The research question
Air pollution is experienced locally, yet the evidence used to describe it is often coarse. My doctoral research examines how the physical city—its streets, buildings, vegetation and patterns of activity—interacts with environmental processes to shape the distribution of PM2.5 across space and time.
The work asks not only whether a variable is associated with air quality, but also where, at what scale and through which mechanism that relationship becomes meaningful for urban planning.
Evidence across scales
The research brings together remote sensing, street-view imagery, monitoring data and spatial indicators. Each source describes a different layer of the city. Their value comes from connecting them through a consistent spatial framework rather than treating them as separate datasets.
Analytical workflow
The workflow combines computer vision, spatial clustering, machine learning and mechanistic interpretation. Predictive performance is treated as one part of the evidence—not as a substitute for understanding how urban form and vegetation affect dispersion, exposure and human experience.
What this enables
The resulting framework supports comparisons between urban environments, identifies where additional measurements are most valuable and clarifies which design interventions deserve further testing. It positions environmental modelling as a decision-support practice: rigorous enough for research, and legible enough for planners and designers.