Valutazione delle precipitazioni e dei venti estremi in Europa su scala convettiva

Convection-permitting models (CPMs) mark a significant change in climate modelling by explicitly resolving atmospheric convection at kilometre-scale resolutions (≤4 km). They offer new possibilities for assessing extreme weather in a climate change context. However, their high computational requirements constrain simulation lengths to 10-20 years. This limits the use of traditional extreme value analysis methods. Moreover, the models’ performance in representing atmospheric extremes has not been fully assessed. This thesis evaluates CPM ensembles for estimating extreme events and addresses key limitations by applying alternative methods. Methodologically, this thesis uses the Simplified Metastatistical Extreme Value (SMEV) approach for both precipitation and wind extremes, this being the first time it has been applied to extreme winds. SMEV allows for robust estimation of the extreme return levels from short CPM simulations, since it uses entire datasets of ordinary events instead of just annual maxima, providing statistically reliable extreme quantiles with uncertainty quantification through bootstrapping. The thesis consists of three studies, which focus on precipitation and wind extremes using CORDEX Flagship Pilot Study CPMs. The first study looks at how orographic features affect extreme sub-daily precipitation biases in the eastern Italian Alps, using 174 rain gauge observations and seven CPM ensemble members. The results show systematic elevationdependent biases: a 20-30% underestimation in lowlands (≤100m) where the highest intensities occur, and a 15-25% overestimation in highlands (>1100m) where the inverse orographic effect appears. Inter-model uncertainty (15-25%) is consistently higher than intra-model uncertainty (12-18%), especially for short durations (1-3 hours) that are critical for assessing flash floods. The ensemble does capture orographic effects, but with a weaker magnitude than observed. This indicates a need for elevation-specific bias corrections. The second study evaluates wind spectral properties at turbine heights (100 m). It shows that CORDEX-FPS CPMs naturally maintain the theoretical −5/3 Kolmogorov slope across the 1-12 day−1 frequency range, without needing postprocessing corrections. This contrasts sharply with NEWA (around -2.10 slope) and ERA5 (around -2.45 slope), which show artificial spectral damping that requires correction methods. The superior spectral representation stems from the freely evolving design of CPM simulations, which preserves natural energy cascade processes, unlike constrained models that employ spectral nudging or data assimilation. The third study creates a framework for assessing extreme wind by combining multi-dimensional surface categorisation (climate, roughness, opography) with Principal Component Analysis to evaluate inter-model agreement across central Europe. Analysing three independent CPMs shows strong consensus, with the first principal component explaining 74.2% of variance. This indicates strong climatological signals despite systematic differences in magnitude. Seasonal variations are evident, with higher winter correlations (0.6-0.8) reflecting synoptic forcing and lower summer correlations (0.1-0.4) dominated by convective processes. These advancements demonstrate that CPMs can be effective tools for assessing extreme weather, while also identifying systematic biases and uncertainties that necessitate ensemble approaches. The findings support the development of better climate adaptation strategies, improved wind energy planning, and more effective risk management in complex terrains. They also provide a scientific basis for moving from traditional coarse-resolution models to convection-permitting frameworks for analysing extreme events.