Hydrological records of the Rhône River at Beaucaire from 1816 to 2020: accounting for uncertainty in the discharge estimation and flood frequency analysis

As floods and droughts strongly influenced the navigability of the river and the commercial activity of the region, the Rhône at Beaucaire has always been monitored by the locals. Consequently, one of the French first permanent hydrological gauging stations was installed at Beaucaire at the beginning of the 19^th century. The Lower Rhône valley has remained a vulnerable area in terms of flood risk. A recent archival work (Pichard et al., 2017) allowed to build a continuous daily stage time series from 1816 to 2020. The goal of this work is to establish the corresponding series of discharge, considering the main sources of uncertainty, and to improve the statistical knowledge of the flood occurrence and variability in the Lower Rhône valley. The conversion from stage series to discharge series is not straightforward, especially for the ancient times for many reasons such as ungauged periods, rating shifts and morphogenic floods, metrological or climatic changes. First, the periods with stable stage/discharge relationships were determined by applying the segmentation procedure of Darienzo et al. (2021) to the available gaugings since 1845. This recursive segmentation procedure accounts for both gaugings and rating curve uncertainties, through a Bayesian framework. Then, the uncertain rating curves of each period have been estimated using the BaRatin SPD model (Mansanarez et al., 2019), allowing to include prior hydraulic knowledge and to transfer information across periods. After checking the homogeneity of this exceptionally long discharge series, a flood frequency analysis has been conducted in a Bayesian framework. Some assumptions have been tested to improve the estimation of the parameters of the GeV distribution, such as a regional estimation of the scale parameter, or the introduction of climatic covariates for a time-varying analysis. A key objective of this work was to account for uncertainties at all steps steps of the approach, from the stage measurements and the gaugings to the flood quantiles estimation.