Google is using old news reports and AI to predict flash floods

Google is employing a novel approach to flash flood forecasting by utilizing its Gemini large language model to analyze 5 million news articles, identifying 2.6 million flood reports and creating a geo-tagged dataset called “Groundsource.” This addresses the challenge of predicting flash floods, which are difficult to monitor due to their localized and short-lived nature. The resulting model, built on an LSTM neural network, is now operational in 150 countries, providing risk assessments to emergency response agencies. While the model has limitations, such as lower resolution and a lack of real-time radar data, it is particularly valuable in regions lacking extensive weather-sensing infrastructure. Google hopes to apply this method to forecasting other events like heat waves and mudslides, and experts see it as a creative solution to data scarcity in geophysics.

(Source：TechCrunch)