From 2018 to 2020, INRAE carried out a study on the characterisation of extreme flows on the Rhine on behalf of DREAL Grand-Est, with discussions within the International Permanent Commission for the Development of the Upper Rhine in conjunction with the ICPR – International Commission for the Protection of the Rhine (in German, IKSR – Internationale Kommission zum Schutz des Rheins). This is the result of major differences of opinion on the value of reference flows associated with return periods. The millennial flood (resp. decamillennial flood) in the study by the Technical University of Karlsruhe (IWG) validated by the Standing Committee in May 2010 corresponds to a centennial flood (resp. millennial flood) in certain French studies.

An opinion by Michel Lang on the estimation of reference floods for the downstream section of the Franco-German Rhine was formulated, based on the processing of two long hydrometric series at Basel (1808-2017) in Switzerland and at Maxau (1815-2018) in Germany. The differences stem from methodological choices. In the German study carried out in 2010 by the University of Karlsruhe, flow data from the twentieth century was used, as information from earlier in the nineteenth century was considered too uncertain, given the developments that have taken place on the Rhine since then. In the study carried out by EdF in 2002 on the Fessenheim and Kembs diversion bays, further upstream, the major floods on the Rhine during the 19th century were retained by using the Swiss hydrological yearbooks from 1808 onwards. On the other hand, the probability laws used by the Germans (Log Normal law) and the French (Gumbel root law) have different asymptotic behaviours which have an impact on the extrapolation of the distribution of floods for long return periods.

The approach proposed during the appraisal consisted of using a probabilistic model that takes into account the uncertainties surrounding flood flows. This made it possible, on the one hand, to penalise old data, which have a greater uncertainty than recent data, and, on the other hand, to include in the analysis major events from the 19th century that have not been observed since. Uncertainties in flood flows are of the order of ±5-7% for the recent period, rising to ±45-48% for the oldest flows. Analysis of the flood distribution led to the conclusion that the current level of safety of the dikes downstream of Strasbourg is consistent with legislation, with overtopping only possible beyond a flood with a return period of 10,000 years. The results of the EDF study (2002) at Basel appear to be overestimated, with flood quantiles Q100 and Q1000 significantly higher than the empirical distribution 1808-2017, and a Q1000 value at Basel higher than that estimated at Maxau. This overestimate is partly due to the decision not to take into account the reduction in flood levels after the Jura correction works in Switzerland, from 1890 onwards, and not to correct downwards the floods prior to 1890 in order to use an identical reference condition.