Google Is Using AI to Predict Floods Before They Happen

Google is expanding the use of artificial intelligence to help predict floods earlier, including in regions where traditional monitoring systems are limited.

The company says its flood forecasting technology now covers areas across 150 countries where more than 2 billion people live. Its models can predict some river floods up to seven days in advance, while urban flash-flood forecasts can provide up to 24 hours of warning.

The forecasts are available through Google Flood Hub, which displays areas that may be at risk of flooding.

Using AI where data is limited

Flood forecasting normally relies on years of measurements from river gauges, weather stations and other sensors. Many parts of the world, however, do not have extensive historical monitoring data.

Google is using AI to help fill those gaps.

For urban flash floods, they developed a system called Groundsource, which uses Gemini to analyze historical news reports about flooding. Google says the system reviewed more than 5 million reports spanning roughly 20 years and helped create a dataset containing about 2.6 million flood events.

That information can then be used to train forecasting models in locations where detailed flood records may be unavailable.

More time to prepare

Earlier warnings can give emergency agencies and residents additional time to prepare, avoid dangerous roads, protect property or evacuate when necessary.

Google says organizations are already using its flood forecasting data in real-world situations. In Nigeria, for example, nonprofit GiveDirectly used Google’s forecasting information to help identify communities at risk and distribute financial assistance before flooding arrived.

Google has also made parts of its flood research, datasets and forecasting technology available to researchers and public agencies, and believe similar AI methods could eventually help address data gaps involving other natural hazards, including heat waves and landslides.

AI cannot prevent floods, and local emergency guidance remains essential. But improved forecasting could give communities more time to react during severe weather.