City Clustering based on topology, activity distribution, and mobility

Abstract

Cities increasingly face challenges from urbanization and climate change, with transportation playing a central role. In car-oriented cities, it consumes space and contributes to emissions, making sustainable solutions more urgent. However, the effectiveness of such solutions depends strongly on urban context. To avoid one-size-fits-all strategies, cities can benefit from learning from similar cases, but identifying comparable cities remains a challenge.
Previous studies have classified cities based on individual features like road layout or mobility patterns. While informative, these studies often overlook interdependencies between domains. This research addresses that limitation by integrating road network structure, activity distribution, and mobility behavior into a single analysis using clustering methods.
The main question guiding this research is:
How can the application of multiple clustering methods reveal distinct groups of European cities based on road network, activity distribution and mobility characteristics?
Data was obtained for 32 European cities, capturing indicators from five domains: road topology, population, economic activity, mobility, and congestion. Indicators were standardized, and redundancy was reduced through correlation analysis and Principal Component Analysis (PCA). Clustering was then performed using K-Means, K-Medoids, and Ward’s Method, and evaluated using silhouette scores, Adjusted Rand Index (ARI), and Jaccard Similarity.
The results showed that cities could be grouped meaningfully and consistently using combined structural and behavioral indicators. Two- and seven-cluster results emerged as the most stable. The two-cluster split revealed a broad regional divide, while the seven-cluster result uncovered distinct urban typologies with full agreement across methods.
This study contributes to urban transport research by offering a replicable and holistic clustering approach that uses open data and multiple analytical techniques. It shows that meaningful typologies can be created by combining structural, functional, and behavioral dimensions, helping cities identify similar urban contexts and learn from specific strategies.
Overall, this research demonstrates that multi-domain clustering offers a valuable tool for comparing cities and supporting more targeted, evidence-based planning for sustainable urban transport.