The stability of forest core areas is an important indicator of biotic diversity on a landscape level in the process of deciding whether certain forests could be cleared. Consequently, we decided to estimate the stability of forest core areas in Slovenia between 1975 and 2012, as we are in possession of sufficiently accurate data to make a comparison of forest cover for these 38 years. Four types of changes in forest coverage (stable non-forest, deforestations, overgrowing and stable forest) are explained with an empiric model designed as a decision tree, which takes into account the altitude, the slope of the terrain, ground type, the distance of each cell on a raster map from the forest edge in the base year and the distance of each cell on a raster map from the nearest settlement in the base year. Explanatory models were built using the method of machine learning (Quinlan, 1986) using See5 (www.rulequest.com). The accuracy of the chosen model was 74.65%. In 2012, stable forests covered 53 % of Slovenia; their coverage has remained unchanged since 1975. From all the variables, land-use changes are best explained by the initial distance from the forest edge and partially by the slope of the terrain. The distance from settlements and the altitude, on the other hand, do not play a significant role.
|
