PREDICTION OF RARE-PLANT OCCURRENCE - A SOUTHERN APPALACHIAN EXAMPLE

Ecologically sound efforts to manage or reintroduce populations of rar e species require detailed knowledge of species habitat requirements. However, the fact that such species are rare implies that the data nee ded for habitat characterization are sparse and that species might wel l be absent from favorable sites due to chance aspects of dispersal or mortality. We use four rare plant species endemic to southern Appalac hian high-elevation rock outcrops, to illustrate how nonparametric and parametric logistic regression can yield predictive models of the pro bability that a species will occur, given certain site conditions. Mod els were constructed for each species at two scales: 100-m(2) plots an d 1-m(2) subplots. At the 100-m(2) plot scale, absences beyond the cur rent geographic range were excluded. At the 1-m(2) subplot scale, abse nces from subplots were only included if the species occurred elsewher e on the 100-m(2) plot. Six significant models resulted; no significan t model could be constructed for Solidago spithamaea or Calamagrostis cainii on 1-m(2) subplots. For 100-m(2) plots, the most valuable predi ctors were potential solar radiation, a soils gradient related to avai lable soil iron, boron, and copper, and coarse-scale rock surface text ure, although Geum radiatum occurrences were difficult to predict at t his scale. For 1-m(2) subplots the best predictors were available soil cations, potential solar radiation, the proportion of exposed bedrock , and vegetation height. Along individual gradients response curves we re often similar, but no two species were predicted by identical sets of site parameters. Beyond current range limits, existence of suitable habitat on 100-m(2) plots was demonstrated for Solidago spithamaea, s upporting a view that the range limits of this species are not necessa rily set by availability of suitable habitat. Habitat-based models hav e numerous management applications (such as to guide restoration and r eintroduction efforts as well as to direct searches for additional pop ulations) and provide a framework for future work on species-specific physiological requirements.