Urban expansion in 30 megacities of China: categorizing the driving force profiles to inform the urbanization policy

Currently, land use policy makers in developing countries face the challenge to control urban expansion in a rational and sustainable manner. In this context, understanding the urban expansion process and its determinants is essential for designing land use planning and governance strategies. This paper presents a methodological framework based on data mining techniques for exploring the determinants of urban expansion. We employ the Random Forest regression to apportion the relative importance for 33 determinants of urban expansion across 30 megacities in China. Analysis is facilitated by a dataset of urban land information in period 1993-2012 (derived from time-series nighttime light imageries) as the independent variable, and an equivalent dataset of 33 determinants of four categories (economic, demographic, social and natural) as exploratory variables. Results show that the relative importance of determinants varies with cities but shows some similarities. Greatest contribution is observed for the economic determinants, followed by social determinants. Demographic and natural determinants have comparative relative importance. In general, economic, social, and demographic determinants have increasing contribution to urban expansion, but the natural determinants show declining influence on urban expansion with time. Four clusters are identified by a Self-Organizing Map among the 30 megacities, with respect to the driving forces profiles. Urban expansion in these four clusters is respectively dominated by demographic, social, natural, and economic determinants. Based on these profiles, the land use planners can revised the original 'top-down' land-driven development model and formulate more locality-oriented measures to control urban expansion in an orderly manner.