To what extent are vegetation types visible in satellite imagery

Vegetation was studied in three national parks in SW Finland using PCA transformed and supervised classified Landsat TM imagery. In Seitseminen and Liesjarvi, the images were compared to published vegetation maps, whereas Landsat study of the Baltic Sea Archipelago Park included comparisons with aerial photographs in addition to field observations. The results reveal that satellite imagery contains accurate information on environmental features, which are often difficult or impossible to examine by other means. However, satellite data fails to distinguish many of the vegetation types recognized by the Finnish phytosociological school, though the major physiognomic categories are for the most part discern ible. The increasing number of vegetation studies based on remote sensing requires a re-evaluation of the classification criteria applied in the botanical tradition. Manual interpretation of digitally enhanced images is superior to supervised classification procedures in many botanical works, because the former is free from operator-defined classes which may not necessarily be representative. Many satellite-based mappings that are available at pres ent are of limited value in biological studies because their accuracy is questionable. There is an urgent need for developing unified accuracy validation procedures to meet the needs of image users.