SEARCH

SEARCH BY CITATION

Keywords:

  • Beta-diversity;
  • Classification;
  • Landscape types;
  • Rao's quadratic entropy;
  • Sigma syntaxonomy

Abstract

Aim: This paper introduces a new method for vegetation-based landscape classification. As a case study, we present the landscape classification of Hungary with a total area of 9.3 million ha.

Location: Hungary.

Methods: Data from the MÉTA (Magyarországi Élőhelyek Térképi Adatbázisa: GIS Database of the Hungarian Habitats) vegetation survey were used in our analyses. The basal units of the survey were hexagons with an area of 35 ha, in which surveyors estimated the cover of the various types of the (semi)-natural vegetation. The sample unit in our analyses was a rosette consisting of seven hexagons. The distance between sample units was calculated based on the relative cover of (semi-)natural habitats by Rao's beta diversity, a new distance measure that includes the similarities between habitats. A hierarchical classification was then performed using the UPGMA (unweighted pair group method with arithmetic mean) algorithm.

Results: The optimal number of groups was 41 on the basis of average silhouette. Landscape types dominated by forests or grasslands were separated at the highest dissimilarity level. At lower levels, the division of groups could be attributed to differences in site conditions (dry, mesic, wet, saline). The most common landscape types (more than 400 occurrences) in Hungary are those dominated by zonal forests, degraded treeless habitats, xeric and mesic saline and alkaline habitats, and reedbeds and wet meadows. The rarest types (less than 50 occurrences) are mosaics of edaphic habitats, and coniferous forests.

Conclusions: Landscape classification based on (semi-)natural vegetation may be used for estimating landscape diversity, landscape modeling, selection of study sites, regionalization of local scientific results, and for landscape development planning and nature conservation management. The new distance measure has met our expectations and resulted in a classification with clearly interpretable groups. It is likely that this distance measure may also prove to be appropriate for numerical syntaxonomy, in which the stability of the resulting groups may be increased by taking into consideration the similarities/differences in phytosociological preference of species.