Moisture content distributions of Scots pine logs in the green state were measured by a novel multi-step procedure. After sample preparation, the transverse sections of the wood surfaces were scanned by an automated scanning device with a fiber optical probe connected to a Fourier transform near-infrared spectroscope. In the course of the measurement sequences, several issues were addressed, such as surface drying, measurement geometry, ease of automation and interconnected data handling. The near-infrared (NIR) data were first modeled separately for heartwood and sapwood by means of multivariate partial least squares regression. The models for moisture content were evaluated by root mean square error of prediction, the result being 0.8% for heartwood and 10% for sapwood. The two models were then applied to the NIR data collected from sets of disks cut from nine logs. The results of the calculated moisture contents were evaluated by methods of descriptive statistics, and they indicated clear differences and trends in the distribution of moisture content in transverse or longitudinal regions of a log. Additionally, inter-tree variation in moisture content was detected.