Search

COMEPRO - Comparison of Metrics for Probabilistic Climate Change Projections of Mediterranean Precipitation


Start date: 01.04.2015
Duration: 2,5 Years
Funded by: DFG (Deutsche Forschungsgemeinschaft)
Local project leader: Prof. Dr. Jucundus Jacobeit Dr. Irena Kaspar-Ott
External scientists / cooperations: Prof. Dr. Heiko Paeth, Universität Würzburg, Institut für Geographie und Geologie

Abstract

Climate protection and adaptive measures require reliable estimates for future climate change. Coupled climate models are still the most appropriate tool. However, the climate projections of individual models differ considerably, particularly at the regional scale and with respect to certain climate variables such as precipitation. Thus, significant uncertainties also arise on the part of climate impact research. The model differences result from unknown initial conditions, different resolutions and driving mechanisms, different model parametrizations and the emission scenarios. It must be impossible to determine which model outputs proper future projections, so many climate model simulations are usually linked to an overall image. By implementing this information in probability density functions, the over- and underflow probabilities with respect to certain thresholds of climate change can be determined. Such probabilistic assessments are indeed relevant for planning, since they show besides the mean change also the range of uncertainty of future climatic development paths.

The aim of this project is to derive such probabilistic estimates of future precipitation changes in the Mediterranean region of the multi-model ensemble CMIP3, CMIP5, ENSEMBLES and CORDEX and from a statistical downscaling approach.

The Mediterranean region represents a so-called hot spot of climate change. The analyses are carried out for both seasonal averages as well as extreme events. The methodologically innovative aspect refers mainly to the comparison of different metrics to derive model weights, such as Bayesian statistics, regression models, spatiotemporal filter, the fingerprinting and quality criteria for the simulated large-scale circulation, and the associated changes in the probability densities of a projected future climate. This approach implies that a part of the model uncertainty is inherent in the system, especially due to the unknown initial conditions and uncertain empirical parameters. Moreover, the probability densities are examined on the multi-model ensembles for other phenomena in the process chain for precipitation formation, such as radiation, evaporation, advection and cloud cover, to determine the level at which the divergence between different climate model projections begins.