As tropical storm Bebinca gains strength off northern Taiwan, meteorologists in Taipei are using advanced artificial intelligence (AI) technology to more accurately predict the storm’s path.
This AI-based forecasting represents a significant leap forward for Taiwan, offering improved precision in tracking typhoons compared to traditional methods.
Earlier this year, AI tools, including Nvidia's FourCastNet, Google's GraphCast, and Huawei’s Pangu-Weather, were used to forecast Typhoon Gaemi—the strongest typhoon to hit Taiwan in eight years
The AI models accurately predicted Gaemi’s trajectory as early as eight days before landfall, outperforming conventional forecasting systems, which typically provide less advanced notice.
“People are starting to realize AI delivered stunning performances compared to conventional models,” said Chia Hsin-sing, director of Taiwan Integrated Disaster Prevention of Technology Engineering Consulting Company Ltd.
AI weather forecasts have sparked interest within the meteorological community due to their superior accuracy, offering a valuable tool for predicting dangerous weather patterns.
As Bebinca approaches, Taiwan's Central Weather Administration (CWA) is again relying on AI to track its path. Lin Ping-yu, a CWA forecaster, noted that AI has given forecasters greater confidence in predicting that Bebinca will not make a direct hit on Taiwan. “This [AI] is a good thing for us,” Lin said. “It is like having one more useful tool to use.”
AI-based forecasting software works by analyzing historical weather data, learning the relationships between meteorological patterns, and predicting hundreds of weather variables in minutes—processes that traditionally take longer using conventional methods.
According to the CWA, AI models have proven to be nearly 20% more accurate than conventional methods when predicting storm tracks in the Western Pacific this year.
These AI technologies also helped predict an unusual loop in Typhoon Gaemi’s path, which prolonged its impact on Taiwan.
This insight allowed forecasters to issue a rare warning of 1.8 meters (5.9 feet) of rainfall, which proved accurate. CWA Deputy Head Lu Kuo-Chen noted that AI’s early warnings gave authorities crucial extra time to prepare for the typhoon.
Looking ahead, Taiwan is collaborating with Nvidia on its new generative AI tool, CorrDiff, which aims to provide more precise predictions of typhoon landfall and higher-resolution images of storm dynamics.
Lu expressed optimism about the potential of AI in weather forecasting, stating, "We are seeing the potential."
While AI has shown significant promise, experts like Chia Hsin-sing urge caution, noting that it is still in its early stages. AI has yet to consistently deliver accurate predictions for typhoon strength and wind speeds, areas that require further development. “Was it just good luck?”
Chia asked, referencing AI’s success with Typhoon Gaemi. “We need to give AI a bit more time, but it’s something to look forward to.”
With AI's growing role in weather forecasting, its success in tracking storms and typhoons is being closely watched worldwide, and it may soon become a crucial tool in predicting and preparing for natural disasters.
