Representations of street networks in space syntax towards flexible maps and multiple graphs

Paper in proceeding, 2017

The shift from Axial to Line-segment maps is one of the most important developments in Space Syntax analysis, both theoretically and methodologically. It followed a long line of investigations and discussions within the field of Space Syntax, which addressed critical issues related to the Axial map representation (e.g. Hillier and Iida,2005; Hillier,1999a,1999b; Turner,2001; Steadman,2004, Dalton, 2001). At the same time, it opened up new possibilities by allowing for the use of Road-centre-line maps in Space Syntax analysis; that is, largely available GIS-based segment maps, used in most areas of urban modelling (e.g. Turner,2007; Dalton et al.,2003). Today, several models implement syntactical analysis to Road-centre-line maps, while claiming to form valid alternatives to the Axial map, not least in capturing the perceptual and cognitive affordances presented by the environment. This paper focuses on such alternative models with a twofold aim: first, to help make well-grounded choices in applying syntactical analysis to Road-centre-line maps; and second, to explore the methodological potentials of Line-segment maps, which are created by their flexibility, by being the least aggregated representations of street networks. The paper introduces an experimental software application to push the investigations and exploit these methodological possibilities even further. The models discussed in this paper are: 1) Angular Segment Analysis (e.g. Turner,2007; Hillier and Iida,2005); 2) Natural Streets maps (Jiang and Liu,2007) and Continuity maps (Figueiredo and Amorim,2005); 3) Directional Distance model (Peponis et al.,2008). Based on a systematic comparison of: a) their geometric representation of the street network (the ‘map’), b) the dual- graph they calculate, and c) the measure of distance they use, we argue that they can all be seen as parametrically defined representations, based on a Line-segment map. In other words, with the same Line-segment map one could produce all of them, if a different set of angular parameters was used, to either define the graph elements (nodes, edges), or to calculate distance. These models, parametrically redefine the relation between the ‘map’ and the ‘graph’; thus, challenging the one-to-one relation between the two, that the Axial map was founded upon. The methodological importance of this development goes beyond the specific models described in this paper. If the same Line-segment map can produce different graphs by using a different set of angular parameters, then, for instance, one could easily change the parameters to follow the latest theoretical insights on human cognition, without having to change the map. Going even further, one could develop and test different representational models to explore new theoretical and methodological paths, by using one single map. The experimental tool also presented in this paper is a step in that direction, allowing to test different models using a single software.