How Google's AI Predicts Floods Seven Days Ahead
Google published an interview on August 18, 2026 with Deborah Cohen, a senior staff research scientist who leads flood forecasting on its Climate Crisis Resilience team. Everything below comes from that first party post.
Transcript
Google predicts floods in one hundred fifty countries and can warn about a river flood seven days ahead.
A group of models processes global rainfall river levels and ground conditions. Urban flash floods get twenty four hours of warning. Predictions appear in Flood Hub and Search.
City flash floods had almost no sensor history. So Groundsource had Gemini read five million flood reports. Two point six million past floods now feed a Flood Hub model.
Give Directly sent cash to Kogi Nigeria before water rose. Google reports incomes more than doubled. Those results are self reported not independently checked. Coverage is limited to cities.
Inboxsmith builds AI receptionists that answer calls and messages for small businesses. Please like and subscribe for more news.
Sources
Every claim in this video comes from the top ranking coverage of this topic. The claims and where each one came from:
- Google makes flood predictions in 150 countries where more than 2 billion people live.(Google official announcement, Ask a Scientist: How can researchers use AI to predict a flood?)
- Google uses a collection of models on global data such as rainfall, river levels and ground surface conditions to predict riverine floods up to seven days before they strike.(Google official announcement, Ask a Scientist: How can researchers use AI to predict a flood?)
- The system predicts urban flash floods up to 24 hours before they hit, and the predictions live in Flood Hub and also show up in Search.(Google official announcement, Ask a Scientist: How can researchers use AI to predict a flood?)
- Sensors monitoring flash flooding in urban areas are very unusual, so Google had almost no historical data on urban flash floods.(Google official announcement, Ask a Scientist: How can researchers use AI to predict a flood?)
- For Groundsource, Gemini read over 5 million news reports about flooding spanning a 20-year period, creating a dataset of 2.6 million historical flood events that Google integrated into a new urban flash flood model live in Flood Hub.(Google official announcement, Ask a Scientist: How can researchers use AI to predict a flood?)
- Using Google's Flood Forecasting API, Give Directly delivered cash to people in Kogi, Nigeria before water rose, and Google reports incomes more than doubled.(Google official announcement, Ask a Scientist: How can researchers use AI to predict a flood?)
- The models powering Flood Hub are currently limited to predicting flash floods in urban areas because that is where Google has the best data.(Google official announcement, Ask a Scientist: How can researchers use AI to predict a flood?)