Enhancing urban liveability and sustainability through geospatial large-scale data analytics

The rapid increase in urban populations and the growing availability of detailed, geo-referenced, large-scale data have led to the emergence of a new field known as the science of cities. While current research in this field typically offers a normative perspective, focusing on understanding these urban dynamics, the inclusion of cities as key actor of the 2015 Sustainable Development Goals (SDG 11) emphasises the need to integrate normative analyses with policy-oriented research. This dissertation presents three research contributions focusing on the development of tools to promote sustainable and healthy individual behaviours and provides insights for designing urban interventions that encourage these behaviours. The first contribution investigates factors contributing to gender-specific barriers to cycling in Western cities using large-scale data from the sport-tracking application Strava. Previous academic research on gender differences in cycling has primarily consisted in stated-preferences studies relying on direct data collection, such as surveys. While these methods provide detailed insights into cycling preferences, their generalisability is limited by small sample sizes and geographical scope. By leveraging automatically collected data from a widely adopted sport-tracking application, this study aims to validate and expand upon the findings of earlier studies with an unprecedentedly large dataset, shedding light on potential strategies to improve urban cycling conditions for women, thereby promoting more sustainable urban transport for all. Aligning with the recommendations of the UN SDG 11.7, which calls for cities to provide access to safe, inclusive, and accessible greenspaces for all their residents, the second contribution is a computational framework to measure the accessibility of public greenspaces in urban areas. Due to computational constraints, evaluations of green exposure and accessibility are typically limited to a single metric. By contrast, our framework evaluates three families of green accessibility indicators encompassing metrics proposed in the academic literature as well as in the public policy domain. Through an analysis of population and area rankings generated by different indicators across more than 1,000 cities worldwide, I question the reliability of single-metric assessments in capturing the complexity of green accessibility within urban systems. The findings suggest that a single indicator may inadequately differentiate across areas or subgroups of the population, even when focusing on one form of green accessibility at a time. From a policy viewpoint, this indicates the need to switch to a multidimensional framework capable of organically evaluating a range of indicators at once. To enhance the usability of the computational framework, the associated interactive web interface, ATGreen, provides a range of functionalities designed for both the general public and policymakers specifically. The third contribution continues the broader discussion on enhancing green accessibility but introduces a shift in focus from the structural aspects of green accessibility to strategies aimed at influencing individual behaviour to enhance exposure to nature. To this scope, I introduce a novel routing system called ATGreenGO, designed to recommend nature-enriching walking routes with minimal detours compared to the shortest path, thus compatible with integration within daily routines. Overall, the contributions presented in this dissertation combine advanced statistical and data analysis with digital tool development to enable a more comprehensive approach to the understanding and managing of urban phenomena and to inform the design of urban interventions to achieve the objectives outlined in the SDG 11.