Abstract: In recent years, the study of rocky desertification is often completed by GIS technology, however, GIS has weak point of the dynamic simulation evolution of the rocky desertification. Cellular Automata（CA） has stronger capability of dynamic space simulation. Therefore, based on the CA theory, two CA models are set up respectively by means of standard Moore neighborhood and extended Moore neighborhood with the support of GIS technology. The rocky desertification models are run respectively by the data of Guangxi Du’an between 1988 and 1999. The models’ parameters are adjusted and the results are inspected through comparing the simulation results with the actual data. Results show that two kinds of CA models can reflect the actual trends in rocky desertification, and the CA model with extended Moore neighborhood has higher accuracy and reliability than the CA model with standard Moore neighborhood. This method of rocky desertification on prediction is an effective method. Finally, the rocky desertification’s status of 2012 is forecasted by the CA model with extended Moore neighborhood.

Key words: Cellular Automata（CA）, GIS, Karst rocky desertification

摘要： 近年来，石漠化的研究常借助于GIS等相关技术来完成，但GIS在石漠化的动态模拟演化方面存在不足。而元胞自动机（CA）具有强大的空间动态模拟功能，鉴于此，在GIS技术的支持下，以CA理论为基础建立了分别采用Moore型和扩展Moore型邻域的CA模型，以某地1988和1999年石漠化数据为始端数据运行模型，通过将模拟结果与实际数据进行比较，对模型进行参数调整和预测检验。实验结果表明：两种模型预测的结果均可以基本反映实际石漠化发展趋势，而采用扩展Moore型邻域的CA模型要比采用标准Moore型邻域的CA模型具有更高的准确度和可信度；用此方法对石漠化进行演化预测是有效的。采用扩展Moore型邻域的CA模型对2012年该区石漠化状况进行了预测。

关键词: 元胞自动机（CA）, 地理信息系统（GIS）, 喀斯特石漠化