Abstracts

Diversity of the Vegetation

There is a long tradition of vegetation classification based on floristic composition in Central and Western Europe. Several hundreds of thousands phytosociological releves, i. e. descriptions of vegetation stands with full species lists, are available, a large part of them being computerized in databases. In the past, classifications were usually produced for smaller areas and individual local classifications were synthesized into large-scale classifications. However, classification criteria and algorithms were never clearly indicated, and such classifications were not repeatable. If large databases are available, the preferable option should be formal classifications, i.e. those with clearly described algorithms. Unfortunately, classical methods of formal classification, e.g. cluster analysis or Twinspan, are sensitive to the dataset structure and do not have potential for incorporating external information on species behaviour in wider areas or other environments. Therefore they produce unstable results. This drawback is overcome in the Cocktail algorithm, developed recently by H. Bruelheide. This algorithm produces formal definitions of vegetation units using statistically formed groups of associated species and combinations of these groups by logical operators. Cocktail algorithm is currently freely available in computer program Juice by L. Tichy. The advantages of this algorithm are demonstrated in a case study of Czech dry grasslands. It is shown, that out of the traditionally recognized vegetation units, the algorithm maintains those which are statistically sound, whereas the others are discarded. Formal definitions of vegetation units are included in a computer expert system which automatically assigns releves to vegetation units.

Note. Abstracts are published in author's edition

|Home Page| |English Part|

Go to Home