AI can Forecast Typhoon? AI is changing typhoon forecast

AI meteorological big models such as Huawei, Google GraphCast and others have compared or even surpassed traditional numerical models on Typhoon trackForecast, and the Forecast revolution is taking place.

The rise of the meteorological model.

Starting in 2023, the AI Weather Forecast large-scale model intensely debuted on Typhoon trackForecast:

• Big Meteorological Model.: Forecast at European Centre for Medium-Term Weather Forecast (ECMWF)'s traditional pattern, published in Nature
• Google DeepMind GraphCast10 Days Global Forecast Secondary Results, Multiple Indicators Beyond Traditional Models
• ECMWF AIFSOld brand organization’s own AI model has been put into operation
• China"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s"s&

In 2023 Typhoon Mawar, 2024 forecast for more than Typhoon, CMA (China) has referenced AI big model results.

Where is AIForecast?

• 快Traditional numerical patterns run the number h on supercomputing, AI models forecast 10 days on a single GPU for tens of seconds
• Province: Calculation cost reduced by several quantitative levels, can run ultra-scale collection Forecast
• The path.Study decades of historical atmospheric data, excellent grasp of large-scale guiding air flows

Shortcuts of AIForecast

• Intensity Forecast is weakTyphoon core is only a few dozen kilometers, existing AI model resolution (about 25 km) can’t catch eye wall details, intensity Forecast still relies on traditional methods and Forecast member experience
• Extreme value is smooth.: AI Tending toward Output" Mediated" Results, Possibly Underestimating Extreme Events such as Explosive Enhancement
• Explained poorly.: Reporting correctly not knowing why correctly, reporting wrongly is difficult to attribute - the AI in the business application is currently" reference" and not" replacement"

The Future: Human Machine Collaboration

The mainstream direction is the integration of AI big model + traditional numerical model + Forecast experience: AI is responsible for rapidly giving path probability distribution, numerical model provides physical details, Forecast makes the final decision. typhoon forecast's "24h error 60km" record, is likely to continue to be refreshed with AI added.

In the multi-agency Forecast showcased by TyphoonSays, the country’s official Forecast has increasingly integrated the contribution of AI models – behind every Forecast line you see, there may be AI involvement.