Rapid discrimination of xiaokeng green tea geographical origin using surface-enhanced Raman spectroscopy coupled with chemometrics

Abstract

Xiaokeng green tea (XKGT), named for its origin, brings high economic benefits to the region due to its superior quality, but it is susceptible to fraud. In this study, 131 roasted green teas from different origins in Anhui Province were used as samples to investigate the ability of surface-enhanced Raman spectroscopy (SERS) combined with chemometrics to accurately discriminate green tea origins in narrow regions. First, spherical Ag nanoparticles (AgNPs) were prepared as SERS substrates. A stratified 5-fold cross-validation method was used to divide the modelling samples. The radial basis function neural network (RBFNN), convolutional neural network (CNN), and random forest (RF) models were built, and four different preprocessing methods were compared. The results showed that the optimised RBFNN model with normalisation, as the preprocessing method, had an average prediction set accuracy of 97.69% in distinguishing samples from the Xiaokeng tea area from other tea areas. The RBFNN model was further used to differentiate tea samples from four different origins within Anhui Province, namely Xiaokeng Village (XK), Chizhou City except Xiaokeng Village (CZ), Lu'an City (L'A) and Huangshan City (HS), with an average accuracy of 91.85% for the prediction set. These findings point to the potential of combining SERS with chemometrics as an effective method for discriminating the geographic origins of XKGT in narrow regions.