Modelling spatial patterns of urban growth using multi-agent systems
Student name: Mr Bhartendu Pandey
Guide: Dr P K Joshi
Year of completion: 2012
Host Organisation: TERI University
Abstract: This study explores the application of multi agent systems for development of an urban growth
model. The model considers three types of interactions i.e. agent-agent, agent-environment, and
environment-environment. The model used transition potential surface generated using logistic
regression while considering past changes in the built up areas. The transition potential surface
was treated as a dynamic layer in the model. The model was initialized using this potential
surface considering that the transition probabilities are rather dynamic based on preferences of
the agents and population dynamics. The population growth rate was modelled using system
dynamics modeller. The locational preferences of the urban and rural population and the
interactions between the populations alter the transition potential of the surface during the model
run. The change in the potential is largely dependent on the population dynamics. In order to
determine the operational scale of the urban growth process in the study area, fractal analysis
was carried out using Grid Algorithm. The model was finally calibrated while conducting
sensitivity analysis and conducting behaviour space experiments with multiple simulation runs.
The calibrated model with optimum parameter settings was used to simulate urban growth
during the period (1999-2009). The results were validated with observed satellite derived urban
growth data using ROC statistic. The model was further used to predict urban growth for the
year 2019. Spatial metrics namely, Shannon’s Entropy and Aggregation index were used to
determine the potential of the model to capture spatial patterns of urban growth.
Keywords: Agent based model, Logistic regression, Urban growth, Spatial metrics, Fractal
analysis
