Mountain vegetation mapping in Dovre area, Norway, using Landsat TM data and GIS.

Vegetation mapping by use of satellite data are often divided into two main operations, the pre- and post-classification processes. Experience from producing vegetation maps based on spectral-only classifications, has shown that misclassifications occurs. The aim of the post-classification process is to improve the pre-classified product by use of ancillary data. The mountain areas of Norway are characterized by complex topography. Vegetation maps are though difficult to produce for these areas. In this study two Landsat 5/TM image from 1986 and 1998, covering parts of the Dovre mountain massif in Norway, are classified using unsupervised classification methods. The spectrally classified product is thereafter corrected using several ancillary data layers. Based on the ancillary data the delineation of forest vegetation and the heather vegetation above the woodland limit is more precisely defined. Bogs and mires are easily differentiated from snow-bed communities. The grass- and herb-rich communities in the mountain areas are spectrally much similar to agricultural areas in the lowland; even the floristical composition and content are totally different. By use of digital elevation models the alpine meadows and cultivated land in the lowland are separated into different classes by the use of an altitude threshold. The cost of, and types of corrections we can do in the post-classification process, largely depends on what additional information is available and the quality of this information.