基于近邻传播聚类元胞自动机模型的武汉城市扩散和聚合过程同步模拟

元胞自动机（CA）作为城市时空动态模拟应用最广泛的模型,可以有效模拟填充式和边缘式城市扩张过程,但是在飞地式扩张模拟方面稍显不足。本文提出一种改进CA模型—APCA,在传统CA基础上利用近邻传播聚类（AP）搜寻城市扩散增长的“种子点”,实现城市增长扩散过程和聚合过程的同步模拟。以武汉市为研究区域,使用APCA模拟其在2005—2025年间城市扩张的时空过程。结果显示：① APCA在设置“种子点”数量为1~8个时模拟总体精度均高于Logistics-CA,当“种子点”数量为6时,模拟新增部分精度最高,达到0.5217;② 2015—2025年武汉市飞地型增长面积约为8.67 km²,占新增城市用地总面积比例为6.30%;③ 武汉市1995—2025年间“先扩散后聚合”的城市扩张过程符合城市增长相位理论。APCA在一定程度上了完善了传统二维平面CA框架,将城市扩张模拟维度由面维扩展到点维,为准确展现城市用地空间扩展规律提供参考。

Synchronized simulation of urban diffusional and aggregational process based on the affinity propagation cellular automata:A case study of Wuhan city

Cellular automata (CA) has been widely recognized as an effective approach in the simulation of spatiotemporal dynamics of metropolitan areas, particularly for infilling and edge urban expansion processes. However, the traditional Logistics-CA has its severe drawbacks in simulating outlying expansion, since it evolves primarily according to the status of a set of neighboring cells, failing to identify other potential seeds that could also stimulate urban expansion in a significant way. This paper develops an advanced CA, called APCA, by using Affinity Propagation (AP) to comprehensively search for urban expansion seed points, as well as by realizing a synchronous simulation of diffusional and aggregational processes of urban cell. This paper uses the data of Wuhan, the largest provincial capital in central China, to validate the effectiveness of APCA. By simulating Wuhan's urban expansion dynamics between 1995 and 2025, the APCA (1) identifies that the total areas of outlying expansion amount to 8.67 km², accounting for 6.30% of added urban land of the city; and (2) successfully simulates a process of "first diffusion and then aggregation" in Wuhan, which is in harmony with the Phase Theory of urban expansion. Compared with the traditional Logistics-CA, the overall accuracy of APCA remains higher regardless of the number of seed points as long as they are within 1-8, while the accuracy of APCA reaches its highest (0.5217) when the seed point is set as 6. The APCA contributes to the two-dimensional CA framework by expanding surface-dimension simulation to point-dimension simulation, and thereby facilitates effective and accurate simulations of urban expansion patterns.