COMPSTAT 2016: Start Registration
View Submission - COMPSTAT
A0477
Title: Bivariate regional frequency analysis of extreme precipitation events in the Czech Republic Authors:  Tereza Simkova - Technical University of Liberec (Czech Republic) [presenting]
Jan Picek - Technical University of Liberec (Czech Republic)
Jan Kysely - Institute of Atmospheric Physics AS CR (Czech Republic)
Abstract: Bivariate regional frequency analysis (RFA) has been applied to estimate quantile curves for simultaneous non-exceedance probabilities $p=0.9, 0.95, 0.99, 0.995, 0.999$ and joint AND/OR return values corresponding to 10-, 20-, 50- and 100-year return periods of extreme precipitation events in the Czech Republic. Maximum annual 1- and 5-day precipitation totals measured from 1961 to 2007 at 210 stations were used as the input dataset. When determining the bivariate distribution function of the 1- and 5-day maximum annual precipitation amounts, standard approaches were followed involving the goodness-of-fit test based on the Cram\'{e}r-von Mises statistic, Akaike Information Criterion and estimation of the upper-tail dependence coefficient and considering a wide copula test space. The H{\"u}sler-Reiss copula family has been identified as the most suitable for modelling the link between variables. Only in the northernmost part of the area the Gumbel and Joe copula families have been preferred. It has been found out the homogeneity condition (i.e. that sites in a region have the probability distribution identical apart from a site-specific scale factor) is satisfied for every of the six proposed regions and considered bivariate distribution functions, and therefore the index-flood based bivariate RFA was preferred to use. The resulting models, bivariate quantiles and joint return values provide valuable information and can be useful for hydraulic engineering design and planning.