Author(s): Masoumeh Ebrahimi; Robin Meurice; Sandra Soares-Frazao
Keywords: Breaching; Dam-failure; Embankment dam; Finite-volumes; Overtopping
Abstract: Embankment dams are very common worldwide for they often constitute cheaper alternatives to concrete dams. Yet, the stability of these dams can be jeopardised through different failure mechanisms. Moreover, the point of no return can be reached very rapidly and lead to progressive or sudden failure of the dam depending on the failure mechanism. Not only will the downstream reach then be subject to catastrophic flooding but the collapsed soil could worsen the downstream impacts. Despite the serious consequences that such failures can trigger, the processes behind them remain poorly understood and difficult to model comprehensively. In the context of the 6th Workshop on River and Sedimentation Hydrodynamics and Morphodynamics, an almost blind benchmark test was proposed to compare the performances of different numerical models able to simulate progressive dam failures. This test consisted of the failure by overtopping of an earthen dam with a breach in the middle of its crest and through which the erosion of the dam would progressively develop, vertically and laterally, as the discharge at the inlet of the reservoir was increasing. Despite the expected non-cohesive character of the dam material, a headcut formation followed by interesting circular erosion patterns were observed, both in the horizontal and vertical planes. Although the repose angles of the dam material were said to lie between 30° and 35°, the banks of the resulting breach showed very straight and even obtuse angles due to an unexpectedly high cohesive behaviour. Two depth-averaged finite-volume models were applied to this benchmark test. While the first model (WCM) decoupled the shallow-water equations from Exner’s one, leaving only weak links between the different operators, the second one (CM) solved them altogether. The WCM was able to reproduce most of the reservoir’s water level evolution and is hence supposed to rather accurately predict the evolution of the breach width. Conversely, the CM led to a faster erosion of the dam than expected, which led to a rapid decrease of the reservoir’s water level. Nevertheless, the CM showed fewer oscillations on the downstream slope of the dam once it was eroded and a lesser braided behaviour. Between the braids obtained with the WCM, straight remnants of the dam’s downstream slope appeared as unphysical. A bank-failure operator was hence used to wash these soil heaps out, but this tended to increase the width of the breach, which impacted negatively the correspondence between the experimental and numerical reservoir’s water level. Proper calibration of this bank-failure operator to cope with the apparent cohesion and to avoid unphysical soil heaps should be performed during the calibration stage that will follow the current blind stage. This second step should deliver more in-depth comparisons between the models’ results and the experimental data.