地理元胞自动机模型的模拟精度会受到元胞尺度的影响。以杭州市土地利用变化模拟为例,分析了元胞尺度分别为50m×50m、100m×100m、150m×150m和200m×200m时地理元胞自动机模型的模拟精度,对地理元胞自动机模型的尺度敏感性进行了分析;并从元胞转换规则入手,研究了元胞自动机模型尺度敏感性产生的原因:(1)元胞尺度会对地理元胞自动机模型的模拟精度产生影响,元胞尺度越精细模拟精度越高;(2)元胞自动机模型的尺度敏感性与元胞尺度相关,在有些尺度区间上表现得明显,而在有些尺度区间上表现并不明显;(3)孤立元胞是元胞自动机模型尺度敏感性产生的主要原因。研究表明,随着元胞尺度的增大,元胞空间的孤立元胞增多,这些孤立元胞本身及其周围元胞具有较低的邻域函数值和较小的转换概率值,并影响了地理元胞自动机模型的模拟精度。

Geographical Cellular Automata (GeoCA) is an efficient model to simulate dynamic geographical process, which has been illustrated by a great number of researches. Surprisingly, there are few researches discussed in detail on the scale sensitivity of cell size during the geo-cellular automata modelling process. Among existing researches using GeoCA, cell size is always decided by the resolution of geographical data source. Whether cell size will affect the simulation results of GeoCA or not is a hotspot. If it does, what is the causality of scale sensitivity of GeoCA? How does cell size affect simulation result of GeoCA? And how can we choose a suitable cell size during geo-cellular automata modelling? Few researches have achieved answers for these questions. In order to figure out effects on simulation result of GeoCA caused by cell size and its mechanism, and to provide a principle to choose cell size for GeoCA, we take land use change simulation of Hangzhou City as a case, and choose cell sizes of 50m×50m, 100m×100m, 150m×150m and 200m×200m to analyze the scale sensitivity of GeoCA. And then, by analyzing transfer rule, causality of scale sensitivity is discussed in this paper. The research result shows that (1) simulation results of GeoCA are sensitive to cell size, and the more fine the cell sizes are, the higher the accuracy of the simulation results; (2) scale sensitivity of GeoCA is not uniform in all cellular sizes, it is sensitive in some range of cellular size while it is insensitive in other ranges; (3) scale sensitivity of GeoCA simulation results is by large caused by increasing isolated cells. These isolated cells and cells around them have lower neighborhood function value, and then their transfer probabilities are also lower. With the increase of cell sizes, the number of isolated cells in cellular automata space increases. Because of lower transfer probability of these isolated cells, these increasing isolated cells caused higher error rate of simulation results of GeoCA. So, the number of isolated cells is the main cause for scale sensitivity of cell sizes when we do the geo-cellular automata modelling.