Spatial design strategies and performance of porous pavements for reducing combined sewer overflows

• Published in: Journal of hydrology (Amsterdam) Vol. 607; p. 127465
• Main Authors: Torres, María Nariné, Rabideau, Alan, Ghodsi, Seyed Hamed, Zhu, Zhenduo, Shawn Matott, L.
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.04.2022
• Subjects: cost effectiveness; drainage; GI clustering; GI spatial design; Green infrastructure; pavements; Porous pavements; Simulation-optimization; stormwater; Stormwater modeling; GI clustering; Simulation-optimization; Green infrastructure; GI spatial design; Porous pavements; Stormwater modeling
• ISSN: 0022-1694; 1879-2707
• DOI: 10.1016/j.jhydrol.2022.127465

•Porous Pavements (PP) can achieve up to 75% volume reduction in a CSO outfall.•PP performance is highly sensitive to the location of the features within a CSOshed.•Good-performing PP systems were identified using k-means clustering and optimization. The installation of Green Infrastructure (GI) is a popular strategy for reducing stormwater runoff that contributes to Combined Sewer Overflows (CSOs). However, quantifying the impact of proposed GI systems on CSO discharges is a difficult task that requires the simulation of runoff in a complex network of land parcels, drainage controls, and sewer pipes. In this study, the performance of Porous Pavement (PP) was examined, with a focus on how the spatial design of PP features affects predicted CSO volume. Numerical experiments were performed to explore how simulation-based designs can identify specific subcatchments for cost-effective PP implementation subject to budgetary constraints. Among the alternative design strategies considered, simulation–optimization was effective at finding cost-effective solutions. The experiments showed that PP can achieve substantial CSO reductions across a range of rainfalls, budgets, and CSO drainage characteristics, but the performance is sensitive to the spatial configuration of the features. Good-performing systems were also identified by generating multiple realizations via a randomized clustering strategy, with additional improvement provided by the more computationally expensive optimization process.