Modern geospatial hazard research relies on complex software developed by interdisciplinary teams. But as projects evolve and developers change, understanding the system architecture can become difficult. In this work, we explore how UML activity diagrams and visual design documentation can help bridge the gap between domain experts and software developers. Using an AI-based geospatial hazard application as a case study, we evaluated whether design diagrams help researchers understand system workflows, trace data pipelines, and identify opportunities for extending the software. Our experience suggests that visual system documentation can significantly improve collaboration, onboarding, and long-term maintainability in interdisciplinary scientific software projects.