Espectroscopia de infravermelho próximo para predição de propriedades da madeira de híbridos de Eucalyptus spp.

• Autor(a) principal: Milagres, Flaviana Reis
• Data de Publicação: 2013
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: LOCUS Repositório Institucional da UFV
• Texto Completo: http://locus.ufv.br/handle/123456789/588
• topic: Eucalyptus; Celulose kraft; NIRS; Eucalyptus; Kraft pulp; NIRS; CNPQ::CIENCIAS AGRARIAS::RECURSOS FLORESTAIS E ENGENHARIA FLORESTAL::TECNOLOGIA E UTILIZACAO DE PRODUTOS FLORESTAIS

Wood is considered the most important raw material for the pulp and paper industry. Brazilian pulp and paper mills are continuously seeking production increases due to increasing demand for wood. The mills seek to identify the best trees and to optimize productivity and wood quality. Rapid acquisition of information about wood properties is required for decisions to be taken in the selection of trees for forest plantations. Some tree selection techniques have been used for this purpose, among which near-infrared spectroscopy (NIRS) is one of the most promising. The NIRS technique has been of increasing interest in research, especially since improvements in instrumentation and data analysis have emerged. Among the advantages that may be mentioned are speed and simplicity of sample preparation, low cost, and non-use of chemical reagents. This study is divided into four papers for characterization and prediction of Eucalyptus spp. by NIRS. Chapter I, "Using Near-Infrared Spectroscopy for the Prediction of Wood Quality of Eucalyptus Hybrids from Three Sites in Brazil", aims at the development of models to predict the physical-chemical and pulping properties of the Eucalyptus spp. hybrids from the Genolyptus project. A total of 193 samples from three locations (Rio Grande do Sul, Minas Gerais, and São Paulo) were divided between calibration and prediction groups, and thus global models were generated as well as models for each site separately. The best models were selected based on the coefficient of determination, the ratio of performance to deviation (RPD), and the standard error of prediction (SEP). Partial least squares (PLS) regression was used to correlate laboratory data with spectral data collected from samples of whole tree chips. Among the models of yield, S/G ratio, and extractives, the individual models for each location and the global models showed high correlations. Among the models of basic density, lignin, and carbohydrates, the global models and some of the local models showed high coefficients of determination. In general, and pulping properties with sufficient accuracy, and some specific models may be used when precision is required for breeding programs. In Chapter II, "Near-Infrared Spectroscopy for Prediction of Eucalyptus Spp. Wood Quality Using Increment Core Sampling", the goal was to use NIRS to develop predictive models of the physical, chemical, and pulp properties of the wood. For this, two sampling methods were used: destructive (whole-tree chip sampling) and nondestructive (increment core sampling). The coefficient of determination for the increment core samples was lower than those of models developed with spectra of the whole tree, but some properties, such as density, total lignin, and S/G ratio can be used in the selection of Eucalyptus spp. hybrids using increment core spectra. Chapter III, "Samples Selection for Near-Infrared Spectroscopy evaluate spectra sampling techniques using three algorithms: CADEX, DUPLEX, and a selection method based on laboratory analysis (Ranking). Different sizes of groups were used for calibration models and validation of lignin content, S/G ratio, and pulp yield. The statistical results showed that the use of 75% of the data for the calibration group was sufficient to develop the models. The Ranking and DUPLEX techniques showed similarities in the NIRS models developed and it is possible to reduce costs and time in the selection of NIRS spectra for analysis. The purpose of Chapter IV, "The Influence of Age on Near-Infrared Spectroscopy Models for Prediction of Wood Physical and Chemical Properties of Eucalyptus Spp.", was the development of NIRS models for the basic density, lignin, and extractives content of the wood of different age groups. A total of 136 samples aged 3 7 years were used to analyze basic density, total lignin, and extractives contents and NIRS spectra collection of milled wood chips. The models were developed by PLS regression and tested by external validations with individuals of different ages. The results for the coefficient of determination of calibrations were 0.84 to 0.89 for basic density, 0.72 to 0.88 for lignin total, and 0.92 to 0.98 for wood extractives. The values of ratio of performance to deviation (RPD) did not show significant predictions when the models had different ages from the samples used in the prediction, except for basic density models with ages of 3 and 7 years, which had the ability to predict properties of samples aged 4 and 6 years.