The application of near infrared (NIR) spectroscopy to the green wood of radial samples (simulated increment cores) and the development of calibrations for the prediction of wood properties are described. Twenty Pinus taeda L. (loblolly pine) radial strips were characterized in terms of air-dry density, microfibril angle (MFA), and stiffness. NIR spectra were obtained in 10-mm steps from the radial longitudinal and transverse faces of each sample and used to develop calibrations for each property. NIR spectra were collected when the wood was green (moisture content ranged from approximately 100% to 154%) and dried to approximately 7% moisture content. Relationships between measured and NIR estimates for green wood were good; coefficients of determination (R2) ranged from 0.79 (MFA) to 0.85 (air-dry density). Differences between calibrations developed using the radial longitudinal and transverse faces were small. Calibrations were tested on an independent set. Predictive errors were relatively large for some green samples and relationships were moderate; R2p ranged from 0.67 (MFA) to 0.81 (stiffness). Dry wood calibrations demonstrated strong predictive relationships with R2p ranging from 0.87 (air-dry density) to 0.95 (stiffness). NIR spectroscopy has the potential to predict the air-dry density, MFA, and stiffness of 10-mm sections of green P. taeda wood samples.

In the plywood industry, the assessment of the flexural static modulus of elasticity (MOE) and specific gravity (SG) of veneers is important for quality control. The rapid assessment of these properties by near infrared (NIR) spectroscopy was examined using 312 Pinus spp solid wood veneers collected from a Brazilian plywood plant. Partial least squares (PLS) regression calibrations (based on 75% of the samples) for MOE and SG were obtained using all available wavelengths (PLS) and only statistically significant wavelengths [jack-knifing PLS (j-kPLS)]. Correlation coefficients (r) for the calibrations ranged from 0.80 to 0.83, while the ratios of performance to deviation ranged from 1.67 to 1.78. When applied to a separate test set (25% of the samples), statistics similar to those reported for the calibrations were obtained. For both MOE and SG, the j-kPLS calibrations performed marginally better than calibrations based on the full spectrum. The models constructed with only statistically significant wavelengths may be more robust as wavelength selection avoids any unnecessary information for the prediction of MOE and SG. This study demonstrates that NIR spectroscopy could potentially be used for in-line assessment of the MOE and SG of pine veneers.