China Meteorological Administration Launches Three AI Meteorological Large Model Systems

TapTechNews June 19th news. Yesterday (June 18th), the China Meteorological Administration announced that it has launched three AI meteorological large model systems, namely 'Fengqing', 'Fenglei', and 'Fengshun'.

Global medium and short-term forecast system 'Fengqing'

This model features strong physical integration and interpretability of the atmosphere. While achieving efficient calculation, it can provide a physical interpretability basis for the prediction results and automatically mine the physical evolution including the inner weather systems.

TapTechNews cites the test results. The model has a global available forecast days of 10.5 days, exceeding the mainstream meteorological forecast large models in Europe and America, especially in the longer forecast validity period, it has a more obvious advantage.

Artificial intelligence nowcasting system 'Fenglei'

This model closely combines the two scientific paradigms of data-driven and physical-driven, significantly improving the forecasting ability of radar echoes on the kilometer scale from 0 to 3 hours and realizing seamless integration of deep learning and physical laws.

The China Meteorological Administration has constructed a set of 'data - computing power - platform' full-process short-term nowcasting system, which can generate radar echo extrapolation products from 0 to 3 hours every 6 minutes within 3 minutes, achieving a 25% improvement in strong echo forecasting skills.

Artificial intelligence global sub-seasonal - seasonal prediction system 'Fengshun'

The 'Fengshun' large model innovatively introduces intelligent generation technology of ensemble perturbation based on flow dependence, and also incorporates the key processes of air-sea interaction, mainly targeting the climate prediction problems with greater uncertainties beyond 15 days.

This system has been operationally deployed on the intelligent computing platform of the China Meteorological Administration and conducts large-sample predictions of 100 ensemble members daily, forming deterministic and probabilistic forecast test products for global basic elements and extreme events in the next 60 days, showing certain advantages in the prediction of global precipitation.

Related reading:

'Using AI to forecast the weather conditions from 0 to 15 days in the future, the China Meteorological Administration released a demonstration plan for meteorological forecast large models'.

Likes