https://hdl.handle.net/10012/21105

Urbanization significantly alters natural hydrological processes, leading to increased flood risks in urban areas. The potential damages caused by flooding in urban areas are widely recognized, making it crucial for urban residents to be well-informed about flood risks to mitigate potential losses. Flood maps serve as essential tools in this regard, providing valuable information that aids in effective planning, risk assessment, and decision-making. Despite floods being the most common natural disasters in Canada, many Canadians still lack access to high-quality, up-to-date flood maps. The occurrence of recent major flood events across the country has sparked renewed interest among government officials and stakeholders in launching new flood mapping initiatives. These projects are critical for enhancing flood risk management across communities.
Traditional flood hazard mapping methods, based on deterministic approaches, often fail to account for the complexities and uncertainties inherent in urban flood dynamics, especially under rapidly changing climate conditions. Uncertainty affects every stage of flood mapping, influencing accuracy and reliability. Recognizing this, recent studies advocate for stochastic approaches to explicitly incorporate these uncertainties. However, there is a lack of industry-standard tools that allow for a convenient and comprehensive analysis of uncertainty, making it challenging to routinely incorporate uncertainty into flood hazard assessments in practice. This underscores the need for a robust framework to model flood uncertainty. While various models have been proposed to address the uncertainty, many remain conceptual or lack the necessary automation. Despite no "perfect models", the Storm Water Management Model (SWMM) and the Hydrologic Engineering Center’s River Analysis System (HEC-RAS) are widely used for urban hydrology and channel hydraulics modeling, respectively, due to their robust physics-based approaches. Both SWMM and HEC-RAS models have been enhanced with commercial and open-source extensions, built on algorithms written in various programming languages, to improve their utility, particularly for automating workflows to handle complex urban flood scenarios. While SWMM has more robust extensions, most available HEC-RAS extensions are designed for one-dimensional (1D) steady state models, which lack the complexity needed for accurate urban flood modeling. The release of HEC-RAS 6.0, which allows for two-dimensional (2D) unsteady flow modeling and incorporates structures like bridges and weirs, marks a significant advancement for urban flood modeling. The current research was motivated by the perceived benefits of designing such extensions for automating workflows in recent versions of SWMM and HEC-RAS, as well as automating the coupling of these two models in a stochastic framework to facilitate the integration of uncertainty into existing flood hazard mapping workflows. This thesis introduces the SWMM-RASpy framework, a novel automated stochastic tool built using the open-source Python programming language. SWMM-RASpy integrates SWMM's detailed urban hydrologic capabilities, such as dual-drainage modeling, with HEC-RAS's 2D unsteady hydraulic modeling, coupled with stochastic simulations through Latin Hypercube Sampling (LHS) to analyze the uncertainty in flood hazard mapping. The framework was demonstrated on the Cooksville Creek watershed, a highly urbanized area in Mississauga, Ontario, known for its susceptibility to flooding. An entropy map was successfully produced for the case study site, which better reflects the uncertainty of flooding and could be used to develop tailored flood planning and preparedness strategies for different zones within the site.
This thesis also presents a detailed application of the SWMM-RASpy framework to assess flood hazards, with a specific focus on topography-based hydraulic uncertainties in the watershed, particularly surface roughness variability, which affects pedestrian safety during flood events. The study highlights that traditional hazard models, which focus mainly on residential buildings, do not adequately account for the risks to pedestrians who are a significant source of fatalities in flood events, especially in densely populated urban areas with high mobility. Three flood hazard metrics were developed and used to evaluate the flood risks to pedestrians given the uncertainty surrounding surface roughness: FHM1, based on inundation depth; FHM2, combining depth and velocity; and FHM3, incorporating depth, velocity, and duration. Key findings from the assessment indicate that surface roughness significantly affects pedestrian hazard estimation across the floodplain, making it a critical factor in flood hazard management. The FHM2 metric, which incorporates depth and velocity, was found to be highly sensitive to roughness variation, potentially leading to the misclassification of hazardous zones as safe and vice versa. The inclusion of velocity in the hazard assessment, while improving accuracy, also increased variability, emphasizing the need for a balanced approach in flood risk evaluations. In contrast, the FHM3 metric, which includes flooding duration, showed minimal sensitivity to surface roughness uncertainty. The research also suggests that confidence maps, produced as part of the analysis and accounting for estimated uncertainties surrounding the hazard metrics propagated from surface roughness, can offer a more reliable alternative to traditional deterministic hazard maps. Lastly, the study, through this analysis, emphasizes the importance of combining grid-level and zonal-level analyses for a more comprehensive understanding of flood hazards at different scales, thereby supporting more robust flood risk assessments.
This thesis extends the application of the SWMM-RASpy framework to assess the impacts of climate change on flood hazards within the Cooksville Creek watershed. It examines how projected changes in rainfall intensity from Global Climate Models (GCMs) affect flood risks, particularly for urban buildings, as well as the importance of incorporating uncertainties from these projections into flood hazard assessments. The same hazard metrics used for pedestrian hazard assessment, FHM1, FHM2 and FHM3, were used to evaluate building hazards. The study predicts a significant increase in flood hazards within the watershed, with a substantial expansion of inundation areas affecting up to 40% more buildings when uncertainties are considered. The analysis shows that without considering uncertainties, FHM1 and FHM3 predict a higher number of damaged buildings than FHM2, with FHM1 predicting the highest number of affected buildings. This suggests that relying solely on FHM1 to estimate building hazards may be sufficient in similar climate change scenarios, although further investigations are needed. However, when uncertainties are included, FHM2 shows a greater increase in the number of buildings at risk compared to FHM1 and FHM3, due to the larger uncertainty associated with velocity versus depth and duration. This underscores the need to incorporate uncertainty into flood hazard assessments to ensure a more comprehensive understanding of potential future damages.
Overall, this study has made significant contributions to the field of urban flood hazard assessment by developing a robust method for incorporating and analyzing uncertainties, thereby supporting more effective flood management and resilience planning. Future research should apply the SWMM-RASpy framework to other watersheds and investigate additional hydrologic and hydraulic variables to further improve flood risk assessments.

https://hdl.handle.net/10012/21105