Google researchers describe a decade-long effort to use AI for flood forecasting, expanding from river floods to urban flash floods with the goal of providing early warnings to billions in 150 countries.

A key hurdle was the scarcity of urban flash flood data, and Groundsource fills this gap by enabling the urban flood model within Flood Hub.

Early applications include Give Directly using the Flood Forecasting API to deliver proactive aid in Kogi, Nigeria, enabling evacuations, protecting assets, and reinforcing food security and livelihoods.

Looking ahead, the plan is to extend predictions to rural, urban coastal flooding and other disasters, with Groundsource potentially supporting heat waves and mudslides.

Groundsource is an AI-driven method that builds a high-quality urban flood data archive by processing over 5 million flood-related news reports, creating a dataset of 2.6 million historical flood events across more than 150 countries.

Google has open-sourced its hydrology framework, Groundsource dataset, and Flood Forecasting API to empower national meteorological services and researchers to customize forecasts and advance flood resilience.

Flood Hub uses AI to integrate global weather, river, and land data to predict riverine floods up to seven days ahead and urban flash floods up to 24 hours ahead, with outputs shown on a map and accessible via a Floods API.