The next chapter in flood resilience: Open sourcing Google’s hydrology framework

Details

Google Research has released its AI hydrology modeling framework on GitHub, making available the same model architecture used to power riverine flood forecasts on Google's Flood Hub. National Meteorological and Hydrological Services and operational forecasters can now integrate the technology into their own workflows with locally-sourced data. The framework is designed for two audiences: hydrological scientists who want to extend Google's work with new models, data sources, and approaches; and operational forecasters who need to incorporate regional data to generate actionable flood warnings for specific areas. Google has built state-of-the-art AI flood forecasting models over several years with a stated goal of giving frontline responders more lead time before floods arrive. Floods are among the most deadly and economically damaging natural hazards globally. Open-sourcing this framework represents a transfer of frontier AI capabilities — previously only accessible through Google's hosted Flood Hub product — to the broader scientific community and public sector agencies that operate independently of Google infrastructure.