Regional growth curves and improved design value estimates of extreme precipitation events in the Czech Republic

Abstract

Regional analysis was used to improve estimates of the probabilities of extreme precipi- tation events in the Czech Republic. The delimitation of 4 homogeneous regions was based on statisti- cal procedures (cluster analysis of site characteristics and subsequent tests for regional homogeneity). The regions also reflect climatological differences in precipitation regimes and synoptic patterns that cause heavy precipitation. The Generalized Extreme Value (GEV) distribution was identified as the most suitable distribution for modelling maximum annual 1 to 7 d precipitation amounts according to the L-moment ratio diagram and goodness-of-fit tests. Only in the northeast region (which is most prone to the occurrence of extremely high precipitation totals) was the Generalized Logistic (GLO) dis- tribution preferred. The regional approach considerably lessened the between-site variation of esti- mates of the shape parameter of the GEV/GLO distribution compared to at-site procedures, and the estimates of high quantiles (e.g. 50 yr return values) were more reliable and climatologically consistent in individual regions. Different growth curve shapes are characteristic of the 4 regions, the between- region variability being larger for multi-day than 1 d events. Particularly noteworthy is the heavy tail of distributions of 5 and 7 d annual maxima in the northeast region, reflected also in the inapplicability of the general 4-parameter kappa distribution in regional homogeneity tests. A nonparametric statistical test on the value of the tail index supports a hypothesis that data in this flood-prone area may be drawn from a distribution with a very heavy tail for which L-moments do not exist.