An urban growth boundary model using neural networks, GIS and radial parameterization: An application to Tehran, Iran

Urban growth boundaries (UGBs) are common tools employed by planners to constrain urban expansion in order to increase density of urban services and protect surrounding rural landscapes. Planners could use models that estimate future urban growth boundaries based on those factors that drive urban expansion. Unfortunately, few models have been developed that simulate the urban growth boundary. This paper presents an urban growth boundary model (UGBM) which utilizes artificial neural networks (ANN), geospatial information systems (GIS) and remote sensing (RS) to simulate the complex geometry of the urban boundary of Tehran, Iran. Raster-based predictive variables are used as inputs to the ANNs parameterized using vector routines. ANNs were used to train on seven predictor variables of urban boundary geometry for Tehran: roads, green spaces, slope, aspect, elevation, service stations, and built-area. We show that our UGBM can predict urban growth boundaries with urban area with 80-84% accuracy. The model predicts urban boundaries in all cardinal directions equally well. We use the model to predict urban growth to the year 2012. We summarize the use of UGBs in planning around the world and describe how this model can be used to assist planners in developing future urban growth boundaries given the need to understand those factors that contribute toward urban expansion. (C) 2010 Elsevier B.V. All rights reserved.