Most of current research in wireless networked embedded sensing approaches the problem of application design as one of manually customizing network protocols. The design complexity and required expertise make this unsuitable for increasingly complex sensor network systems. We address this problem from a parallel and distributed systems perspective and propose a methodology that enables domain experts to design, analyze, and synthesize sensor network applications without requiring a knowledge of implementation details. At the core of our methodology is a virtual architecture for a class of sensor networks that hides enough system details to relieve programmers of the burden of managing low-level control and coordination, and provides algorithm designers with a clean topology and cost model.We illustrate this methodology using a real-world topographic querying application as a case study.