关键词: 香茶；品质因子；产地判别；随机森林回归

Abstract: In this study, 50 samples of Zhejiang Xiangcha tea, a representative of bulk green tea, from four varieties grown in three regions in Zhejiang province were collected and investigated. Correlation analysis among physicochemical indexes, catechin components, sensory factors and local wholesale prices was performed. The characteristic quality factors were selected and geographical origin and varietal discrimination were conducted by linear discriminant analysis (LDA) and random forest regression (RF-R). Moreover, this study attempted to establish a reference wholesale price prediction model. The results showed that the contents of free amino acids and caffeine were significantly correlated with several sensory factors, which were the key physicochemical factors for the formation of the flavor quality of Xiangcha tea. The contents of free amino acids and catechins and the ratio of phenol to ammonia were the origin-related factors. There were highly significant varietal differences in the overall sensory score, appearance, and sensory score for taste of Xiangcha tea, and significant differences in the aroma and the sensory score of brewed tea leaves. LDA demonstrated clear clustering of tea samples from each county and the feasibility of discriminating tea samples from different counties but not from different varieties. The reference wholesale price prediction mode developed by RF-R had good fitness (R2 = 0.867, mean absolute error = 7.907). The relative importance of factors for price fitting was ranked as follows: appearance > overall sensory score > aroma > infusion color > sensory score of brewed tea leaves > taste > amino acids > water extract > phenol to ammonia ratio > caffeine > tea polyphenols. The results of this study will guide the quality factor analysis, geographical origin traceability, varietal discrimination and trade reference pricing of tea.

Key words: Xiangcha tea; quality factors; geographical origin identification; random forest regression