engleski

Spatial variation of the matching between regional and local classifications: an example of Central European grasslands

The expectations towards broad-scale classifications of phytosociological data are often twofold: they are believed to reveal the most important broad-scale patterns of the vegetation and, at the same time, to test the validity of previously known patterns which were detected during local studies. We studied the relationship between regional and local patterns obtained from such classifications. We compiled a data set of 8033 relevés representing Arrhenatheretalia elatioris and Brometalia erecti grasslands of eight countries in Central Europe (Austria, Croatia, Czech Republic, Germany, Hungary, Poland, Slovakia, Slovenia). This data set was classified by beta-flexible agglomerative method, which served as a ’regional classification’ in the analysis. From this data set, local subsamples containing 50 plots were drawn randomly with a spatial constraint in order to represent variation on smaller areas. Each local subsample was classified by the same method as was the regional data set. The partition divergence measure was calculated between pairs of relevés based on the dendrogram of the regional and of the local classification. The correlation between partition divergence values of the relevés was applied to quantify the degree of matching between the regional and local classifications. We analysed the spatial distribution of the correlation values and tried to explain differences between areas of strong and weak matching between regional and local classifications. For a more detailed testing we ran analyses with reduced regional coverages as well (combining only one or two countries). Correlation values of areas in the western part of the Western Carpathians were the lowest, while high correlations were found at the periphery of the study area. Analyses with reduced extents showed that the position of the areas with the highest and the lowest matching may change if the regional coverage is altered. We found that both the geometry and special features of the study area can affect the matching between regional and local classifications. The centres of study areas tend to show weaker matching than peripherial parts because of the similarity patterns within broad-scale floristic gradients. Local classifications of floristically unique areas and areas with low beta diversity matched well with the regional clustering.

vegetation classification

nije evidentirano