World Meteorological Organization

Background

AI/ML has rocked the weather world over the past year, with deep-learning based numerical weather prediction (NWP) emulators demonstrating notable skills in weather forecasts while requiring much lower computational costs than traditional NWP. Similar advancements are now also being seen for other timescales (e.g., climate) and more components of the Earth system (e.g., oceans and hydrology). These factors could democratize access to forecast information and benefit Members that previously lacked the computational resources for in-house traditional numerical prediction models. Developments are advancing at an unprecedented rate, spurred by the leading role of the private sector, compared with the usual pace of academic research. AI/ML is challenging established pathways for research, bypassing the research to operation cycle that has been the well-worn practice in the field to date. A few National Meteorological and Hydrological Services have taken these new approaches in step and are on-boarding them within their development streams. More generally, Members are turning to different WMO programmes for help in navigating this evolving landscape.  

Terms of Reference

Under the guidance of the RB, the TT-AI4Wx will build on the previous work conducted by the RB Task Team on Exascale, Data handing and AI (2020-2022) to advise on current challenges with the high-performance computing and associated big data volumes that underpin numerical predictions. The TT-AI4WX aims to identify opportunities to leverage existing WMO programmes and activities to help address immediate questions on AI/ML for Members and identify potential synergies across the WMO sponsored and co-sponsored research programmes, including the Global Atmosphere Watch (GAW), World Climate Research Programme (WCRP) and World Weather Research Programme (WWRP). 

The scope of the TT-AI4Wx is to identify within existing WMO programmes, AI/ML activities and pertinent activities that can be easily applied to AI/ML systems, and which have the potential for benefits to Members. The TT-AI4Wx will identify opportunities for the RB to leverage these activities and continue to monitor and communicate AI/ML advances and their relevance to the research community, and to Members. The plan should recognize the need for agility even when the WMO mechanisms are not working at the same speed as the AI/ML developments. 

The TT-AI4WX will: 

1. Generate a list of existing activities within WMO programmes and regional associations;
2. Review activities which can be expanded with limited resources to include AI/ML systems;
3. Identify main gaps and science questions of immediate relevance from the community of experts making up the TT; and
4. Develop a report on the above including recommendations on immediate actions. 

Deliverables

• compendium of relevant literature for the benefit of all Members
• TT-AI4Wx Interim Report (Endorsed by Research Board)