Preferential centrality as a multi-regional model for spatial interaction and urban agglomeration
Understanding how transportation networks affect regional development has been a long-standing challenge for modellers in several disciplines, both in research and practice. Approaches span between light-weight accessibility and centrality models, to data-heavy land use-transport interaction models. Centrality models have been increasingly employed to support spatial planning on the city-scale, where such techniques are attractive due to their low requirements of socio-economic and demographic data, while they also maintain representations of essential features such as accessibility. However, it has been less clear if such approaches can be successfully extended from the urban to the regional scale. In this paper we demonstrate how a recently introduced centrality measure – preferential centrality – can be used as a modelling framework on the multi-regional scale, while retaining high intra-urban spatial resolution. Centrality is calculated on a zonal level, with local plot characteristics and network travel times as input. Preferential centrality is calculated similarly to Google PageRank and eigenvector centrality, but with preferential growth as an additional component that represents local agglomeration processes. To examine the explanatory power of this approach, we compare computed centrality with empirical land taxation values, using the southern half of Sweden as a case study area. Using a static accessibility model as benchmark, we find that the preferential model has a higher capacity to reproduce empirical patterns, with regard to geographical correlations as well as for probability distributions. Our findings suggest that preferential centrality analysis can have practical value in urban and regional planning contexts, for example when assessing the geographical distribution of impacts from transport infrastructure investments.