Near infrared spectroscopy as a tool to discriminate tannins from Amazonian species

Abstract

Near infrared spectroscopy (NIR) is a tool capable of providing efficient results for organic molecules of different materials. We developed a predictive model using Fourier Transform NIR Spectroscopy to distinguish the types of tannins in different forest species in the Amazon. Samples were obtained from different regions of the State of Amazonas/Brazil, and tests for tannins were performed, including obtaining NIRS spectra. The assembly of spectral data matrices versus analytes of interest was crossed with the results of traditional analyses. In addition, a calibration and validation set was constructed for condensed tannins, hydrolyzable tannins, and samples with no tannins. Finally, the performance of classification models was evaluated for sensitivity, identification index, and errors. The condensed tannin classes were detected in 63% of the species studied, followed by 34% of the species not containing tannin. The discriminant analysis produced groupings of classes, with a hit sensitivity index >90%. The developed model can be applied in studies of ecology, forestry and chemotaxonomy, with a focus on phenolic compounds such as tannins. The proposed methodology has advantages over the reference methods, reflected as a lower need for sample preparation, shorter analysis time, no use of reagents, and, consequently, no generation of waste.