CLIMATE RESILIENCE OF CRITICAL INFRASTRUCTURE: Network Modelling for Flood Resilience Assessment and Recovery Optimisation of Urban Rail Transit

A majority of studies on transport resilience assessment employ network modelling to quantify the changes in topological attributes under random or targeted failure scenarios. This approach is insufficient to capture the multidimensionality of infrastructure resilience and falls short in representing the disruption scenarios caused by climate-related hazards. To advance this technique towards practical climate resilience analysis for informed decision support on infrastructure disaster risk management, this research is dedicated to address the research question “How to advance complex network modelling to incorporate physical, operational, social, and economic dimensions to assess and enhance the resilience of URTS exposed to flood risk?”. It demonstrates the designed methodologies through a case study of 15 lines of the London URTS under 30-year, 100-year, and 1,000-year flood risk scenarios. For URTS flood resilience assessment, this research advances network modelling to capture the dynamic service delivery of URTSs at the disruption moment and during recovery. It begins by generating service disruption scenarios through the identification of flooded stations and tracks that would become non-functional in the event of flooding. This is accomplished utilising flood depth maps with carefully designed procedures, including identifying URTS flood entry points, determining their flood depths, classifying URTS flooding incidents, and capturing diverse practical disruption effects through network modelling. Subsequently, it models the dynamic recovery process and evaluates the operational performance of service delivery during this period. On this basis, resilience is quantified by the total unsatisfied travel demand throughout the disruption duration, where revenue loss, recovery time, and delays are also assessed. The findings of the London URTS case study indicate that on average, there would be approximately 1.8 million, 4.2 million, and 7.2 million instances of unsatisfied travel demand under the 30-year, 100-year, and 1,000-year flood risk scenarios, correspondingly leading to revenue loss around £3.1 million, £6.8 million, and £11.4 million. These results demonstrate the significant potential impacts of plausible flood risks and underscore the urgent need to enhance system resilience. For URTS flood resilience enhancement, this research develops novel applications of advanced algorithm-based decision models to optimise recovery resource scheduling, with the goal of minimising the above flood impacts on URTS operations. It integrates a genetic algorithm into the URTS flood resilience assessment model and yields satisfactory optimisation results. In the London URTS case study, revenue loss can be reduced by 10.9%, 10.7%, and 6.7% across the three flood scenarios, corresponding to savings of approximately £337K, £708K, and £760K, along with decreased unmet travel demand of 197K, 404K and 470K. Insights gleaned from these results inspire a potentially cost-effective disaster risk management strategy focusing on effective recovery. Subsequently, to enable real-time decision-making optimisation and explore superior scheduling solutions, this research further develops a deep reinforcement learning model to address the same problem. While the case demonstration remains incomplete, this work has revealed challenges associated with implementing this method in a real-world context and lays the groundwork for future model tuning. To the best of the author's knowledge, this research represents the state-of-the-art in URTS flood resilience analyses. It offers multifaceted contributions to the climate resilience analysis of complex infrastructure systems. Despite the remaining challenges and limitations, this research has extended the boundaries of what can be achieved and uncovered what needs to be done to effectively enhance the ability of transit agencies to anticipate, assess, respond to, and recover from flood crises in a changing climate, thereby minimising direct and indirect impacts on passengers, rail transit operations, and broad society.