Regional modelling of extreme hydrological events

  • Rosbjerg, Dan (Project Manager)
  • Madsen, Henrik (Project Participant)

    Project Details


    In the research project the so-called partial duration series (PDS) method has
    been further developed for use in both at-site and regional extreme value modelling of floods and rain storms etc. The method uses all measurements above a certain threshold contrary to the annual maximum series (AMS) method which only takes the annual maximum values into account. A complete at-site analysis using a flexible type of distribution for the threshold exceedances has been carried out, and recommendations for choice of an optimal combination of exceedance distribution and estimation method for the model parameters have been presented. The regional index-flood model, originally developed for AMS, has been extended to apply also for PDS data, and it has been shown that also when regional data sets are available, PDS should be preferred to AMS in most cases. Furthermore, it has been shown that delineation of homogeneous regions can be done in a much more unambiguous fashion if based on PDS data. Finally, the regional PDS index-flood method has been combined with empirical Bayes procedures. It has been shown that, when generalized least squares (GLS) regression is used for assessment of prior information about the PDS parameters, regional heterogeneity and intersite dependence between the stations in the region can adequately be accounted for. All together the theoretical framework as well as practical use of the PDS method has been substantially advanced.
    Effective start/end date01/02/199331/12/1998