ALBERT

Description: Reliable estimate of future changes in precipitation extremes at sub-daily temporal scales is of fundamental importance for improving risk assessment and management and increasing our resilience to climate change. This is particularly true in mountainous regions as they are vulnerable to natural hazards triggered by short-duration extreme precipitation such as flash floods, landslides and debris flows. High-resolution convection-permitting models (CPMs) can represent the convective processes, key to short-duration extremes, better than coarser resolution regional models, even in complex-orography areas. Here, we derive future projections of extreme return levels of short-duration precipitation in a mountainous region using an ensemble of CPM simulations. We focus on the northeastern Italian Alps, where previous studies already reported significant changes in observed precipitation extremes over the last decades. Using a six-model CPM ensemble CORDEX-FPS (spatial resolution ∼3 km, 8.5 RCP scenario), we estimate the future changes, and related uncertainty, in return levels of 1 to 24 h precipitation up to 1% yearly exceedance probability (100yr return time), for near (2041-2050) and far (2090-2099) future. We estimate return levels from the available 10yr-time-period simulations using a state-of-the-art non-asymptotic statistical approach which was already applied successfully to the short time periods of CPM simulations. Future changes appear to vary spatially also depending on the rainfall duration and the time period (near versus far future). We provide a physical interpretation of these patterns and evaluate the model uncertainty in relation with the natural variability of the ensemble. These results provide reliable and useful information for risk management stakeholders.