Urban Typologies: Methodology

 

Methodological overview

Clustering is a statistical technique used to identify and group similar observations based on selected characteristics. The core idea is to maximise similarity within each group (or cluster) while ensuring distinctiveness between groups. This approach enables the detection of underlying patterns in data and simplifies complex information by organising it into interpretable and meaningful segments. In contexts where large datasets contain numerous variables, clustering helps distil this complexity into coherent profiles, making it easier to draw insights and support decision-making.

 

For our analysis, we used the well-established and computationally efficient K-means clustering algorithm to partition the data into K groups (MacQueen, 1967). Since K-means is sensitive to initial conditions, we improved performance by initialising with results from hierarchical clustering using Ward's method on a random data subset (Ward, 1963; Milligan, 1980). To address sensitivity to data scaling, we applied min-max normalisation – rescaling all variables to the [0,1] range – prior to clustering, as recommended by Milligan and Cooper (1988). A key step in clustering is selecting the number of clusters, into which the dataset will be partitioned. We used a method based on cluster stability (Ben-Hur, 2002), which assumes that robust clustering should yield similar results under varying initial conditions and data perturbations. For more insights see Kok et al. (2016).

 

In our analysis, we applied clustering to examine European NUTS-3 regions from four different perspectives related to urban planning in times of climate change. Each clustering exercise focused on a distinct set of indicators expressed by variables. The typology Capacity for Action includes socio-economic and demographic indicators, grouping cities based on the resources and capacities they can mobilise for adaptation or mitigation measures. The Climate Hazards typology uses environmental risk and exposure indicators to cluster regions facing similar climate-related threats. The Contributions to Mitigation typology identifies areas according to the primary sources of CO2 emissions and their potential for renewable energy deployment. Lastly, the Urban Morphology typology captures the structural characteristics of urban areas within the regions, grouping regions with comparable spatial and built environments. An overview of the indicators used for the four different typologies, including original data sources and key processing steps, is provided below.

 

To understand the adaptation opportunities and challenges unique to each cluster, it was essential to examine how clusters differ from and relate to one another. We compared the average indicator values and analysed their spread across clusters to identify defining characteristics. In addition, we looked at the geographical distribution, which often revealed spatial patterns and added contextual insights beyond the data. Together, these elements helped us build clear profiles of each cluster, highlighting their unique traits and interrelationships.

 

Study area

The study area includes NUTS-3 regions that part of the EU27, in addition to the NUTS-3 regions of Belfast (UKN06) and Istanbul (TR100). Excluded from the study area are the following:

  • French Oversee Territories (FRY10, FRY20, FRY30, FRY40, FRY50)
  • Azores and Madeira (PT200, PT300)
  • Canary Islands (ES703, ES704, ES705, ES706, ES707, ES708, ES709)
  • Spanish autonomous regions of Ceuta and Melilla (ES630, ES640)

 

You can access the overview of indicators and processing steps here. [ADD LINK TO METHODOLOGY TABLE]

 

References

Ben-Hur, A., Elisieeff, A., & Guyon, I. (2002). A stability based method for discovering structure in clustered data. Pacific Symposium on Biocomputing, 2002, 6–17.

Kok, M., Lüdeke, M., Lucas, P., Sterzel, T., Walther, C., Janssen, P., Sietz, D., & de Soysa, I. (2016). A new method for analysing socio-ecological patterns of vulnerability. Regional Environmental Change, 16(1).

MacQueen, J. (1967). Some methods for classification and analysis of multivariate observations. In Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability (Vol. 1, pp. 281–297). University of California Press.

Milligan, G. W. (1980). An examination of the effect of six types of error perturbation on fifteen clustering algorithms. Psychometrika, 45, 325–342.

Ward, J. H. (1963). Hierarchical grouping to optimize an objective function. Journal of the American Statistical Association, 58, 236–244.