Next-Generation Modeling of Cloud Feedbacks and Climate Change using AI: Implications for Alternative Energy
| Reference number | |
| Coordinator | Lunds universitet - Institutionen för naturgeografi och ekosystemvetenskap |
| Funding from Vinnova | SEK 5 634 189 |
| Project duration | November 2020 - October 2025 |
| Status | Completed |
| Venture | AI - Leading and innovation |
| Call | AI in the service of climate |
Important results from the project
Most goals of the project were met. However, objectives 3 and 4 were not completed, which prevented SMHI from becoming involved. The neural network was constructed and tested offline. Adequate accuracy was demonstrated offline. There were ten publications and the postdoc´s report that acknowledge Vinnova support (see list attached). Five of these document aspects of Tasks 1 and 2 in the original proposal. The report by the chief postdoc is at https://doi.org/10.5281/zenodo.17367602
Expected long term effects
The project demonstrated that neural networks can, in principle, represent the ´memory´ of natural clouds and the microphysics of aerosol–cloud interactions in climate models. Although implementation challenges prevented stable simulations, it strengthened national competence in AI-based climate modeling and built lasting links between machine learning and meteorology. The experience will guide future interdisciplinary projects and the development of next-generation climate models.
Approach and implementation
The project aimed to develop and test an AI-based cloud parameterization using a single-column model as a test-bed. The plan involved embedding our cloud model in a global model as a super-parameterization to create the training dataset. Activities were scientifically relevant, but implementation was affected by organizational and technical challenges that delayed progress. Key components were completed, providing experience and lessons for future interdisciplinary climate–AI projects.