Quantifying the evidence of climate change in the light of uncertainty exemplified by the Mediterranean hot spot region

Abstract Climate change projections are subject to uncertainty arising from climate model deficiencies, unknown initial conditions and scenario assumptions. In the IPCC reports and many other publications climate changes and uncertainty ranges are usually displayed in terms of multi-model ensemble means and confidence intervals, respectively. In this study, we present a more quantitative assessment and statistical testing of climate change signals in the light of uncertainty. The approach is based on a two-way analysis of variance, referring to 24 climate models from the CMIP3 multi-model ensemble, and extents over the 21st century. The method also distinguishes between different climate variables, time scales and emission scenarios and is combined with a simple bias correction algorithm. The Mediterranean region has been chosen as a case study because it represents an assumed hot spot of future climate change, where temperature is projected to rise substantially and precipitation may decrease dramatically by the end of the 21st century. It is found that future temperature variations are mainly determined by radiative forcing, accounting for up to 60% of total variability, especially in the western Mediterranean Basin. In contrast, future precipitation variability is almost completely attributable to model uncertainty and model internal variability, both being important in more or less equal shares. This general finding is slightly depending on the prescribed emission scenario and strictly sensitive to the considered time scale. In contrast to precipitation, the temperature signal can be enhanced noticeably when bias-correcting the models' climatology during the 20th century: the greenhouse signal then accounts for up to 75% of total temperature variability in the regional mean.