Environmental drivers and spatial scaling of species abundance distributions in Palaearctic grassland vegetation

Species abundance distributions (SADs) link species richness with species abundances and are an important tool in the quantitative analysis of ecological communities. Niche‐based and sample‐based SAD models predict different spatial scaling properties of SAD parameters. However, empirical research on SAD scaling properties is largely missing. Here we extracted percentage cover values of all occurring vascular plants as proxies of their abundance in 1725 10‐m(2) plots from the GrassPlot database, covering 47 regional data sets of 19 different grasslands and other open vegetation types of the Palaearctic biogeographic realm. For each plot, we fitted the Weibull distribution, a model that is able to effectively mimic other distributions like the log‐series and lognormal, to the species–log abundance rank order distribution. We calculated the skewness and kurtosis of the empirical distributions and linked these moments, along with the shape and scale parameters of the Weibull distribution, to plot climatic and soil characteristics. The Weibull distribution provided excellent fits to grassland plant communities and identified four basic types of communities characterized by different degrees of dominance. Shape and scale parameter values of local communities on poorer soils were largely in accordance with log‐series distributions. Proportions of subdominant species tended to be lower than predicted by the standard lognormal SAD. Successive accumulation of plots of the same vegetation type yielded nonlinear spatial scaling of SAD moments and Weibull parameters. This scaling was largely independent of environmental correlates and geographic plot position. Our findings caution against simple generalizations about the mechanisms that generate SADs. We argue that in grasslands, lognormal‐type SADs tend to prevail within a wider range of environmental conditions, including more extreme habitats such as arid environments. In contrast, log‐series distributions are mainly restricted to comparatively species‐rich communities on humid and fertile soils.