Detection and prediction of spontaneous afforestation using multispectral satellite data and GIS methods
2002
Kobler, A. | Kusar, G. | Hocevar, M. (Slovenian Forestry Inst., Ljubljana (Slovenia))
Various options for streamlining forest cover change detection for the period 1975 - 1995 at the regional level were tested, based on Landsat TM satellite data and on existing maps and GIS layers. Using a spatial regression model, spontaneous afforestation was forecasted for the next 20 years. Forest change was determined with map differencing, using maps from 1975 and 1995. The 1975 forest border was acquired from legacy maps. The 1995 forest border was derived by classification of a Landsat TM satellite image. Machine learning was used in some steps of classification in order to streamline the procedure as much as possible. Overall, thematic accuracy of the 1995 map was 82 %. When nomenclature was aggregated to only forest / non-forest, the accuracy increased to 91 %. Half of the mapped forest border was within 14 m of the true location. The average departure from the true location was 26,6 m. Between 1975 and 1995, 37 % of agricultural land was lost to spontaneous afforestation. The regression model explains 57 % of the afforestation variability. In the next 20 years, the most heavily afforested municipality will be Divaca (52 % afforestation of the present agricultural land), followed by Postojna, Cerknica, and Pivka (36 % to 41%), according to the current trend.
Afficher plus [+] Moins [-]Mots clés AGROVOC
Informations bibliographiques
Cette notice bibliographique a été fournie par University of Ljubljana
Découvrez la collection de ce fournisseur de données dans AGRIS