A cellular automaton-based model was developed to predict patterns in the spread of rural residential development in 80 km2 of the rural countryside near Toronto, Canada. The model was executed on a computer-based geographic information system (GIS), utilizing a cell size of one 1 ha and a grid of 80 by 100 cells. The rules of the model were derived from literature concerning planning policies, and from the environmental and social conditions in Puslinch Township, Wellington County, Ontario. Operating within the time period 1955-1983, the model was run in two scenarios: (1) a static set of rules based on conditions in 1955; and (2) rules that changed as conditions or policies within the township changed. Both scenarios were compared with measured data. Scenario 1 reproduced the tendency for houses to be developed in high and medium density clustering, but these clusters did not follow the same spatial patterning of the measured data. Scenario 2 resulted in somewhat less clustering of houses, but the spatial patterns of those clusters were similar to measured data. The model demonstrated some replicative and predictive validity, and possesses strong structural validity. It has the potential to be run "into the future" to predict the outcome of policy decisions. © 1993 Academic Press. All rights reserved.
