Identification of Dark Tea ( Camellia sinensis ( L . ) ) Origins According to Chemical Composition Combined with Bayes Classification Pattern Recognition

As one of the six major teas in China, dark tea is mainly produced in Yunnan, Hunan, Sichuan, Hubei and Guangxi provinces of china. At present, identification geographical of teas mainly depends on the sensory evaluation, because of lacking the quantitative discriminate method. In this study, 38 dark teas were taken, which were collected from five regions. And the main chemical compositions of tea samples were detected according to international standard. Using SPSS18.0 statistical software to reduction dimension, then chose four compositions (GA, EGC, caffeine, total catechins) as the principal component factors, by using Bayes discriminate analysis method, we established the quantitative discriminate model, which could identify the dark teas from different regions. The results show that the Bayes discriminate analysis can be used to discriminate the 38 samples from five regions and the correct rate could be reached 100%, which means the methods established is reliable.


INTRODUCTION
Dark tea is one of the six major teas in China, which initial processes are fixing, rolling, piling and drying.As the key step of forming dark tea characteristics, piling is based on the activity of microorganism, promotion of a series of complex biochemical changes through biochemical dynamics (extracellular enzyme), materialized power (microbial heat combined with moisture of green tea) and the combined effect of microbial metabolism itself (Wan, 2003;Zengsheng et al., 1991;Fulin et al., 2006).Therefore, the common feature of dark tea manufactured from different regions is its old and rough raw material, coarse shape, old stems and pile discoloration.In that case, it is difficult to discriminate by sensory evaluation without the compression molding.
At present, Fourier Transform Infrared Spectroscopy (FTIR) has been used to identify different regions of dark teas.5 typical dark tea samples were analyzed by Fourier Transforms Infrared Spectroscopy (FTIR) and their spectral characteristics were also compared (Xin-He et al., 2012).Infrared absorption spectrometry was used to compare the infrared absorption spectral differences of 37 dark tea samples from 6 different regions.And the characteristic peaks of the FTIR fingerprinting of dark tea samples were recognized and compared (Yan and Gui-nian, 2014).Therefore, different recognition methods of the fingerprint for solar dried Pu'er raw tea were investigated in order to identify Pu'er raw tea from different places (JingMing et al., 2010).The electronic tongue (He et al., 2009) and micellar electrokinetic capillary Chromatography (Kodama et al., 2007) was also used to identify tea of different regions.But these methods above all are qualitative, not give us quantitative discriminant method on the region of tea discrimination.

MATERIALS AND METHODS
Tea samples: Thirty eight kinds of representative dark tea samples manufactured in five regions were collected, in which 18 Yunnan Pu'er tea samples, 4 Hunan Fuzhuan tea samples, 6 Guangxi Liubao tea samples, 5 Hubei Qingzhuan tea samples and 5 Sichuan Kangzhuan tea samples.

Data processing methods:
Principal component analysis: This technique can be applied to generalize varieties of information in large amount of measurement data, which is a effective method for the dimensionality reduction (Wei-ping et al., 2007;Duda, 2003;Shushen and Zhongsheng, 1999).PCA is based on the derivation of linear combinations of the variables to produce new variables called Principle Components (PCs) that are uncorrelated (Wold et al., 1987).

Bayes discriminant analysis:
Bayes discriminant is a probability discriminant analysis in statistical classification field and begins with a prior distribution over all possible models by using probability statistical knowledge classification: W = b 0 +b 1 x 1 +b 2 x 2 +…+b p x p Based on the distribution density functions, by calculating the probability of the X sample from a general in the observation of the conditions of X sample, discriminate which classification the X sample belongs to Wei (2011).

Quantitative analysis of chemical composition:
High-Performance Liquid Chromatography (HPLC) was employed to analyze the quantization of the dark tea catechins and caffeine.Gallic acid, caffeine as well as various kinds of catechin in tea samples could be well separated (Fig. 1a).Peak symmetry and well abundance showed this method was a reliable and effective approach for routine analysis of dark tea.The content of +C and EC were too low to detect in some of dark teas.

Results from chemical compositions of dark tea samples:
High-Performance Liquid Chromatography (HPLC) was employed to analyze the beneficial components of those dark tea samples.The contents of main components (GA, EC, +C, ECG, EGC, EGCG, caffeine and theanine) were shown in Table 1.In dark tea samples, the amounts of total catechins ranged between 0.31 and 3.32%, while the amounts of caffeine in different samples ranged from 1.25 to 3.62%.The highest content of total catechins was shown in Hunan which was 3.32%.The lowest content of total catechins was shown in Yunnan which was 0.31%.EGCG and EGC were the most abundant catechins in the majority of the dark tea.The levels of EGCG ranged from 0.01% (Yunnan) to 1.32% (Hunan) and the levels of EGC ranged from 0.13% (Yunnan) to 0.86% (Guangxi).The main catechins in Hunan were EGCG and EGC which accounted for about 39.8 and 25.6%.In Hubei dark teas, EGC and EGCG accounted for about 41.5 and 35.8%.Theanine varied from 0.01 to 0.14% and the levels of GA ranged from 0.14% (Hubei) to 0.47% (Yunnan).
Table 4 showed the classification result of five regions.It was found that the discriminate rate was 100%, while the cross-validated rate was 92.1%, which demonstrated that Bayes discriminate analysis had high and robust classification performance.

CONLUSION
HPLC was utilized to measure main chemical composition simultaneously in samples from five dark tea origins.The tea sample contents were used as chemical descriptors to identify the origins of dark tea.Bayes analysis was applied as a pattern recognition tool to develop identification models.The overall results sufficiently demonstrated that it was feasible to realize identification of dark tea Origins according to chemical composition combined with Bayes Classification Pattern Recognition.We conclude that Bayes pattern recognition can be applied to quantitatively identify the dark tea origins based on their chemical component.

Table 1 :
Contents of main components in different area dark tea

Table 2 :
Contribution rate and cumulative contribution rate of each principal component

Table 3 :
Degree of correlation of original variables and the principal componentComponent

Table 4 :
Result of classification from different regions