Portable sensors are required for rapid and nondestructive measurements of wood properties in the field to ensure optimal use of the fiber. We tested a handheld near-infrared (NIR) spectrometer to estimate moisture content (MC) and basic specific gravity (BSG) of quaking aspen (Populus tremuloides Michx.) and balsam poplar (Populus balsamifera L.) frozen and thawed logs. Partial least square (PLS) regression general models were developed to estimate MC and BSG by considering the influence of the following factors: log conditions (frozen and thawed wood), species, and types of wood (sapwood and heartwood). In addition, the influence of MC was also considered when estimating BSG. Including the two types of wood in a general model improved MC prediction (R 2 p ¼ 0.83, RMSE p = 11.1%) while including the two species improved BSG prediction (R 2 p ¼ 0.57, RMSE p = 0.022). Similar accuracies were obtained for BSG prediction from green (R 2 p ¼ 0.35, RMSE p = 0.027) and oven-dried wood (R 2 p ¼ 0.43, RMSE p = 0.027). PLS discriminant analysis was applied successfully to NIR spectra to sort the wood according to their MC, BSG, species, and wood type with overall accuracies of 86%, 65%, 98%, and 79%, respectively.