Abstract. Relative abundance distributions (RADs) are an important feature of community structure, but little is known of the factors affecting which type of RAD is observed in a particular community. We examined the influences of species richness and of spatial scale on the RAD of plant communities.
The effect of species richness was examined by analysing simulated communities generated under the Broken stick model, the Sequential breakage model, and a randomized version of Niche pre-emption model. In all cases, when there were few species in the community the data only occasionally gave the best fit to the model that was used to generate it. With 40–65 species, a best fit was obtained for the correct model in about 75 % of cases, almost irrespective of the model.
Effects of spatial scale were examined in data from four dune slacks and from two semi-arid grasslands, by analysing biomass values at a range of sample sizes. The model that best fitted the whole sample differed between the four slacks and between the slacks and the semi-arid grasslands. The change in which model of RAD fitted best, as sample size was reduced, varied between sites and between habitat types. At the smallest sample sizes, the Zipf-Mandelbrot model often fitted, and in the slack sites the Broken stick also, though neither fitted (in the vegetation examined) at larger spatial scales.
It is concluded the RAD is affected by species richness and by spatial scale, in ways that currently do not enable simple prediction. RADs can theoretically give information on the processes such as resource partitioning, immigration and competition that have structured the community, but they are a blunt tool for this purpose.