Developing low-likelihood climate storylines for extreme precipitation using ensemble boosting

Heavy precipitation events as the one in western Germany and the Benelux countries in July 2021 destroy the local infrastructure and numerous fatalities. Due to the lack of long homogenous climate data and methodological framework, it is uncertain how intense precipitation extremes could get. We address these questions by developing storylines of the rarest precipitation events. We here generate large samples of reinitialized heavy rainfall events starting from the most extreme events in an initial condition large ensemble for the near future, carried out with CESM2. In an approach referred to as ensemble boosting, we first reinitialize the most extreme 3-day precipitation events to estimate how anomalous they could get. We find that the most extreme precipitation events can be substantially exceeded in the boosted ensembles for different regions across the world. Second, we evaluate whether the model can reproduce analogues of the precipitation event in July 2021 and re-initialize these events to analyze how this event type could have evolved and whether it could have become even more intense. In doing so, the ensemble boosting method provides storylines of heavy rainfall development beyond the observational record, which can be used to generate worst-case scenarios and stress test the socioeconomic system.