Modeling urban land use changes in Lanzhou based on artificialneural network and cellular automata

摘要

This paper presented a model to simulate urban land use changes based on artificial neural network (ANN) and cellularautomata (CA). The model was scaled down at the intra-urban level with subtle land use categorization, developed withMatlab 7.2 and loosely coupled with GIS. Urban land use system is a very complicated non-linear social systeminfluenced by many factors. In this paper, four aspects of a totality 17 factors, including physical, social-economic,neighborhoods and policy, were considered synthetically. ANN was proposed as a solution of CA model calibrationthrough its training to acquire the multitudinous parameters as a substitute for the complex transition rules. A stochasticperturbation parameter v was added into the model, and five different scenarios with different values of v and thethreshold were designed for simulations and predictions to explore their effects on urban land use changes. Simulationsof 2005 and predictions of 2015 under the five different scenarios were made and evaluated. Finally, the advantages anddisadvantages of the model were discussed.