Google’s Groundsource AI Predicts Flash Floods with Gemini

Google is introducing Groundsource, a new AI-powered methodology leveraging the Gemini model to transform public reports into a historical dataset for predicting urban flash floods. This initiative aims to provide critical early warnings, particularly in regions lacking advanced weather-sensing infrastructure.

Turning News Reports into Actionable Data

Groundsource addresses the historical data gap that has hindered accurate flash flood prediction. By analyzing millions of public reports, the system has compiled a dataset of over 2.6 million historical flood events spanning more than 150 countries [1]. Google Maps is then used to pinpoint the geographic boundaries of each event, creating a focused dataset for flash flood modeling.

24-Hour Advance Notice

Using this dataset, Google has trained a new model capable of predicting flash flood risks in urban areas up to 24 hours in advance [2], [3]. This capability represents a significant advancement in flood forecasting and expands Google’s existing flood coverage, currently reaching 2 billion people in over 150 countries [2].

Flood Hub and Data Sharing

These urban flash flood forecasts are now available within Google’s Flood Hub, alongside existing riverine flood forecasts [2]. Google is similarly sharing this data with emergency response agencies to enhance preparedness and mitigation efforts.

Potential for Broader Applications

The Groundsource methodology isn’t limited to flash floods. Google suggests the same AI approach could be applied to build historical datasets for other natural disasters, such as landslides or heat waves, to accelerate global crisis resilience [1].

Limitations and Future Development

Currently, the model identifies risks across a 20-square-kilometer area [3]. It also does not integrate local radar data, which may limit its precision compared to systems like the US National Weather Service’s flood alert system. However, Groundsource is specifically designed for regions lacking advanced weather-sensing infrastructure.

“By turning public information into actionable data, we aren’t just analyzing the past — we’re building a more resilient future for everyone towards our goal that no one is surprised by a natural disaster,” stated Google [1].