A Bayesian Approach to ‘Community Belonging’ and Segregation in a Divided City

This paper presents an alternative approach to the measurement of segregation that uses Bayesian statistics to combine information on residents’ perceptions (PGIS) and behaviour (GPS tracking) in order to generate probabilistic surfaces of ‘community belonging’ for each of the main communities in the study area. Because these surfaces are based upon both perception and behaviour, they are not limited to either residential or activity-space segregation and they also avoid several problems associated with traditional, census based analyses such as homogeneity within areal units and the modifiable areal unit problem. These surfaces are then assessed for segregation and scale sensitivity using a modified version of the lacunarity metric for spatial heterogeneity, in order to demonstrate how this approach has the potential to give new insights into the use and segregation of space, which is illustrated using a case study in North Belfast, Northern Ireland.