Vegetation of mountain areas affected by SO2 pollution (Krugn6 hory Mts., Czech Republic) was analysed using multivariate methods. Communities with prevailing species CaIamagrostis villosa, a rhizomatous grass expanding into deforested sites, were sampled by Braun-Blanquet relev6 method. Canonical correspondence analysis was used to assess the effect of environmental variables (soil removal, deforestation, and shading). To test the effects of light, moisture, soil acidity and nitrogen, mean sample indicator values were correlated a posteriori with sample axes on ordination scores. Light, soil acidity, moisture, and site history (in terms of past deforestation and soil removal applied in reclamation procedures) were found to be the main factors responsible for the community composition. Nitrogen level had not a significant effect on the community composition. When analyzing the whole data set, i.e. including also remnants of natural spruce forests, light was the factor affecting at most the composition of communities. Within the bare spot vegetation, if treated separately, the highest variation was found along the soil acidity/moisture gradient. The effect of soil removal was only obvious at early successional stages. Species diversity increased with moisture and decreased with soil acidity. Species exhibiting Sand/or R-strategy are successful on extremely acid soils whereas forbs present in bare spots appear to be supported by disturbances.