Modelling urban sprawl with the optimal integration of Markov chain and spatial neighborhood analysis approach

Abstract

The Markov chain method has been applied to develop dynamical models for land use patterns from the point of time series since early time. Later, spatio-temporal transition models were established to study the spatial change in the land use by taking the spatial information into account. However, the methods of the spatial neighborhood effect and the appropriate number of neighbors for spatial analysis are still under research. Our effort is to find the best way to study urban sprawl by integrating the time series approach and the neighborhood effect. In this article, two spatio-temporal models were developed, by combining weighted distance approach and direct neighborhood approach with Markov chain approach respectively. As a case study, based on classified TM images in 1995, 1996, 1997 and 2001, we simulated land use transition of Shunyi Country near to Beijing City in China with both two simulation models using 8, 48, 120 and 224 neighbors respectively in 1996 and 1997. Comparing the simulated images with classified images, the results showed the spatio-temporal model with 48 or 80 neighbors of weighted distance neighborhood approach is the best for modeling urban sprawl.