Urban Land Use Nonlinear Structure And Its Evolution

Urban land use structure and its evolvement is one of the main contents of urban spatial structure study. In this paper, based on the analysis of study progress in domestic and foreign countries and the non-linear theory, Fractal, Cellular Automata and Artificial Neural Network are integrated to set up the Fractal simulation modelof urban land use structure, the Self-Organizing Maps Network model of Urban land use variation and Artificail Neural Network model of urban land use structure based on the CA (LCA-ANN). With these models, the land use static structure, dynamic evolvement and prediction of the central downtown of Shanghai and the chosen sample area are made to gain more evaluable conclusions.In chapter 1, the background of the study is mentioned, and the study progress of urban land use non-linear in domestic and foreign countries is introduced. In the end, the advantage and weakpoint of these models are summarized.In chapter 2, the spatio-temporal database is set up, and the way how to build the spatio-temporal database is explained, such as the data source of this paper, the enhancement of the SPOT image, the coding of the land use and the digitalization of the satellitic image and so on.In chapter 3, based on the fractal map of the land use, three fractal dimensions of land use spatial structure, which are boundary dimension, radial dimension and information dimension, are summarized. Then the central downtown of Shanghai is chose as the study object to calculate the dimensions of land use spatial structure. The result shows that these three dimensions are different. They reflect respectively three aspects of the land use spatial distribution, which are complexity, centrality and equality.In chapter 4, Self-Organizing Map Network Cluster Model about urban land use variation is set up to study land use variation. With the two-period data of the chosen sample area, the land use variation pattern of sample area is studied. The result shows that SOM model can well reflect the variation law of sample area land use, and realize the spatial display of the land use variation law. In the end, author summarized the determining principle of Network structure in this model.In chapter 5, to simulate and predict land use structure, CA and ANN are integrated to set up the LCA-ANN model. In this model, land use cellular spatio-temporal database is built. ANN is used to learn the transfer rule and accomplish the prediction of land use structure. With the two-period data of the chosen sample area, the prediction of the land use structure in 2006 is made. The result shows LCA-ANN model can well achieve theprediction of the land use structure, but this model has a bug of information waste. In the end, the improve measure is also pointed out.In chapt 6, author summarized the study acheivement and discussed the models, which are used in this paper.