Nerual Network Based Model For Simulating Land Use Changes

Land use and land cover changes are crucial and urgent research subjects in global environment changes research. Simulating and predicting land use changes is the core part in LUCC assessment and local analysis. Thesis puts forward a method to simulate land use changes by using artificial neural network (ANN). The architecture of ANN is the important part in the thesis, such as selecting the method for pretreatment of inputs, confirming the number of hidden layer, selecting the method for initializing the parameters of ANN and the method of selecting study area. Thesis constructs ANN models to simulate land use changes by trial and error on comparing the methods of constructed ANN.Thesis simulates and predicts land use changes in Taian city of two counties Taishan and Daiyue by using constructed ANN. Two RS images (1987,2002) are converted into land use images by using ERDAS. ANN inputs that can affect land use changes are generated by using GIS. The simulation was executed by using the land use in 1987 as initial grid. Simulating land use patterns in 2002. The simulation is quite acceptable with overall accuracy of 76% by using comparing the simulated and actual land uses in 2002.At last, thesis puts forward the characters and limitation in simulating land use changes by using ANN. And puts forward the methods that integrating GIS and ANN.