DETERMINATION OF E. benthamii PROPERTIES BY INFRARED SPECTROSCOPY AND PLS

Resumo

Considering the interest in the Eucalyptus benthamii species and the search for alternatives to determine some physical properties in a fast and reliable way, the objective of this study was to build multivariate calibration models through the near infrared spectroscopy (NIR) technique and partial least squares regression (PLS) for wood basic density (BD) and chip moisture (U). Trees were sampled in five diametric classes and used to evaluate DB, as a function of 3 ages and 3 production regions of E. benthamii. We considered only plantations at 7 years of age in the harvest phase and chip samples considering pre-defined intervals of ten days, from the date of tree harvest to 90 days, in two seasons of the year (winter and spring) for moisture analysis. For DB and U, NIR models were calibrated and validated using partial least squares (PLS) regression. Calibration models were evaluated by correlation coefficients (R²), root mean square error (RMSE), and variance performance ratio (RPD). The models for near infrared (NIR) spectroscopy showed R² values ranging from 0.60 to 0.68 for basic density and 0.72 for moisture. The best model for DB observed was the one that considered only the DAP samples. It was concluded that the NIR technique was suitable for estimating the properties of moisture and wood density in this evaluated species, E. benthamii.