ABSTRACTS 'Extreme Discharges' - CHR-KHR

The flood routing in the Rhine from Maxau till Lobith is done with the 1-d hydrodynamic model SOBEK on a
hourly basis (refs). A description of this SOBEK model is given by Lammersen (2002,2001). The boundary conditions
are provided by the discharge calculated by HBV for the tributaries and the part upstream of Maxau.
4. CASE STUDY AND RESULTS
The discharge of the river Rhine at Lobith with an average return period of 1250 year has been examined using
both the observed and generated annual maximum discharges. The maximum likelihood and Bayesian methods
have been used.
The maximum likelihood estimation method, applied to observed discharges at Lobith during the period 1901-
2002, resulted in seven probability distributions which could not be rejected. On the other hand, the analysis
based on generated annual maximum series lead up to four distributions which could not be rejected.
The Bayesian method applied to observed annual maximum discharges resulted in a reasonable fit. Nevertheless,
The Bayes weights computed are meagre. As a matter of fact, the highest Bayes weight is about 23 % at
most corresponding to the Weibull distribution. Moreover, applying this estimation method to generated annual
maximum discharges for a period of 1000 years, more decisive Bayes weights are calculated. In that case the
Gamma distribution fits excellent with a Bayes weight of 80 %. This suggests that the observed and generated
series are not identically distributed.
5. DISCUSSION
The method consisting of a statistical rainfall generator combined with hydrological and hydraulic models is
promising. The fact is that the method generates automatically more homogenous annual discharges. Also, we
notice that the method provides opportunities to study the effects of different measures against flooding. On the
other hand, the basic assumption of the method is rather empirical. Indeed, the method is based on observed
and homogeneous rainfall data during a period of about 40 years.
Finally, With regard to the observed and generated annual maximum discharges, the remarkable things are that
the calculated design discharges and Bayes weights are incoherent. All in all, at the moment it is not known
whether the combined method consisting of a rainfall generator and hydrological and hydraulic calculations,
generates annual maximum discharges reliably. Therefore, a decisive validation of the method is inevitable. It
should be mentioned, though, that also series of generated maximum discharges are available for which flooding
takes place in Germany (upstream from Lobith). These floods occur at extreme discharges larger then the observed
ones, thereby yielding a lower design discharge. A further investigation of the problem whether the observed
and the generated maximum discharges are identically distributed should also incorporate series with
floods taking place in Germany.
6. ACKOUWLEDGEMENTS
The daily precipitation, temperature and sunshine duration data were made available by the following institutions:
German Weather Service (DWD), Service de la météorolgie et de l’hydrologie de Luxemburg, Météo
France and the Swiss Meteorological Institute through the international Commission for the Hydrology of the
Rhine Basin (CHR/KHR). The generated time series with date from the rainfall generator were provided by the
Royal Dutch Meteorological Office (KNMI) and the HBV model for the Rhine was supplied by the Federal Institute
of Hydrology (BfG).
33