The objective of this study was to develop and evaluate a digital classification methodology for producing thematic cartography for the National Reserve in Valdivia using multispectral Landsat TM images. A supervised classification method with a maximum likelihood decision rule was selected for processing the satellite data. The proposed methodology permitted a separation of five different types of land uses, which corresponded to old native forest, secondary growth forest, plantation, shrub and grassland. Overall, the classification accuracy was about 64%, and the accuracy for the different forest types was about 77%. The results obtained can be considered acceptable, and are promising for the preliminary mapping of areas covered with native forests.