Abstract
The purpose of the study was to explore and compare three different methods for modelling potential natural vegetation (PNV), a hypothetic natural state of vegetation that shows nature's biotic potential in the absence of human influence and disturbance. The vegetation was mapped in a south-central Norwegian mountain region, in a 34.2 km2 area around the village of Beitostølen, in 2009. The actual vegetation map (AVM) formed the basis for the development of PNV using three different modelling methods: (1) an expert-based manual modelling (EMM), (2) rule-based envelope GIS-modelling (RBM), and (3) a statistical predictive GIS-modelling method (Maxent). The article shows that the three modelling methods have different advantages, challenges and preconditions. The findings indicate that: (1) the EMM method should preferably be used only as a supplementary method in highly disturbed areas, (2) both the RBM and the Maxent methods perform well, (3) RBM performs especially well, but also Maxent are more objective methods than EMM and they are much easier to develop and re-run after model validation, (4) Maxent probably underestimates the potential distribution of some vegetation types, whereas RBM overestimates, (5) the Maxent output is relative probabilities of distribution, giving higher model variation than RBM.
Acknowledgements
We thank Finn-Arne Haugen, Rune Halvorsen, Vegar Bakkestuen, Anders K. Wollan, Sejal O'Donnel, Terje Gobakken, Pablo Dourojeanni and the reviewers for valuable comments or technical assistance. The article has been supported by the Research Council of Norway (project number 189977) and is a part of the scientific project Cultural Landscapes of Tourism and Hospitality (Cultour), administrated by the Norwegian Forest and Landscape Institute.