Google Research has introduced a new AI-powered system called Groundsource that transforms large amounts of public information into structured sets of data to better predict natural disasters.
The tool analyses millions of historical reports to build a detailed network of past events, starting with urban flash floods. Groundsource addresses the idea that many natural disasters lack high quality historical data, which makes it difficult to predict new ones.
By using Gemini AI to process unstructured information, the system identifies over 2.6 million flood events across 150+ countries, mapping them with precision using Google Maps. This data set is then used to train predictive models that are able to forecast urban flash floods up to 24 hours in advance, with the results being available from Google’s Flood Hub platform.
This approach is meant to make reports turn into actionable insights for researchers, communities, and even governments, to help prepare for potential disasters. Groundsource is showing that AI can be used for more than generating content, it can be used to structure complex existing information in uncertain times.
ADD WORKGROUP