Rareness Amplified INtensification of Extreme rainfall (RAINE): how the worst events get worse the fastest

Floods associated with extreme precipitation cause tremendous damage and losses every year, and are projected to become more frequent and more severe with climate change in most land regions. Events of much higher intensities than previously observed can cause unforeseeably large impacts due to their unprecedentedness. The changing occurrence probability of such “surprise events” is closely related to changes in the statistical distribution of extreme precipitation: while a constant scaling with temperature (such as Clausius-Clapeyron) causes a constant fractional increase for all return levels, strong increases in the variability of extreme precipitation (distribution width) lead to relatively stronger intensification of the most extreme events. The latter change is indicative of increasing high-impact surprise event probabilities. Regions where rare extremes exhibit a faster relative intensification than moderate extremes (skewed intensification) are subject to RAINE: Rareness-Amplified INtensification of Extremes. In other words, RAINE means the worst events get worse the fastest.

We present a statistical framework based on extreme value theory to diagnose RAINE in annual maximum daily precipitation (Rx1d) from observations and simulations. We focus in particular on results from the 10-member high-resolution (0.25° atmosphere/land and 0.1° ocean) CESM1 ensemble (MESACLIP, historical+RCP8.5), which has been shown to simulate Rx1d quite accurately. We identify a strong RAINE-effect for most of the global land over the 21st century under RCP8.5. We categorise the data based on region and Rx1d-causing weather phenomenon, and find that a physical scaling based on vertical updraft and relative humidity explains the RAINE pattern. Different seasons, regions and phenomena feature different relative contributions of vertical updraft and relative humidity to RAINE, which can be linked to different environmental conditions and climate change effects governing Rx1d changes. In observations, robust distributional changes are difficult to detect due to high variability of extreme precipitation. Our combined statistical and physical characterisation of RAINE can help explain and constrain uncertainties in future risks posed by unprecedented extreme precipitation.