Rapid discrimination of Anji Baicha origin using field-portable spectroradiometer

Anji Baicha, named according to its origin, is one of the most expensive green teas available and is vulnerable to the risk of fraud. In this study, the discriminative power of spectroradiometer data and statistical techniques were analyzed using 267 Anji Baicha samples. Different spectral preprocessing techniques, five classification methods, and chemically driven wavelengths were compared and validated using ten-fold cross-validation. The results showed that the optimized SVM model relies on SNV and the 2nd derivative of Savitzky-Golay as preprocessing steps and has an average accuracy of 98.9% in distinguishing core and other production areas. Further distinguishing between the six production areas within Anji County and the seven production areas outside Anji County was also tested. The average accuracy rates were 72.1% and 82.5% within and outside the county, respectively. These findings point to the possibility of a spectroradiometer combined with chemometric approaches as a screening technique for Anji Baicha origin differentiation. •267 Anji Baicha samples were analyzed using a portable spectroradiometer.•Reliable results were obtained using a support vector machine classifier.•Chemically driven wavelength selection improves model interpretability.•Classification results are highly dependent on geographical distance.•Geographical origin has a greater effect on spectral data than harvest year.