As Tropical Storm Bebinca heads toward northern Taiwanese waters and gains strength into a potential typhoon, weather forecasters in Taipei are using a new and so far successful method — artificial intelligence (AI) — to help track its path.
AI-generated forecasts, some of which are powered by software from technology giants including Nvidia, whose chips are made by Taiwan’s domestic semiconductor champion TSMC, have so far outperformed traditional methods in predicting typhoon tracks.
In July, an AI-based weather model used for the first time helped Taiwan better forecast the path and impact of Typhoon Gaemi, the most powerful storm to hit the island in eight years, bringing record-breaking rainfall.
The new technology impressed Taiwanese forecasters, as they predicted a direct strike from Gaemi eight days before it made landfall—far better than traditional methods, which remain the mainstay of forecast planning.
“People are starting to realize that AI has really achieved some amazing performance compared with traditional models,” said Chia Hsin-sing, director at weather service provider Taiwan Integrated Disaster Prevention of Technology Engineering Consulting Co. Ltd.
Bebinca is now being monitored using the same AI tools, including by Lin Ping-yu, a forecaster at Taiwan’s Central Meteorological Administration (CWA), who said the AI ​​has given him greater confidence that there will be no direct attack.
“This (AI) is a good thing for us. It’s like having another useful tool,” Lin said.
Proposed AI weather programs include Nvidia’s ForecastNet, Google’s GraphCast and Huawei’s Pangu-Weather, as well as a deep learning-based system by the European Centre for Medium-Range Weather Forecasts.
“This is a very hotly anticipated competition. We will soon know who is winning,” Chia said.
According to forecasters and academics, such AI models are being used in other regions as well to predict storms and cyclones with good accuracy.
The AI-based software is trained using historical weather data to learn the cause-and-effect relationships between meteorological systems and can forecast hundreds of meteorological variables several days in advance – a process that takes only a few minutes to complete.
According to data compiled by the CWA, for all storms that formed in the western Pacific through mid-September this year, the AI’s accuracy in predicting the storm’s path over a three-day period was about 20% higher than traditional models.
Ahead of Gaemi, AI helped authorities forecast an unusual loop in its path that prolonged its impact on Taiwan and prompted them to immediately issue a rare warning for rainfall of up to 1.8 metres (5.9 feet), which later proved accurate, said Lu Kuo-chen, deputy head of the CWA.
“(AI) has increased forecasters’ confidence,” Lu said, adding that the early warning gave officials extra time to prepare.
Lu also hopes to partner with Nvidia, which this year announced a generative AI tool called Cordiff that aims to more accurately forecast hurricane landfall locations and provide high-resolution images inside storms.
“We see potential in this,” Lu said.
However, so far experts say AI tools are not capable of providing quality forecasts about a hurricane’s more detailed impact, such as its strength and winds, and the new technology needs more time to solidify its lead over traditional methods.
“Was it just good luck?” Chia said, pointing to the AI’s impressive performance on Gamee. “We should give the AI ​​a little more time. This is something we are looking forward to.”
(This story has not been edited by NDTV staff and is auto-generated from a syndicated feed.)
