23 OCT 2025 (THU) 14:35 - 14:55

Transit-Oriented Development as Urban Public Realm: Vibrancy, Diversity, and Safety

Ms FAN Zhuangyuan 

( Supervisor: Prof Becky P.Y. Loo )

Abstract:

Transit-oriented development (TOD)’s success has been largely measured by ridership and built environment quality. This dissertation reconceptualizes TOD as a dynamic public realm and extends TOD’s performance measure to the social qualities it promotes in a public space—vibrancy, diversity, and safety. Four complementary studies advance this agenda. The first study demonstrated how to measure the TOD as activity space using billions of mobile place visits records from 4,290 rail stations in the U.S. It leverages activity-based, semi-unsupervised clustering algorithm to uncovers 13 “rhythms of places”.

The second study examines the metro extension history of Hong Kong in three decades and explains the relationship between change of social segregation and the expansion of public transit. We found that metro growth simultaneously redistributes low-income residents away from station catchments yet broadens citywide activity-space exposure to diverse groups. Built upon the findings from the second study, the third study extends to five global cities, Boston, Chicago, Hong Kong, London, and Sao Paulo. Using travel survey from over 200,000 residents, the study demonstrates that activity-space place embeddings, constructed based on the local transit network, predict people’s daily exposure to different amenities (restaurants, retails, cafe, offices, hospitals, sports, etc.) with up to 81% accuracy, spotlighting the pivotal role of urban form and mobility networks in bridging the social segregation.

Lastly, looking from a global perspective, the fourth study applies a deep-learning based computer vision models to 26.8 millions of street view images from 106 cities in the world, and classified their street space into six visual clusters. Controlling for socioeconomic and policy factors, we found that cities with 1% increase in bus presence see a 0.35% drop in road deaths per 100, 000 residents. Taken together, these studies reposition TOD and public transit as engines for daily social interaction and public safety. Methodologically, they demonstrate how large-scale mobile traces, longitudinal and cross-sectional travel surveys, representation learning, and street-level imagery can be fused to evaluate TOD and public transit performance. The dissertation aims to inform policies that synchronize land use, transit planning, and streetscape design to cultivate vibrant, inclusive, and safer cities.