element bioavailability, the physiological chemical differences among cultivars, etc. 5 8,18 However, it is difficult to deduce an overall conclusion on the difference of element content in tea leaves among the eight cultivars owing to the limitation of single-variable statistical analysis. Thus elemental fingerprint analysis, based on pattern recognition procedures and element composition in tea leaves, was applied to achieve a more reliable tea cultivar differentiation. PCA analysis The first four principal components (PC1 4), with eigenvalues >1, explained 82.7% of the total variability among the 15 variables in the original data, where PC1, PC2, PC3, and PC4 contributed 32.8%, 18.6%, 16.4%, and 15.0% of the total variance, respectively. The Cu, P and Pb contents showed the highest weights in PC1; Mg and Zn content loaded highly in PC2; Co and Fe content were with highest weights in PC3; and Ca and Mn were the dominating features contents in PC4. Element grouping is less clear in the other components, which contain less of the total variance. Thus, PCA analysis suggests that clustering and trends around PC1 3 represented most of the variance, and differences observed in Figure 1. PCA scores of eight tea cultivars. the element pattern of eight cultivars are primarily due to the characteristic element fingerprint of Cu, P, Pb, Mg, Zn, Co, Fe, Ca, and Mn presented for each cultivar. Examining a three-dimensional scores plot (Fig. 1) in the space defined by the first, second and third principal components, a clear separation among tea cultivars was found. On the score plot, BY and JK clusters are on the left lateral plane of the cubic, LJ, MX and ZH clusters are located on the right lateral plane of the cubic, and YS and MZ clusters lie on the parameter plane defined by PC1 and PC2. It is noteworthy that JF is in one case close to the YS cluster, the other two being on the corner of the left lateral plane. For LJ, one case is far away from the other two, located at the cross-plane between the left and right lateral plane. HCA analysis HCA of samples was performed using the selected chemical descriptors as variables, the squared Euclidean distance as similarity measurement and Ward s method as amalgamation rule. In this case, similar results to PCA analysis were obtained after the application of HCA, shown as a dendrogram (Fig. 2). As can be seen from Fig. 2, HCA shows a clear discrimination between samples from the eight cultivars. A first cluster comprising JF, BY, and JK was obtained, followed by a second cluster of YS, MZ, MX, ZH, and LJ. A third cluster was composed of BY and JK. A fourth cluster included MX and ZH, followed by two clusters (fifth and sixth clusters) containing YS and MZ, and YS, respectively. Finally, a seventh cluster consisted only of JF. Thus it can be concluded that eight cultivars can be clustered together for each, clearly using HCA based on the element content in tea leaves. LDA analysis In order to find an operative classification rule for discriminating tea cultivars, supervised learning pattern recognition techniques must be applied, such as stepwise LDA. This assumes a prior knowledge c 2009 Society of Chemical Industry J Sci Food Agric 2009; 89:

4 Elemental fingerprint of eight tea cultivars Table 3. Element concentration in leaves of eight tea cultivars, presented as means (standard error of mean) (n = 3) on a dry weight basis Element (mg kg 1 ) JF MZ LJ ZH YS MX JK BY Al (54.42)g (16.65)b (2.42)d (27.16)c (10.26)e (10.22)a (9.9)e (5.16)f B (1.68)e (2.38)cd (0.44)b (0.63)a (0.52)ab (0.19)a (0.17)c (0.06)d Ca (81.32)cd (29.98)ab (5.53)bc (9.69)cd (21.64)a (6.74)ab (16.71)b (11.2)d Cd 0.07 (0)abc 0.05 (0)d 0.08 (0.01)a 0.07 (0)ab 0.06 (0)cd 0.06 (0)bc 0.03 (0)e 0.03 (0)e Co 0.15 (0.01)cd 0.11 (0.01)d 0.35 (0.03)a 0.11 (0.01)d 0.13 (0)cd 0.16 (0.01)c 0.11 (0.01)d 0.25 (0)b Cr 0.43 (0)b 0.48 (0.03)b 0.97 (0.26)a 0.58 (0.12)b 0.4 (0.06)b 0.45 (0.01)b 0.42 (0.15)b 0.39 (0.07)b Cu (0.28)c (0.13)e (0.22)a (0.05)e (0.06)b 9.68 (0.07)f (0.05)b (0.19)d Fe (2.59)d (0.98)c (0.2)a (1.28)b (0.7)e (1.26)a (0.76)e 92.2 (0.35)c K (474.18)d (56.09)e (7.21)c (184.64)g (123.63)e (117.78)f (46.24)b (107.12)a Mg (44.77)f (18.84)e (4.91)e (1.89)d (6.78)c (14.61)b (9.95)a (9.49)a Mn (25.66)e (2.55)a (3.96)c (3.53)e (4.64)b (1.38)b (3.19)d (6.41)c Na (2.12)ab (3.85)d (2.69)b (6.5)d (4.68)d (5.05)c (5.47)a (2.2)c P (48.55)e (3.09)f (17.29)c (6.99)g (13.35)d (1.22)g (4.46)a (17.93)b Pb 2.28 (0.07)g 2.55 (0.04)f 2.71 (0.05)de 4.33 (0.09)b 2.83 (0.03)d 5.61 (0.03)a 2.56 (0.03)ef 3.59 (0.05)c Zn (0.72)bc (0.12)c (0.57)bc (0.51)b (0.16)c (1.35)b (0.67)bc (1.21)a For each parameter, different letters within a row indicate difference between cultivars with Duncan s multiple range test (P < 0.05) J Sci Food Agric 2009; 89: c 2009 Society of Chemical Industry

