Fig. 3: Regional probability distributions of flood magnitudes of different flood generation processes.

a–d Kernel density estimates of flood magnitudes of annual maxima Qmax scaled by mean maximum annual flood of all flood events for each study catchment) aggregated for each region and flood generation process. The right tails of probability density functions are only shown up to the value of 4 (four times the mean maximum annual floods) and displayed in full in the corresponding insets. In all panels, the value of 1 on x-axes corresponds to the mean maximum annual flood. The significance of flood magnitude distributions differences for different generation processes is tested with the two-sided Kolmogorov–Smirnov test (α = 0.01) with a correction for pairwise comparisons by using a procedure based on the false discovery rate (see Supplementary Table 1). In the NORTH only the differences between the distributions of Snowmelt and Rain.Snow flood magnitudes are not significant. In the ATL only the distribution of Rain.Dry floods is different from all others. In the CENTR and in the MED regions the distributions of magnitudes are significantly different among all four flood generation processes apart from the pair Rain.Dry and Snowmelt floods in the latter region. e–h Semi-log plots of the exceedance probability of flood magnitudes (F(x), calculated using Weibull plotting position of floods individually for each generation process) for all study catchments visualising particularly the right tails of regional distributions. The extent of x axes depends on the range of the recorded flood magnitudes in the catchments of corresponding region.