The company says the system can forecast urban flash floods up to 24 hours in advance, potentially giving authorities and residents valuable time to prepare, News.Az reports, citing foreign media.
The tool, called Groundsource, converts publicly available information into a detailed historical record of disaster data that can be used to improve flood prediction models.
According to Google, Groundsource uses AI to analyse large volumes of public data and identify patterns linked to past flooding events. The system focuses particularly on flash floods in urban areas, where forecasting has traditionally been difficult due to limited historical data.
***
The company explained in a blog post that the new forecasts are available through Google Flood Hub, which already provides river flood predictions for more than 2 billion people across over 150 countries. Groundsource expands these capabilities by creating an open-source dataset that can help improve research and disaster preparedness worldwide.
Users can access the tool directly on the Flood Hub website, where data for more than 150 countries is available. The platform compiles historical flood information, including water-level records, to help visualise flood risks.
Google says the initiative is intended to support a wide range of users, including communities, policymakers, researchers and scientists. By providing open data and forecasting tools, the company hopes to help cities better prepare for disasters and reduce the impact of floods.
Sundar Pichai said in a post on X that Google trained a new flood forecasting model capable of predicting flash floods in urban areas up to 24 hours ahead of time.
Pichai added that the Groundsource system uses Gemini to identify more than 2.6 million historical flood events across over 150 countries, helping address gaps in flood-related data.
He also said the dataset is being made open source to support global research, while urban flash flood forecasts are now live on Flood Hub to help communities stay safer during extreme weather events.
