Weather researchers are using artificial intelligence (AI) systems to improve existing weather prediction methods. But, experts say the AI tools currently face limitations and should be used along with traditional prediction methods to be most effective.
AI systems trained to predict, or forecast, weather events are now being used by many government agencies and organizations worldwide. Such systems aim to produce weather predictions faster and at a lower cost than traditional forecasting methods.
One weather predicting system that has shown promise is the Google-financed GraphCast method. This machine learning-based system trains directly on weather data that has already been collected and examined. Such methods have demonstrated an ability to outperform traditional forecasting systems.
The system works by combining past weather predictions with modern forecasting models to provide the most complete picture of weather and climate.
In Europe, the European Center for Medium-Range Weather Forecasts (ECMWF) has been using AI prediction tools since January. The organization provides detailed weather forecasts four times per day to nations across Europe.
The ECMWF technology is called the Artificial Intelligence/Integrated Forecasting System (AIFS). The group describes the system as a “data-driven” forecasting model. It is designed to make many predictions quickly, including for extreme events involving powerful storms and heatwaves.
AI-supported data from the ECMWF correctly predicted intense rains last month across parts of Europe that resulted in widespread flooding. But while the predictions were right, destruction caused by the flooding could not be avoided.
Experts told Reuters this is largely because it is still difficult to gather and fully utilize some collected weather data. In addition, there is a need to strengthen and improve current AI models used to predict weather.
Andrew Charlton-Perez is a professor of meteorology – the scientific study of weather processes – at the University of Reading in Britain. He told Reuters, “In some cases and for some variables, AI models can beat physics-based models, but in other cases vice versa.”
Charlton-Perez said one problem is that the effectiveness of an AI model is based on the information it is given. Weather disasters can be harder to predict if there is too little data to enter into AI systems. This can also be true if extreme events happen repeatedly at different times of the year or in different areas.
Charlton-Perez said he thinks the best use of AI-based weather forecasts would be to use them in combination with traditional weather predicting tools. This, he noted, could utilize AI data to produce weather predictions based on large sets of information collected from multiple sources.
Thomas Wostal is with the weather observatory GeoSphere in Austria. He told Reuters his group’s models correctly predicted 300 to 400 millimeters of local rains in September. And records show that same amount actually fell in the storms.
But scientists say even in cases where predictions are correct, effective communication is needed to get the information out to communities and local officials so they can effectively prepare.
Shruti Nath is a research assistant in weather prediction and climate at Britain’s Oxford University. She told Reuters, “I think what happened with (the recent floods) … is that it’s so rare – a one in 150- to 200-year event – that even if the weather models capture it, there’s a reasonable degree of uncertainty.”
Nath said AI-supported forecasts need to be clearly communicated to the public in a way that warns of the severity and possible destruction of extreme events. This way, people might see the importance of taking action before severe weather happens in order to prevent costly cleanup and recovery efforts.
I’m Bryan Lynn.
Reuters reported this story. Bryan Lynn adapted the report for VOA Learning English.
_____________________________________________
Words in This Story
utilize – v. to use something in an effective way
variable – n. a number, amount or situation that can change
vice versa – adv. used to say that what you have just said is also the true in the opposite way