Planning Stormwater Retention for Resilience Against Extreme Rainfalls

Climate changes inflict prolonged and intensified rainfalls on cities around the world. Unfortunately, existing efforts hardly meet the urgent need for climate-adaptive stormwater management. Urban stormwater infrastructure is typically planned based on empirically predetermined rainfall scenarios on the intensity-duration-frequency (IDF) curve, which fails to capture the worst-case scenarios that should be planned against. This paper identifies the worst-case rainfalls that cause the most severe flooding loss through robust optimization, of which the uncertainty set is constructed using IDF curves and historical rainfall statistics. Our analysis reveals that not all rainfalls on the same IDF curve are equal in terms of the incurred flooding loss. If planners fail to account for the city’s intrinsic infrastructure capabilities, they may pick a wrong 50-year rainfall to target and end up barely surviving a two-year rainfall. In contrast, we provide city planners with interpretable risk mitigation guidelines and infrastructure plans that are more robust to misspecification of rainfall scenarios. Using three planning models (loss-based, resilience-based, and systematic-cost-based) with case studies (based on Toronto, Manhattan, and New Orleans), we show that stronger infrastructure generally shifts the worst-case scenario toward shorter, more intense storms, whereas a higher return period shifts the worst-case scenario toward longer, lower-intensity events. Moreover, cities should be savvy in striking a balance. Although green infrastructure offers faster absorption with multiple co-benefits, planners should prioritize gray infrastructure when managing severe rainfall events and/or facing tight budget constraints. In short, cities must move rapidly to contain flood loss as climate risks intensify.

Funding: W. Qi acknowledges the support from the National Natural Science Foundation of China [Grants 72242106, 72188101, and 72521001]. S. Liu acknowledges the support from the Natural Sciences and Engineering Research Council of Canada [Grant RGPIN-2022-04950], University of Toronto School of Cities, Management Data and Analytics Lab, and Dobson Business & Climate Project Research Grant (Rotman School of Management). A. Zhang acknowledges the support from the Natural Sciences and Engineering Research Council of Canada [Grants RGPIN-2024-05910, DGECR-2024-00094], and the Einwechter Centre for Supply Chain Management at the Lazaridis School of Business and Economics.

Supplemental Material: All supplemental materials, including the code, data, and files required to reproduce the results, are available at https://doi.org/10.1287/opre.2025.2161.