Introduction:

In 2023, Huawei Corporation in China released the Pangu-Weather model that became the first
global artificial intelligence (AI) model to outperform the European Centre for Medium-Range
Weather Forecasts’ (ECMWF) state-of-the-art numerical weather prediction (NWP) system in
medium-range forecasts. It achieved this while being orders of magnitude faster. This break-
through spurred a surge of research and development efforts of using AI in meteorological
research and forecast applications across Asia.
The 2025 Asia Oceania Geosciences Society (AOGS) meeting provided a valuable op-
portunity for AI researchers and meteorologists from across the region to share progress,
exchange ideas, and discuss challenges and future directions. Among several AI-related
sessions at the conference, this summary will highlight key takeaways from the special
plenary session SS01 (https://www.asiaoceania.org/aogs2025/public.asp?page=special_sessions.
asp) and the associated atmospheric sciences session AS58, both of which focused on me-
teorological applications of AI.
The sessions brought together computer scientists, weather and climate researchers,
and forecasters from five Asian operational meteorological services—located in Beijing,
Busan, Hong Kong, Singapore, and Taipei—broadening the discussion to span theory, re-
search, implementation, and downstream applications. The rapid pace of AI development
in China was reflected in the field of weather forecasting. In a recent national competi-
tion, China Meteorological Administration selected 14 models from the submissions of
35 academic, government, and private industry institutions. Developers of three of these
selected models—FengWu, FuXi, and TianXing—were among the invited speakers at the
special plenary session.

pdf : Chang et al. (2025)