Modeling Land Conversion in the Colombo Metropolitan Area Using Cellular Automata

Abstract This paper proposes a Cellular Automata (CA) model to evaluate the urbanization patterns arising from the regulation of urban growth on paddy lands in the Colombo Metropolitan Region (CMR). Most of the historic map data available for the CMR before 1990 are temporally sporadic and spatially incomplete. As an alternative to maps, classified remote sensing data are used to analyze the urbanization process. Logistic regression is applied to derive factors of urbanization and the various relationships among them. The relation between ′urban′ and ′non–urban′ serves as an explanatory variable. The factors explaining that relationship are calculated by exploratory logistic regression analyses. The probability calculated from the statistical model is used for CA transition with a random number. Several growth patterns are simulated based on a range of transition thresholds to test the CA model. Status quo growth and several growth control scenarios are simulated for the period from 1987 to 2002 based on an optimum threshold. The simulation result of the status quo growth is evaluated with several evaluation methods. The level of agreement between the estimated result from the status quo model and the actual data is 62%, while the multi–scale goodness–of–fit method produces highly accurate values for the given range of resolutions.