China’s AI Weather Models Prove Vital During Typhoon Dolphin
China is leveraging artificial intelligence to enhance weather prediction accuracy and speed, particularly for extreme events like Typhoon Dolphin. Systems such as Fengwu, Pangu, and Fuxi now operate alongside traditional numerical models, offering faster forecasts with comparable or superior accuracy for track predictions. While AI excels in speed and cost efficiency, it currently lags behind conventional methods in predicting storm intensity and long-term climate shifts, necessitating a hybrid approach for reliable decision-making.
Key points
- Chinese AI models including Fengwu, Huawei’s Pangu, and Fudan University’s Fuxi are being tested during typhoon season to improve forecast speed and accuracy.
- Fengwu, developed by Shanghai AI Laboratory and industrialized by Techwind, reportedly outperformed Google’s GraphCast in approximately 80% of evaluated weather variables and extended skillful forecasts beyond 10 days.
- During Typhoon Dolphin, Fengwu predicted the landfall time and location within 30 minutes and 20 miles of the actual event five days prior, demonstrating high precision for track forecasting.
- AI models offer significant advantages in processing speed and lower computing costs compared to traditional supercomputer-based numerical models.
- Despite these gains, AI systems currently struggle with predicting storm intensity and long-term climate events like El Niño, requiring years of further research to build public trust.
- The technology is being used to help local and national governments, farmers, and fishermen make informed decisions regarding evacuations, flooding, and transport disruptions.
Background
This development follows recent advancements in global AI weather modeling, such as DeepMind’s WeatherNext, which demonstrated improved hurricane lead times and open-source potential in 2025 and 2026. The current focus on China’s systems highlights a competitive landscape where nations and tech giants race to integrate AI into meteorological operations to handle intensifying extreme weather events.
Why it matters
As extreme weather events become more frequent and severe, the ability to provide accurate, timely forecasts is critical for public safety and economic stability. AI models offer a scalable, cost-effective complement to traditional methods, potentially saving lives and reducing economic losses by providing earlier warnings for evacuations and infrastructure protection. However, the reliability gap in intensity and long-term climate predictions underscores the need for continued hybrid modeling approaches.
What to watch
Expect continued integration of AI and traditional models in operational weather forecasting. Chinese developers will likely focus on improving AI accuracy for storm intensity and long-term climate predictions to build public trust. Global competition among tech companies and meteorological agencies will intensify as nations seek to leverage AI for climate resilience and disaster preparedness.
Want the full story? Read the original reporting
Read on nbcnews.com