AI summary
Summary of article:
There are differences between AI and traditional models in the path forecasting of Typhoon Qiangbi. Both research and practical cases show that AI can be more accurate in path decision, but intensity and speed prediction still need to be combined with traditional models.
Weather forecasts over the past few days have seen significant differences in the predicted path of Typhoon Qiangbi through traditional models of supercomputer calculations and artificial intelligence (AI) models. Some analyzes have shown that AI has overtaken supercomputers in the ability to predict the path of typhoons, but it is still inadequate in predicting the intensity and speed of typhoons.
According to the typhoon warning issued by the Central Meteorological Administration at 5 pm on the 23rd, Typhoon Qiangpi is currently located about 140 kilometers southeast of Yilan, and is moving north-northeast at a speed of 10 rpm and 17 kilometers per hour.
south korea JoongAng Ilbo reported that since the formation of Typhoon Changbi, traditional supercomputer-based weather models and AI models have diverged in predicting the typhoon. The AI model is predicting that Typhoon Qiangbi will turn sharply north-east before touching Taiwan and then head towards Japan.
However, the European Center for Medium-Range Weather Forecasts (ECMWF) and other conventional forecasting models predict that Typhoon Qiangbi will move westward until it passes northeastern Taiwan or reaches the sea off mainland China before turning.
The US Joint Typhoon Warning Center has adopted AI forecasts. The center said the AI model has proven accurate in predicting the northward turn of typhoons with atmospheric conditions similar to Typhoon Kiangbi in the past.
Research shows that AI has outperformed supercomputers in predicting the path of typhoons. In most cases, AI models have smaller errors in predicting typhoon paths than traditional numerical models, according to an analysis of typhoons over the past two years by the Jeju National University Typhoon Research Center.
ECMWF’s comprehensive forecast system is one of the world’s top numerical weather prediction models. Its average error in typhoon track forecasts up to five days in advance is 205.5 kilometers, which is greater than the less than 200 kilometers of most AI models.
“AI has shown a significantly higher accuracy advantage than traditional numerical models in long-term predictions. If predictions from multiple high-performance AI models are combined, typhoon predictions can be made more accurate,” said Moon Il-ju (transliteration), director of Jeju National University’s Typhoon Center.
However, AI models have also proven unreliable in predicting the intensity and speed of hurricanes. Models may underestimate the intensity of the storm due to limited data. Experts say hybrid systems that leverage AI and supercomputers will be key to improving weather forecasting.
Summary FAQ
-
AI models have always predicted that Qiangpi will turn significantly north before approaching Taiwan, and then move towards Japan; Traditional numerical models mostly predict it will continue to move westward, possibly passing through northeastern Taiwan or approaching mainland China waters before turning.
-
Because Jeju National University analyzed typhoon data in the past two years and found that in most cases the path error of the AI model is smaller than the traditional numerical model, and the average error of the first 1 to 5 days of ECMWF is also higher than that of most AI models.
-
The article pointed out that although AI has advantages in path prediction, its judgment about the intensity and speed of the hurricane is still unreliable, and it may underestimate the intensity of the hurricane due to limited training data. Therefore, experts advocate combining AI with supercomputers.
