Ferramenta alternativa para a taxonomia vegetal: assinatura espectral da madeira de árvores caídas naturalmente na Amazônia Central

Abstract

Alternative tool for plant taxonomy: Fingerprint of wood from natural fallen trees in the Central Amazon The infrared spectrum reveals a significant range of information about organic molecules present in various materials, including plant tissue. This spectrum carries the sample's "fingerprint" important information for integrative taxonomy studies. The challenge of taxonomic recognition of fallen trees in tropical forests, which often lack flowers, fruits, and leaves during field collection, motivated this study. The main objective was to evaluate and develop predictive models using Fourier transform near-infrared spectroscopy (FT-NIR) as an alternative tool for the botanical discrimination of wood from natural fallen trees from the Lecythidaceae and Myristicaceae families occurring in the Central Amazon. During a forest inventory in the Alto Solimões region (Amazonas, Brazil), arboreal individuals were diagnosed on the forest floor, known as natural fallen trees. A significant number of trees, such as máta-matá, tauari, and ucuúba, were identified by parabotanists and collected for the study. Samples were taken from the trunk for anatomical identification and obtaining NIR spectra (10,000 to 4,000 cm-1). A total of 1,836 spectra were collected and used in multivariate analysis (calibration, validation, and testing). Investigation of absorption bands in the NIR region and the use of chemometrics facilitated the interpretation of spectra with a high identification rate. The application of the Mahalanobis distance algorithm to the spectral data of Lecythidaceae wood identified seven species (Couratari sp., Couratari guianensis, C. stellata, Eschweilera coriacea, E. odora, Lecythis usitata and Holopyxidium sp.) distributed in four genera. In turn, linear discriminant analysis (LDA) was able to discriminate Myristicaceae wood into two genera: Iryanthera ulei and Virola (V. elongata, V. pavonis, V. surinamensis, Virola sp1 and Virola sp2). NIR spectroscopy, combined with multivariate models on the xylem tissue of fallen trees, emerges as a valuable tool to facilitate botanical recognition, especially in scenarios with limited availability of fertile plant material. This approach not only proves useful in floristic studies but also plays a vital role in characterizing the properties of the wood raw material, thus contributing to optimizing its use in line with the principles of economic botany. Integration of these techniques can offer crucial insights for decision-making in forest management plans, indicating alternative uses for this biomass often neglected by the forestry sector.

Keywords: Fallen trees, alternative taxonomy, Amazonian wood, multivariate models, near-infrared spectroscopy