Uncertainties in the assessment of future temperature and precipitation extremes in Central Europe

Abstract

In the context of global warming, changes in climate and weather extremes are of particular relevance, although their assessment is subject to many sources of uncertainty. In this study, we address to what extent the estimate of projected temperature and precipitation extremes in Central Europe is sensitive to i) the choice of a theoretical extreme value distribution, ii) to a random sample of given data and iii) to the initial conditions of a model experiment. When evaluated with empirical extreme values, the Gumbel distribution is outperformed by the other considered extreme value distributions, especially for temperature. The above-threshold distribution is characterized by the highest flexibility and the best fit at many grid boxes. Projected changes in temperature and precipitation extremes are more sensitive to the chosen statistical model than to initial conditions. The discrepancies are largest in mountainous regions. Using the best statistical fit at every model grid box reveals a mostly significant tendency towards warmer temperature extremes and more intense heavy precipitation, particularly in the Alpine region.