In our era, the rapid increase of parallel programming coupled with high‐performance computing (HPC) facilities allows for the use of two‐dimensional shallow water equation (2D‐SWE) algorithms for simulating floods at the “hydrological” catchment scale, rather than just at the “hydraulic” fluvial scale. This approach paves the way for the development of new operational systems focused on impact‐based flash‐floods nowcasting, wherein hydrodynamic simulations directly model the spatial and temporal variability of measured or predicted rainfall on impacts even at a street scale. Specifically, the main goal of this research is to make a step to move toward the implementation of an effective flash flood nowcasting system in which timely and accurate impact warnings are provided by including weather radar products in the HPC 2D‐SWEs modelling framework able to integrate watershed hydrology, flow hydrodynamics, and river urban flooding in just one model. The timing, location, and intensity of the street‐level evolution of some key elements at risk (people, vehicles, and infrastructures) are also discussed considering both calibration issues and the role played by the spatial and temporal rainfall resolution. All these issues are analyzed and discussed having as a starting point the flood event which hit the Mandra town (Athens, Greece) on the 15 November 2017, highlighting the feasibility and the accuracy of the overall approach and providing new insights for the research in this field.

In this paper attention is first focused on a comparative analysis of three hydraulic models for overland flow simulations. In particular, the overland flow was considered as a 2D unsteady flow and was mathematically described using three approaches (fully dynamic, diffusive and kinematic waves). Numerical results highlighted that the differences among the simulations were not very important when the simulations referred to commonly used ideal tests found in the literature in which the topography is reduced to plane surface. Significant differences were observed in more complicated tests for which only the fully dynamic model was able to provide a good prediction of the observed discharges and water depths. Then, attention is focused on the fully dynamic model and in particular on the analysis of two numerical schemes (TVD-MacCormack and HLL) and the influence of the grid size. Numerical tests carried out on irregular topography show that, as the grid size decreases, the performance of the HLL scheme becomes closer to that of the TVD-MacCormack scheme in shorter computational times at least for high rainfall intensity.