Wavelength-specific information in benchtop and handheld Vis-NIR and UV-Vis spectroscopy: the case of Coffea arabica

UV-Vis and NIR spectroscopy are widely used for analysis in agri-food applications. On-site analysis enabled by handheld spectrometers offers critical advantages therein. However, cost-efficient, miniaturized instruments, frequently feature narrow wavelength regions. Despite direct impact on analytical performance, literature lacks deeper insight into how the analytical information from distinct wavelength regions is correlated with each other. This knowledge is important for both fundamental reasons (i.e., interpretability) and practical scenarios, when selective spectral regions are available. This study aims to gain deeper insight into the physicochemical nature underlying the broad wavelength range of UV-Vis to NIR in the case of Coffea arabica. The wavelength-specific data from a full-range handheld Vis-NIR (350-2500 nm), a benchtop NIR (1000-2500 nm), and a benchtop UV-Vis spectrometer (230-450 nm) was compared for differentiating Arabica cultivars. Non-linear classification and Two-Dimensional Correlation Spectroscopy (2D-COS) served as the basis for interpretation of the spectral regions. Principal Component Analysis (PCA) in combination with quartimax rotation helped to assign spectral contributions more clearly. All three spectrometers effectively discriminated between cultivars, with the benchtop NIR achieving the highest accuracy (93-98 %), followed by the handheld Vis-NIR (89-93 %) and the UV-Vis (79-93 %). 2D-COS analysis identified key spectral regions and the correlations existing between them. Elucidated were both chemical and physical factors, e.g., electronic transitions and vibrational absorptions of key chemical constituents of Arabica coffee, effects of roasting and color, or moisture, specific in different wavelength windows.