Battling Climate Change Induced Flooding: Rapid Flood Modeling, Optimization, Uncertainties and Metrics
Climate change is likely to magnify flood damages. The hydrodynamic models used to test mitigation strategies are complex and computationally intensive. As an alternative, rapid flood models(RFM) are often used in large-scale watersheds or when many simulation runs are required (e.g. multiple climate scenarios). However, the low-complexity nature of RFM makes this approach less accurate than hydrodynamic models. In addition, the literature shows that the lack of proper metrics to evaluate flood impacts on urban infrastructure. In addition, there are uncertainties associated with future flooding predictions, which impairs floodplain managers and urban planners to develop adequate flood protection and mitigation systems. Therefore, this dissertation aims to:(1) advance the current state-of-art of RFM; (2) propose new metrics to improve measuring flood impacts on urban infrastructure; and (3) provide a systematic evaluation method to test mitigation strategies under climate uncertainties. First, a rapid flood model using a conditional generative adversarial network (cGAN-Flood) was developed. This model performed satisfactorily in accuracy for catchments located outside the training dataset. Then, this dissertation evaluated the impacts of climate change on the transportation infrastructure in San Antonio, Texas. New metrics, inspired by hydrological footprint residency (HFR), were developed specifically for roads (HFRR)and bridges (HFRB). These new HFR metrics combine flood extent and duration, both essential for evaluating how flood affects the transportation sector. Finally, a multi-objective optimization was performed to minimize implementation costs of Low Impact Development while maximizing infiltration and minimizing peak flow. The robustness of optimized solutions under different budgets was calculated for several climate scenarios to facilitate decision-making.