Increasing rate of progress in hardware and artificial intelligence (AI) solutions is enabling a range of software systems to be deployed closer to their users, increasing application of edge software system paradigms. Edge systems support scenarios in which computation is placed closer to where data is generated and needed, and provide benefits such as reduced latency, bandwidth optimization, and higher resiliency and availability. Users who operate in highly-uncertain and resource-constrained environments, such as first responders, law enforcement, and soldiers, can greatly benefit from edge systems to support timelier decision making. Unfortunately, understanding how different architecture approaches for edge systems impact priority quality concerns is largely neglected by industry and research, yet crucial for national and local safety, optimal resource utilization, and timely decision making. Much of industry is focused on the hardware and networking aspects of edge systems, with very little attention to the software that enables edge capabilities. This paper presents our work to fill this gap, defining a reference architecture for edge systems in highly-uncertain environments, and showing examples of how it has been implemented in practice.