Discriminação de espécies de Eucalyptus sp. por espectroscopia no infravermelho próximo (NIR)

• Ano de defesa: 2023
• Autor(a) principal: Bonfatti Júnior, Eraldo Antonio
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Mato Grosso; Brasil; Faculdade de Engenharia Florestal (FENF); UFMT CUC - Cuiabá; Programa de Pós-Graduação em Ciências Florestais e Ambientais
• Palavras-chave em Português: Quimiometria; Química da madeira; Espectro; Reflectância; CNPQ::CIENCIAS AGRARIAS::RECURSOS FLORESTAIS E ENGENHARIA FLORESTAL; Chemometrics; Wood chemistry; Spectrum; Reflectance
• Link de acesso: http://ri.ufmt.br/handle/1/5734

The identification of forest species can be difficult, time consuming and expensive, especially when dealing with species of the same genus, requiring technologies that can provide accuracy and agility in this process. The present study aimed to discriminate three species of the genus Eucalyptus of important economic interest in the southern region of Brazil: Eucalyptus dunnii, Eucalyptus benthamii and Eucalyptus saligna, planted in the city of Canoinhas, state of Santa Catarina. For this purpose, four different tree components were collected: wood, bark, branches, and leaves. These tree components were reduced to powder to determine the chemical composition of the lignocellulosic biomass (holocellulose content, total lignin content, total extractives content and ash content), in addition to collecting the near infrared spectrum with Fourier Transform. On these spectra eight mathematical transformations were tested: first derivative (1stD), second derivative (2ndD), light scattering correction (MSC), normal signal standardization (SNV), MSC+1stD, MSC+2ndD, SNV+1stD and SNV+2ndD. Principal component analysis (PCA) was used for species discrimination. The three species evaluated showed distinction in the chemical composition of all tree components. The only mathematical transformations that were successful in discriminating species using PCA were MSN and SNV, due to the mitigation of the scattered light effect in obtaining near infrared spectra. The results of these treatments were similar and, besides the discrimination of species, allowed the identification of the peaks responsible for this discrimination. Considering the tree components, the discrimination of species was more accurate when the wood and branches were used, indicating that the components with higher extractive content can be detrimental to the results of near infrared spectroscopy analysis.