Neglecting Spatiotemporal Rainfall Variability Underestimates Flood Hazard and Risk

This project investigates the effects of commonly held assumptions around rainfall and probability in flood hazard and risk modeling. Flood hazard and risk models are produced through computer simulations, in which natural systems are simplified to make the simulation computationally feasible. One such simplification is the "design storm" approach to simulating rainfall, in which a synthetic storm is constructed for a specified duration and return period (e.g., the 24-hour, 100-year storm), with precipitation represented as uniform in space, and highly idealized in time. The return period associated with the resulting flooding and damage are then assumed to be equal to that of the storm (e.g., a 100-year storm would produce 100-year flood and damage events). We investigate the effects of these assumptions on flood hazard maps and risk datasets by comparing those produced through design storm approaches against those produced via stochastic storm transposition (SST), an alternative, probabilistic approach that preserves realistic space-time structures in rainfall and allows for nonlinear relationships between the frequency of rainfall, hazard, and risk.

This research provides insight into the validity of common flood hazard/risk modeling assumptions, with implications for the accuracy and utility of existing flood hazard and risk information. It is broadly directed toward audiences interested in flood hazard and risk, including hydrologists, modelers, engineers, floodplain managers, planners, and city managers. The datasets produced by this research, as provided herein, allow for validation of the results and conclusions drawn from this research, and enable future studies investigating e.g., the potential building-scale economic impacts of flooding, optimal flood risk reduction and management strategies, and other studies leveraging high-resolution flood hazard and risk data.

本研究旨在探究洪水灾害与风险建模中，围绕降雨与概率的普遍预设假设所产生的影响。洪水灾害与风险模型通过计算机仿真构建，为使仿真具备计算可行性，需对自然系统进行简化。此类简化方法之一为设计暴雨（design storm）降雨模拟法：针对特定时长与重现期（如24小时一遇、100年一遇暴雨）构建人工暴雨，将降水在空间上设为均匀分布，且在时间维度上高度理想化。随后默认所得洪水与灾害的重现期与该暴雨的重现期一致（例如100年一遇暴雨将引发100年一遇的洪水与灾害事件）。本研究通过对比设计暴雨法与随机暴雨移置法（stochastic storm transposition, SST）的建模结果，探究上述假设对洪水灾害图与风险数据集的影响；后者作为一种替代的概率化建模方法，可保留降雨的真实时空结构，并允许降雨频率、灾害与风险之间存在非线性关联。
本研究可阐明常见洪水灾害/风险建模假设的合理性，对现有洪水灾害与风险信息的准确性与实用性具有参考价值。其受众覆盖所有关注洪水灾害与风险的群体，包括水文学家、模型研发者、工程师、洪泛区管理者、规划人员与城市管理者。本研究所生成的数据集（即本文提供的数据集）可用于验证本研究的结果与结论，同时支持后续相关研究：例如评估洪水对建筑单体的潜在经济影响、优化洪水风险降低与管理策略，以及其他利用高分辨率洪水灾害与风险数据开展的研究。