The functionality of spatio-temporal analysis and modeling is a drive for Geographical Information Systems (GIS) to further applications in various applied fields and Digital Earth plan. However, currently commercial GIS systems lack those capabilities for spatiotemporal distribution, prediction, and simulation of spatiotemporal data. So more effective and powerful modeling methods need to be developed Cellular Automaton Model(CA)provides GIS with a bottom-to-top spatiotemporal modeling frame, which is composed of a fourfold: ceils, states, neighbors and rules. The simplicity and flexibility make CA able to simulate a variety of behaviors of complex systems. In order to meet the needs of spatiotemporal simulation under GIS environment, the standard CA is extended to GIS Context about the concept of cells, system time and transition rules. Based on these extensions, severa1 factors affecting urban evolution are simulated on a virtual city. Then Urban Landose Evolution Simulation and Prediction Model (LESP) was established by integrating GIS with ECA, Finally, to take Baotou City as an example, LESP model was used to simulate Baotou City expansion process. The initial time is 1992 when land use data were acquired by aerial photography. The change data collected by differential GPS are used for model calibration. The appropriate extents of Baotou City development in 1998. 2008 and 2020 are simulated The result of 1998 is consistent with the surveyed data with GPS. Thus, the established model can be used as a virtual lab for decision support in urban planning and land policy-making.