This study seeks to estimate the mechanical properties of thermally modified wood (TMW) using near-infrared (NIR) spectroscopy to measure 80 samples in three-point bending tests. Near-infrared spectra collected from the transverse, radial, and tangential sections of wood, coupled with chemometric techniques, were used to predict the mechanical properties of southern pine wood, from which NIR models were constructed based on partial least squares and corresponding cross-validation. The coefficient of determination between NIR transverse section spectra, as well as two mechanical properties of wood samples, modulus of rupture (MOR) and modulus of elasticity (MOE), were above 0.92 and greater than values for other sections. Spectral data from the transverse sections were richer than data from other sections, and thus, a model based on transverse sections better predicts the mechanical properties of wood. A close relationship between the values for mechanical properties (MOE and MOR) and the NIR spectra of thermally modified southern pine wood can be demonstrated, which provides the potential to predict the mechanical properties of untreated and thermally modified wood.
