The challenge of identifying and conserving valuable ecosystems close to human settlements in a northern area. An approach based on field- and satellite data.

Abstract

The rich broadleaved forests of North Norway have high species diversity. Mappings of biodiversity have been undertaken in the two municipalities Målselv and Bardu, but these mappings are far from exhaustive. This study examines classification methods for mapping rich broadleaved forests with the use of Landsat ETM+ images, and with vegetation indices as ancillary data. Three classifications were made; one supervised (on a July image) and two unsupervised (on the July image and a September image). Of these, the unsupervised classification of the July image had the best Overall Accuracy at 60.59 % and a Kappa coefficient of 0.4262. It seems that it is somewhat difficult to differentiate between the various rich broadleaved forest types with the use of Landsat ETM+ images, with their medium resolution, and a per-pixel classification. But with the added use of a tresholded NDVI it is possible to discern richer forest types in the study area, and to some degree imply what kind of forest we might expect to find based on the best classifications. I have compared my findings with the earlier biodiversity maps, and on this background I suggest that a new, and more thorough, mapping of the region is carried out.