5 Y Chen et al. Figure 3. LDA analysis of eight tea cultivars Figure 2. HCA dendrogram of eight tea cultivars. of the number of classes as well as the class membership of each sample in a training set. The variables included in the analysis are determined with a stepwise LDA, using a Wilk s lambda selection criterion and an F-statistic factor, to establish the significance of changes in lambda when a new variable is tested. Considering the tea samples with known cultivars, stepwise LDA was applied to develop a set of discriminating functions, derived from values of nine quantitative variables (concentration of Al, B, Ca, Cu, Fe, K, Mn, P, and Pb in tea leaves), capable of discriminating tea cultivars. As can be observed, the classifications of the eight tea cultivars were 100% accurate in total. Figure 3 shows the different groups of samples in the plot of discriminant functions 1 and 2. As can be seen, all samples were correctly classified. LJ, BY and JK were located at the upper right, MX, MZ and ZH were at the upper left, and YS and JF lay at the lower center. These results suggest that the element fingerprint of teas is related to cultivar. Furthermore, the content of Mn and Al, which are considered to be an accumulated elements in tea plants, can be significant predictors of tea cultivars. BPNN analysis There are many types of network architectures, but the type used most commonly for classification purposes is BPNN. In this study, BPNN was applied to classify the eight tea cultivars based on the element content of tea leaves. In our case, the BPNN network topology was plus bias, consisting of 15 neurons (the content of elements) in the input layer, a hidden layer with 12 neurons, and an output layer with eight neurons corresponding to the eight tea cultivars. Training objects were taken at random. A learning rate η = 0.1 andamomentumµ = 0.6 were kept constantduring training. Byselecting the maximum epoch number as 1100, the results showed that classification accuracy is 100% for each cultivar of teas. The composition of organic molecules such as amino acids, polyphenols, volatile flavour compounds and caffeine 15,19 21 have successfully been used to determine tea varieties and authenticity. However, the instability of many organic molecules after harvest and the time consumed by the analysis, which may last for almost an hour per sample for, high-performance liquid chromatographic analysis of physiological amino acids, for example, is a major drawback of this kind of analysis. Thus analysis of the elemental composition is an attractive alternative because elements are stable and may be stored for years without the elemental composition changing significantly. 22 It is reported that PCA showed not only that the elemental fingerprints of these two additional genotypes (Otira and Barthos) were similar to those of the others, but also that the soil in which the plant had been growing could be accurately predicted on the basis of the PCA scores from the genotypes. 22 Chen found that the samples could be clustered reasonably into five groups, and the elemental distribution characteristics are related to the breeds of Panax quinquefolium L. 23 Pattern recognition procedures such as display methods (PCA and CA) and supervised techniques (LDA and ANN) were applied to differentiate black, green, and Oolong teas. 16 LDA and ANN analysis of metal content provided the best results in the classification of tea varieties. These chemometric procedures were also useful for distinguishing between Asian and African teas and between the geographical origin of different Asian teas. 14 The results of elemental chemometrics analysis indicated that a distinction between African and Asian and between Chinese and other Asian teas could be made. 9 In the present study, PCA analysis suggests that clustering and trends around PC1 3 represented most of the variance, and differences observed in the element pattern of eight cultivars are primarily due to the characteristic element fingerprint of Cu, P, Pb,Mg,Zn,Co,Fe,Ca,andMnpresentedforeachcultivar.Inthe three-dimensional scores plot space defined by the first, second and third principal components, a clear separation among tea cultivars was found. Also, eight cultivars can be clustered together for each clearly, using HCA based on the element content in tea leaves. The classifications of eight tea cultivars by stepwise LDA were 100% accurate in total and showed that the content of Mn and Al can be significant predictors of tea cultivars. BPNN analysis c 2009 Society of Chemical Industry J Sci Food Agric 2009; 89:

