Development and Application of Geographical Cellular Automata Modelling Framework Based on GIS
FENG Yong-jiu,TONG Xiao-hua,LIU Miao-long,HAN Zhen
2010-01-01
Abstract:In this paper,a novel GIS-based geographic CA modelling framework,SimUrban,is proposed to simulate and project urban expansion and development.Based on object oriented programming technology,the framework is developed by using Visual Studio.NET in GIS environment,which enables the wide usage of RS images and GIS data.The proposed framework composes three parts:1)data acquisition,using SimUrban,ArcGIS,and Envi,the vector topographic maps and RS images are integrated;2)new transition rules of CA,such as partial least squares regression method,least square support vector machines,particle swarm optimization,kernel principal component analysis,and kernel logistic regression method,are allowed to imbedded;3)using this framework,urban expansion and land use/cover changes are simulated with assessment of the results by the error matrix method.The urban development and growth of Jiading District,Shanghai,from 1989 to 2006 was successfully simulated in SimUrban with the CA model based on principal component analysis(PCA).Simulation accuracy of non-urban category is 82.04%,urban category is 70.23%,the overall accuracy is 76.55%,and the Kappa coefficient is 52.58%.The implementation of Jiading District has demonstrated that the modelling framework can generates patterns match well with the actual urban morphology.