Author(s): Vincent Wolfs; Victor Ntegeka; Maria Bermudez; Patrick Willems
Keywords: Computational efficiency; Computational hydraulics; Neural networks; Surrogate modelling; Urban drainage model; Urban flooding
Abstract: Detailed 1D/2D models have become standard practice for urban flood modelling. However, many applications require computationally fast simulation models. Due to their prolonged calculation times, these1D/2D models are unsuited for such applications. This research compares three modelling approaches with different levels of complexity and simulation times: (1) a highly detailed 1D/2D model, (2) a 1D/1D model with two different flood cone parameterizations, and (3) a newly developed surrogate dual drainage model. The three approaches are tested and compared on a Belgian case study. Results show that the surrogate dual drainage model can emulate the results of highly detailed models with calculation time reductions in the order of magnitude of 105.