基于稳定同位素比率差异的西湖龙井茶产地溯源分析

通过测定不同产区扁形茶稳定同位素比率，借助化学计量学工具探索建立西湖龙井茶产地区分模型。从山东、四川、浙江、贵州等扁形茶主产区代表性取样，采用稳定同位素质谱测定样品中碳、氢、氧、氮、锶、铅、镉稳定同位素比率，对其进行方差分析，采用逐步线性判别（FLDA）、决策树C5.0和神经网络方法构建模型进行产地区分。结果表明，不同产地扁形茶中稳定同位素比率表现出产地差异。δD、δ¹⁸O受陆地效应和高程效应影响较大，并且二者有较好的正相关关系，⁸⁸Sr/⁸⁶Sr比值随地区纬度增加而增大，在不同产地扁形茶间差异尤为显著。FLDA和BP-ANN模型对外部样本的预测能力相对较差，分别为76%和84%；决策树算法模型回代验证准确率为91.35%，对外部样本的预测准确度达到92%。结果表明，在西湖龙井茶产地判别中，采用基于稳定同位素比率的决策树算法模型可取得较好的效果。

Geographical Tracing of the West Lake Longjing Tea Based on the Stable Isotope Ratios

The West Lake Longjing tea is a famous specialty with long history in Zhengjiang province and its price is higher than any other flatten‐shaped tea .However , traders tend to sell counterfeit West Lake Longjing tea in the market to acquire higher economic values .Otherwise ,plants from different geographical locations have distinct isotope ratio ,w hich form the element fingerprints due to the variation of climate and soil properties .Thus ,it is necessary to do the geographical tracing of the West Lake Longjing tea based on the stable isotope ratios .In this study ,we collected the flatten‐
shaped tea samples from the Hangzhou ,Shaoxing and Lishui of Zhejiang ,Qingchuan of Sichuan ,Liping of Guizhou and Rizhao of Shandong where the flatten‐shaped tea were widely produced .And the stable isotope ratios (C ,H ,O ,N ,Cd ,Pb ,Sr ) in the flat‐ten‐shaped tea samples were also assayed .Differences of the stable isotope ratios in the flatten‐shaped tea from different producing areas were obtained by the variance analysis . The comparision of the stable isotope ratio values of West Lake Longjing tea and other flatten‐shaped tea producing from different areas were completed .Fisher linear discrimi‐nant analysis (FLDA ) ,decision tree and back propagation artificial neural network (BP‐ANN) were used to establish the origin recognition model .The stable isotope ratios in flatten‐shaped tea produced from different areas showed the difference followed with the different producing areas .For theδD andδ18 O in samples ,maybe influenced by the land effect and the elevation effect ;a positive correlation was found between δD and δ18 O ;The ratio of 88Sr/86Sr was increased following the latitude’s increasing of the four prov‐inces .Compared the recognition ability among those models of FLDA ,decision tree and BP‐ANN ,all the recognition ability were above 76% .As applied by external validation samples ,decision tree model was suitable for geographical origin traceability of different regions ,while the accuracy rate was 92% .So the decision tree model was more suitable for geographical tracing of West Lake Longjing tea .