Cellular Automata Urban Growth Simulation and Evaluation - A Case Study of Indianapolis

The objective of this paper is to develop and implement a Cellular Automata (CA) algorithm to simulate urban growth process. Indianapolis city, Indiana is selected as a case study to simulate its urban growth over the last three decades. Urban growth simulation for an artificial city is carried out first. It evaluates a number of urban sprawl parameters including the size and shape of neighborhood besides testing different types of constraints on urban growth simulation. The results indicate that circular-type neighborhood shows smoother but faster urban growth as compared to nine-cell Moore neighborhood. Also, CA rules definition is critical stage in simulating the urban growth pattern in a close manner to reality. Next step includes running the developed CA simulation over classified remotely sensed historical imagery in a developed ArcGIS toolkit. A set of crisp rules are defined and calibrated based on real urban growth pattern. Uncertainty analysis is performed to evaluate the accuracy of the simulated results as compared to the historical ground truth images. Evaluation shows promising results represented by the high average accuracies achieved. The average accuracy for the predicted growth images 1987 (for a period of 5 years) and 2003 (for a period of 10 years) is over 80 %. Modifying CA growth rules over time to match the growth pattern changes is important to obtain accurate simulation. This modification is based on the urban growth relationship for Indianapolis over time as can be seen in the historical images. The feedback obtained from comparing the simulated and ground truth images is crucial in identifying the optimal set of CA rules for reliable simulation and calibrating growth steps.