LEAF-BASED SPECTROSCOPY AND SPECTRAL MODELS FOR AMAZONIAN TREE SPECIES DISTINCTION

Sampling trees in a natural environment can be used in studies ranging from floristic composition and phytogeography to vegetation management and growth modeling. Relying on hyperspectral approach, this study aimed to differentiate spectral libraries of four Amazon tree species. Based on vegetation indices results to reduce data volume, the principal component and cluster analysis defined the best indices to differentiate spectrometry of leaves of Amazonian trees. These combined methods brought satisfactory results, with PC1 highly related to the variability of the vegetation indices results (99.37%). Adopting this approach in hyperspectral data at the leaf level and well-defined classes results in good responses. We emphasize the importance of using combined vegetation indices, with greater contributions by indices developed for quantization or absorption of electromagnetic radiation by chlorophyll, which are based in the visible region.