Enhancing urban flood management : the issues of data availability, model configuration, and robustness

Abstract

Urban areas are highly susceptible to flooding, necessitating effective strategies for urban flood management. This thesis centers on enhancing urban flood management through three principal avenues: data accessibility, model configuration, and model robustness. For data availability, one major obstacle is the limited availability of timely and reliable information regarding the flood event and its consequences. However, the growing volume of social media data offers an invaluable and rapidly accessible untraditional source. This thesis thus developed a standardized and optimized workflow to assess the impacts of urban flooding by extracting and analysing social media data, as well as identifying the intensive public response areas. Using the 2020 Chengdu rainstorm-induced flooding in China, social media data provide spatial flood information and 232 flood sites with geological locations. The spatiotemporal analysis of social media revealed the process of flooding and enables quick determination of severely affected areas, demonstrating the potential of social media to support urban flooding impact assessment and storm-flooding numerical modelling by obtaining extra valuable flood data. For model configuration, the complexity of urban surfaces and limited subsurface drainage information pose great challenges to modelling urban rainfall-runoff and flood hydraulics. This thesis developed an integrated hydrodynamic model framework incorporating both surface and subsurface drainage processes. To properly involve drainage flow effects, we presented three methods to quantify the drainage flow process with limited pipe data. The model framework was verified against six benchmark cases, showing that both hydrological and hydraulic processes can be well reproduced by the model with NSE all larger than 0.5. The model was also successfully applied to simulate rainstorm-induced flood events in a rural-urban catchment, the upper Shenzhen River catchment. Results further demonstrated the robust capability of the model in the quantification of flow exchange between surface and subsurface systems to some extent even in the absence of drainage information. For model robustness, the scarcity of high-precision data, such as topographic data, rainfall data, and validation data, intensifies the uncertainty associated with models. This thesis developed a bottom-up approach for urban flood hazard mapping at multiple levels (grid-kilometer-district), built upon the integration of the developed flood modelling with data acquisition from open sources. The developed model was applied to flooding in Chengdu in August 2020 with the support of the identified social media flood data. The multilevel hazard mapping approach developed here shows less sensitivity to the data input quality and model uncertainty, indicating relatively higher reliability. Overall, this thesis confirmed the capacity of social media data to support urban flood management; optimized modelling configuration to achieve robust and accurate simulations even with limited availability of detailed sewer data; proposed a multilevel urban flood hazard mapping method to mitigate the negative impact of data scarcity and quality on hazard modelling. These findings and methodologies have the potential to strengthen overall resilience and minimize the negative impacts of flooding in urban areas.