Fig. 4: Regional flood anomalies and shifts in flood generation processes in four European regions.

a Regionally aggregated flood-rich and flood-poor periods. The height of bars corresponds to the portion of catchments where at least one flood-rich (blue colours) or flood-poor (red colours) anomaly was detected. Flood-rich anomalies are plotted in front of flood-poor anomalies (see Supplementary Fig. 9a for a stacked version). Proportions of anomalies with different return periods (2, 5 and 10 years) are indicated correspondingly FR2, FR5 and FR10 (shades of blue) for flood-rich periods and FP2, FP5 and FP10 (shades of red) for flood-poor periods. b Deviations of frequencies of flood generation processes during periods with opposing prevailing anomalies (corresponding to the blue and red time windows indicated on the bottom of the panel a, see also Methods and Supplementary Note 5 for the sensitivity analysis of the definition of these time windows on the test results) compared to the mean composition of the whole study period in all catchments. Error bars show confidence intervals of differences for each flood generation process for a significance level of α = 0.05. The significance of these differences is evaluated using a χ² test with false discovery rate correction among regions (** indicates significant differences with α = 0.05 of false discovery rate correction; exact p-values for each individual χ² test are provided in Supplementary Table 2) for each period with opposing prevailing anomalies indicated by the blue and red time windows.