Revisiting probabilistic extreme-event attribution under multiple forcings and nonlinear responses to global warming

Probabilistic extreme-event attribution is based on a well-established methodology that quantifies the influence of anthropogenic climate change on weather-related extreme events. Most commonly, a generalized extreme value (GEV) distribution is fitted to observational, reanalysis, and/or climate model data with global mean surface temperature (GMST) as covariate. This results in a smooth evolution of the GEV parameters, with a prescribed linear GMST dependence of its mean parameter for temperature extremes and an exponential GMST dependence of its mean and spread parameters for precipitation extremes. In reality, however, the smooth (quasi-linear) GMST dependence can be broken by a complex interplay of how multiple forcings evolve and how extreme events respond. Introducing aerosol optical depth (AOD) as additional covariate in aerosol-affected regions has recently been shown to improve the representation of the historical evolution of heat-wave statistics, increasing confidence in expected future changes. Moreover, some event-based storyline simulations of extreme precipitation events exhibit an intensification that seems to level off at high (e.g., around +3°C) warming levels.

Here we use AWI-CM1 CMIP6 historical and scenario ensemble simulations to further explore such nonlinear dependences of heat and precipitation extremes in selected continental regions of the northern extratropics. Specifically, we compare GEV distributions fitted separately for distinct warming levels with those obtained from covariate-based fits. We find cases with pronounced nonlinearities that can be accounted for only partly by adding AOD as a second covariate. Our results underscore the need for careful interpretation of current probabilistic extreme-event attribution, particularly when extrapolating into the future, and highlight the importance of continued methodological development.