PREDICTION OF WOOD DENSITY USING NIR SPECTROSCOPY: RADIAL VERSUS CROSS-SECTIONAL SURFACE DATA COLLECTION

摘要

It has been demonstrated that data from SilviScan and NIR can be used to calibrate and predict wood properties. Most of the predictions are based on NIR spectra collected from the radial-longitudinal surface of the wood. However, predictions from the crosssectional surface are more practical in the field. In this work, wood density of balsam fir (Abies balsamea), black spruce (Picea mariana) and the combined samples were calibrated and predicted based on SilviScan and NIR results from both the radial and cross-sectional surfaces. The results showed that predictions from cross-sectional surface were as good as those from radial surface. The predictions for the combined samples were better than those for the single species samples. For the combined samples, the R2 of prediction were 0.89(based on radial surface) and 0.90 (based on cross-sectional surface), respectively. The best model from cross-sectional surface would require the NIR spectra of the full log, while only 3 spots of each strip would be needed for the radial surface.