The efficacy of species-area relationship to indicate fragmentation effects varies with grain size and with heterogeneity

Island species-area relationships (ISAR), relating fragment area with fragment-level species richness, and 'species density'-area relationships (D-SAR), relating fragment area with the number of species within a standardized sampling area (species density), are both commonly used as indicators of fragmentation effects on diversity. While numerous mechanisms underlie a positive ISAR, only a small subset of these underlies a positive D-SAR; thus, a positive D-SAR can be used as an indicator for the action of these particular mechanisms. Most frequently, a positive D-SAR is interpreted as a support for the habitat heterogeneity hypothesis of ISAR, which relates species richness with habitat heterogeneity. In this study, we examine two often-neglected aspects concerning the application of D-SAR as an indicator for fragmentation effects. First, the detection of a significant D-SAR, and thus its usefulness as an indicator for some underlying mechanisms, is grain-size dependent. Second, the derivation of D-SAR as an indicator of habitat heterogeneity is often implicitly based on a seldom tested assumption of a parallel within-island species-area relationship. We used a grain size-dependent hierarchical uniform sampling of plant species density and species richness in a fragmented semi-arid agroecosystem to address the following questions: a) Can fine-scale heterogeneity account for D-SAR in Mediterranean plant communities?; b) Do the detection of a significant D-SAR and its slope vary with grain size?; and c) Do within-fragment SARs have uniform slope or does the slope vary with fragment size? In a system of fragments with a relatively uniform habitat, we found a significant SAR, but a non-significant D-SAR regardless of the grain size, which varied from 0.0625 m(2) to 225 m(2). These results contrast with those from a heterogeneous fragmented system where both SAR and D-SAR were significant and where D-SAR was grain size-dependent. We also found that the slope of within-fragment SAR decreased with fragment size, in contradiction to their frequent depiction in the literature as being parallel. Our study supports the notion that D-SAR is mainly generated by the processes considered by the habitat heterogeneity hypothesis. Furthermore, it provides evidence that habitat heterogeneity may account for significant D-SAR even when within-fragment SAR do not share a common slope. Our study emphasizes the need for clarity in matching theoretical prediction with actual empirical measures, while studying fragmentation effects on biodiversity.