Somodi I, Molnár Zs, Czúcz B et al: Implementation and application ... (2017)

Somodi Imelda, Molnár Zsolt, Czúcz Bálint, Bede-Fazekas Ákos, Bölöni János, Pásztor László, Laborczi Annamária, Niklaus E. Zimmermann
Implementation and application of Multiple Potential Natural Vegetation models – a case study of Hungary
Journal of Vegetation Science 28(6): pp. 1260-1269.
Angol nyelvű összefoglaló: 

Multiple Potential Natural Vegetation (MPNV) is a framework for the probabilistic and multilayer representation of potential vegetation in an area. How can an MPNV model be implemented and synthesized for the full range of vegetation types across a large spatial domain such as a country? What additional ecological and practical information can be gained compared to traditional Potential Natural Vegetation (PNV) estimates?
MPNV was estimated by modelling the occurrence probabilities of individual vegetation types using gradient boosting models (GBM). Vegetation data from the Hungarian Actual Habitat Database (MÉTA) and information on the abiotic background (climatic data, soil characteristics, hydrology) were used as inputs to the models. To facilitate MPNV interpretation a new technique for model synthesis (rescaling) enabling comprehensive visual presentation (synthetic maps) was developed which allows for a comparative view of the potential distribution of individual vegetation types.
The main result of MPNV modelling is a series of raw and rescaled probability maps of individual vegetation types for Hungary. Raw probabilities best suit within-type analyses, while rescaled estimations can also be compared across vegetation types. The latter create a synthetic overview of a location’s PNV as a ranked list of vegetation types, and make the comparison of actual and potential landscape composition possible. For example, a representation of forest vs grasslands in MPNV revealed a high level of overlap of the potential range of the two formations in Hungary.
The MPNV approach allows for viewing the potential vegetation composition of locations in far more detail than the PNV approach. Rescaling the probabilities estimated by the models allows easy access to the results by making potential presence of vegetation types with different data structure comparable for queries and synthetic maps. The wide range of applications identified for MPNV (conservation and restoration prioritisation, landscape evaluation) suggests that the PNV concept with the extension towards vegetation distributions is useful both for research and applications.

ÖBI kiemelt publikáció