<p>We study the future frequency of atmospheric environments leading to severe thunderstorms over Europe under different climate change scenarios in CMIP5 models. Our method is founded on dynamical system theory, which makes it possible to detect future atmospheric configurations that are close analogues of past events in the class of interest.</p><p>We rely on the EM-DAT and the European Severe Weather Database to select an ensemble of past events leading to significant damage or disruption, including severe thunderstorms, hail storms, derechos and tornadoes, between 1950 and 2020. We consider the geopotential height field at 500 hPa in ERA5 data as a dynamical proxy of the corresponding configurations. Then, we leverage extreme value theory to detect close dynamic analogues in the output of CMIP5 climate projection models under two emission scenarios, namely RCP4.5 and RCP8.5.</p><p>First of all, we assess possible differences between the trends in severe weather frequency due to different radiative forcing. Then, we study the spatial structure of such trends, highlighting regions where the occurrence of such phenomena could see a sharp increase or decrease. Finally, we estimate potential future impact of such phenomena where the information about economic damage is available in EM-DAT.</p><p>&#160;</p>