Aim To compare theoretical approaches towards estimating risks of plant species loss to anthropogenic climate change impacts in a biodiversity hotspot, and to develop a practical method to detect signs of climate change impacts on natural populations.
Location The Fynbos biome of South Africa, within the Cape Floristic Kingdom.
Methods Bioclimatic modelling was used to identify environmental limits for vegetation at both biome and species scale. For the biome as a whole, and for 330 species of the endemic family Proteaceae, tolerance limits were determined for five temperature and water availability-related parameters assumed critical for plant survival. Climate scenarios for 2050 generated by the general circulation models HadCM2 and CSM were interpolated for the region. Geographic Information Systems-based methods were used to map current and future modelled ranges of the biome and 330 selected species. In the biome-based approach, predictions of biome areal loss were overlayed with species richness data for the family Proteaceae to estimate extinction risk. In the species-based approach, predictions of range dislocation (no overlap between current range and future projected range) were used as an indicator of extinction risk. A method of identifying local populations imminently threatened by climate change-induced mortality is also described.
Results A loss of Fynbos biome area of between 51% and 65% is projected by 2050 (depending on the climate scenario used), and roughly 10% of the endemic Proteaceae have ranges restricted to the area lost. Species range projections suggest that a third could suffer complete range dislocation by 2050, and only 5% could retain more than two thirds of their range. Projected changes to individual species ranges could be sufficient to detect climate change impacts within ten years.
Main conclusions The biome-level approach appears to underestimate the risk of species diversity loss from climate change impacts in the Fynbos Biome because many narrow range endemics suffer range dislocation throughout the biome, and not only in areas identified as biome contractions. We suggest that targeted vulnerable species could be monitored both for early warning signs of climate change and as empirical tests of predictions.