The chemical constituents of tea leaves are influenced by many factors, such as cultivar, soil, climate, harvest season and manufacturing process. The effects of harvest season on the tea metabolome are significant. A pattern recognition methods for discrimination of green teas with seasonal variations was developed based on UPLC-Q-TOF/MS combined with chemometrics. Supervised principal component analysis (PCA) explained 100 % of the total variance (50.2 % and 49.8 %, respectively). Orthogonal signal correction- orthogonal partial least squares-discriminant analysis (O2PLS-DA) can obtain excellent predictive power with R2X, R2Y and Q2 of 0.923, 0.976 and 0.974, respectively. The polyphenols compositions in tea infusions of green tea were analyzed. Thirty-nine polyphenols were detected and quantified, and among them, fifteen polyphenols were screened as good markers for season identification of green teas. The present strategy also provides great potential for quality evaluation of other foods.