This paper presents a new method to simulate complex land use systems by integrating support vector machine（SVM）

cellular automata

and GIS.Recently

cellular automata（CA） have been increasingly used to dynamically simulate urban growth and land use.There are many issues that should be solved in the simulation of this type of complex systems.One major problem is how to define transition rules using training data.Linear boundaries are often used to retrieve transition rules which define the probability of state conversion.However

many geographical phenomena are very complex and transition rules should be defined using nonlinear boundaries.In this study

a CA model based on the support vector machine（SVM） is developed using Visual Basic and ArcObjects of GIS.The GIS provides both data and spatial analysis functions for constructing SVM-CA model.Training data is conveniently retrieved from remote sensing and GIS database for calibrating and testing the model.The SVM method is used to transform the data from nonlinear boundaries in the original space to linear boundaries in the Hilbert space.The nonlinear transition rules can then be defined by using the functions of SVM.The SVM-CA model can be applied to the simulation of urban development.Complex global patterns can be generated from the local interactions with the SVM-CA model.This paper demonstrates that the proposed model can overcome some of the shortcomings of the existing CA models in simulating complex urban systems by using the nonlinear transition rules.The model has been successfully applied to the simulation of urban development in Shenzhen city of the Pearl River Delta.