A data driven approach to urban area delineation using multi source geospatial data.

Area of Science:

• Geographic Information Systems (GIS)
• Urban Studies
• Data Science

Background:

• Traditional urban delineation methods often lack precision, relying on basic road data aggregations.
• Existing approaches struggle with noise and irrelevant data, leading to inaccurate urban area identification.

Purpose of the Study:

• To introduce a data-driven, bottom-up approach for precise urban delineation.
• To enhance urban clustering by integrating feature engineering with DBSCAN.
• To provide a scalable and replicable model for urban studies and planning.

Main Methods:

• Utilized OpenStreetMap (OSM) data across various categories.
• Applied feature engineering to refine data selection and reduce noise.
• Integrated feature engineering with the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm.

Main Results:

• Achieved a 5% improvement in average accuracy for urban delineation.
• Demonstrated effective noise reduction and mitigation of DBSCAN pitfalls through feature engineering.
• Validated the model using nighttime light data and Zipf's law, confirming a strong fit (p-value=0.99).

Conclusions:

• The proposed method offers a significant improvement in urban delineation precision and methodology.
• Feature engineering is crucial for refining spatial clustering and overcoming data limitations.
• The study provides a robust, scalable, and data-driven model for delineating urban areas, supporting informed urban planning and policy-making.