As autonomous systems become ubiquitous, assurance that their designers have considered all of the relevant "edge-cases" is ever more important. Many global projects seek to address the rigorous generation of such scenarios for testing and certification. We strive to extend their efforts and propose deriving concrete driving scenarios (for testing autonomous systems) from higher-level "digital twin" models representing the physical world. When encoded as infinite state machines, these models become much more expressive and "natural" for encoding than individual scenarios. Furthermore, they are susceptible to formal verification and reasoning, empowering the designers to tackle even the most complex domains.