China is accelerating the use of artificial intelligence in weather forecasting as stronger storms and increasingly disruptive extreme-weather events place greater pressure on authorities to provide faster and more precise warnings.
Fengwu, Pangu and Fuxi Challenge Traditional Forecasting
China has developed several systems that use historical weather observations rather than relying entirely on the computationally intensive physics simulations that have dominated forecasting for decades.
Reuters identified Shanghai AI Laboratory’s Fengwu, Huawei’s Pangu and Fudan University’s Fuxi among Chinese systems capable of producing forecasts much faster than conventional models while matching or surpassing them on some accuracy measures.
NBC News explains that traditional numerical forecasting relies on supercomputers to simulate atmospheric physics, while AI models learn patterns from vast archives of historical weather observations and can generate forecasts in a fraction of the time.
That speed could prove particularly valuable during typhoon season, when changing storm tracks can influence decisions on evacuations, flooding preparations and transportation disruptions.
Fengwu Forecast Typhoon Dolphin Within 30 Kilometers
Fengwu has attracted attention for both medium-range forecasts and tropical cyclone tracking.
Fengwu’s developers said the model outperformed Google’s GraphCast across roughly 80% of evaluated weather variables and extended skillful global medium-range forecasting beyond 10 days.
The system also delivered a striking forecast during Typhoon Dolphin.
Fengwu predicted Dolphin’s mainland China landfall five days in advance to within 30 minutes of the eventual time and 30 kilometers, or 19 miles, of the location.
Sun Zhi, chief technology officer of Techwind, which handles Fengwu’s industrial applications, connected those improvements directly to public decision-making.
Reuters quoted Sun as saying that as extreme weather increases, local governments, national authorities, ordinary citizens, farmers and fishermen all need better information to make decisions.
AI Still Struggles to Predict Storm Intensity
The technology is not yet ready to replace conventional forecasting.
NBC News said AI systems increasingly complement traditional models because they are faster and require less computing power, but they still trail conventional forecasts in predicting storm intensity.
The systems remain largely untested for predicting major longer-term climate developments such as El Niño events far in advance.
Sun said years of scientific research would still be needed before people could reasonably trust AI systems making predictions about climate events 18 months ahead.
That limitation explains why meteorologists are more likely to combine AI and traditional forecasting than choose between them. AI can rapidly identify patterns from enormous historical datasets, while physics-based systems remain important for understanding the evolving physical characteristics of storms.
China Joins a Global AI Weather Race
China is not alone in pursuing the technology.
Reuters identified Google’s GraphCast and GenCast, Nvidia-backed FourCastNet and the European Centre for Medium-Range Weather Forecasts’ AI Forecasting System, or AIFS, among the leading international AI forecasting projects.
The immediate value may be measured less by whether AI completely replaces supercomputer forecasting than by whether it gives people additional time to act.
When a typhoon is approaching, even modest improvements in predicting where and when it will arrive can influence evacuation orders, flood preparations and transport planning. China’s growing investment in AI forecasting therefore reflects a wider shift in artificial intelligence: from generating text and images toward systems whose predictions can shape real-world decisions during increasingly dangerous weather.
