Examining urban functions along a metropolitan gradient. A geographically weighted regression tells you more

Urban expansion in advanced countries is the result of inherent transformations in settlement forms and socioeconomic functions leading to new metropolitan configurations. Assessing latent shifts from mono-centric structures to more fragmented, low-density and spatially-decentralized models was complicated by emerging processes of urban sprawl. In these regards, the present study proposes an original approach to the analysis of metropolitan configurations using a Geographically Weighted Regression (GWR) to investigate changes in the spatial distribution of specific urban functions over time. This framework was applied to long-term (1920–2010) evolution of three functions (population density, building density and per cent share of built-up area in total municipal area) in a Mediterranean city (Athens, Greece) moving from a mono-centric configuration to a more dispersed settlement model. Local GWR parameters (adjusted R 2 , intercept, slope, standard residuals) estimated at 4 time points (1920, 1950, 1980 and 2010) constituted the input of a Multiway Factor Analysis reconstructing the intimate dynamics of urban growth in the study area. Local regression slopes and intercepts evidence a non-linear expansion path with the highest spatial polarization in urban and rural districts observed in the early 1980s. Model’s goodness-of-fit increased progressively over time moving from central to peripheral locations. Results of this study contribute to a comparative analysis of the relationship between urban functions and distance from inner cities.