Files
Abstract
We tested the applicability of NIR-based wood property calibrations for eucalyptus plantations in central Brazil to trees in southern Brazil. Prediction errors for the southern samples were too high. As we added southern samples to the calibrations, prediction statistics for most wood properties improved considerably. We also collected increment cores and drill shavings from trees in central Brazil and collected NIR spectra from both. The aforementioned calibrations were used to predict the wood properties of the trees, and milled shavings were found to provide biased wood property estimates. Next, we collected bleached eucalyptus handsheets from a pulp mill in Esprito Santo. Mechanical and physical properties of the handsheets were measured and NIR calibrations were created. Calibration and cross-validation statistics were poor for all properties. Finally, we collected unbleached eucalyptus pulp samples, measured their carbohydrates, and created NIR calibrations. Calibration and prediction statistics were excellent for both xylose and rhamnose, but poorer for other carbohydrates.