Accelerating Aerosol Hygroscopicity Prediction with Deep Learning and HPC Deployment
Aerosol hygroscopicity strongly influences clouds, radiation, and air quality, but physics-based simulations are computationally expensive. This project develops a machine-learning emulator to predict Ƙsu and Ƙca directly from atmospheric composition fields, as a step toward an AI-based aerosol process component for integration into high resolution global climate models within the DestinE Climate Digital Twin initiative.
Authors
- Harri Kokkola (ORCID: https://orcid.org/0000-0002-1404-6670)
- Billy Braithwaite
- Anton Laakso (ORCID: https://orcid.org/0000-0001-7055-5691)
- Radek Halfar (ORCID: https://orcid.org/0000-0003-1453-0101)
- Juha Tonttila (ORCID: https://orcid.org/0000-0001-6033-1862)
- Antti Vartiainen (ORCID: https://orcid.org/0009-0004-7184-4750)
- Arup Sarkar
Institutions
- VSB - Technical University of Ostrava (CZ)
- University of Eastern Finland (FI)
- Finland University (FI)
- CSC - IT Center for Science (Finland) (FI)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-07-17
- DOI
- https://doi.org/10.5281/zenodo.21414090
- Primary Topic
- Atmospheric aerosols and clouds
- Type
- article
- Field-Weighted Citation Impact
- 0.00