Espectroscopia no infravermelho próximo na predição de características da madeira em clones de Eucalyptus

Abstract

In the search for new options for fast, simple and reliable characterization of the wood properties, the non-destructive evaluations stand out. Among them, the near-infrared spectroscopy technique (NIRS) is cited. This study aimed to apply the NIRS technique to predict the basic density, chemical and anatomical characteristics of the wood in Eucalyptus clones for the production of charcoal. Thirty - nine clones of Eucalyptus spp., at the age of 81 months, were used from a clonal test, located in the municipality of Curvelo, in the state of Minas Gerais. The basic density, chemical and anatomical analyzes of wood were determined by conventional method. The near-infrared spectra were acquired via optical fiber and integration sphere in the solid and ground wood by the Bruker MPA spectrometer. The calibrated models for the properties of the wood were evaluated according to their performance regarding the acquisition path and sample type. A simple correlation analysis was performed between the characteristics measured in the wood. The results showed low coefficients of variation among the clonal materials regarding their properties. The best calibration obtained by the NIRS technique was performed for the basic density, in the solid wood by the spectral average per clone, with determination coefficient of 0.89 and deviation performance ratio of 2.07, indicating that the average per clone improved the Model performance. The models predicted for the chemical properties presented low performances, as well as the anatomical properties of the wood and core / sapwood relationship. The correlations that presented expressive results were fixed carbon content and volatile materials, insoluble lignin content and total lignin, fiber wall fraction and lumen diameter and wall fraction and cell wall thickness of the fiber.