Fast arrival-time modelling of dike breach floods for effective disaster management

Abstract

Modelling of flood arrival times after a dike breach is essential for efficient flood disaster management, shaping evacuation strategies and effective disaster response. However, common hydrodynamic models remain too computationally expensive for real-time flood forecasting, while current surrogate models have so far not been able to simulate flood arrival times for dike breaches. In this study, we develop a fast conceptual model for this purpose, based on a digital elevation model and a linear regression equation for flood propagation velocity. The linear regression was derived from a sensitivity analysis of hydrodynamic simulations of idealized dike breaches, where we identify the most important hinterland characteristics that determine the flood propagation velocity. The sensitivity analysis shows that hinterland slope in the propagation direction is most important for the propagation velocity, while breach discharge is most important close to the breach. The fast conceptual model computes flood arrival times by applying this linear regression along the local drainage direction (steepest slope) path leading away from the breach, taking peak breach discharge as input. The model achieves accurate results for two case study dike breaches along the Rhine river near the Dutch–German border, especially in the first 48 h after the breach. The mean absolute arrival time error is about 2 to 4 h, and computation time is less than a second. We conclude that this model can serve as a fast disaster preparation and response tool to support flood disaster management, for determining evacuation strategies and uncertainty analysis of possible breach discharge scenarios.