Impact-based flood forecasting reframes the core question of disaster risk from "how intense is this hazard?" to "who and what will be affected, and how badly?" Despite growing policy momentum behind this shift, operational implementations using fully open geospatial stacks remain scarce in the literature — particularly in areas where exposure data is fragmented, hazard layers arrive in inconsistent formats and resolutions, and institutional capacity constrains both tooling choices and maintenance overhead.
This talk presents a reproducible, production-oriented pipeline for automated impact-based flood forecasting built on three open tools: Python for orchestration and operational logic, PostgreSQL for data management, and QGIS for precalculating facility exposure to high-resolution probabilistic hazard layers. The workflow applies configurable thresholds derived from river basin warning levels to produce categorized impact assessments for schools and critical infrastructure at both facility and administrative unit levels.
We discuss the key engineering and institutional decisions that dominate real deployments but rarely appear in conceptual frameworks — including performance tradeoffs driven by high-resolution hazard data, and how to embed automated workflows within government operations in ways that build local ownership. We are honest about where the pipeline succeeds, where it struggles, and what we would do differently. Attendees working on climate risk, disaster preparedness, or humanitarian response will leave with a transferable framework and a realistic picture of what open geospatial tools can deliver in heterogeneous data environments.