Spatial planning-constrained modeling of urban growth in the Yangtze River Delta considering the element flows

Abstract

ABSTRACT Urban growth models that consider spatial control and factor flows are important for the integrated management of urban agglomerations, and to date, no simulation method has been available to consider both thoroughly. We proposed a new cellular automaton (CAGBDT) model based on gradient boosting decision trees (GBDT) to simulate urban growth in the Yangtze River Delta (YRD) area of China. In this model, GBDT was used to retrieve land transition rules by integrating multiple weak learners and a feature expansion approach was employed to remove high correlations among urban growth drivers and then expand the factor features. We applied the final urban pattern as the dependent variable and selected nine drivers as the independent variables of urban growth in YRD. The simulation results show that the overall accuracies exceeded 89% and the figure-of-merits (FOMs) exceeded 27%, about 10% higher than other similar simulations in YRD, indicating the strong ability of CAGBDT to simulate urban growth in YRD. Although the inclusion of inter-city material and information element flows in the model improves the accuracy by only 1%, it reveals the different development patterns in Hangzhou Bay in the south of YRD and the Taihu Lake basin in the center of YRD. By considering urban scenarios under different strategies of spatial control, it shows that the simulated FOMs declined by 5% with the stronger enforcement of spatial regulations, reflecting that the actual urban growth in fully non-developable areas and 20% partly developable areas of YRD has already violated the regulations. The results can help urban planners and local authorities to develop solutions for the development of a high-quality YRD urban agglomeration. The proposed model is applicable to diagnose urban land-use change elsewhere, especially in rapidly developing cities that need balancing urban growth and ecological protection.