Incorporating spatial autocorrelation and settlement type segregation to improve the performance of an urban growth model. (September 2020)

• Record Type: Journal Article
• Authors: Xu, Chao; Pribadi, Didit O; Haase, Dagmar; Pauleit, Stephan
• Is Part Of: Environment & planning. Volume 47:Number 7(2020)
• Page Start: 1184
• Page End: 1200
• Publication Date: 2020-09
• Subjects: Spatial autocorrelation -- autologistic regression -- Markov chain -- cellular automata -- settlement density; City planning -- Periodicals; Urban ecology (Sociology) -- Periodicals; 307.11605
• DOI: 10.1177/2399808318821947
• Languages: English
• ISSNs: 2399-8083

Abstract

As rapid urbanization and population growth have become global issues, urban growth modeling has become an essential tool for decision-makers to understand how urban growth works in overall dense environments and to assess the sustainability of current urban forms. However, in urban growth models (particularly when incorporating quantitative approaches to include driving factors of urban growth), spatial autocorrelation may influence the overall model performance. In this paper, an empirical study was conducted in the region of Munich, and an integrated urban growth model was tested to explain current urban growth. The modeling contributes to advances in the state of the art by combining a range of driving factors using autologistic regression with a transition probability matrix from the Markov chain method in a cellular automata model simulation. The autologistic regression employed here addresses the impact of spatial autocorrelation compared to ordinary logistic regression. Furthermore, this study compared modeling of overall settlement growth with modeling high- and low-density settlement types separately. Incorporating spatial dependency into the model through application of autologistic regression showed improvements when compared to the ordinary logistic regression model. The Kappa indexes were higher when separately modeling the two types of settlement density compared to modeling overall settlement growth since the driving factors of settlement growth of different densities might be different. From an urban planning perspective, this novel autologistic regression-Markov chain-based cellular automata model is a powerful tool that offers an opportunity for planners and government authorities to gain a more precise understanding of the different urban growth processes that might occur in an urban region similar to the one tested here. It should allow for a better assessment of the potential costs, benefits, and risks of corresponding planning strategies.