Automated identification of urban substructure for comparative analysis

Area of Science:

• Urban planning
• Geographic information science
• Computational topology

Background:

• Neighborhood definitions are crucial for urban planning but often lack scalability and appropriate aggregation patterns.
• Existing methods may not adequately capture the complex spatial relationships within cities.

Purpose of the Study:

• To propose a generalized and scalable method for identifying urban neighborhoods using topological data analysis.
• To apply this method to understand neighborhood formation, social mixing, and urban infrastructure design.

Main Methods:

• Utilized topological data analysis (TDA) to identify barrier-enclosed neighborhoods.
• Employed readily available building parcel data, requiring no prior domain knowledge.
• Analyzed data from three diverse American cities: Houston, New York, and San Francisco.

Main Results:

• The TDA method successfully identified neighborhoods consistent with historical definitions across the studied cities.
• A consistent scale of physical isolation driving neighborhood emergence was observed in all three cities.
• Differences in connectivity were found: Houston showed less disconnection at smaller scales but more at larger scales compared to New York and San Francisco.

Conclusions:

• The proposed TDA approach provides a scalable and data-driven method for neighborhood identification in urban planning.
• Physical isolation plays a key role in neighborhood formation across different urban contexts.
• Urban morphology and connectivity vary significantly between cities, impacting planning strategies.