Askume North, Sept 13 – As Tropical Storm Bebinca moves toward waters north of Taiwan and intensifies into a possible typhoon, weather forecasters in Taipei are turning to a new method so far successful to help track its path: the use of artificial intelligence.
Forecasts generated by artificial intelligence, some of which are powered by software from tech giants such as Nvidia Corp NVDA.O whose chips are made by Taiwan’s domestic semiconductor champion TSMC Corp 2330.TW , have so far helped predict the storm’s path.
In July, a weather model based on artificial intelligence was used for the first time to help Taiwan better predict the path and impact of Typhoon Jamie. It was the strongest typhoon to hit the island in eight years, bringing record rainfall.
The new technology impressed Taiwanese forecasters as they predicted a direct hit from Jimmy eight days before landfall, far better than traditional methods that remain the mainstay of forecasting planning.
“People are starting to realize that artificial intelligence has really achieved some amazing performance compared with traditional models,” said Jia Xinxing, director of weather service provider Taiwan Integrated Disaster Prevention Technology Engineering Consulting Co.
Babinka is now using the same AI tools for surveillance, including Lin Pingyu, a forecaster at Taiwan’s Central Weather Bureau (CWA), who said the AI gives him more confidence that a direct attack will not occur.
“This (artificial intelligence) is a good thing for us. It’s like having another useful tool,” Lin said.
Proposed AI weather projects include Nvidia’s ForecastNet, Google’s (GOOGL.O) GraphQL and Huawei’s (HWT.UL) Pangu-Weather, as well as a deep learning-based system from the European Centre for Medium-Range Weather Forecasts.
“It is a very tough competition and we will soon know who will win,” Zee said.
According to forecasters and academics, such AI models are also being used in other sectors to predict storms and cyclones with high accuracy.
Artificial intelligence-based software is trained using historical weather data to understand the cause-and-effect relationships of weather systems and can predict hundreds of weather-related variable days ahead – a process that can take years to complete.
For all storms that formed in the Western Pacific through mid-September this year, artificial intelligence was about 20 percent more accurate than traditional models in predicting the storm’s three-day path, according to data compiled by the CWA.
CWA deputy director Lu Kuo-chen said that ahead of Gaemi, artificial intelligence helped officials predict an unusual loop in its path, which amplified its impact on Taiwan and prompted them to immediately issue a rare rainfall of up to 1.8 metres (5.9 ft) alarm.
“[Artificial intelligence] increases forecasters’ confidence,” Lu said, adding that the warning gave officials extra time to prepare.
Lu also hopes to work with Nvidia, which this year launched a generative artificial intelligence tool called Chordify that aims to predict more accurate hurricane locations and provide high-resolution imagery of storm interiors.
“We see potential in this,” Lu said.
However, experts say that so far, artificial intelligence tools are not able to provide high-quality predictions of more detailed impacts of hurricanes, such as intensity and winds, and it will take longer for new technologies to gain advantages over traditional methods.
“Is it just luck?” Chia asked, pointing to the AI’s excellent performance on GameEye. “We should give the AI a little more time. That’s what we’ve been waiting for.”
