In March 2026, Google announced Groundsource, an AI-powered methodology that transforms public disaster data into a high-quality data archive—starting with urban flash floods in cities. The move complements existing Flood Hub forecasts and marks a step toward broader AI-driven disaster prediction for public safety.
HOW FLOOD HUB DISTINGUISHES RIVER FLOODS FROM URBAN FLASH FLOODS
Google describes Flood Hub as a tool that uses AI to pull data from around the world and generate prediction alerts on a map. Initially focused on river floods, Flood Hub has evolved to include urban flash floods through the Groundsource approach. The system relies on two AI models: the Hydrologic Model, which forecasts water flow using weather and land conditions, and the Inundation Model, which uses streamflow data to predict affected areas. TechStaged has also covered OpenAI and CodeAI Partner to Prepare the First AI Generation.
DATA POWERING THE FORECAST: FROM REAL-TIME RIVERS TO GLOBAL ARCHIVES
The Flood Forecasting initiative draws on real-time river data and a growing set of publicly available sources to predict floods up to seven days ahead for rivers and up to 24 hours for urban flash floods. Groundsource expands this with a global data pipeline that incorporates large-scale textual data to build a historical flood dataset and improve urban predictions.
GROUNDSOURCE: BUILDING A GLOBAL DISASTER DATA ARCHIVE
Groundsource reads millions of reports to assemble a dataset of historical flood events. Specifically, it used millions of news reports to create a dataset of 2.6 million historical flood events across more than 150 countries, forming the basis for the new urban flash flood model live in Flood Hub.
OPEN ACCESS AND PRACTICAL IMPACT
The hydrology framework powering these tools has been open sourced to support National Meteorological and Hydrological Services and other agencies to integrate their data with the model. The Groundsource dataset and the Flood Forecasting API are publicly available to support future research and applications. In a real-world example, Give Directly used the Flood Forecasting API to support pre-emptive cash transfers in Kogi, Nigeria, helping communities evacuate and protect assets ahead of flooding.
WHAT’S NEXT FOR FLOOD HUB AND GROUNDSOURCE
Google notes that current models are most developed for urban flash floods, with ongoing research to extend predictions to rural, coastal, and other flood types. The company is also exploring extensions of Groundsource to other disasters like heat waves or mudslides as part of broader AI-driven hazard forecasting.
RELATED COVERAGE
- OpenAI and CodeAI Partner to Prepare the First AI Generation
- OpenAI launches ChatGPT for Teens with stronger protections and parental controls
- Operation Blue Skies launches as AI-driven contrail mitigation trial over the North Atlantic
- OpenAI backs 14 AI policy projects to broaden opportunity and resilience in the Intelligence Age
- Software articles
SOURCES
- News from Google: Ask a Scientist: How can researchers use AI to predict a flood? Published · Primary source
