Application of near-infrared spectroscopy to predict microfibril angle of 14-year-old Pinus patula

Abstract

An investigation was conducted to test the feasibility of near infrared spectroscopy (NIR) as a tool for predicting the microfibril angle (MFA) of solid wood samples of Pinus patula. Thirty 14 year-old trees were selected from three compartments located in KwaZulu-Natal, South Africa with three different site indices. Radial strips obtained at breast height were scanned at 5 and 10 mm intervals in a radial-longitudinal surface using a NIR spectrophotometer. The same strip samples were scanned using Silviscan-2 to determine their corresponding MFA profiles. Calibration models were developed to predict MFA from NIR spectra using partial least squares (PLS) regression. The results showed that good correlations between NIR spectra and MFA values were achieved when data from all three sites were combined as well when considered individually. All correlation coefficients were greater than 0.9, highlighting the reliability of NIR technique to predict MFA for Pinus patula. When calibration equations developed using data from individual sites were tested on the other data sets, predictions ranged from 65 to 80% and standard errors of prediction from 3.4 to 3.9 degree. Calibration statistics and prediction ability improved slightly using data from scans at 10 mm resolution compared to 5 mm resolution.