Climate change is expected to increase the frequency and magnitude of river floods ¹. Floods not only cause damage by inundation and loss of life ^2,3 but also jeopardize infrastructure because of bank failure and riverbed erosion processes that are poorly understood. Common flood safety programmes include dyke reinforcement and river widening ^{4, 5, 6, 7, 8–9}. The 2021 flood in the Meuse Basin caused 43 fatalities and billions of dollars of damage to infrastructure ¹⁰. Here, on the basis of analysis of the Meuse flood, we show how uneven widening of the river and heterogeneity of sediment deposits under the river can cause massive erosion. A recent flood safety programme widened the river ¹¹, but created bottlenecks where widening was either prevented by infrastructure or not yet implemented. Riverbed erosion was exacerbated by tectonic uplift that had produced a thin top gravel layer above fine-grained sediment. Greatly enhanced flow velocities produced underwater dunes with troughs that broke through the gravel armour in the bottlenecks, exposing easily erodible sands, resulting in extreme scour holes, one more than 15 m deep. Our investigation highlights the challenges of re-engineering rivers in the face of climate change, increased flood risks and competition for river widening space, and calls for a better understanding of the subsurface.

Sustainable river management often requires long-term morphological simulations. As the future is unknown, uncertainty needs to be accounted for, which may require probabilistic simulations covering a large parameter domain. Even for one-dimensional models, simulation times can be long. One of the acceleration strategies is simplification of models by neglecting terms in the governing hydrodynamic equations. Examples are the quasi-steady model and the diffusive wave model, both widely used by scientists and practitioners. Here, we establish under which conditions these simplified models are accurate. Based on results of linear stability analyses of the St. Venant-Exner equations, we assess migration celerities and damping of infinitesimal, but long riverbed perturbations. We did this for the full dynamic model, that is, no terms neglected, as well as for the simplified models. The accuracy of the simplified models was obtained from comparison between the characteristics of the riverbed perturbations for simplified models and the full dynamic model. We executed a spatial-mode and a temporal-mode linear analysis and compared the results with numerical modeling results for the full dynamic and simplified models, for very small and large bed waves. The numerical results match best with the temporal-mode linear analysis. We show that the quasi-steady model is highly accurate for Froude numbers up to 0.7, probably even for long river reaches with large flood wave damping. Although the diffusive wave model accurately predicts flood wave migration and damping, key morphological metrics deviate more than 5% (10%) from the full dynamic model when Froude numbers exceed 0.2 (0.3).

Sustainable river management often requires long-term morphological simulations. As the future is unknown, uncertainty needs to be accounted for, which may require probabilistic simulations covering a large parameter domain. Even for one-dimensional models, simulation times can be long. One of the acceleration strategies is simplification of models by neglecting terms in the governing hydrodynamic equations. Examples are the quasi-steady model and the diffusive wave model, both widely used by scientists and practitioners. We established under which conditions these simplified and often more efficient models are accurate.

Sustainable river management can be supported by models predicting long-term morphological developments. Even for one-dimensional morphological models, run times can be up to several days for simulations over multiple decades. Alternatively, analytical tools yield metrics that allow estimation of migration celerity and damping of bed waves, which have potential for being used as rapid assessment tools to explore future morphological developments. We evaluate the use of analytical relations based on linear stability analyses of the St. Venant-Exner equations, which apply to bed waves with spatial scales much larger than the water depth. With a one-dimensional numerical morphological model, we assess the validity range of the analytical approach. The comparison shows that the propagation of small bed perturbations is well-described by the analytical approach. For Froude numbers over 0.3, diffusion becomes important and bed perturbation celerities reduce in time. A spatial-mode linear stability analysis predicts an upper limit for the bed perturbation celerity. For longer and higher bed perturbations, the dimensions relative to the water depth and the backwater curve length determine whether the analytical approach yields realistic results. For higher bed wave amplitudes, non-linearity becomes important. For Froude numbers ≤ 0.3, the celerity of bed waves is increasingly underestimated by the analytical approach. The degree of underestimation is proportional to the ratio of bed wave amplitude to water depth and the Froude number. For Froude numbers exceeding 0.3, the net impact on the celerity depends on the balance between the decrease due to damping and the increase due to non-linear interaction.

The Waal river in the Netherlands is experiencing a bed degradation of 2-3 cm per year. This is most probably the result of river training measures in the past, as well as the construction of dams in the upstream part. Bed degradation causes hindrance for shipping and may lead to groundwater issues in the floodplains. River widening (e.g. by large scale floodplain lowering) affects the slope of the river bed and hence reduces bed degradation. In this study, we analyze the morphological changes as a result of river widening, using a 1D morphological model (with mixed size sediment) with which we performed long-term (100 years) simulations for a long river stretch (order of 100 km) . First results show that the model is able to reflect the changes in slope correctly, but that only river widening itself is not sufficient to stop the degradation. Additional simulations with upstream sediment nourishments show that still substantial volumes are needed to stabilize the bed. These preliminary results may already be used for policy preparations but fine-tuning with a 2D model is necessary, before actual management decisions can be taken.