The development of a fast-computing emulator to predict dike failure during severe flood events

Flooding is one of the most damaging and frequent natural hazards in the world and it is expected that flood events will affect even more people in the future due to climate change, land use change and population growth. The use of proper evacuation schemes has the potential to reduce the consequences of a flood event, i.e. reducing the damage and the number of life-losses and cattle-losses, in areas at risk significantly. Typically, sophisticated two-dimensional (2D) hydraulic models are used to simulate flood propagation of severe flood events to inform flood management decisions. Although these models have proven to be accurate in predicting flood wave propagation and inundation extents in areas with complex dynamic interactions, these models cannot be used for short-term flood forecasting due to the high computational demands and long simulation times. For this reason, emulators (e.g. artificial neural networks (ANNs)) have gained much attention in recent years. In this study, we set up neural networks that are able to predict the dike failure locations during flood events and corresponding outflow hydrographs, based on discharge measurements at an upstream gauge station. These outflow hydrographs can be used to predict the potential inundated areas, enabling evacuation of the people in the areas at risk on time.