Predicting the expansion of urban boundary using space syntax and multivariate regression model

Abstract

Urban boundaries (UBs) are of great significance for urban planners to constrain urban expansion and protect the surrounding rural landscapes. However, existing studies mainly focus on evaluating land use suitability and modeling land conversion; thus, the prediction of UBs is indistinct and often even failed. The current study presents an urban growth boundary model (UGBM) using space syntax and multivariate regression model. The UGBM is established on the basis of the location of UBs and regards the layout of traffic network as a crucial factor influencing the pattern of UBs. The independent variables of the multivariate regression model are obtained from morphological variables, and the dependent variable is the distance to the UBs. As the morphological variables are highly correlated with the aggregation degree of human activities and traffic flows and an overwhelming majority of human mobility is found inside the UBs, we assume that UBs can be predicted using such variables to extend the UBs from urban physical development to contain the dimension of human mobility and activities. The simulation of UBs in the fast-growing town of Cotton Lake in Guangdong, southern China, was implemented. We compare the UB simulation of the proposed UGBM with a null UGBM without incorporating predictor variables. The results show that the proposed UGBM performs better than the null UGBM using quantity and location metrics of percent area match. We argue that space syntax has a great potential in simulating the expansion of UBs.