High-impact Mediterranean precipitation extremes to increase faster than moderate extremes in the CORDEX future projections

An increase in the intensity of daily precipitation extremes is among the most robust responses to anthropogenic climate change. However, while many studies have focused on moderate extremes corresponding to the mean of annual maxima, or their median which corresponds to a return period of 2 years, high-impact extreme precipitation events are related to less studied events with much longer return periods (e.g. 100 years, or longer). The physical and statistical study of these events is hampered by the difficulty in building robust statistics in climate records only a few-decades long. In particular, it is still poorly understood whether moderate and high-impact precipitation extremes may intensify at the same rate, or whether differences may arise due to, for instance, changes in the frequency or meteorology of the driving weather events, in their seasonality, or in the balance between convective and stratiform precipitation.

We address this question by exploring the projected changes in tail heaviness of daily  precipitation extremes in 63 single-member simulations from the EURO-CORDEX ensemble, run at 12km resolution, in the RCP8.5 scenario. Tail heaviness (TH) is here defined as the ratio between the quantiles corresponding to the 100-year return period relative to the 2-year return period. Due to the difficulty in evaluating long return periods from single-member simulations, we first use the 50-member initial condition CRCM5 regional large ensemble, for which statistics can be accurately estimated, to test the ability of extreme value theory (GEV distribution) and Simplified Metastatistical Extreme Value theory (SMEV) in estimating changes in TH.

The results show that SMEV has a smaller root mean squared error than GEV in estimating changes in TH from 30-year long climate records extracted from the CRCM5 ensemble, proving it a better methodology for this purpose. When SMEV is applied to the CORDEX ensemble, a likely (66% to 90% of models) increase in TH is found in the Mediterranean region, while small and non robust changes are found in Central and Northern Europe. The robustness of the Mediterranean response is not detectable using GEV. The increase in TH is shown to constitute a sizable contribution to the increase in the 100-year level of Mediterranean precipitation extremes. A reduction in the number of precipitation events partly balances the increase in the 2-year return period, but has little impact on the 100-year return period, contributing to its faster relative intensification. 

We conclude that while in Central and Northern Europe the rate of change in moderate (2-year) and high-impact extremes cannot be distinguished from estimation uncertainties, great care is needed in the Mediterranean region, where the risk of exposure to high-impact precipitation events due to climate change may be increasing faster than what perceived based on the trends of moderate extremes