6 Elemental fingerprint of eight tea cultivars results showed that classification accuracy is 100% for each tea cultivar. In brief, it can be seen that the eight tea cultivars can be effectively discriminated according to their elemental composition. Each cultivar presented a distinctive element fingerprint, as was indicated by HCA, PCA, LDA, and BPNN analysis. CONLUSION This study has shown that the ranges of element concentrations in leaves of the eight cultivars were in good agreement with the levels obtained in previous studies and the level of most elements in tea leaves was significant different among cultivars. The classifications of eight tea cultivars were 100% accurate in total by principal component analysis (PCA), hierarchical cluster analysis (HCA), linear discriminant analysis (LDA), and back-propagation neural networks (BPNN) analysis. In conclusion, each cultivar presented a distinctive element fingerprint and the elements in tea leaves can be significant predictors in differentiating tea cultivars. ACKNOWLEDGEMENTS We appreciate Dr Xiaochang Wang of Zhengjiang University Tea Research Institute for collecting samples and Dr Yuanzhi Shi of Key Laboratory of Tea Chemical Engineering, China Ministry of Agriculture, for sample analysis. This work was supported by the Program of National Natural Science Foundation of China ( , , ). REFERENCES 1 Editorial board of China tea varieties, China Tea Varieties. Shanghai Scientific & Technical Publishers, Shanghai (2001). (in Chinese). 2 Willson KC and Clifford MN (eds), Tea: Cultivation to Consumption. Chapman & Hall, London (1992). 3 China Ministry of Commerce, Guide Manual for China Tea Export pdf [7 October 2007] (in Chinese). 4 YajunY (ed.), China Tea Cultivation. Shanghai Scientific & Technical Publishers, Shanghai (2005). (in Chinese). 5 Han WY, Zhao FJ, Shi YZ, Ma LF and Ruan JY, Scale and causes of lead contaminationinchinesetea. Environ Pollut 139: (2006). 6 Jin CW, Du ST, Zhang K and Lin XY, Factors determining copper concentration in tea leaves produced at Yuyao County, China. Food Chem Toxicol 46: (2008). 7 Ruan JY and Wong MH, Accumulation of fluoride and aluminium related to different varieties of tea plant. Environ Geochem Hlth 23:53 63 (2001). 8 Yemane M, Chandravanshi BS and Wondimu T, Levels of essential and non-essential metals in leaves of the tea plant (Camellia sinensis L.) and soil of Wushwush farms, Ethiopia. Food Chem 107: (2008). 9 Marcos A, Fisher A, Rea G and Hill SJ, Preliminary study using trace element concentrations and a chemometrics approach to determine the geographical origin of tea. J Anal Atom Spectrom 13: (1998). 10 Moreda-Pineiro A, Marcos A, Fisher A and Hill SJ, Evaluation of the effect of data pre-treatment procedures on classical pattern recognition and principal components analysis: a case study for the geographical classification of tea. J Environ Monit 3: (2001). 11 Moreda-Pineiro A, Fisher A and Hill SJ, The classification of tea according to region of origin using pattern recognition techniques andtracemetaldata. J Food Compos Anal 16: (2003). 12 Kang HN, Yang MF, Chen B, Han C, Wang L and Wang XR, Trace element determination in teas and discrimination analysis for tea. Rock Miner Anal 25:22 26 (2006). (in Chinese). 13 Liu X, Van Espen P, Adams F, Yan SH and Vanbelle M, Classification of Chinese tea samples according to origin and quality by principal componenttechniques. Anal Chim Acta 200: (1987). 14 Fernandez-Caceres PL, Martin MJ, Pablos F and Gonzalez AG, Differentiation of tea (Camellia sinensis) varieties and their geographical origin according to their metal content. J Agric Food Chem 49: (2001). 15 Valera P, Pablos F and Gustavo González A, Classification of tea samples by their chemical composition using discriminant analysis. Talanta 43: (1996). 16 Herrador MA and Gonzalez AG, Pattern recognition procedures for differentiation of Green, Black and Oolong teas according to their metal content from inductively coupled plasma atomic emission spectrometry. Talanta 53: (2001). 17 Han WY, Shi YZ, Ma LF and Ruan JY, Arsenic, cadmium, chromium, cobalt, and copper in different types of Chinese tea. BEnviron Contam Toxicol 75: (2005). 18 Shi YZ, Ruan JY, Ma LF, Han WY and Wang F, Accumulation and distribution of arsenic and cadmium by tea plants. JZhejiang Univ Sci B 9: (2008). 19 Winterstein M and Finger A, Classification of Darjeeling teas using multivariate data analysis. Deut Lebensm-Rundsch 94: (1998). 20 Borse BB, Rao LJM, Nagalakshmi S and Krishnamurthy N, Fingerprint of black teas from India: identification of the regio-specific characteristics. Food Chem 79: (2002). 21 Pongsuwan W, Bamba T, Yonetani T, Kobayashi AK and Fukusaki E, Quality prediction of Japanese green tea using pyrolyzer coupled GC/MS based metabolic fingerprinting. J Agric Food Chem 56: (2008). 22 Husted S, Mikkelsen BF, Jensen J and Nielsen NE, Elemental fingerprint analysis of barley (Hordeum vulgare) using inductively coupled plasma mass spectrometry, isotope-ratio mass spectrometry, and multivariatestatistics. Anal Bioanal Chem 378: (2004). 23 Chen JH, Xie MY, Fu BQ, Yang MF and Wang XR, Principal component analysis and cluster analysis of inorganic elements in PanaxQuinque folium.l.spectrosc Spectral Anal 26: (2006). (in Chinese) J Sci Food Agric 2009; 89: c 2009 Society of Chemical Industry