Google expands AI flood forecasting to cover 100 countries, enhances accuracy

Google has announced a major expansion of its AI-driven flood forecasting system, extending coverage to 100 countries and increasing the global population served to 700 million — up from 400 million across 80 countries.

This growth is supported by advancements in Google’s AI technology and the availability of more labelled data to enhance the forecasting model’s accuracy.

In a blog update, Google explained the improvements made to the system, which now includes an upgraded model architecture and new data layers. This includes 250,000 “virtual gauges,” which simulate flood risks based on geological and atmospheric data, allowing for more precise predictions of riverine floods globally. However, only regions with reliable historical data will have flood predictions displayed, ensuring accuracy.

Google has also developed a new API, currently in development, to give researchers and partners streamlined access to data. The data is accessible through the Google Runoff Reanalysis & Reforecast (GRRR) datasets and can help researchers and local authorities make informed decisions on flood management.

The system’s forecast range has been extended from five days to seven, improving early warning capabilities for communities and public agencies. Google believes the enhanced AI-based system will help authorities in high-risk areas improve disaster response, protect lives, and manage infrastructure more effectively during flood events.