The impact of dune evanescence on the Po River conveyance during flood events

Flood modelling plays a pivotal role in flood forecasting, defence structures design, and risk assessment. Therefore, reliable model descriptions are essential to provide accurate results. Traditionally, flood models consider fixed-bed conditions and assume that resistance coefficients remain constant throughout the event, regardless of hydrodynamic conditions. Despite its success under relatively stable conditions, this approach can fail when the riverbed undergoes rapid morphodynamic changes. Varying hydrodynamic regime, especially under severe flood conditions, influences bedform development, thus appreciably altering the flow resistance, which in turn affects water levels and flow velocity.

For example, in the terminal reach of the Po River, the largest river in Italy, fixed-bed models with constant resistance coefficients have been observed to overestimate the water level by up to 2 meters during flood events. Sand dunes, clearly visible in the bathymetric survey, are widely distributed in the main river channel, but their height and length, which are key factors in determining the total flow resistance, can vary considerably depending on the hydrodynamic regime. Moreover, recent studies have shown that transitioning from sand dune to upper-stage plane bed is possible even at Froude numbers lower than previously thought (i.e., F<0.8). Thus, among the potential causes of the water level overestimation, we identified the changes in flow resistance induced by bedform evolution as one of the most probable.

To test this hypothesis, we developed a 2D finite element hydrodynamic model of the terminal 200-km long reach of the Po River. To quantify flow resistance in the momentum equations the model employs the Gauckler-Strickler formulation and the spatial distribution of the associated resistance coefficient was calibrated using water level and discharge data collected at multiple locations and during multiple flood events. In general, the results of the fix-bed, constant-resistance model were good up to a certain discharge, above which the relevant, systematic overestimation of water levels was confirmed.

We enhanced the 2D hydrodynamic model by dynamically updating the resistance coefficient within the main river channel as a function of bedform evolution. The dune height and length were computed dynamically following the Van Rijn model as a function of the current local flow conditions, introducing a single additional calibration coefficient that scales the height of the dunes. By using this dynamical roughness predictor, combined with an additional resistance component to account for dissipative mechanisms not explicitly addressed by the model, a strict match was achieved between measured and modelled water levels at five different gauging stations along the terminal reach of the Po River under low, intermediate, and severe flood conditions. Despite the challenges in accurately capturing these processes, the results demonstrate that accounting for the dynamic contribution of bedforms can significantly improve the reliability of flood predictions, offering a more robust tool for managing flood risks in complex river systems.