The use of k-nearest neighbour (k-nn) approaches for the creation of bespoke neighbourhoods has become more common in segregation research in recent years. The reasons are manifold but include increased availability of high-resolution data, increasing computational power and the development of software designed to process huge numbers of k-nn commands. In this chapter, we present and test a new geo-computational add-on that has been introduced in the latest version of EquiPop (EquiPop Flow). An important novelty is that bespoke neighbourhoods do not necessarily need to grow radially until they reach a designated k-value but can make use of user-defined networks to grow at different speeds at different locations, such as following street and transportation infrastructure. We compare the geographical compositions of two different k-nn based bespoke neighbourhood techniques and discuss the pros and cons of expanding traditional k-nn computations to include data on infrastructure. Results indicate that infrastructure-integrating bespoke neighbourhoods are considerably better in depicting neighbourhoods, especially in areas with complex geographies that restrict mobility in some directions. However, the increase in computational time and complexity in setting up a network k-nn model makes a traditional radial growth approach attractive in areas where variation in connectivity between locations is limited.
John Östh, Aura Reggiani and Giacomo Galiazzo
In much accessibility research, arbitrary estimates of the distance sensitivity parameters have been used to represent the distance decay parameters in potential accessibility models. These estimates might be considered arbitrary since the choice of value and the choice of the distance decay function is often motivated by statistical indicators of the goodness of fit on spatial flows, given the fact that measures of ‘real’ accessibilities are missing. Starting from these considerations, in this chapter we introduce a new approach, the half-life model originating from the natural sciences, to estimate distance decay parameters. This method is compared with two conventional approaches originating from spatial economic science for the computation of distance decay parameters: the unconstrained and the doubly constrained spatial interaction models. The emerging distance decay parameters will be then considered in the construction of accessibility indicators based on the potential accessibility introduced by Hansen in 1959. In this context, both the mean and the median distance will be taken into account in order to identify MAUP-related issues. The exploration of these three approaches focuses on empirical analyses of accessibility in Sweden at the municipal level for 1993 and 2008. All the emerging accessibility indicators are compared in order to analyse similarities and differences in the hierarchical accessibility levels of the Swedish municipalities. The chapter concludes with some methodological and empirical remarks on the adoption of these three approaches, in the light of possible forecasts and related policy analyses.