Integrating logistic regression with ant colony optimization for smart urban growth modelling

Abstract

Urban growth does not always strictly follow historical trends; the government may reshape urban growth patterns with considerations of ecological conservation or other plans. Both urban dynamic rules and landscape characteristics are the two main factors influencing the spatial patterns of cities, and obtaining an optimized spatial pattern is very important for sustainable urban growth. Therefore, in this study, we integrated logistic regression (LR) with the ant colony optimization (ACO) model to analyze the optimal scenario for smart urban growth. The LR model was used to discuss the relationship between urban patterns and environmental variables such as topography, development centers, and traffic conditions. Then, the urban growth probability was generated using the parameters obtained from LR. The ACO model was further integrated to optimize urban land allocation, which can meet the requirement of high growth probability, and a connected and compacted landscape pattern. This can solve the problem of urban land only being allocated by LR from being distributed fragmentarily in the space. With this integrated model, Guangzhou City, a rapidly developing area in China, was selected as a case study. The urban patterns derived from LR, as well as a simulation scenario using logistic regression-based cellular automata (LR-CA), were used in the comparison. Six landscape metrics were chosen to validate the performance of this proposed model at the pattern level. The results show that the LR-ACO model has a better performance in urban land allocation. This study demonstrated that models that couple dynamic rules and planning objectives can provide plausible scenarios for smart urban growth planning.