Determination of Optical Parameters and Moisture Content of Wood with Visible–Near Infrared Spectroscopy

We used the Kubelka-Munk theory equations for calculating the absorption coefficient (Kλ), the scattering coefficient (Sλ), the transport absorption (σλa), the reduced scattering coefficient [σλs(1 – g)] and the penetration depth (δλ) from visible-near infrared reflectance spectra acquired over thin samples of quaking aspen and black spruce conditioned at three different moisture levels. The computed absorption and scattering coefficients varied from 0.1 mm−1 to 4.0 mm−1 and from 5.5 mm−1 to 10.0 mm−1, respectively. The absorption coefficients varied according to the absorption band, but the scattering coefficients decreased slowly towards high wavelengths. The sample moisture content was then estimated using the partial least squares (PLS) regression method from the Kλ and/or Sλ spectra, and the resulting PLS models were compared to those obtained with raw and transformed [multiplicative scatter corrected (MSC), first and second derivative] absorption spectra. The best PLS models for black spruce, quaking aspen and both species were obtained when only the 800–1800 nm range was used with the raw or MSC spectra. They led to a root mean square error of cross validation (RMSECV) of 1.40%, 1.09% and 1.23%, respectively, and to a coefficient of determination (R2CV) higher than 0.94. We also found that the Kλ spectra between 800 nm and 1800 nm can provide PLS models having an acceptable accuracy for moisture content estimation (R2CV = 0.83 and RMSECV = 2.32%), regardless of the species.