Estimativas de propriedades e identifica??o da madeira com uso da espectroscopia no infravermelho pr?ximo - NIR

• Ano de defesa: 2020
• Autor(a) principal: Pace, Jos? Henrique Camargo
• Orientador(a): Latorraca, Jo?o Vicente de Figueiredo
• Banca de defesa: Latorraca, Jo?o Vicente de Figueiredo, Moulin, Jord?o Cabral, Paes, Juarez Benigno, Gomes, Fernando Jos? Borges, Castro, Jonnys Paz
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal Rural do Rio de Janeiro
• Programa de Pós-Graduação: Programa de P?s-Gradua??o em Ci?ncias Ambientais e Florestais
• Departamento: Instituto de Florestas
• País: Brasil
• Palavras-chave em Português: modelos preditivos; qualidade da madeira; ensaios n?o-destrutivos; Mata Atl?ntica
• Palavras-chave em Inglês: predictive models; wood quality; non-destructive methods; Atlantic Forest
• Área do conhecimento CNPq: Recursos Florestais e Engenharia Florestal; Qu?mica; Probabilidade e Estat?stica
• Link de acesso: https://tede.ufrrj.br/jspui/handle/jspui/5625

The near infrared spectral technology (NIR) is widely applied in the forestry sector, assisting in industrial processes, for instance, in the prevention of wooden structures, identification of wood, among others. The NIR used in the identification of wood, can assist in combating the illegal wood trade, especially Atlantic Forest native species. The technique of spectroscopy is combined with techniques of multivariate statistics, and this combination is known as Chemometrics. In this sense, the general objective of this research was to use the near infrared spectral technique to assist in predictions related to the use of wood, such as estimates of forest biomass, response to the xylophages attack and identification / differentiation of the wood species. Thirteen native species from the Atlantic Forest were used, from homogeneous plantations with 3 x 2 m spacing. Using chemometrics, statistical models were developed to estimate wood properties and identify species, through the translation of spectra through Principal Component Analysis (PCA), Partial Least Squares Regression (PLS), and Discriminant Function Regression (PLS-DA). In the separation and identification of the species, the model determination coefficients generated by the PLS-DA was 0,94 in the cross calibration and in the prediction of independent samples 0,92, with a prediction error of about 7%. Therefore, the use of predictive models based on the NIR spectra, was significant for both the discrimination and the identification of Atlantic Forest species analyzed. The estimation of aerial biomass by means of non-destructive methods was also studied to know the accumulated carbon stock. For this purpose, seven different dry biomass estimation methodologies were used, directly (destructive and non-destructive) and indirect (statistical models). The destructive method included rigorous weighing and cubing and the non-destructive method used basic density values estimated by the NIR spectroscopy technology and the X-ray densitometry (QTRS-01X Tree Ring Scanner). The models generated by PLS, for basic density estimated by NIR and X-ray, can be used to estimate aerial part biomass for the studied species. They presented different strategies for carbon accumulation, demonstrating the importance that forest management has on native species of the Atlantic Forest. Predictive models to distinguish the stages of biodegradation of four native species of white wood were also studied. Biodeterioration tests and chemical characterization were carried out to recognize the material. The statistical evaluation of mass loss was by the non-parametric test (Kruskal-Wallis). The species showed high natural resistance, being categorized as highly resistant to the fungus attack and presented a relationship with the extractive content. The predictive models corroborated with the results of the mass loss, discriminating with 100% precision in the presence of the fungus in the wood. Further studies are needed to expand the database, bringing robustness to the preliminary models generated by the PLS-DA, and increasing information on species with timber potential in the Atlantic forest, adding more justification to the management of this biome in relation to its economic viability with the conservationist.