Evaluating sub-daily extreme precipitation from an ensemble of convection-permitting simulations: the role of topography.

Past studies have shown that in orographically complex terrain, observed extreme precipitation intensity is impacted by elevation in different ways at different durations. Convection-permitting climate models (CPMs) are receiving increasing attention thanks to the more realistic representation of extreme sub-daily precipitation compared to coarser climate models. Two almost still unexplored themes concern: i) CPMs' ability to represent the observed relationship between precipitation and topography and ii) how the model ensemble uncertainty depends on elevation. To address these questions, we evaluate sub-daily extreme precipitation from an ensemble of eight CPM members (reanalysis-driven simulations) on topographically diverse terrains. We use observed data from rain gauges as benchmark. The analysis is conducted over the Eastern Italian Alps, where a strong relationship between precipitation sub-daily extremes and topography is observed (Dallan et al., 2023). We apply a non-asymptotic statistical approach (Simplified Metastatistical Extreme Value, SMEV) to estimate extreme precipitation return levels and assess their intra-model and inter-model uncertainties using a bootstrapped samples method. It is shown that the ensemble mean describes in a realistic way the precipitation extremes, with fractional standard errors of the mean-over-the-ensemble return levels ranging between 0,16 (24 hrs duration, 2 yrs return time) to 0,41 (1 hr duration, 100 yrs return time). We found that, compared to rain gauges, CPMs systematically underestimate extreme return levels in lowlands, whereas overestimate them at higher altitudes. Nevertheless, the CPMs can capture the relationship between rain depth and elevation, which is particularly important for 1-3 hrs duration. While the intra-model uncertainty decreases systematically with elevation at all durations, a more complex behaviour is observed for both inter-model and total uncertainty. These findings help to characterize the impact of elevation on the ensemble of CPM simulations, which is particularly required for the applications of these simulations for adaptation to future flood risk.

REFERENCES
Dallan, E., Marra, F., Fosser, G., Marani, M., Formetta, G., Schär, C., & Borga, M. (2023). ID56. How well does a convection-permitting regional climate model represent the reverse orographic effect of extreme hourly precipitation? Hydrology and Earth System Sciences, 27(5), 1133–1149. https://doi.org/10.5194/hess-27-1133-2023.