Evaluation of discharge extremes in the Meuse river and her tributaries
In July 2021, extreme discharges caused extensive flooding in many tributaries of the Meuse, driven by an intense precipitation event. At these locations, this event was the highest observed on records and much higher than previous extremes observed. This event also happened in summer, which makes the event even more rare. Frequency analysis based on observed time series in the basin have a high degree of uncertainty to estimate the return period of such an event, because the observed time series are only a few decades long. To estimate large return periods of extreme discharge, the GRADE method was developed in the Netherlands. This method statistically extrapolates observed weather time series using a statistical weather generator that temporally resamples the time series and generate much longer weather time series. The resulting weather time series are used with a hydrological models to generate very long synthetic time series of discharge events. This can extend record lengths but cannot generate more extreme daily rainfall events than the one observed and outliers have a large impact on the resampling scheme, becoming too frequently resampled. Finally, it cannot generate time series at subdaily time scales because the method is based on daily weather time series, while the peak discharge event may be subject to shorter time scale processes. Even though some of the statistical uncertainty is reduced by this method by increasing the length of the record, it does not generate the physics of extreme events not observed previously, such as extreme summer events.
This report investigates improvement on the GRADE approach. Whereas GRADE uses daily observed weather variables, here we base the generation of synthetic discharges on a physically-based climate model, which results in meteorological time series at full spatial and temporal resolution for the Meuse basin. Furthermore, we capture the physical processes leading to extreme discharges through a physically-based hydrological model. In this way, the time series of meteorological variables representing the current climate are extended to a much longer record length of 1,040 years, generating physically plausible weather systems that can lead to extreme discharge but have not been captured by observations.