Google expands AI flood forecasting to 100+ countries, reaching 700 million people

Google announced a major expansion of its Flood Hub platform, extending AI-powered river flood forecasting to more than 100 countries and bringing coverage to around 700 million people, up from 460 million. The free public tool gives communities advance warning of riverine floods, with forecasts up to seven days ahead in many areas.

The expansion rests on model improvements. Google reports that a new version, trained on more labeled data and three times as many locations with a more robust architecture, now achieves at a seven-day lead time the accuracy the previous model delivered at five days. The system is designed to work even in data-scarce regions, learning from global patterns of rainfall and river behavior rather than relying solely on dense local sensors, which many vulnerable areas lack.

“The free public tool gives communities advance warning of riverine floods, with forecasts up to seven days ahead in many areas.”

Google also moved to open the technology to others. The company began offering forecasts to researchers and partners through an API pilot and a dataset called the Google Runoff Reanalysis and Reforecast, alongside an expert data layer on Flood Hub with close to 250,000 forecast points across more than 150 countries. The platform was built with input from governments, NGOs and UN partners, and Google.org has worked with Red Cross societies and Yale researchers to build offline alert networks for communities without reliable internet.

There are real limits. A forecast only saves lives if it reaches people in time and they can act on it, which depends on local alert systems, infrastructure and trust. The model covers riverine flooding, not every flood type, and accuracy varies by region. But for hundreds of millions of people in flood-prone areas, earlier and more reliable warnings are a tangible benefit, and sharing the underlying data invites others to build on the work rather than depend on a single provider.