The rise of AI weather forecasting as China emerges as a leader in race to improve predictions

Chinese-designed AI weather models, including Fengwu from the Shanghai AI Laboratory, Huawei's Pangu, and Fuxi from Fudan University, are emerging as leaders in the race to improve weather prediction, delivering forecasts significantly faster than traditional systems while matching or surpassing them on some accuracy measures. Fengwu, developed by the Shanghai AI Laboratory and commercialized by Techwind, surpassed Google's GraphCast across roughly 80% of evaluated meteorological variables and extended effective global medium-range predictions beyond 10 days, according to its creators. Techwind CTO Sun Zhi said AI models can already project typhoon trajectories with high accuracy, but they still trail classic models on storm intensity and need years of research before being trusted for long-term climate predictions.

The rise of AI weather forecasting as China emerges as a leader in race to improve predictions Researchers say AI can generate forecasts much faster than conventional systems while matching or surpassing them on some measures of accuracy - Bookmark As meteorologists tracked Typhoon Dolphin's path toward China, a new generation of artificial intelligence weather models worked alongside traditional forecasting systems. It’s a sign of China's emergence as a leading player in the race to improve weather prediction. Among a select group of AI /topic/ai -driven platforms, Chinese-designed systems such as Fengwu from the Shanghai AI Laboratory, Huawei's Pangu, and Fuxi from Fudan University can deliver forecasts significantly quicker than classic models while equalling or beating them across specific precision metrics, experts say. Traditional meteorology /topic/meteorology has spent decades relying on supercomputer-based numerical frameworks designed to simulate atmospheric physics. Conversely, AI models analyze historical observation data to identify underlying patterns, delivering predictions in a fraction of the time. This technology is undergoing real-world trials during East Asia's typhoon season, where incremental enhancements to path predictions assist officials in organizing evacuations, managing transport disruptions, and preparing for floods. The growth of AI-powered forecasting has opened a fresh competitive field spanning technology firms, academic institutions, and national meteorological bodies, with China establishing itself as a key competitor. Internationally, prominent AI forecasting systems include Google's GraphCast and GenCast, the Nvidia-backed FourCastNet, and the European Centre for Medium-Range Weather Forecasts' platform, known as AIFS. Fengwu gained significant notice after its creators announced it surpassed GraphCast across roughly 80% of evaluated meteorological variables while pushing effective global medium-range predictions beyond 10 days. "With more extreme weather, people need information to make decisions, both local governments, the national government, also the average person, farmers and fisherman," said Sun Zhi, the CTO of Techwind, the company responsible for Fengwu's industrial applications. "So we want to help provide better information so people can make decisions." Though artificial intelligence platforms serve as an increasingly valuable supplement to standard forecasting due to reduced computational expenses and high processing speeds, they are not expected to entirely supplant traditional physics-based models in the near future. According to Sun, AI applications can already project typhoon trajectories—five days prior to Dolphin's landfall, Fengwu determined the location and timing of its arrival on mainland China to within 30 km 19 miles and 30 minutes. However, these systems continue to trail classic models when evaluating storm intensity and remain unproven regarding major climatic developments. "If we predict a climate change event 18 months in advance, people won't believe it," Sun said. "They need to know it's reliable. We need to do years of scientific research before people trust us when we say there will be an El Nino event or we say the changing temperature on the sea's surface will affect the breeding cycle of fish." As technology developers work to refine these forecasting tools, the side-by-side reliance on both computational techniques is expected to persist for some time.