Although the species-area relationship (SAR) is of central importance to ecology, we still have little understanding of the ecological mechanisms that determine the rate at which species number rises with area. When considering how different mechanisms operate in conjunction with one another, ecologists generally assume that each mechanism is most influential over a contiguous scale range (e.g., colonization dynamics could structure the SAR at intermediate scales, but we would not expect colonization to be influential at small and large but not intermediate scales). The three most commonly cited explanations for the SAR are: a sampling effect (larger areas sample more individuals and thus more species), habitat heterogeneity (larger areas have more habitats), and dispersal/extinction dynamics (island biogeographic theory). Here I present the results of a 3-yr experiment in which I evaluated the relative importance of these three explanations for “mainland” (vs. island) SARs. I established a series of plots on serpentine grassland and applied manipulations to reduce either dispersal from plants inside the plots, dispersal from outside the plots, or spatial heterogeneity in the soil, plus all two-way combinations of these manipulations. SARs were calculated for all vascular plant species in each plot in each year.
The results indicate that sampling does contribute to the positive slope of the SAR, but sampling by itself does not explain the rise in species with area or the functional shape of the SAR. Reduction of dispersal within plots influenced disjunct portions of the SAR, but reduction in dispersal from external sources and soil homogenization were less influential. The two-way manipulations suggest important interactions between dispersal and spatial heterogeneity, and that the seed bank has a role in structuring a portion of the SAR. The general conclusion is that the influence of processes affecting diversity can rise, fall, and then rise again as a function of spatial scale. Thus, models combining mechanisms in a linear fashion, as has been done in the past, may not capture all of the variation we see in the SAR.