Urban expansion simulation based on constrained Artificial Neural Network cellular automata model

With development of metropolis, it has been widely accepted the urgent need to simulate and forecast urban growth. In this paper, we proposed a constrained-CA-Urban-model relies on remote sensing (RS) techniques, geospatial process of geographic information systems (GIS) and artifical neural network (ANN). A three-layer back-propagation (BPNN) is set up for acquisition of transition rules in terms of urbanization probabilities for CA model. As an example, urban expansion maps are extracted from three TM satellite imageries (1991, 2001, and 2007) in Chaoyang District at Beijing. Neighborhood, distance, and constrain variables are considered. Evaluation of results indicates this model is effective in simulation for urban expansion. The experiment results in study area demonstrated this model is feasible and convenient for CA simulation and forecast for urban system.