An evaluation of the use of regional climate model data applied to extreme precipitation in the Meuse basin

In July 2021, unprecedented extreme precipitation led to enormous summer discharges that were never observed before, causing disastrous flood events in parts of the Meuse basin and stressing the importance of adequate extreme estimates for flood resilience. Conventional techniques for extreme discharge estimation, such as statistical extrapolation or the GRADE Weather Generator, have clear limitations which are attributed to the restricted length of the observational input data. This study introduces an alternative approach, in which long synthetic meteorological data spanning more than 1000 years are employed in a hydrological model, generating a dataset from which large hydrological extremes can directly be derived. The main aim of this research is to evaluate the meteorological RACMO dataset for this specific purpose and to provide insight in the advantages of the synthetic RACMO dataset compared to observations, with focus on meteorology.

The climatology of RACMO precipitation means and extremes in the Meuse basin and its tributaries compare very well with observations, making the dataset useful for hydrological computations. Furthermore, the extensive length of the RACMO dataset reveals a range of extreme values that was previously unanalysed. This provides new insight into the tail of the distribution of annual precipitation extremes, particularly in the curvature of GEV distributions which is described by the shape parameter. This shape parameter has a large influence on the value of the high extremes estimated by a GEV distribution and can better be estimated with the use of the longer RACMO dataset.

The statistical uncertainty in the estimation of extreme precipitation is strongly reduced by the use of RACMO data: roughly a factor 4 for daily extremes and a factor 10 for hourly extremes. Further statistical analysis shows that the GEV shape parameter from observed precipitation is more robust with a long dataset. This is reflected by the spatial inconsistency of the shape parameter of observations and the more spatially consistent RACMO shape parameter. Furthermore, the shape parameter increases substantially by averaging over time and slightly when averaging in space.

There is a clear distinction between the GEV distributions of summer and winter precipitation extremes, suggesting the existence of a double population. Depending on the time step (and therefore the annual dominance of summer or winter events), such a double population may influence the GEV of annual maxima. The existence of a double population, particularly in summer precipitation, is well-known in literature and is said to be related to dewpoint temperature. The existence of a double population is difficult to obtain from observations, but can have an enormous impact on the return values of summer extremes, such as the event of July 2021. The translation of summer extremes of precipitation into extreme discharges depends on the rainfall-runoff response of the considered catchment and on its catchment characteristics.