Desert Fishes Council, Death Valley, CA, November 2003.

Rarity, fragmentation, and the scale-dependence of extinction risk in desert fishes.

W. Fagan & P.J. Unmack

Theoretical efforts and small-scale experiments have given rise to the widespread belief that the fewer occurrences a species has or the more fragmented its distribution is, the more vulnerable that species should be to extinction. Lacking, however, are large-scale studies exploring the connection between these aspects of spatial rarity and local extinction risk across many species. We present a landscape-level, biogeographic test of this widely assumed linkage. Using a unique dataset detailing the occurrence patterns of native freshwater fishes of the Sonoran Desert (compiled by W.L. Minckley), we obtained several measures of spatial rarity for each of 25 species. Some of these rarity measures were scale-dependent, and one, the "scale-area slope" was independent of spatial scale. This slope statistic, which characterized the degree to which species' ranges were historically fragmented, proved a consistently strong predictor of extinction risk, and reached a maximum of predictability at intermediate scales. At the 100km scale, historic range fragmentation explained over 90% of the among-species variance in realized extinction risk, and desert fish species with the most fragmented historic distributions were more than nine times more likely to be currently absent from a given stream reach than were species with the most continuous distributions. In contrast, the number of reaches occupied (as defined on a series of hierarchical scales) was a significant predictor of extinction risk only if fragmentation had not already been accounted for and decreased in importance as scale increased. These findings have three major implications. First, they underscore what a strong link exists between spatial distribution and vulnerability to extinction. Second, they clarify that the link exists even at the landscape-level and across an entire biogeographic fauna. Last, they demonstrate how extinction risk can be a scale-dependent phenomenon that is affected by aspects of species' distributions operating at both finer and coarser scales.