Impact assessment of climate change on species' distribution is necessary to reduce ecological risks; however, it is well known that projected future climates vary among General Circulation Models (GCMs). To reveal the degree of prediction uncertainty, we compared (1) the future climate simulations (2081-2100) of 20 GCMs based on the SRES A1B scenario and (2) future potential habitats of representative plant species of Japan predicted by species distribution models (SDMs).

Future climate data of 20 GCMs were obtained from the 3rd phase of Coupled Model Intercomparison Project (CMIP3), and four climatic variables (warmth index, monthly mean of the daily minimum temperature of the coldest month, and summer and winter precipitation) were calculated for each GCMs. The presence/absence records of target species were used as a response variable of SDMs.

Future thermal variables were consistently increased. However, future precipitation clearly differ among the 20 GCMs, e.g. summer precipitation of 11 GCMs were increased whereas those of the rest were maintained or decreased. Depending on the variance of the future climate variables, obvious uncertainty exist in the predicted potential habitats for plant species. These results suggest that the use of multi-climate data is necessary for understanding the climate change impact on plant distributions.