Author(s)

M. Holický, K. Jung & M. Sýkora

Abstract

Damage of construction works due to flooding in 1997 and 2002 in the Czech Republic initiated investigations of structural failures and reassessment of available data for discharge extremes. In this study, hydrological data for 166 annual maximum discharges of the Vltava River in Prague since 1827 are analysed using various statistical methods. Moment characteristics of the measurements – the mean, standard deviation and skewness – are estimated and the enhancing effect of an exceptional observation in 2002 is detected. The annual maxima are described by two- or three-parameter lognormal distributions and the extreme value distributions of the type I and II. Standard statistical Kolmogorov and chi-square tests are applied to assess goodness of fit of the theoretical models. It appears that a two-parameter lognormal distribution may be the most suitable theoretical model. Assuming this distribution, extreme discharges corresponding to characteristic and design values are estimated. It is shown that the partial safety factor estimated from the measurements significantly differs from the recommended value of 1.5. The discharge in 2002 corresponds to an exceptionally long return period. It is concluded that statistical methods provide a valuable background for evaluation and prediction of discharges. However, the presented analysis should be further improved to include non-statistical aspects that influence discharges such as the effects of water management and deforestation. Keywords: discharge, probabilistic assessment, extremes, failure, statistical methods.

Keywords

discharge, probabilistic assessment, extremes, failure, statistical methods.