B. Papp szerk.: Studia Botanica Hungarica 39. 2008 (Budapest, 2008)

Dobolyi, K., Erős-Honti, J.; Botta-Dukát, Z.: Habitat preference of Linum dolomiticum (Linaceae)

Quercuspubescens woodlands and rarely in rock steppe. Some parts of the habitat are considered either as transitions among these plant communities or various stages of their degraded forms. Monitoring of the total distribution area of Linum dolomiticum has been ongoing since 2001. Results of the monitoring proved that the distribution of this species shows a peculiar pattern: in certain patches the plant occurs in rather high density, but these alternate with large "empty" areas where it does not occur at all (DOBOLYI 2001-2006). On the basis of these results and the field experiences the following questions arise: - Does Linum dolomiticum show any preference to any of the surrounding vegetation types? - Are there any other species in its range that have similar phytosociological preferences to Linum dolomiticum} - Does Linum dolomiticum belong to any species-coalition? To find answers for these questions, phytosociological relevés from the distribution area were analysed by multivariate methods; the faithful, constant and dominant species were determined. MATERIAL AND METHODS The present study is based on 277 phytosociological relevés of 2 m x 2 m size taken on the distribution area of Linum dolomiticum so that the complete area of distribution and all types of vegetation are represented. Linum dolomiticum is present in 137 relevés, absent in 140 relevés. Nomenclature of the plant names follows HORVÁTH et al. (1995). Numerical classification of the relevés was done by the method proposed by BOTTA-DUKÁT et al. (2005): at first principal co-ordinates analysis (PCoA) (LEGENDRE and LEGENDRE 1998) based on relativised Manhattan distances was performed to determine the main gradients in the data set. Possible number of ordination axes in such analysis equals the number of relevés minus 1, but usually only the first few ordination axes contain interprétable ecological information, while the others contain largely noise. To establish the number of interprétable axes, the eigenvalues were compared with random expectations based on the broken-stick model (JACKSON 1993, LEGENDRE and LEGENDRE 1998). In our case the first five axes proved to be significant. The co-ordinates along the significant axes of PCoA were used instead of the raw data as input for the classification. In this way, we were able to avoid the low robustness of agglomerative classification