Environmental scientists, land managers, and policy actors are increasingly presented with high-stakes high-uncertainty problems stemming from human-ecosystem interactions. To help address these problems, scientists frequently use models that produce enormous geospatial and temporal datasets that are constantly modified and often seek input from communities outside their discipline. To assist scientists—as well as others who interface with policymakers—in generating insights from this complex and changing data, this research examines a co-production effort where ecologists, environmental scientists, computer scientists, software engineers, and social scientists collaborate on the development of domain-specific software as part of the Visualization of Terrestrial and Aquatic Systems (VISTAS) project. Findings from this case study suggest that visualization is critical for communicating science to both experts and non-experts, and integral to many aspects of the science production pipeline and policymaking process, including data exploration and model validation. Additionally, there is evidence among our collaborators that this software co-production process not only resulted in useable and useful application features for visualization and data analytics, but also influenced the way that scientists approach their research activities. This research has potential implications for both the production of domain-specific software and the co-production of scientific knowledge, highlighting the challenges of embarking on a software co-production effort that spans multiple years with changing project personnel and shifting priorities. Recommendations for policies on how to promote and support such research activities both within this domain context and other institutional contexts are also discussed.