Improving historic flood reconstruction using a detailed 1D-2D coupled hydraulic model approach

The 1374 flood event is considered the largest flood of the last millennium in the Rhine River. Herget and Meurs (2010) used a simple one-dimensional approach to reconstruct peak discharges of historic Rhine River flood events at Cologne, Germany, including the 1374 flood event. They found a 1374 peak discharge of around 23,200 m3 /s with an uncertainty range of between 18,800 m3 /s and 29,000 m3 /s. Recently, using one-dimensional−two-dimensional (1D-2D) coupled hydraulic models to solve complex hydraulic has become popular thanks to its advantages regarding the accuracy of model results and computational efficiency (Dasallas et al., 2020, Leandro et al., 2016). This study sets up a 1D-2D coupled hydraulic model in HEC-RAS for a study area stretching from near Andernach to Haus Burgel (Germany) to reconstruct the discharge magnitude of the 1374 flood event. A high-resolution palaeoDEM for the Lower Rhine catchment (van der Meulen et al., 2020) and hydraulic roughness for
landscape classes corresponding to the palaeo situation were used as the input data for this model. We performed an uncertainty analysis with different river bed levels and roughness values to estimate the influence of these uncertainties on the reconstructed peak discharge. The upstream discharge wave was also varied, with discharge peak values ranging from 12,000-24,000 m3 /s. The simulated flood water levels were then compared with the 1374 flood marks at Cologne and Haus Burgel. Based on this comparison, the discharge magnitude was determined to be between 12,800-21,400 m3 /s, with a best estimate between 14,000-18,300 m3 /s. These best estimate values were used in a flood frequency analysis to determine the design discharges corresponding to different return periods (Bomers et al., 2019). A significant reduction of 2000 m3 /s in the design discharge was found corresponding to a 100,000 year return period compared to the previous 1374 peak discharge estimations (Herget and Meurs, 2010).