On June 16, the promise of Artificial Intelligence (AI) to revolutionize forecasts and help build resilience to more extreme weather and climate impacts is high on the agenda of the 79th Session of the World Meteorological Organization (WMO)' s Executive Council. 

Starting from 2024, the China Meteorological Administration (CMA) has collaborated with domestic and international meteorological departments, universities, and research institutions to strengthen the AI-abled application in meteorological monitoring, forecasting, warning and services, providing vigorous technological support for the United Nations' Early Warnings for All (EW4ALL) initiative and meteorological disaster prevention and mitigation.

4 AI-abled weather models

Al Nowcasting Model (CMA-AIM-Nowcast-Fenglei)

CMA and Tsinghua University jointly build “Fenglei”, which can provide information on the generation, dissipation and evolution of the convective storms in response to common issues in global nowcasting.

It is possible to generate a national radar echo extrapolation product of 1 kilometer for 0-3 hours every 6 minutes on Tianqing system within 3 minutes.

The Technical Architecture of Fenglei

Al Global Weather Forecast Model (CMA-AIM-GFS-Fengqing)

CMA and Tsinghua University set up a research team to establish "Fengqing".

This model achieves a deep integration of atmospheric physics and data-driven methods, enabling efficient computation while providing a physically interpretable basis for prediction results.

lt can generate global forecasts of atmospheric variables such as precipitation, wind, and temperature with a 25-kilometer resolution every 6 hours for the next 15 days within 3 minutes, providing detailed and accurate weather information for people's daily lives, travel, and tourism.

"Fengqing" Forecasting Products and Service Application Scenario on the SWAN Platform

AI Global Climate Forecasting Model (CMA-AIM-S2S-Fengshun)

CMA cooperated with Fudan University and the Shanghai Academy of Artificial Intelligence Science, and build “Fengshun” based on AI.

The model runs daily to provide 100 ensemble members for the next 60 days with significantly higher computational efficiency, which also takes atmosphere-ocean coupling into consideration, thus enhancing the prediction skills of Madden–Julian Oscillation (MJO).

The Schematic Diagram of the Structure of “Fengshun”

AI Space Weather Forecasting Model (CMA-AIM-SW-Fengyu)

CMA jointly researched and developed "Fengyu" with Nanchang University and Huawei Technologies Co Ltd.

The model includes a full-chain AI model for space weather, covering multiple physical regions such as solar wind, magnetosphere and ionosphere. It also is equipped with core functions such as ionospheric disturbance prediction, geomagneticstorm response, and solar wind forecasting.

The Technical Architecture of "Fengyu"

Cloud-based Early Warning System

Focusing on the 4 pillars of EW4ALL- Disaster risk knowledge, Detection, observation, monitoring, analysis, and forecasting, Warning dissemination and communication, and Preparedness and response capabilities, Cloud-based Early Warning System incorporates AI technology and meteorological forecasting models to support the development of disaster warning capabilities in developing countries.

The System has been applied in countries including Pakistan, Ethiopia, and the Solomon Islands.

The design diagram of the System

Urban Multi-hazard Early Warning Toolbox

Urban Multi-hazard Early Warning Toolbox is equipped with a comprehensive set of institutional tools, algorithmic models, business system tools, and case studies based on 4 pillars of EW4ALL. It also features 5 major application scenarios: data intelligent compression, forecast intelligent analysis, multi-scenario assessment coordination, automatic collection of disaster situation reports, and disaster prevention knowledge dissemination.

Currently, the Toolbox has been in trial application in 12 countries and regions, covering multiple disaster scenarios such as urban flooding, typhoons and rainstorm.

The forecaster of Papua New Guinea National Weather Service used the Toolbox to overlay satellite data and quickly identify high-risk areas. Source: Shanghai Meteorological Service

Editor: JIANG Zhiqing