采用双约束空间聚类方法对元胞空间进行分区，在此基础上对不同的分区分别求取元胞转换规则，从而提高 元胞自动机的模拟精度。以杭州市土地利用变化为例，采用本文提出的基于双约束空间聚类的分区元胞自动机模型对 研究区域2000年—2005年的土地利用变化进行模拟，并利用逐点对比法和Moran I指数对模拟结果进行精度评估。结果 表明：(1)采用双约束空间聚类算法对元胞空间进行分区，可以保证同一分区内的元胞既在空间上邻近，又具有相对一 致的非空间属性信息，分区效果较好；(2)与不分区元胞自动机模型和基于空间聚类的分区元胞自动机模型相比，双约 束空间聚类元胞自动机模型具有较高的模拟精度，尤其是在空间形态和整体结构上具有较好的模拟效果。

To overcome the disadvantage of non-partitioned geo-cellular automata (GeoCA) modeling, a partitioned GeoCA based on dual-constraint spatial cluster is discussed in this paper. Comparing with the traditional spatial cluster method, dualconstraint spatial cluster concerns not only spatial distance but also the likelihood of attributes between objects. Under the partitioned GeoCA based on dual-constraint spatial cluster modeling framework, cellular space is departed into several partitions by dual-constraint spatial cluster. And then, a set of transfer rules corresponding to each partition are calculated accordingly. Later on, Hangzhou City is selected as the study area and the Partitioned GeoCA modeling based on dual-constraint spatial cluster is employed to simulate the land use change during the period between 2000 and 2005. The study area of Hangzhou is divided into 5 partitions by using K-Means cluster algorithm according to both spatial distance and attribute likelihood, and C5.0 decision tree is employed to obtain the cellular transferring rules for each partition for the study area. In each partition, corresponding cellular transferring rules are employed to drive GeoCA to run simulation. Finally, confusion matrix and Moran I index are calculated to evaluate the accuracy of simulated results by using the GeoCA model. Results show that the accuracy by using the partitioned GeoCA modeling based on dual-constraint spatial cluster owns higher simulation accuracy compared with those simulated based on non-partitioned GeoCA simulation and the partitioned GeoCA based on the traditional spatial cluster. Besides, when Moran’s I index is employed to evaluate the accuracy of simulation result; the result of partitioned GeoCA modeling based on dual-constraint spatial cluster is closer to the real land use pattern. The empirical study shows that a more accurate simulation result can be achieved by using the partitioned GeoCA modeling based on dual-constraint spatial cluster.