Applications and middleware popular in the study of wireless sensor networks (WSNs) often involve monitoring dynamically changing environments. As the complexity and scale of these applications increase, so does the need for effective comparative analysis. Deterministic benchmarks for head-to-head comparisons, as well as stochastic tests modeling the unpredictability of environmental phenomenon are needed. Approaches employed for ad hoc network simulation are not sufficient, since the network is typically modeled solely in terms of network topology, without a separate model for the physical environment within which the network is deployed. We propose a simulation architecture in which various cellular automata models representing dynamic physical environments can be developed. Our architecture allows environment scenarios to evolve independently of simulation models for network protocols and topology. To verify the architecture, we implement cellular automatons to model different physical systems: the FHP rule describes the motion of particles traveling in a discrete space colliding with each other; the OFC rule describes stress loading and rupture cycles; and a mathematical CA model describing the spread of a fire, incorporating weather (wind) and land topology conditions. We use these rules to effectively model scenarios such as spreading of gas, earthquakes, bridge or building rupture, and forest fires.