|ABSTRACT||Effective adaption and mitigation measures against the impact of anthropogenic global warming require reliable estimates of future climate change. Coupled general circulation models (CGCMs) are still the most appropriate tool to assess future changes. However, the climate projections of individual models differ considerably, particularly at the regional scale and with respect to certain climate variables such as precipitation. Model differences result from unknown initial conditions, different resolutions and driving mechanisms, varying model parameterizations and emission scenarios. It is very challenging to determine which model properly simulates future climate conditions.
The aim of this study is to derive optimized probabilistic estimates of future precipitation changes in the Mediterranean region from the multi-model ensembles of CMIP3 and CMIP5. The analyses are carried out for the meteorological seasons in eight Mediterranean sub-regions, based on the results of principal component analyses.
To derive model-specific weights which affect the probabilities of the projected precipitation changes, atmospheric predictor variables for the statistical downscaling of Mediterranean precipitation are analyzed. Important key predictors like sea level pressure, wind components, atmospheric layer thickness, and air humidity are investigated for observation-based/reanalysis and modeled data. The ability of each analyzed CGCM to represent the observed predictor-predictand relationships leads to specific model weights implying that models which are closer to the observed results get higher weights than models with less realistic results. To fulfill the requirements of temporal consistency and transferability to scenarios, the model weights are calculated and then validated in independent historical time periods.
Finally, the weights are applied to the projected precipitation changes, whereby the greater (lower) weight of better (worse) models entails a more reliable precipitation change signal in the Mediterranean region.|