关键词: 土地利用变化, 元胞自动机, 尺度敏感性, 正交试验设计, 遥感图像处理

Abstract:

Land use change is a core field of the global environmental change and it has turned into worldwide research focus. Accurate prediction of land use change is of great importance from the standpoint of theory and practice. Cellular Automata (CA) model has unexampled advantages in creating spatial modeling approaches for simulating land use changes, which has been applied and proved by a great number of researches for decades and all achieved satisfactory effects. However, along with the attention of scale effect in geoscience, it is found that the scale effect in CA model is reflected on the transition rule, the cell size, the neighborhood size and the neighborhood type. That is to say, CA model is sensitive to these four factors. Different scales of these four factors work together and remarkably impact the accuracy of the results simulated by CA model. Among those existing researches analyzing the sensitivity of CA model in land use change simulation, single factor is always the research priority. But they are just far from comprehensive and the research results are not very precise since all these factors, without question, will affect directly to the performance of CA model. What is more, the complex relationships between those factors need to explore as well. To cure the above problems, an orthogonal experiment design method is proposed to analyze the scale sensitivity of CA model to these influential factors. What is more, the existing problems of the correlation between these factors were proved for the first time in the article. Under this scheme, Wuhan, the capital of Hubei Province was taken as an example to analyze the sensitivity of CA model by comparing the difference of the simulated results. Using the special method of range analysis in orthogonal experiment design, it helped to find the influence of different factors to the Cellular Automata model and the optimum combination of factors and also test whether there exist some correlation between any two factors of the cellular automata model. Specifically, when it comes to the sensitivity with which CA model reflects those factors, the sequence decreasingly was the transition rule, the cell size, the neighborhood size and the neighborhood type. In addition, these factors and their respective optimum level together make up the optimum combination of factors to CA model. Besides, the results also revealed that there existed remarkable interactions between the transition rule and the cell size, and between the transition rule and the neighborhood size. In general, this study provides a comprehensive basis for scale sensitivity analysis of CA model in land use change simulation and fills the gaps of the existing methods in exploring this problem.

Key words: land use change, Cellular Automata model, scale sensitivity, orthogonal experiment design, remote sensing image processing