This paper deals with the design of an efficient object model for propagation phenomena. It is applied to the phenomenological model developed at the University of Corsica, within the context of simulation of vegetation fires. The objective is to simulate large-scale fire propagation, and on the longer term to develop a decision aid tool to guide forest firemen and managers. Based on both cellular automata and discrete event specification (DEVS) formalisms, a new kind of model, called active-DEVS, is specified. Modeling methods based on enhanced cellular automata facilitate both spatial dynamic expression of propagation phenomena, and parallel architectures exploitation. However, such environments usually lack the ability to integrate easy component modifications. The DEVS formalism makes it possible to exploit the cellular models efficiently whatever their dimensions, and to reduce simulation times considerably. A simulation framework is developed to implement and compare active-DEVS model and classical discrete time system specification (DTSS) models. This framework relies on designs patterns, and thus keeps a modular, elegant and adaptable design.