Lessons learned from building Confluence: Cloud-based global river discharge estimation software for the SWOT satellite

We present Confluence, a cloud-based workflow that produces river discharge products on demand by processing the entirety of NASA/CNES Surface Water and Ocean Topography (SWOT) mission hydrology data. Confluence integrates dozens of established science algorithms into a unified framework, leveraging cloud resources to meet demands for RAM, storage, and CPU time beyond the capacity of local clusters. Built on AWS Batch, Step Functions, and other AWS services, Confluence accesses input data from 15 sources in multiple formats, orchestrates algorithm processing, detects and recovers from failures, and reports on workflow states. It produced global SWOT discharge products in approximately 4.5 days at a cost of ∼$6,000. Confluence’s design includes a community of practice to guide development, with over 50 hydrologists contributing to results validation. This architecture, together with its community engagement model and cost-efficient execution strategies, offers a generalizable solution for science teams translating complex algorithms into operational cloud-based workflows.