General strategy for extracting vegetation classification from large phytosociological databases
Department of Botany,
Masaryk University, Kotlarska 2, CZ-611 37 Brno, Czech Republic
A strategy for extracting formally defined vegetation units from large phytosociological databases is presented, as developed for the current project of vegetation survey of the Czech Republic. In enables to reproduce traditionally recognized vegetation units in a formal way and to create formal definitions of these units which can be incorporated in a computer expert system. New relevés can be assigned to the units of existing classification a posteriori.
This strategy includes the following steps: (1) Geographically stratified selection of relevés from a database containing all vegetation types of a large area; such selection eliminates unduly influence of oversampled areas on the results; (2) Quality assessment of the traditionally recognized associations and alliances by statistical determination of their diagnostic species, using calculations of species fidelity to these vegetation units; (3) Cluster analysis of relevés assigned to the traditionally recognized classes, in order to assess whether the major discontinuities correspond to the traditional associations and alliances; (4) Defining groups of species that tend to occur together, using goodness-of-fit statistics applied to the geographically stratified data set; (5) Defining unequivocal definitions of associations by logical combinations of species groups or dominance of individual species (Cocktail method), with emphasis on sharply delimited vegetation units as defined in steps 2 and 3; (6) Assignment of relevés to the associations, using their formal definitions; (7) Determination of diagnostic, constant and dominant species for associations, based on statistical analysis of the assigned relevés; (8) Assignment of so far unassigned relevés based on calculation of their similarity to constancy columns of associations.