Authors

Archy Guo, Kunjal Bhatta, Ruoyu Li

Research Question

How do different types of urban amenities co-locate, and how do these patterns vary across spatial, demographic, and behavioral contexts? Additional sub-questions include:

• What are the statistically significant colocation patterns among different POI types (e.g., grocery stores, clinics, cafes)?
• How can we model these relationships as a dependency or influence network?
• How do amenity clusters correlate with social or demographic typologies (e.g., income, age, lifestyle segments)?
• Can we benchmark amenity configurations across neighborhoods or cities to inform urban design, investment, or policy?
• How can open-source data (OSM, OS, Overture, Locomizer) be harmonized to create a reusable and replicable spatial intelligence tool?

Background

Urban life is shaped by the spatial organization of amenities, where we live, work, eat, learn, and socialize. Yet our understanding of how these amenities cluster and co-depend remains fragmented, especially across cities with diverse spatial and socioeconomic fabrics. This project proposes a geospatial toolkit that identifies and visualizes clusters of amenities — spatial groupings of co-located points of interest (POIs) as well as colocation networks that model the statistical dependencies between amenity types.

Using a combination of open-source geospatial data (OpenStreetMap, Ordnance Survey, Overture Maps), the toolkit will classify POIs using hierarchical categories, detect clusters via spatial algorithms (DBSCAN, HDBSCAN), and reconstruct co-location or dependency networks through measures like mutual information, conditional probabilities, or edge-weighted graphs. Demographic overlays and persona segmentation (via census or proxy datasets) will enable user-centric benchmarking: e.g., what amenity configurations are typical around schools in high-income vs. low-income neighborhoods? What "urban DNA" characterizes emerging areas of the city?

The toolkit will initially focus on London as a testbed but will be designed for global extensibility. It will allow planners, designers, researchers, and civic actors to explore questions of equity, access, and systemic resilience through the lens of amenity distribution.

Methodology

The project will follow a modular and data-driven approach, structured around four core components:

1. Data Collection & Standardization: Aggregate POI data from OpenStreetMap, Ordnance Survey, and Overture Maps, using consistent spatial identifiers; Enrich spatial units (hex bins) with demographic data from ONS; Classify amenities into hierarchical types using a unified ontology.
2. Amenity Clustering: Apply spatial clustering algorithms (e.g., DBSCAN, HDBSCAN, kernel density estimation) to detect localized clusters of amenities. Optionally, use graph community detection (e.g., Louvain) if POIs are treated as nodes with proximity-based edges.
3. Dependency Network Construction: Build a co-location matrix capturing the frequency or conditional probability of one amenity type appearing near another. Use network modeling techniques (e.g., mutual information networks, edge-weighted graphs, Bayesian networks) to visualize and analyze these dependencies.
4. Benchmarking & Typology Mapping: Compare amenity clusters and dependency patterns across demographic or spatial typologies (e.g., affluent vs. deprived areas, urban core vs. periphery); Develop persona-based overlays (e.g., student hubs, aging communities, creative clusters) to explore how amenity landscapes support (or exclude) different populations.

Deliverables

• Apply spatial clustering and network analysis to real-world urban amenity data
• Integrate and clean open-source geospatial datasets (e.g., OSM, ONS, Ordnance Survey)
• Construct and interpret amenity colocation and dependency networks
• Benchmark urban areas based on amenity configurations and demographic overlays
• Communicate spatial insights through clear visualizations and maps
• Develop reproducible, open-source workflows for urban spatial analysis

Data Sources

• Locomizer: Human mobility data
• OpenStreetMap and Overture Maps Foundation: Points-of-interest (POI) data
• UK Office for National Statistics: Census data and administrative boundaries