Taiwan uses artificial intelligence to predict storm paths

As the ferocious Tropical Storm Bebinca slowly approaches the northern coast of Taiwan, its winds are increasing in strength, posing a direct threat to the lives of local residents. Faced with this urgent situation, weather forecasters in Taiwan have quickly deployed a brand new and promising forecasting tool: artificial intelligence.

With the ability to process large amounts of meteorological and hydrological data in a short period of time, modern AI models are being used to analyze the detailed parameters of Typhoon Bebinca, from wind speed, direction of movement to expected rainfall. As a result, forecasters can make more accurate predictions about the storm’s trajectory, intensity and potential impact area, thereby supporting authorities and local people to plan effective responses.

A revolution is underway in the field of weather forecasting. Thanks to the remarkable advances in AI technology and the enormous computing power of leading chips, weather forecasts are now significantly more accurate than traditional methods, especially in predicting the path of storms. Leading technology companies such as Nvidia (USA), along with the "king" of semiconductor chips TSMC (Taiwan), are making important contributions to this revolution.

This past July saw a major turning point in weather forecasting in Taiwan. For the first time, AI-based weather forecasting models were used to track Typhoon Gaemi, which was considered the strongest storm to hit the island nation in eight years. With the help of AI, forecasters were able to more accurately predict the storm’s trajectory and destructive power, helping to significantly reduce the damage.

This new technology has really shaken up the Taiwanese weather forecasting world. Thanks to its superior forecasting ability, the system "saw" Typhoon Gaemi eight days before it made landfall, much earlier than traditional methods. This has opened a new era in weather forecasting, allowing scientists to issue earlier and more accurate warnings.

Chia Hsin-sing, Director of Taiwan Integrated Disaster Prevention of Technology Engineering Consulting Company Ltd., affirmed: "Compared with traditional models, AI has proven its superior ability in the field of weather forecasting. The results we have achieved are truly amazing."

Typhoon Bebinca is being closely monitored by weather experts, including Lin Ping-yu, a forecaster at Taiwan’s Central Weather Bureau (CWA). With the help of AI, forecasters like Lin have become much more confident in their ability to predict the storm’s path, and believe the risk of a direct hit is very low.

“AI is really a great helper for us,” said Lin Ping-yu. “It’s like a sharp tool added to our weather forecasting toolkit, helping us work more efficiently.”

Modern AI weather forecasting models are increasingly diverse and powerful. Some typical names include Nvidia’s FourCastNet, Google, GraphCast, Huawei’s Pangu-Weather, and the deep learning system of the European Centre for Medium-Range Weather Forecasts. These tools have been and are contributing significantly to improving the accuracy of global weather forecasts.

AI weather forecasting models are not only used in Taiwan, but are also widely used to predict storms and tornadoes in many regions around the world. Scientists and hydrometeorologists have noted the incredible accuracy of these tools.

AI software is trained on large amounts of historical weather data. Through this process, the AI ​​learns to recognize complex relationships between weather factors and is able to accurately predict hundreds of future weather variables. What’s amazing is that the entire forecasting process takes just a few minutes to complete.

According to CWA statistics, for all storms that have appeared in the Western Pacific this year as of mid-September, the AI ​​model's ability to predict the storm's path within the next 3 days has outperformed traditional methods, with nearly 20% higher accuracy.

Before Typhoon Gaemi made landfall, AI detected an unusual pattern in the typhoon’s movement, predicting that the storm would have a prolonged impact on Taiwan. Thanks to this accurate forecast, CWA promptly issued a rare warning of up to 1.8 meters of rainfall, a decision that has proven to be wise. According to CWA Deputy Director Lu Kuo-Chen, AI played a key role in helping local authorities respond to the typhoon in a timely manner.

“Thanks to AI, forecasters were more confident in predicting the unusual trajectory of Typhoon Gaemi,” said Lu Kuo-Chen. “Early warnings based on AI forecasts have given authorities more time to deploy disaster prevention measures.”

Lu Kuo-Chen also expressed his expectations for the collaboration with Nvidia. The company has just launched the CorrDiff generative AI tool, which promises to significantly improve the accuracy of predicting the location of typhoons. In addition, CorrDiff is capable of creating more detailed images of the inside of typhoons, helping us better understand their structure and evolution.

Despite significant advances, current AI tools still have certain limitations. They cannot provide detailed forecasts of factors such as storm intensity and wind speed. Experts say it will take time for AI technology to mature and completely outperform traditional forecasting methods.