Multiparameter probability distributions for heavy rainfall modeling in extreme southern Brazil

Abstract

Abstract Study region The study was conducted in the Rio Grande do Sul state – Brazil. Study focus Studies about heavy rainfall events are crucial for proper flood management in river basins and for the design of hydraulic infrastructure. In Brazil, the lack of runoff monitoring is evident, therefore, designers commonly use rainfall intensity–duration–frequency (IDF) relationships to derive streamflow-related information. In order to aid the adjustment of IDF relationships, the probabilistic modeling of extreme rainfall is often employed. The objective of this study was to evaluate whether the GEV and Kappa multiparameter probability distributions have more satisfying performance than traditional two-parameter distributions such as Gumbel and Log-Normal in the modeling of extreme rainfall events in southern Brazil. Such distributions were adjusted by the L-moments method and the goodness-of-fit was verified by the Kolmogorov–Smirnov, Chi-Square, Filliben and Anderson–Darling tests. New hydrological insights for the region The Anderson–Darling and Filliben tests were the most restrictive in this study. Based on the Anderson–Darling test, it was found that the Kappa distribution presented the best performance, followed by the GEV. This finding provides evidence that these multiparameter distributions result, for the region of study, in greater accuracy for the generation of intensity–duration–frequency curves and the prediction of peak streamflows and design hydrographs. As a result, this finding can support the design of hydraulic structures and flood management in river basins.