NIR spectroscopy and chemometrics-based traceability of specialty Brazilian green canephora coffee

There is a need to develop efficient analytical methods to recognize the origins of coffee beans, especially from large producers such as Brazil, which offers high value-added Geographical Indication (GI) coffees. However, the challenge is not only the need for fast and clean techniques but also understanding how sample preparation and data treatments directly affect the performance of the applied technique. In this study, Near Infrared (NIR) Spectroscopy was combined with Partial Least Squares Discriminant Analysis (PLS-DA) to assess the ability to discriminate green coffee samples with recognized GI (Robusta Amazônico from Rondônia and Conilon from Espírito Santo), examining the influence of sample presentation (ground or whole bean) and spectral pre-processing. The results demonstrated that NIR performed with high efficiency for both whole beans and ground green coffee, achieving 100 % correct prediction. The most effective pre-processing was the combination of the 1st derivative of Savitzky-Golay and Multiplicative Scatter Correction (MSC). This suggests that the technique can be used for rapid discrimination in green coffee trading, with the direct analysis of natural whole beans being much more advantageous, as it avoids milling, which requires liquid nitrogen and a specific mill. Thus, NIR coupled with PLS-DA is a non-invasive, easy-to-operate, low-cost, and sensitive technique that can be applied directly to intact canephora coffee samples.