Potential vegetation models are of great importance as a baseline of vegetation ecology. They can form the basis of climate change modelling and can assist effective nature conservation and habitat restoration as an esimation of what vegetation could cover a currently unvegetated surface. Somodi et al. (2012) argued for the usefulness of Potential Natural Vegetation (PNV) estimations and introduced the concept of multiple Potential Natural Vegetation, which assesses PNV in a probabilistic setting.
We developed models of the PNV of Hungary, those are based on data of actual natural vegetation from the MÉTA (Landscape Ecological Vegetation Mapping of Hungary) database, which contains among others presence/absence observations for each vegetation type of Hungary at the scale of 35 ha. Ecologically relevant explanatory variables (including climate and soil conditions) were calculated for the whole extent of Hungary, than models were built using the gradient boosting algorithm (GBM) and applied to the existing environmental conditions covering the full country.
The multiple probabilistic assessment (1) represents the potential variation within vegetation types in a single location, (2) supports an informed decision of nature conservation on which vegetation types are among the sustainable ones at a location and also (3) supports restoration with a range of potentially selfsustainable vegetation types. The multiplicity also allows for representing rare or uncertain vegetation types, e.g. forest steppes in Hungary.
Developments and R-script model runs were performed on a high capacity linux R server of EcoInfLab. The project was supported by OTKA PD 83522.
