Towards a sustainable urban transport system : applications of spatial analysis and machine learning methods

Abstract

Urban transport systems are pivotal in shaping modern cities, evolving to address the complex demand of urbanization. Despite the progress in developing frameworks and indicators to guide sustainable objectives, there remains a critical need for micro-scale studies that delve into the spatial-temporal dynamics of urban transport systems to inform evidence-based policy interventions. This thesis responds to this gap, focusing on the multifaceted nature of urban transport issues, through analysing and enhancing the dimensions of convenience, safety, resilience, and equality. This thesis adopts a research design grounded in geography to explore spatial-temporal dynamics, utilizing a blend of spatial analysis and machine learning methods. Through a series of empirical studies in the context of Hong Kong, this thesis harnesses innovative data sources and analytical techniques to enhance walkability in terms of convenience, quantify the dynamics of pedestrian flows, evaluate the impact of traffic crash disruptions to the urban transport system, and address the mobility needs of people with disabilities. The first study assesses pedestrian route directness across the city for over three million home-destination pairs. The results highlight significant spatial disparities, with new towns and suburbs facing longer walking detours compared to the urban core, thereby challenging the prevailing vehicle-oriented urban design. The findings advocate for a pedestrian-oriented approach to urban planning, emphasizing the need for a critical threshold of pedestrian route directness to guide interventions. Building upon promoting walking or walkability, the second study pioneers the measurement of pedestrian exposure (to the risk of traffic crashes) to enhance pedestrian safety analysis. It explores the utility of unconventional data in capturing pedestrian dynamics and advances the analysis of pedestrian safety. In addition, this study also establishes models to identify pedestrian-vehicle collision risk factors under different road conditions. The third study offers a methodology for quantifying travel delays resulting from road crashes. The analysis uncovers the determinants of crash-induced travel delays, highlighting the influence of traffic dynamics and built environment factors. It provides an estimation of the economic cost associated with these delays. Overall, the study contributes to the understanding of urban transport system resilience and the need for effective management strategies to mitigate the impacts of non-recurrent traffic events. The final study examines the impact of the COVID-19 pandemic on the travel patterns of people with disabilities using paratransit services. The study analyses the variations in Rehabus service usage during the pandemic, pinpointing vulnerable individuals and regions most susceptible to the disruptions. It highlights the crucial role of Rehabus in providing necessary transport services, especially for medical needs even during the pandemic. This thesis pioneers a paradigm shift in urban transport research, advocating for sustainability through the lens of people-oriented planning and design. Moving beyond the conventional environmental narrative, it emphasizes the importance of addressing the nuanced needs of diverse population groups to achieve a truly sustainable urban transport system. It breaks new ground by demonstrating the value of unconventional data sources and advanced computational techniques in urban transport research.