Ensemble boosting of extreme precipitation in the Alps

The Alps, with their complex topography, important geographic location and varying climatic influences have become a highly vulnerable region. Especially extreme precipitation and its associated impacts - from floods to landslides - are directly amplified by this distinct local environment.

Because observational timeseries are rather short and sample only limited locations, the impact-producing extreme tail of the precipitation distribution remains largely unexplored. In addition, the non-stationarity of the climate system makes data from a past climate less useful for gaining insights into current and future conditions. Coarse resolution global climate models can be used to produce long simulations including rare extreme events, but important processes such as topographic forcing and deep convection are poorly resolved, which limits physical interpretability. A different approach is needed to produce robust and actionable climate information on the local scales required for stress testing, early warning, adaptation and risk mitigation.

We suggest expanding the method of Ensemble Boosting into the realm of high-resolution modeling. We employ a global ICON setup with 10-20 km grid spacing with a two-way nested kilometer-scale European domain. Our initial goal is to simulate the 2013 Northern Alps flooding using ERA5 initial conditions. We asses lead time sensitivities for reinitializing simulations to optimize for variability and intensity within the boosted ensemble. We expect to produce physically consistent, interpretable and realistic storylines based on a historic extreme precipitation event in the Alps. These storylines enable us to assess driving processes and test physical limits of extreme precipitation in today’s climatic conditions.

With the current focus on a specific region and event we want to exercise a proof of concept embedded in a user-oriented framework. Next steps include producing a catalogue of extremes sampling across event types with the goal to physically constrain the extreme tail of precipitation distributions to reduce uncertainty in extreme value estimation, and to estimate return periods. Further applications of our approach could also be focused on climate projections or pseudo global warming simulations to gain insights into possible extremes in future climates.