PDS-MODELING AND REGIONAL BAYESIAN-ESTIMATION OF EXTREME RAINFALLS

Since 1979 a country-wide system of raingauges has been operated in De nmark in order to obtain a better basis for design and analysis of urb an drainage systems. As an alternative to the traditional non-parametr ic approach the Partial Duration Series method is employed in the mode lling of extreme rainfalls. The method is applied to two variables: th e total precipitation depth and the maximum 10-minute rain intensity o f individual storms. On the basis of the atsite modelling a regional a nalysis is carried out. It is shown that the previous assumption of sp atial homogeneity of extreme rainfalls in Denmark cannot be justified. In order to obtain an estimation procedure at non-monitored sites and to improve at-site estimates a regional Bayesian approach is adopted. The empirical regional distributions of the parameters in the Partial Duration Series model are used as prior information. The application of the Bayesian approach is derived in case of both exponential and ge neralized Pareto distributed exceedances. Finally, the aspect of inclu ding economic perspectives in the estimation of the design events is b riefly discussed.