元胞自动机在城市模拟中的不确定性研究

he issues of data errors, error propagation and model uncertainties are important but often neglected in urban CA models. This paper has examined and addressed some of these issues by carrying out experiments with GIS data. Many model errors are related to model configurations, i.e. how to define a proper model to reflect the real process of urban development. This study demonstrates that some of them, however, are quite unique to CA: 1) data source errors will be reduced during simulation because of the averaging effects of neighborhood functions; 2) simulation errors will decrease with time because the availability of land suitable for urban development will be decreased in constrained urban CA as the urban areas grow in size; 3) the number of time steps (iterations) can cause different spatial patterns and simulation closer to actual development can be achieved with the increase in time steps; and 4) the major uncertainties of simulation are mainly found at the edge of simulated urban areas. These characteristics are quite different from those of general GIS modeling. The study shows that errors and uncertainties of urban CA are less severe than what one would normally expect from a simulation model. The uncertainties of the simulation will be reduced if more amounts of land are developed and the uncertainties are mainly located at the urban fringe. The findings of the study can help urban modelers and planners to understand more clearly the characteristics of errors and uncertainties in urban CA so that they can be used more effectively in urban planning.