Google's New AI System Predicts Urban Flash Floods 24 Hours in Advance Using 2.6 Million Historical Flood Events
Summary
Google launches Groundsource, a Gemini-powered AI system that analyzes 2.6 million historical flood events across 150+ countries to predict urban flash floods up to 24 hours in advance, with forecasts now available on Google's Flood Hub and the dataset released open-source for global researchers.
Key Points
- Google launches Groundsource, a new Gemini-powered AI methodology that analyzes decades of public reports to identify over 2.6 million historical flood events across more than 150 countries, creating a high-quality dataset for disaster prediction.
- Using this dataset, a new AI model is now capable of predicting urban flash floods up to 24 hours in advance, with forecasts made available through Google's Flood Hub alongside existing riverine flood forecasts that already cover 2 billion people.
- The Groundsource methodology holds potential beyond floods, with plans to apply the same AI-driven approach to other natural disasters like landslides and heat waves, while the dataset is being released as an open-source benchmark for researchers and partners worldwide.