Search

In this study, proton-transfer reaction mass spectrometry (PTR-MS) and near-infrared spectroscopy (NIRS) were compared for the authentication of geographical and farming system origins of Brazilian coffees. For this purpose, n = 19 organic (ORG) and n = 26 conventional (CONV) coffees from distinct producing regions were analyzed. Overall, differences (p ≤ 0.05) in 44 and 71 ion intensities were observed between the main producing regions and farming systems, respectively. Principal component analysis was not effective in illustrating differences between the coffees according to the farming system or geographical origin using neither PTR-MS nor NIRS data. However, when the PLS-DA was applied, which produced the best performing models compared to several other chemometric techniques, the farming system was adroitly differentiated. The fact that the classification performance (>80%) was independent of the data acquisition method used gives NIRS an edge over PTR-MS in the differentiation of the farming system because of its rapid analysis and cost. Differentiating geographic location of coffee was rather complex. The PTR-MS calibration models showed slightly better PLS-DA classification rates compared to the NIRS models (69% vs. 61%, respectively), which is even more evident when the alternative classifier is used (LDA-kNN, 69% vs. 39%, respectively). Coffee samples from either Minas Gerais (MG) or Sao Paulo (SP) were differentiated from the other regions. In conclusion, our study provides information on alternative rapid analysis coupled with chemometric techniques to differentiate the farming system and trace the geographical provenance of Brazilian specialty coffee. [ABSTRACT FROM AUTHOR]