B. Papp szerk.: Studia Botanica Hungarica 30-31. 1999-2000 (Budapest, 2000)

Csontos, Péter, Lőkös, László; Molnár, Katalin: Numerical analysis of lichen zones in Komárom, NW Hungary

Multivariate methods were applied for data evaluation. Based on the attribútum duality concept (cf. PODANI 1997), both the sampling sites and species were grouped by hierarchical classification methods (cluster analysis), of which results related to the sampling sites are discussed in this paper. Several similarity functions and coefficients are known to measure the floristic similarity of the sampling sites (based on presence/absence data). Since the average species number of our sites was low compared to the species pool of the whole sample, it is reasonable to omit "mutual absences", when the similarity matrix is calculated (TAMÁS 1997). Among indices using only 3 cells (a, b, and c) of the contingency table, we used the Jaccard-index: J t.) - a/(a+b+c), where a is the number of species present both in samples i and j, b is the number of species present in sample i only, and c is the number of species present in sample j only. Jaccard-index is the unweighted member of a subgroup of indices often producing topologically identical results (PODANI 1997, TAMÁS 1997). Three of the clustering algorithms were used: the complete linkage-, the group average- and the global optimization method (PODANI 1989, 1993). It is recommended to use various algorithms in a study for better representation of the general trends in the analysed data set. The SYNTAX program package were used for computation (PODANI 1991, 1993). Localities with only one species, and species occurred at only one locality were removed from the data set. Thus a binary matrix of 36 species and 65 localities served as raw data set for the multivariate analysis (see Appendix). RESULTS AND DISCUSSION Results of the global optimization method The dendrogram produced by the global optimization method is shown in Figure 2. The two major groups of objects are formed by sites where the majority of the species were found on tree barks (marked by number 1 ), and sites where saxicolous species dominated the samples (marked by number 2). Further groups can be distinguished, if the dendrogram is cut at 0.736 level of dissimilarity (see broken line on Fig. 2). One of the epiphytic subgroups (number 3) contains the sampling sites of wooded areas relatively far from the built-up areas. The other subgroup (number 4) is formed by localities in small wooded patches and allées