Floristic variation, chorological types and diversity:: do they correspond at broad and local scales?

The relationships between biogeographical patterns and local-scale patterns based on microscale features, such as topoclimate, are well known in plant biogeography. Here we present a method of determining this correspondence using constrained ordination and correlations. We examined compositional gradients at two different scales, biogeographical chorotypes, and diversity. Compositional data (124 taxa x 113 plots) were sampled at four regularly spaced sites in south-eastern Spain. Longitude (LONGI) was used as a spatial variable representing an east-west climate gradient, together with a radiation index (RADIN), elevation, and a disturbance indicator. All factors correlated with the compositional gradients, but the local-topoclimate factor (RADIN) and the broad-scale factor (LONGI) were most important. These two, spatially independent factors were both correlated with the two first ordination axes, and therefore should relate to the same general trend in species-turnover. There was a significant Spearman's rank correlation between the species order along these two gradients. This is interpreted as an ecological self-similar pattern, i.e. coenoclines repeating at different scales. A consistent order of species along local- and broad-scale coenoclines may indicate that similar operational factors act at several scales, here related to moisture and temperature. The distribution of Mediterraneo-Macaronesian, Mediterraneo-Saharo-Arabian and Ibero-Maghribian species confirmed the correspondence between the broad- and local-scale gradients. The former group decreases in number with increasing aridity along both gradients, whereas the two latter groups increase. A discordant pattern was found with south-eastern Iberian endemics, but this may be explained by several of them being edaphic (saxicolous) specialists. There is a significant decrease in species richness with high radiation, but the expected increase along the longitudinal gradient from west (dry) to east (moist) was not statistically significant. This may be due to the correspondence between high richness and disturbance, both occurring in the middle of the broad-scale gradient.