【摘要】 城市扩展模拟可为城市可持续发展与国土空间规划提供参考。智能体模型(ABM)与元胞自动机(CA)结合可兼顾城市空间增长的自组织性和不同决策主体的决策过程,人工神经网络(ANN)可描述智能体与城市扩展之间复杂的非线性关系。该文基于ANN-ABM-CA耦合模型,在构建CA转换规则时基于ABM刻画人类决策行为的影响,并采用ANN挖掘不同类型的智能体在城市扩展过程中的偏好差异,同时考虑宏观和微观层面的智能体决策行为,结合城市扩展的10个驱动因素,模拟武汉市主城区2005-2015年的扩展情况,结果表明:1)相比传统的ANN-CA模型,ANN-ABM-CA模型模拟性能更优,从宏观与微观相结合的角度更好地解释了城市扩展的驱动机制,OA值为97.46%,Kappa系数为0.9176,FoM值为0.4375,结果可靠且合理;2)不同收入层级的居民智能体对城市扩展的决策偏好不同;3)武汉主城区城市扩展模式主要为边缘型扩展,洪山区西南部有少部分填充型扩展、东南部出现飞地型扩展,与实际扩展情况相符。更多还原

【Abstract】 The human decision-making behavior plays an important role in urban expansion process.However, it is often ignored in the cellular automata(CA) modeling of urban expansion.To overcome this limitation, this study adopts the agent-based model(ABM) to characterize the effects of human decision-making behaviors and couples it with the artificial neural network(ANN) to derive the transition rules of CA models.The ANN-ABM-CA model has the ability to couple the self-organization and human decision-making behaviors in the simulation of urban expansion process, and can provide supports for the sustainable development of the city.Meanwhile, the decision-making behaviors of macro agent(government) and micro agent(residents from three income levels) have been built, which can explain the driving mechanism of urban expansion in a better way.Then the urban expansion of main urban area of Wuhan from 2005 to 2015 was simulated with 10 driving factors.The results show that: 1) The overall accuracy(OA) value is 97.46%,Kappa coefficient is 0.9176 and the figure of merit(FoM) value is 0.4375,the accuracy of simulation has been significantly improved comparing with the traditional ANN-CA model.2) Residents of different income levels have different development preferences for urban expansion.3) The urban expansion pattern of Wuhan′s main urban area in simulation results is mainly marginal expansion, with a small part of filling expansion in the southwest of Hongshan District and enclave expansion in the southeast of Hongshan District, which is consistent with the actual expansion situation.更多还原