Quantitative comparison of cities: Distribution of street and building types based on density and centrality measures

It has been argued that different urban configurations - planned vs. organic, treelike vs. grid like - perform differently when it comes to the intensity and distribution of pedestrian flows, built density and land uses. However, definitions of urban configurations are often rather abstract, ill-defined and at worse end in fixed stereotypes hiding underlying spatial complexity. Recent publications define morphological typologies based on quantitative variables (e.g. Barthelemy, 2015; Serra, 2013a; Gil et al., 2012; Berghauser Pont and Haupt, 2010) and solve some of these shortcomings. These approaches contribute to the discussion of types in two ways: firstly, they allow for the definition of types based on multiple variables in a precise and repeattable manner, enabling the study of large samples and the comparison between both cities and regions; secondly, they frame design choices in terms of types without being fixed and so open up for design explorations where the relation between the variables can be challenged to propose new types. This paper explores the typologies defined by Serra (2013a) and Berghauser Pont and Haupt (2010) further, as these target two of the most important morphological entities of urban form, namely the street network and the building structure. The purpose is to gain a better understanding of how types are composed and distributed within and across different cities. The method is based on GIS and statistical modeling of four cities to allow for a comparative analysis of four cities: Amsterdam, London, Stockholm and Gothenburg. For the street network, we process the Road-Centre-line maps to obtain a clean network model, then run segment angular analysis to calculate the space syntax measures of betweenness at different metric radii, defining the “centrality palimpsest” (Serra, 2013a). For the building structure, we process elevation data to obtain building height, then run accessible density analysis for all building density metrics (FSI, GSI, OSR, L) using the Place Syntax Tool (Berghauser Pont and Marcus, 2014). The street and building types are defined using cluster analysis (unsupervised classification), following a similar approach to Serra (2013a). The result is a typology of street (´paths´) and building types (´places´), with different profiles of centrality and density across scales. The spatial distribution and frequency of these types across the four cities gives an objective summary of their spatial structure, identifying common as well as unique traits