PCAPS ORCAS at EGU25, 27 April–2 May 2025, Vienna, Austria
Our PCAPS Observational Requirements in the Context of AI prediction Systems (ORCAS) Task Team had the chance to share our work advancing data-driven approaches to sea ice prediction at the EGU General Assembly 2025, which took place from 27 April to 2 May in Vienna, Austria, with more than 18,000 in-person participants in attendance. Task Team Co-Chairs Lorenzo Zampieri (ECMWF) and Clare Eayrs (KOPRI, a member of the PCAPS steering group) presented a poster that highlighted the goals and challenges of the ORCAS initiative.
Eayrs, C. and Zampieri, L.: Observational Requirements in the Context of AI prediction Systems - a PCAPS ORCAS Task Team, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14272, https://doi.org/10.5194/egusphere-egu25-14272, 2025.
Our time at EGU25 was marked not only by engaging discussions during our poster session, but we also used the opportunity to network and create greater awareness about PCAPS. Lorenzo also chaired the Cryosphere session “Advances in sea-ice modelling: developments and new techniques” and presented his research on “Coupling approaches for data-driven Earth system models” in the highly popular and overflowing Earth and Space Science Informatics session on “Machine Learning and Digital Twins for Earth System Observation and Prediction”. In a fun science communication session, Clare presented progress on the GOAT initiative (GOAT- Getz-Ocean interactions, sentinel of Antarctic Transition to a warming climate) that was featured in a recent PCAPS blog post.
Clare Eayrs and Lorenzo Zampieri presenting the ORCAS poster at EGU25. Photo credit: InWoo Park.
As climate change continues to drive rapid and sometimes unexpected changes in the polar regions, environmental risks due to increased human activity are increasing, and so is the demand for user-tailored forecasting services. Sea ice prediction has traditionally relied on physics-based numerical models, which are computationally intensive and sensitive to initial conditions. AI models, in contrast, are demonstrating comparable or superior performance while using fewer computational resources. These systems open the door for faster, more frequent predictions, enhancing decision-making capabilities for communities, operations, and research in the polar regions.
One of the most pressing challenges, however, is how to effectively train and validate AI models using real-world observations. Observational integration is essential not only for model training and initialisation, but also for evaluating physical realism and building scientific credibility. The ORCAS Task Team is addressing this challenge by evaluating data-driven model performance across diverse datasets, identifying which types of observations are best suited for training or validation, and supporting the design of future observing systems, including those planned for initiatives like Antarctica InSync and the fifth International Polar Year.
Drumming up interest in our poster during a 2-min talk held the previous day. Photo credit: Yeon Choi
A particularly exciting aspect of EGU25 was the strong presence of early career researchers working at the intersection of AI and polar science. The PCAPS ORCAS team is dedicated to fostering this emerging community and encouraging new voices and collaborations in this fast-evolving field.
Clare Eayrs attendance was supported by Korea Institute of Marine Science & Technology Promotion (KIMST) funded by the Ministry of Oceans and Fisheries (RS-2023-00256677; PM23030).