元胞自动机(CA)是一种时间、空间、状态都离散,空间的相互作用及时间上的因果关系皆局部的网格动力学模型,其“自下而上”的研究思路,强大的复杂计算功能、固有的平行计算能力、高度动态以及具有空间概念等特征,使得它在模拟空间复杂系统的时空动态演变方面具有很强的能力.在城市空间动态变化的模拟研究方面,CA模型已应用到除非洲、南极洲的所有大洲的城市模拟研究当中.CA模型和GIS的集成,一方面增强GIS的空间模型运算及分析能力,另一方面,GIS提供的强大空间处理能力可以为CA模型准备数据和定义有效的元胞转换规则以及对模拟结果进行可视化.同时CA模型还可以与神经网络、主成分分析、遗传算法、模糊逻辑以及其他研究方法相结合,以增强其在城市空间变化模拟研究方面的能力.将CA与MAS技术相结合,建立一个能够模拟多个不同参与因子(自然系统)、不同决策者(人文系统)共同影响下的城市发展模型,以此来模拟与预测城市发展的真实状况,将是CA模型在城市空间变化模拟与预测研究中的未来发展趋势.
Cellular automata (CA) is not only a discrete system in time, space and state, but also a local grid dynamics model of both spatial interaction and consequence in time. Because of its "bottom-up" research method, powerful complex computation function, inherent parallel computation capability, high dynamic feature, as well as spatial concept, it has powerful capability in modeling the spatio-temporal dynamic evolution of the spatial complex system. CA is regarded as a spatio-temporal dynamic method, and has capability of modeling two dimensional spatial evolution process. It is therefore widely used in many fields of geography. In the aspects of modeling research of urban spatial dynamic change, CA is used in urban modeling research in all seven continents except Africa and Antarctica. Particularly, CA is used in the urban growth modeling by more and more scientists after the 1990s. The integration of CA and GIS, on the one hand, can well enhance the spatial model computation and analysis capability of GIS. On the other hand, the powerful spatial processing capability of GIS can prepare necessary data and define valid cell transformation rules for CA and visualize the modeling results. In addition, some researches indicate that CA can be directly developed by raster GIS. Although the urban spatial change can be modeled by the integration of CA and GIS, each scientist in this field has different integrated methods. CA can be combined with neural network analysis, main component analysis, genetic algorithm, fuzzy logic, and other analysis methods as well, which make its capability of urban spatial change modeling more powerful.