A multi‐scale analysis of fragmentation effects on remnant plant species richness in Phoenix,Arizona

Aim Understanding complex ecological phenomena, such as the determinants of species richness, is best achieved by investigating their properties at different spatial scales. Factors significantly affecting the number of species occurring at one scale may not impact on richness at other scales. While this scale dependence has become increasingly recognized, there still remains a need to elucidate exactly how richness is structured across scales, and which mechanisms are influential for determining this important community property. This study explores how woody plant species richness varies in a fragmented system at multiple scales, and which factors are primarily responsible for these patterns. Location The study area is located in the Sonoran Desert within the bounds of metropolitan Phoenix, Arizona, which is the locus of the Central Arizona–Phoenix Long‐Term Ecological Research (CAP‐LTER) site. Methods Estimates of local and fragment plant species richness were generated from field data collected from 22 sites. Independent variables describing fragment sites were also calculated, including area, habitat heterogeneity, density of individuals, mean elevation, and extent of isolation. Structural equation modelling, multiple regression, and analysis of covariance were used to assess the contribution of independent variables to richness at the fragment and local scales. Results Fragment species richness was significantly influenced by area, though not isolation, habitat heterogeneity, mean elevation, or density of individuals. Local richness was not significantly related to fragment area, but was positively related to fragment richness, plant density, and elevation. Main conclusions The fragment species–area effect resulted from larger remnants supporting higher numbers of individuals at comparable densities, increasing richness through either passive sampling of progressively less common species and/or lower extinction rates among larger populations. Without using multi‐temporal data it is not possible to disentangle these mechanisms. We found that patterns evident at one scale are not necessarily apparent at other scales, as elevation and density of individuals significantly affected richness at the local scale but not at the fragment scale. These results lend support to the concept that mechanisms influencing the species richness of natural communities may be operable only within certain domains and that relevant scales should be specified.