Assessing Climate Model Performance in Simulating Temporal Clustering of Extreme Precipitation in Europe

Understanding the temporal clustering of heavy precipitation is important for flood risk assessment in Europe, but current climate models show limited skill in reproducing these events. This study evaluates ten CMIP6 models against the E-OBS dataset for the period 1981–2020, focusing on consecutive days of extreme rainfall defined above the 95th percentile. Cluster features were measured using mean cluster length and maximum consecutive wet days, and model skill was assessed with correlation, mean absolute error, and Nash–Sutcliffe Efficiency. The results show that models capture large-scale spatial patterns but underestimate persistence by 15–30% in Mediterranean and Alpine regions. In contrast, performance is better in maritime climates, where correlations reach 0.55–0.65 and mean absolute error is close to one day. Sensitivity tests indicate that thresholds and linking rules strongly affect cluster metrics. Spatial analysis further highlights systematic underestimation of persistence in southern Europe. These results point to the need for better representation of persistence, convective rainfall, and orographic effects in climate models. Although the analysis is limited by the coarse resolution of observations and the small set of models, it provides useful evidence for improving model evaluation and supports more reliable flood risk management under climate change.