Several lessons about the process of calibration were learned during development of a self-modifying cellular automaton model to predict urban growth. This model, part of a global change research project on human-induced land transformations, was used to predict the spatial extent of urban growth 100 years into the future. The context of the prediction was to evaluate urban environmental disturbances such as land use conversion, urban heat island intensification, and greenhouse gas generation. Using data for the San Francisco Bay area as a test case, methods were developed, including interactive and statistical versions of the model, animation and visualization tools, automated testing methods, and Monte Carlo simulations. This presentation will enumerate, analyze, and discuss the lessons learned during the extensive process of model calibration. Experience with the methods developed may have broader use in assisting the rigorous calibration for other CA models, and perhaps those coupled environmental models with an extensive spatial data component. These methods are now under test as the project moves to a new data set for the Washington, D.C.-Baltimore area.