Top News

Google's AI can now predict flash floods 24 hours ahead
14 Mar 2026

In a major breakthrough, Google has announced an artificial intelligence (AI) model that can predict urban flash floods up to 24 hours in advance.

The innovative system, powered by the company's Gemini technology and a new method called Groundsource, has been trained on millions of historical events to enhance disaster forecasting.

The dataset is now available on Google's Flood Hub platform.

Groundsource methodology
Innovative approach

The Groundsource methodology was developed to overcome the persistent problem of unreliable datasets on flash floods.

Unlike river floods, which have been systematically tracked, urban flash floods that occur rapidly have remained unstructured due to a lack of quality data.

To tackle this issue, the Groundsource system has analyzed decades of publicly available data using Gemini and identified over 2.6 million historical flood events across 150 countries.

Datasets created using Google Maps
Data structuring

The Groundsource system also leveraged Google Maps to define the exact boundaries of each event.

This has helped the researchers create structured datasets focused on flash floods in urban areas.

The datasets created using Groundsource are now available through Flood Hub, Google's own platform providing information on flood warnings and risks.

Flood Hub now covers 2 billion people
Expanded coverage

Google's Flood Hub, which already provides river flood predictions covering two billion people in over 150 countries, has now added urban flash flood predictions to its offerings.

The company hopes this expanded coverage will help communities prepare for extreme weather events better than before.

The Groundsource methodology could also be used for other natural disasters in the future, further enhancing AI-based early warning systems globally.