Google is using artificial intelligence to predict flash floods. This new system analyzes millions of news reports from around the world. It helps improve forecasts for sudden, dangerous floods.
Flash floods are very deadly weather events. They kill over 5,000 people each year. Predicting them is hard because they happen fast and in small areas.
AI Reads the News
To get more data, Google researchers used the Gemini AI model. They analyzed about 5 million news articles globally. This helped them find information on 2.6 million flood events.
They created a database called Groundsource. This database turns news data into usable scientific information. It is used to train flood prediction models. This is one of the first times Google used large language models. They turn news text into data for science.
This project is a first. Google uses big AI language models. They change news text into numbers. These numbers can be studied by scientists.
Training a Model for Disaster Prediction
Next, researchers used a special AI network. It's called a Long Short-Term Memory network. This model analyzed global weather data. It estimated the chance of flash floods in certain areas.
This model is now on Google's Flood Hub platform. The platform shows flood risks in cities. It covers 150 countries worldwide. Google shares this prediction data. Emergency teams use it to prepare for disasters faster.
Helpful Tool for Less Resourced Nations
The AI model still has limits. It can spot risks in areas about 20 square kilometers. It is less accurate than advanced systems. Examples include the US National Weather Service. But the main goal helps countries without good weather monitoring.
Many countries lack advanced weather radars. They also lack accurate historical climate data. This AI model helps fill that information gap.
Future Applications
Researchers think this tech can be used for other hard-to-measure events. This includes heat waves or mudslides. It works by changing text from reports into data. AI can then analyze this data.
This work fits a growing trend in weather forecasting. Deep learning techniques are used to build better prediction models. Climate change makes these challenges greater.