Multi-scenario based urban growth modeling and prediction using earth observation datasets towards urban policy improvement

Abstract

Urbanization is a growing challenge for city planners and policymakers who are continuously focusing on computer-based statistical models, and machine learning for a sustainable and livable city. The main objectives of this article were to develop a robust artificial intelligence-based hybrid geo-simulation model to support multi-scenario urban growth modeling for urban policy improvement. In this study, earth observation datasets, Artificial Neural Network-Multilayer Perceptron coupled with Markov Chain (MLP-Markov) and Cellular Automata and Markov Chain (CA-Markov) were applied and the best performance was measured for urban growth modeling. The result shows that the urban land use was 25.79, 31.40, 45.19, 89.22 and 147.96 square km in 1971, 1981, 1991, 2001 and 2011 which has been predicted for 2021, 2031, 2041 and 2051 based on the planned and unplanned development scenarios. The predicted urban land use of the planned development scenario is 242.10, 312.69, 363.80 and 400.72 square km while 242.91, 314.31, 366.23 and 403.98 square km of the unplanned development scenario during 2021, 2031, 2041 and 2051. The uncertainty result shows that overall agreement (84.99%) and other indices are higher, and disagreement is lower (15.01%) for MLP-Markov than the CA-Markov for the urban land use prediction. The hybrid geo-simulation models were tested over multiple urban planning indicators to understand urban growth patterns and related scenarios. The result shows that the geo-simulation model is extremely sensitive to the complex pattern of urban growth and disperse indicators over space and time. This study provides a promising guideline for urban planners and conservation scientists to implement a robust artificial intelligence-based hybrid geo-simulation model for compact, organized, and integrated land use-transportation development.

Topics & Concepts