Vegetation maps represent the composition and distribution of plant communities of an area and thereby are a significant source of information. Retrospective mapping can provide an insight into the past and provide a basic point according to which it is possible to determine the course and intensity of the environmental changes. This research examines the possibilities of supplementing the old field vegetation maps from the second half of the 20th century that cover the areas of Nature parks Medvednica and Žumberak-Samoborsko gorje and National park Krka, using the supervised classification of Landsat archive satellite images in a geographic information system environment. The results suggest that “maximum likelihood” algorithm was more reliable than the “minimum distance” algorithm in all analyzed areas. The accuracy analysis of the classification process showed very poor matching of classified classes and control classes of habitat maps for the area of Nature park Medvednica whose Kappa values are 0.121 for “maximum likelihood” and 0.065 for “minimum distance” algorithm, while the most reliable area was Nature park Žumberak-Samoborsko gorje with Kappa values of 0.716 for “maximum likelihood” and 0.616 for the “minimum distance”. For the northern satellite image of the National park Krka area, the Kappa values are 0.620 for “maximum likelihood” and 0.456 for the “minimum distance” and for the southern satellite image 0.511 for “maximum likelihood” and 0.245 for the “minimum distance” algorithm. With increase of mapped classes, obtained accurracy decreased. The results of this research indicate limited possibilities of using Landsat archive imagery for the purpose of retrospective mapping.