Modelling urban expansion with cellular automata supported by urban growth intensity over time

The simulation of urban expansion has become an important means to assist urban development planning and ecological sustainable development. However, the spatial and temporal heterogeneities of urban expansion has been a major challenge for modelling urban expansion. This study designed three features from the perspective of spatiotemporal heterogeneity to improve the accuracy of CA model. The new features cover the trends effects of long time-series data on urban expansion, urban spatial growth intensity based on urban growth kernel estimation and allocation probability of the newly generated urban cells from global neighbourhood effects. Finally, urban expansion in Huizhou, China, was simulated and predicted. The experimental results show that the new features can effectively reduce the prediction error for the total amount of urban growth with a deviation of about 2%, and the overall accuracy of urban expansion is as high as 0.93. The features designed in this paper are shown to be effective and can be applied to urban simulations and scenario prediction with various models.