Google AI Extends Flood Forecasts to 5-7 Days Across 80+ Countries

Google Research has published a study in Nature showing how machine learning can significantly improve global flood forecasting—especially in regions where accurate local data is scarce.

Floods are the most common natural disaster, causing roughly $50 billion in annual damages worldwide. Nearly 1.5 billion people (19% of the world's population) face substantial flood risks. But reliable forecasts require data that many regions simply don't have.

"We extended the reliability of currently-available global nowcasts, on average, from zero to five days," Google noted, "and improved forecasts across regions in Africa and Asia to be similar to what's currently available in Europe."

The key challenge: most flood forecasting systems rely on local gauge data—measurements from river monitoring stations. But 80% of the world's rivers lack adequate monitoring. Google used ML to predict floods in "ungauged watersheds" by learning patterns from the limited available data and applying them globally.

The system now provides real-time river forecasts up to seven days in advance across over 80 countries through Flood Hub (available on Google Search, Maps, and Android). The evaluation was conducted in collaboration with the European Centre for Medium-Range Weather Forecasts (ECMWF).

"Upgrading early warning systems to make accurate and timely information accessible can save thousands of lives per year," Google noted.