Rapid Prediction of Different Wood Species Extractives and Lignin Content Using Near Infrared Spectroscopy

Abstract

The feasibility of using Fourier transform near infrared spectroscopy (FT-NIR) to rapidly determine the lignin and extractive content of various wood species (including softwoods and hardwoods) was investigated. Partial Least Square regression analyses were performed to describe the relationships between the data sets of wet laboratory chemical data and the FT-NIR spectra. The selection of relevant wavenumbers combined with the appropriate data pre-processing methods produced satisfactory prediction models. The test statistics (R² , RMSECV, RMSEP, RPD) improved compared with the models over the wave number range 7500 cm⁻¹ to 4000 cm⁻¹. Automatic selection was superior to manual selection. The predicted lignin and extractive content models, using the full cross-validation in the appropriate wave number ranges (in cm⁻¹) of 5450.1 to 4246.7, 6102.1 to 4597.7, 6252.4 to 4246.7, and 6252.4 to 6098.1 using the spectral data preprocessing methods of the straight-line subtraction, minimum–maximum normalization, and first derivative + vector normalization, were established. The high R² values were 0.9838, 0.9809, and 0.9625, respectively. The low RMSECV values were 0.425%, 0.452%, and 0.185%, respectively. RPD values were 7.86, 7.25, and 5.17, respectively. Predictions were very good, with R² of 0.9775, 0.9751, and 0.9521; RMSEP of 0.418%, 0.403%, and 0.206%; and RPD of 6.78, 6.7, and 4.57 for the lignin, 1% sodium hydroxide extractive, and ethanol-benzene extractive models, respectively.

Keywords:

• FT-NIR
• wood
• extractives
• lignin