2D and 3D Computational Modeling of Surface Flooding in Urbanized Floodplains: Modeling Performance for Various Building Layouts

Abstract

Understanding the strengths and limitations of the modeling capacity of surface flooding in urbanized floodplains is of utmost importance as such events are becoming increasingly frequent and extreme. In this study, we assess two computational models against laboratory observations of surface urban flooding in a reduced‐scale physical model of idealized urban districts. Four urban layouts were considered, involving each three inlets and three outlets as well as a combination of three‐ and four‐branch crossroads together with open spaces. The first model (2D) solves the shallow‐water equations while the second one (3D) solves the Reynolds‐averaged Navier‐Stokes equations. Both models accurately predict the flow depths in the inlet branches. For the discharge partition between the outlets, deviations between the computations and laboratory observations remain close to the experimental uncertainties (maximum 2.5 percent‐points). The velocity fields computed in 3D generally match the measured surface velocity fields. In urban layouts involving mostly a network of streets, the depth‐averaged velocity fields computed by the 2D model agree remarkably well with those of the 3D model, with differences not exceeding 10%, despite the presence of helicoidal flow (revealed by the 3D computations). In configurations with large open areas, the 3D model captures generally well the trajectory and velocity distribution of main surface flow jet and recirculations; but the 2D model does not perform as well as it does in relatively channelized flow regions. Visual inspection of the jet trajectories computed by the 2D model in large open areas reveals that they substantially deviate from the observations.