Modelling urban spatial structure using Geographically Weighted Regression

  • Noresah M S, Ruslan
  • Published 2009


This paper examines urban spatial structure in terms of urban built-up area of the rapidly developing city of Sungai Petani Malaysia. We developed a model of urban built-up area using Geographically Weighted Regression (GWR) to estimate the strength of the relationship between urban builtup area and factors associated with urban change. Ordinary regression models yield only a single estimate of the relationships. In comparison, GWR allows an estimate of the spatial variation of this relationship. In this study twenty explanatory variables describing access and proximity, neighbourhood, zoning and physical factors, were hypothesized to influence the change in the built-up area and analysed using GWR to allow for spatially varying relationships across the study area. The dataset includes the amount of urban built-up area that has increased from 1992 to 2002 and 20 spatial variables. For the period of 1992-2002, approximately 158 percent of land had been converted to urban use. The spatial variables are generated using geographical information system techniques and calibrated into GWR model. The results of the GWR model are compared to global model. The use of GWR has increased the strength in the relationship especially in terms of the goodness-of-fit statistics (R^2) from 0.29 (OLS global model) to 0.63 (GWR), with individual GWR models ranging from 0.0 to 0.99. Maps of the residuals show that the GWR model fits better in the central region of the study area than the outer region. This is partly due to the better accessibility in the central than the outer region. A Monte Carlo test of the GWR model found that 17 of the 20 explanatory variables displayed significant spatial non-stationarity.

5 Figures and Tables

Cite this paper

@inproceedings{S2009ModellingUS, title={Modelling urban spatial structure using Geographically Weighted Regression}, author={Noresah M S and Ruslan}, year={2009} }