Classification and authentication of tea according to their harvest season based on FT-IR fingerprinting using pattern recognition methods

The potential of FT-IR spectral fingerprinting was investigated to classify tea samples based on the harvest season (May and September). Tea samples were collected from five geographical regions (north of Iran) during the harvesting period 2019-2020. Principal component analysis (PCA), principal component analysis-linear discriminant analysis (PCA-LDA) and partial least square-linear discriminant analysis (PLS-LDA) were employed in order to assess the feasibility of discrimination of tea samples based on their harvest season using their FT-IR spectral data. The results showed that the tea samples from two harvest seasons can be identified based on FT-IR spectral fingerprints. All calibration samples were correctly classified (100.0 %) by the PCA-LDA and PLS-LDA models using leave-one-out cross validation. The mean sensitivity and specificity (for prediction set) were both 98.6 % for PCA-LDA model and 100.0 % for PLS-LDA mode. A high percentage of correct clas-sifications for the training set shows the strong relationship between the FT-IR spectral fingerprinting and the harvest season, while the satisfactory results for the prediction set demonstrates the ability to identify the harvest season of an unknown tea sample based on its FT-IR spectral data.

Automated summary

The potential of FT-IR spectral fingerprinting was investigated for classifying tea samples based on their harvest season. The results demonstrated that the two harvest seasons of tea samples can be identified accurately using FT-IR spectral fingerprints. Classification and prediction of the harvest season of tea samples were achieved using PCA-LDA and PLS-LDA models.