Taiwan Uses AI to Predict Typhoons

Tracks as Storm Bebinca Nears

Taiwan uses AI to predict typhoon tracks, improving accuracy and preparation as tropical storm Bebinca nears. Learn how AI models like GraphCast and FourCastNet are changing meteorology.

As tropical storm Bebinca approaches northern Taiwan, meteorologists are turning to artificial intelligence (AI) for more accurate forecasts. Leveraging cutting-edge AI tools, Taiwan is now better equipped to predict typhoon paths and prepare for potential disasters.

AI Outperforms Traditional Forecasting Models

Taiwanese weather forecasters have noticed that AI-generated predictions are more accurate than conventional models. Tools powered by tech companies like Nvidia, whose chips are manufactured by TSMC, have shown a significant edge in reliability.

For instance, when Typhoon Gaemi hit Taiwan in July, AI-based meteorological systems predicted its path earlier and more precisely than traditional forecasting methods. The storm, Taiwan’s most severe in eight years, brought historic rainfall levels—yet AI helped the country prepare in advance.

Early Success with Typhoon Gaemi

AI-enabled models were able to forecast a direct hit from Typhoon Gaemi up to eight days before it made landfall. This level of accuracy stunned Taiwanese forecasters. Conventional systems could not match the performance of the AI tools during this event.

According to Chia Hsin-sing, Director at Taiwan Integrated Disaster Prevention of Technology Engineering Consulting Company Ltd.,

“People are starting to realise AI indeed delivered some stunning performances compared to conventional models.”

AI in Action: Tracking Bebinca

The same AI models that accurately predicted Gaemi’s trajectory are now being used to track Tropical Storm Bebinca. According to Lin Ping-yu, a forecaster at the Central Weather Administration (CWA), AI tools have significantly increased their confidence in forecasting that Bebinca will not directly hit Taiwan.

“We should be happy about this. It’s like having another practical instrument at my disposal,” Lin said.

Prominent AI weather systems include:

• Google’s GraphCast

• Huawei’s Pangu-Weather

• Nvidia’s FourCastNet

• The European Centre for Medium-Range Weather Forecasts’ deep learning models

Real-Time Learning and Faster Predictions

These AI tools analyse vast amounts of historical weather data to identify patterns and cause-and-effect relationships. Unlike traditional methods that rely heavily on physics-based models, AI can generate predictions in just minutes.

According to CWA statistics, AI-based models were approximately 20% more accurate than traditional models in forecasting typhoon paths within a 3-day window in the Western Pacific during the 2023 typhoon season.

Preparing for Extreme Weather with AI

AI not only improved prediction accuracy but also enhanced preparedness. Before Gaemi struck, the CWA was able to forecast an unusual loop in its path, extending the storm’s impact. Thanks to this, officials issued a rare warning for rainfall exceeding 1.8 meters (5.9 feet), a forecast that proved remarkably accurate.

Lu Kuo-Chen, CWA’s deputy head, noted:

“AI boosted the confidence for forecasters to make that prediction.”
“The early warning gave extra time for authorities to carry out preparations.”

Future Partnerships and Possibilities

Looking ahead, Lu is hopeful about CWA’s collaboration with Nvidia. In 2024, Nvidia introduced a generative AI tool called CorrDiff, designed to predict more accurate typhoon landfalls and provide high-resolution storm imagery.

“We are seeing the potential,” Lu added.

However, experts caution that AI still has limitations. While it excels in predicting paths, it currently struggles with precise forecasts of a storm’s wind strength and intensity. These details remain vital for disaster preparedness.

A Promising Yet Developing Tool

While AI showed impressive results with Gaemi, some experts remain cautious. Chia noted:

“Was it just good luck? We must give AI some further time. It’s something to anticipate.”

Even so, the integration of AI into Taiwan’s weather forecasting systems marks a pivotal shift. By combining human expertise with machine intelligence, Taiwan is now better positioned to face the increasing threat of severe weather.

#TaiwanWeather, #AITyphoonForecast, #BebincaStorm, #TyphoonPrediction, #NvidiaFourCastNet, #TSMC, #GraphCast, #PanguWeather, #CWA, #ClimateTech,