Abstract
The main two approaches currently used for urban flood modelling are the one-dimensional and the two-dimensional approach. Despite a more realistic description of the physical processes, the use of the two-dimensional approach suffers from large computational cost because of the strong variability in the urban geometry that requires fine meshes and data. The present PhD is devoted to alternate approaches to the classical two-dimensional approach.<br />In a first part, a conceptual discharge distribution relation is developed for one-dimensional models of street networks. The relationship is implemented into a one-dimensional kinematic routing model to simulate the flooding of the Richelieu district in Nîmes. An empirical distribution relation has been developed for four branch crossroads using two-dimensional flow simulations for a wide range of crossroad geometries.<br />The second part is dedicated to the development of a macroscopic, two-dimensional flow model. The urban zone is characterized at a large scale via average properties such as the porosity. Extended urban districts can then be represented using coarse meshes, allowing for short computational times. Besides porosity, a local head loss tensor must be introduced into the shallow water equations to account for the influence of urban singularities. An empirical formulation for this tensor has been developed using two-dimensional simulations over a synthetic street network. The macroscopic model has been successfully validated against scale model measurements.<br />Furthermore, an approximate-state Riemann solver has been developed for the two-dimensional shallow water equations. The Riemann invariants are expressed as functions of the flux components and account for the influence of the source terms. This allows the fluxes at the interfaces between the computational cells to be calculated directly. A comparison with classical solvers such as the HLL/HLLC solver shows a substantial improvement in the numerical solution.