Executive Summary
Google has introduced Groundsource, a new Gemini-powered methodology designed to create high-quality historical datasets for natural disasters. Its first application addresses the data gap in urban flash floods by analyzing millions of public reports and using Google Maps to create a new dataset. This dataset has been used to train a model capable of forecasting urban flash floods up to 24 hours in advance, with predictions now available on Google's Flood Hub.
Key Takeaways
* Product Name: Groundsource, a new AI-powered methodology.
* Core Function: Uses the Gemini model to analyze decades of public reports and Google Maps data to create a historical dataset of disaster events, starting with urban flash floods.
* New Capability: A new predictive model, trained on the Groundsource dataset, can forecast urban flash floods up to 24 hours in advance.
* Scale: The initial dataset includes over 2.6 million historical flood events from more than 150 countries.
* Availability: Forecasts are now accessible in Google’s Flood Hub. The Urban Flash Floods model and dataset are available as part of the Google Earth AI family.
* Future Application: The Groundsource methodology has the potential to be applied to other natural disasters, such as landslides and heat waves.
Strategic Importance
This initiative enhances Google's crisis response capabilities and AI for social good narrative, turning unstructured public data into a valuable, proprietary asset for predictive modeling. It expands the utility of Google's public safety tools and strengthens its position in the geospatial AI field.