Sensor networks based on the de facto standard Berkeley TinyOS platform are changing the way environmental information is collected in the field. One such network has been designed, deployed, and tested in order to determine where ephemeral streams (small, temporary channels of runoff) form during precipitation events. This small, proof-of-concept test network was designed around a generic nondeterministic finite state machine component, which was built to be re-used in later environmental sensor network applications. A simplistic broadcast mechanism was devised to provide collective sampling interval changes to adapt to environmental conditions. In this paper, the design and testing of the ephemeral stream detection network are discussed, along with design features that can be re-used in later applications. Improvements for a later deployment of a larger, operational ephemeral stream detection network are also described.