Résumé
In recent decades, African cities have been confronted with a series of floods linked to the rapidand poorly managed urbanisation, the intensification of heavy rains and the failure of therainwater drainage system. Faced with this situation, protection against the risk of flooding is amajor development issue. Increasing interest is being shown in its characterisation andforecasting. This study, carried out in the urban periphery of Dakar, aims to propose a modellingof flows and overflows of structures at fine resolution (5m) for rainfall intensities of differentreturn periods. Two methodological steps are combined to achieve this objective: 1) theconstruction of the urban drainage topology taking into account the modifications induced bythe building, the artificial channels and the retention basins which lead to modifications of thedirections of the natural drainage; 2) the simulations of the flows in real time or in project mode,by a parsimonious spatial model adapted to the local context coupling a hydrological model(SCS-LR) at the scale of small basins, with a hydraulic model (Kinematic Wave) for thepropagation in the network of the flows produced at the outlet of the basins. The former ensuresthe speed of the calculation, and the latter provides precise information on the behaviour of thenetwork in the event of flooding. The overflow points of the network are identified by thedifference between the maximum flow and the capacity of the network to evacuate floods. Theresults of the simulations show that the drainage network evacuates volumes of runoff producedfor rainfall with a low return period (2-10 years) but appears to be overflowing for rainfall witha rare frequency (20-100 years) with overflow rates sometimes exceeding 24 m3/s. This modelprovides sufficiently informative simulations to assess the effectiveness of the schemes inproject mode, in a context of limited data. The model also provides boundary conditions forapplying more complex hydraulic models to determine the impact of overflows on limited areas.Finally, the calculation times are short enough to be compatible with real-time monitoring orforecasting. The method seems to be flexible and adaptable to different contexts, and facilitatedby the development of global open-source platforms, and the possibility of accessing the spatialdata of cities.