Change Detection - Cellular Automata Method for Urban Growth Modeling

This paper describes an on-going work on the development and implementation of cellular automata based urban growth modeling using multitemporal satellite imagery. The algorithm is designed to simulate the historical growth as a function of local neighbourhood structure of the input data. Transition rules in the algorithm drive the urban growth over time. Calibration is introduced in the cellular automata model. Spatial and temporal calibration schemes are used to improve the prediction accuracy. Spatially, the model is calibrated on a township basis to take into account the effect of site specific features, while the temporal calibration is set up to adapt the model to the changes in the growth pattern over time. Calibration provides the optimal values for the transition rules to achieve accurate urban growth modeling. The paper discusses at the end a proposed automatic rule calibration method using genetic algorithm. The aim is to optimize the transition rule values. Prediction accuracy is selected as the fitness function. A set of strings are used as initial population over which the genetic algorithm runs till convergence. The cellular automata model is tested over city Indianapolis, IN, USA to model its urban growth over a period of 30 years. Besides the land use data derived from the satellite imagery, population density is used. Results indicate good accuracy on a township basis for short term (5 years) and long term (11 years) prediction. The model succeeds in adapting to the dynamic growth pattern. Genetic algorithm shows promising potential in the calibration process.