基于CA模型的城市边缘区土地利用演变模拟——以广州市花都区为例

研究目的：探索土地利用的多地类CA模拟方法,并掌握城市边缘区土地利用变化规律。研究方法：选取典型城市边缘区 —— 广州市花都区为研究区域,利用C#语言结合ArcEngine GIS平台编程进行花都区土地利用演变CA模拟研究。以不同时间研究区土地利用图为基础数据,比较研究期内各地类的变化数量与方向以确定地类转换之间的优先级,并确定各地类的转换概率阈值。然后利用蒙特卡罗方法结合控制因素进行判断,最终确定元胞的转化状态。研究结果：模拟结果表明,2000 — 2005年间,新增建设用地分布在除北部山区以外的所有区域,但主要集中在城市中心区域,农用地主要分布在西南部和东北部,而且破碎化程度越来越高。在流向上,仍然有不少数量的农用地（主要为耕地和林地）转向建设用地。与实际情况相比,模拟的数量精度为84.8%,位置精度为71.3%。研究结论：研究结果表明此方法便于理解与操作,同时模拟精度较高。

Simulation of the Land Use Evolution in Urban Fringe Based on CA Model: A Case Study of Huadu District, Guangzhou City

The purpose of this paper is to explore the methods of CA simulation regards to multi-land types and to grasp the essential rules of the land use change in urban fringe. As the typical urban fringe area, the land use structure evolution in Huadu District in Guangzhou city was simulated via CA model based on the integrated programming of C# Language and ArcEngine GIS platform. The steps of simulation included: (1)to compare the quantities and directions between each type of land use change in order to assess the priority among various land changes and to set the probability threshold based on the current land use map of the Huadu District in different periods; (2 )then that tosimulate the cellular state, i.e., changed or not, can be determined via the judgment based on Monte Carlo method combined with related factors controlling. The results indicate that the stimulation of the land use change from 2000 to 2005 showed that the newly built-up construction areas were distributed in most areas except the northern mountain area; furthermore, they were gradually aggregated in the central area of the city. The agricultural lands were mainly distributed in northwest and northeast areas, and the degree of farmland fragmentation became even higher. In terms of the direction of land use change there were still some conversions from agricultural use to construction use, which mainly are cultivated land and forest land. Comparing with the reality, the accuracies of the stimulation are 84.8% in terms of amount and 71.3% in terms of location. Thus, it is concluded that the proposed method in the paper is easy to be understood and feasible to be implemented, meanwhile, the accuracies of simulation are relatively higher.