Abstract
Natural hazards are a major threat to humankind and likely to increase in frequency and intensity due to climate change. It is therefore urgent that cities become more resilient to urban flooding by taking flood risks into account throughout the entire process from the early design phase of districts to maintaining the existing city.
In order to enable the urban planners to take flooding into consideration, suitable simulation tools need to be available. Software dealing with this topic are usually based on simplifications of the Navier-Stokes equations, leading to 1d or 2d approximations like the 1d Saint-Venant equation or the 1d/2d Shallow water equation (SWE). While these approaches suffice in predicting where the water will go, which areas will be under water as well as the 2d dynamics, they are not able to capture sudden transient and 3d effects.
In this work, the Smoothed Particle Hydrodynamics (SPH) model is applied to aid. Since the SPH model is computationally demanding, a rather small area can be modelled. Hence, a combination of the different approaches is anticipated in order to estimate the large-scale risks whereas the SPH model can focus on the dynamic process in a smaller and critical area. We present first steps towards such a modelling framework allowing the prediction in flooding scenarios in the built environment. An SPH model developed within and coupled to a Discrete Element Method (DEM) model and parallelized on the GPU to enhance efficient computation is presented. The model is validated using a generic city flooding experiment [1]. It is further applied to a real-world case and compared to results using an SWE approach. An important aspect in our work is the availability of input data for flood simulation and we discuss the importance of uncertainties for assessing simulation results.