News

Mapping the AI weather forecasting - review powered by UPWr-UWr collaboration

29-09-2025

A cross-disciplinary team from Wrocław has published a concise, state-of-the-art review of AI weather models in Engineering Applications of Artificial Intelligence. The article "Beyond the horizon: A comprehensive analysis of artificial intelligence-based weather forecasting models" surveys ca. 40 systems (e.g., GraphCast, Pangu-Weather, FourCastNet), compares data, lead times, and verification practices, and pinpoints gaps.

The study brings together AI/observation experts (Dr Saeid Haji-Aghajany and Prof. Witold Rohm) from the Institute of Geodesy and Geoinformatics (UPWr) with meteorology/climatology researchers (Prof. Maciej Kryza) at UWr’s Department of Climatology and Atmosphere Protection, and computer-science/AI specialists from UWr’s Computational Intelligence Research Group (prof. Piotr Lipiński) - a synergy bridging observations and modelling with cutting-edge ML.

Why it matters:

• Clear map of the field: What AI does well today, where it lags (especially record-breaking extremes), and how to test it fairly.
  
• Practical guidance: Calls for shared baselines, transparent scorecards, and probabilistic, hybrid AI-plus-physics approaches for operation.

This maps the way for future development of AI weather models and its implication for geodesy and remote sensing. It also demonstrates that collaboration between diverse range of specializations are needed to push the boundaries of AI atmosphere modeling and forecastin.

Paper: Beyond the horizon: A comprehensive analysis of AI-based weather forecasting models, Engineering Applications of Artificial Intelligence, vol. 162, Part A, 112335, Haji-Aghajany S., Rohm W., Lipiński P., Kryza M., 2025.