Automatic neighbourhood identification from population surfaces

Abstract

The definition of coherent social neighbourhoods has long been a pursuit of social geographers. There has been renewed interest in the identification of neighbourhood and community boundaries for market analysis and service delivery, with the widespread availability of digital spatially referenced population data. Traditional computational approaches to neighbourhood identification have typically involved aspatial multivariate statistical techniques or statistical regionalization approaches which are fundamentally limited by the predefined boundaries in the spatial data. Incorporation of greater spatial analytical capabilities in GIS software has led to a widespread recognition that the nature of the underlying spatial data model places fundamental limitations on the analyses which can be performed. Surface models of population provide an alternative data structure for the development of explicitly geographical analysis techniques. This paper considers the application of computationally intensive regionalization methods to population surface models, and presents a prototype methodology for the automatic identification of social neighbourhoods without the constraints of preexisting zone boundaries.