Original paper

Combining numerical and traditional approaches to classify Echinops sphaerocephalus invaded communities in the Czech Republic

Petřík, Petr; Dostálek, Jiří; Neuhäuslová, Zdenka

Abstract

Various approaches were used to classify all habitats invaded by the great globe thistle (Echinops sphaerocephalus L.), an invasive alien in the Czech Republic. In this way, all vegetation plots (i.e., 284 relevés) were (i) classified based on the formalized expert system using Cocktail definitions of associations and combined with fidelity index and (ii) arranged deductively (i.e., using a top-down process and traditional expert assessment) within a hierarchical classification system of vegetation supported by the divisive method TWINSPAN. E. sphaerocephalus is considered a diagnostic species of the alliance Onopordion acanthii throughout Europe. However, based on our results, E. sphaerocephalus can also invade both human-influenced and natural communities in less xerophilous habitats. Employing the expert system first, core communities at the level of association characterized by diagnostic species were chosen and then other out-of-rank-syntaxa could be classified deductively. Only 83 relevés (i.e., ca 29 % of all used) were classified using the expert system as 26 various synanthropic communities congruent with Cocktail definitions. Using deductive sorting, all remaining 197 relevés were classified within the framework of two broad vegetation types on the level of four vegetation classes (Artemisietea, Chenopodietea/Secalietea, Festuco-Brometea, Galio-Urticetea) based on abundance of their diagnostic species. To classify ruderal vegetation that have no association-diagnostic species, a combination of the formalized Cocktail-based definitions and diagnostic species-based deductive approach is suggested. However, the successful employment of the deductive method has to be based on the formation of a widely accepted and stable classification system built with formalized and repeatable steps.

Keywords

aliensupervised classificationfidelity conceptdiagnostic speciesdeductive methodruderal vegetation