Cellular automata urban growth model calibration with genetic algorithms

Last few decades witness a dramatic increase in city population worldwide associated with excessive urbanization rates. This raises the necessity to understand the dynamics of urban growth process for sustainable distribution of available resources. Cellular automata, an artificial intelligence technique composed of pixels, states, neighborhood and transition rules, is being widely implemented to model the urban growth process due to its ability to fit such complex spatial nature using simple and effective rules. The main objective of our work is to use genetic algorithms to effectively calibrate, i.e., identify transition rule values, a cellular automata urban growth model that is designed as a function of multitemporal satellite imagery and population density. Transition rules in our model identify the required neighborhood urbanization level for a test pixel to develop. Calibration is performed spatially to find best rule values per township. Genetic algorithms calibration model, through proper design of their parameters, including objective function, initial population, selection, crossover and mutation, is prepared to fit the cellular automata model. Genetic algorithms start processing the initial solution space, through sequential implementation of the parameters, to identify the best rule values using a predefined criterion over the maximum number of iterations. Minimum objective function, representing the total modeling errors, is used to identify the optimal rule values. Each rule set is evaluated in term of urban level and pattern match with reality. Calibration with genetic algorithms proves to be effective in producing the optimal rule values in a time effective manner at an early generation. Proposed calibration algorithm is implemented to model the historical urban growth of Indianapolis-IN, USA. Urban growth results show a close match for both urban count and pattern with reality.