A Classification of Multidimensional Open Data for Urban Morphology
Alexandros Alexiou; Alex Singleton; Paul A. Longley (2016). Built Environment, 42(3), 382-395. DOI: 10.2148/benv.42.3.382
Abstract
Identifying socio-spatial patterns through geodemographic classification has proven utility over a range of disciplines. While most of these spatial classification systems include a plethora of socioeconomic attributes, there is arguably little to no input regarding attributes of the built environment or physical space, and their relationship to socioeconomic profiles within this context has not been evaluated in any systematic way. This research explores the generation of neighbourhood characteristics and other attributes using a geographic data science approach, taking advantage of the increasing availability of such spatial data from open data sources. We adopt a SOM (Self-Organizing Maps) methodology to create a classification of Multidimensional Open Data Urban Morphology (MODUM) and test the extent to which this output systematically follows conventional socioeconomic profiles. Such an analysis can also provide a simplified structure of the physical properties of geographic space that can be further used as input to more complex socioeconomic models.
Extended Summary
This research develops a new geodemographic classification system based on the physical characteristics of built environments rather than traditional socioeconomic data. The study addresses the limitation that existing neighbourhood classification systems largely ignore physical attributes of the built environment and their relationship to social patterns. The research uses Self-Organizing Maps (SOM) methodology to analyse 18 different variables related to urban morphology, including proximity to transport infrastructure, green spaces, retail centres, housing types, and historic buildings. The dataset covers 181,408 Output Areas across England and Wales, utilising exclusively open data sources including Ordnance Survey mapping data, Historic England archives, and Census housing information. The methodology measures both adjacency effects (within 100 metres) and intermediate effects (within 600 metres) to capture how different built environment features influence neighbourhood character. The analysis produces eight distinct neighbourhood types: High Street and Promenades, Central Business District, The Old Town, Victorian Terraces, Railway Buzz, Suburban Landscapes, Countryside Sceneries, and Waterside Settings. Each cluster demonstrates unique physical characteristics that correspond to different urban morphologies. When compared with the Office for National Statistics’ existing Output Area Classification, the new MODUM system shows significant correlation (Cramer’s V = 0.328), particularly for rural areas which achieve over 82% correspondence. The research reveals that physical built environment characteristics systematically relate to socioeconomic patterns, suggesting residential decisions are influenced not just by social homophily but also by urban form preferences. For instance, proximity to railway infrastructure, historic buildings, and green spaces creates distinct neighbourhood types with predictable demographic patterns. The study demonstrates that people’s residential choices reflect preferences for specific physical environments, which traditional geodemographic systems fail to capture. The MODUM classification offers practical applications for urban planning, property valuation, and public service delivery by providing a simplified structure of geographic space based on physical properties. This approach enables more nuanced understanding of neighbourhood dynamics and could enhance existing socioeconomic models by incorporating previously overlooked built environment factors. The methodology is reproducible and updateable, making it valuable for ongoing urban analysis and policy development in the era of big data and open government information.
Key Findings
- The new MODUM classification identifies eight distinct neighbourhood types based on physical built environment characteristics rather than traditional socioeconomic data.
- Physical urban features show significant correlation (Cramer’s V = 0.328) with established socioeconomic geodemographic classifications, particularly for rural areas.
- Residential decisions are influenced by built environment preferences beyond social homophily, including proximity to transport, green space and historic features.
- The methodology successfully uses only open data sources, making it reproducible and suitable for ongoing urban analysis applications.
- Built environment classifications can enhance traditional socioeconomic models by capturing previously overlooked physical neighbourhood characteristics and spatial patterns.
Citation
@article{alexiou2016classification,
author = {Alexandros Alexiou; Alex Singleton; Paul A. Longley},
title = {A Classification of Multidimensional Open Data for Urban Morphology},
journal = {Built Environment},
year = {2016},
volume = {42(3)},
pages = {382-395},
doi = {10.2148/benv.42.3.382}
}