As Typhoon No. 13 (Dolphin) approached China in early August, a new generation Artificial Intelligence (AI) weather model was utilized alongside conventional weather forecasting systems for path monitoring by Chinese meteorological authorities. Reuters reported that this once again demonstrated China's emergence as a major player in the field of AI-powered weather forecasting.

According to Reuters, AI forecasting systems developed in China include "Fengwu" developed by the Shanghai AI Lab, "Pangu" by Huawei Technologies, and "Fuxi" by Fudan University in Shanghai. Researchers say these systems can generate forecasts in much shorter times compared to conventional forecasting systems, and in some accuracy metrics, they perform equally well or even better.

Until now, weather forecasting has relied on numerical prediction models that simulate atmospheric physical phenomena on supercomputers. However, AI models can learn patterns from vast amounts of past meteorological observation data and derive prediction results in significantly shorter times than before. Due to these advantages, the practical application of AI technology is progressing during the typhoon season in East Asia.

Even a slight improvement in typhoon path prediction is expected to aid administrative authorities in formulating flood control measures, evacuation plans, and responding to traffic disruptions. With the development of AI-powered weather forecasting, new competition is emerging among technology companies, research institutions, and meteorological agencies, and China is becoming one of the leading players in this field.

Fengwu is said to have demonstrated superior performance to Google's GraphCast for about 80% of meteorological variables, according to evaluation results published by its development team. It also attracted attention for extending the duration of valid medium-range forecasts to more than 10 days.

The Chief Technology Officer (CTO) of Techwind, responsible for the industrial application of Fengwu, pointed out that "In an environment where extreme weather events are increasing, people need information for decision-making." He added, "This applies not only to local and central governments but also to ordinary citizens, farmers, and fishermen. We want to provide better information and support people's judgments."

Meanwhile, Reuters also reported that "AI systems are gaining importance as a complement to conventional weather forecasting due to their high prediction speed and low computational cost, but they are considered unlikely to completely replace conventional weather models in the near future."

According to the same CTO, AI models have already demonstrated high capability in typhoon path prediction. Five days before Typhoon No. 13 made landfall in mainland China, Fengwu predicted the landfall time with an error of less than 30 minutes and the landfall location with an error of less than 30 kilometers.

On the other hand, it still falls short of conventional models in predicting the intensity of storms and typhoons, and it lacks sufficient track record for predicting large-scale climate change phenomena. Even as AI weather forecasting technology advances, AI models and conventional numerical prediction models are expected to be utilized in parallel for the time being. (Edited by Hinata)