Satellite data were used in developing methods for forest inventory and mapping in a cooperation project involving Finland, Norway and Sweden. A goal was to improve the technical competence in forestry remote sensing in those countries. Landsat TM‐ and simulated SPOT‐imageries were classified using, e.g., filtered input data and contextual classifiers. The relative area distribution of usual forest classes was estimated at an acceptable accuracy. A two‐phase sampling scheme was introduced for compartmentwise (in‐place) inventories. The first phase involves analysis of satellite data, the second phase measurement of field plots. Correlations between satellite‐ and field‐measured values of various forest characteristics were relatively high, close to those obtained using aerial photo interpretation.