Infrastructure Resilience and Management Under Changing Tropical Cyclone Risk

Abstract

Coastal regions are vulnerable to hurricane hazards (i.e., storm surge, strong winds, and severe rains). The consequences may worsen due to climate change and coastal development. Improving coastal resilience is a major challenge to be addressed with knowledge of climatic threats and infrastructure systems behavior. This dissertation attempts to merge science and engineering to assess infrastructure risk and develop risk mitigation strategies that account for climate change effects, infrastructure response, economic feasibility, and policy. It focuses on three main infrastructure systems threatened by hurricanes: power, transportation, and coastal protection systems. Hurricane induced power outage and blackout-heatwave-compound hazard risks are evaluated for Harris County, TX and Louisiana State, employing climate projection of hurricane hazards and heatwaves and a physics-based power outage and recovery process model. It is found that climate change will largely increase the hurricane-blackout-heatwave compound risk. It is also discovered that strategically burying a small portion of distribution network might significant increase power system resilience. Hurricane evacuation on the transportation system is highly complex. A framework for rapidly predicting the hurricane evacuation traffic flow based on hurricane forecasting, evacuation orders, the road network, and population information is developed. The framework is applied to predict Florida’s largest evacuation event -- Hurricane Irma in 2017 -- evaluated/calibrated with game-theory-based reconstruction of the traffic flow for the event. The framework is applicable for evacuation management. For example, the analysis shows that a minor adjustment to the evacuation order could considerably alleviate the traffic congestion during Hurricane Irma. Also, a coupled power system-transportation system analysis shows that if most evacuation cars were changed to electrical vehicles, the power demand may surpass the supply limit. Coastal protection systems safeguard infrastructure against hurricane threats. An reinforcement-learning-based technique is adopted to design adaptive seawall and engineering-policy protection systems (buyout/retrofit/dike), for New York City as an example. The reinforcement-learning approach combats the deep uncertainties in climate projection by systematically observing the climate condition and updating the design. Minimizing the life-cycle cost, reinforcement-learning strategies also have a better controlled tail risk, compared to other strategies. The method may be applied broadly to analyze flexible climate change adaptation strategies.