Geographic Information Science (GIS) has provided the methodological and technical supports for modeling and simulation in the geographical domain. However, research methods on building complex simulations in which agents behave and interact in discrete time and space are lacking. The existing simulation systems/software are application-oriented and do not provide a theoretical (conceptual) view. The simulation theories and methods that exist do not incorporate spatial issues, which are the key to linking GIS with simulation theory and practice. This paper introduces a method for developing a conceptual theoretical framework for a spatial simulation system which can potentially be integrated with GIS. Firstly, based on classical discrete event simulation and fresh agent technology, a simulation theory is proposed, which is represented by a conceptual simulation model using UML-based visual syntax. In this theoretical framework, spatial issues including spatial setting, spatial constraints, spatial effects and spatial awareness are emphasized. Next, a testing scenario in the microscopic traffic simulation domain is set up to examine the feasibility of the simulation philosophy. Finally, the method is evaluated from the aspects of feasibility, uncertainty and applicability.