Is variable plot size a serious constraint in broad–scale vegetation studies? A case study on fens
|Year of publication
|Article in Periodical
|Magazine / Source
|Journal of Vegetation Science
|MU Faculty or unit
|fens; phytosociology; plot size; scale; specialist plants; species–area relationship; vegetation classification; vegetation plot; vegetation survey; wetlands
|Filtering vegetation plot records according to sampling size is an essential methodological step in vegetation studies. In fens, the variation of traditionally used plot sizes seems to limit continental-scale syntheses following the Braun-Blanquet approach. Which plot sizes harbour the analogous number of habitat specialists (i.e., diagnostic/indicator species) and capture the main compositional gradients identically? The data set of fen vegetation plot records was compiled using large databases and categorised into four distinct habitats. For each habitat, semi-log species–area curves of specialists and other species were fitted using generalised additive models (GAM). In addition, we surveyed 72 sites in a series of plot sizes (0.07, 0.25, 1, 4, 16 m2) where we applied, separately for each plot size, Non-Metric Multi-Dimensional Scaling (NMDS) and compared the resulting patterns with Procrustes analysis. Consistently across different fen habitats, the species–area curves of specialists increased steeply up to the plot size of 1 m2, while increasing negligibly in the plot size range of 1–25 m2. In contrast, the species–area curves of other species displayed mostly linear to linear-exponential trends. NMDS ordinations of medium (1 and 4 m2) and large plots (16 m2) were the most congruent, while the patterns captured in the ordination of the smallest plots (0.07 m2) differed most from the others. In fens, plot sizes of at least 1 m2 describe sufficiently the broad-scale pattern in specialists’ diversity as well as the main environmental gradients. The range of plot sizes of 1–25 m2 may be safely merged in broad-scale analyses of fen vegetation without introducing substantial bias, at least when compared with other possible uncertainty sources.