Comparison of three different numerical implementations to model rainfall-runoff transformation on roofs

Abstract

[Abstract:] Roofs represent a high percentage of the impervious surfaces in urban areas, and hence their implementation in urban drainage models is essential for accurate results to be achieved. Current modelling approaches are based on parameters such as a roof's slope and width, its roughness coefficient, and the initial abstraction. In this study, an experimental campaign was performed in order to assess the sensitivity of the roof runoff hydrographs to these parameters. The experimental tests were carried out in a new large-scale urban drainage facility equipped with a rainfall simulator. The experimental tests were replicated numerically using three different levels of model resolution, from a high fidelity representation with a spatial resolution of 5 mm (which can be considered a digital twin) to a lumped representation. Our experimental results show that, for practical purposes, the sensitivity of the outlet runoff hydrograph to the roof slope tested is negligible. The numerical upscaling analysis carried out showed that flat roofs present a slightly different hydrograph behaviour with greater times of concentration than sloped roofs. No significant sensitivity of the outlet hydrograph to the surface roughness coefficient was found. In terms of numerical modelling, the use of a very high spatial resolution for the roof, which implies a high computational cost, does not affect the results significantly compared to the far simpler lumped approach. The current research involves the first thorough experimental and computational analysis of the runoff over roofs to date.