We present the GAIA (Geospatial Artificial Intelligence for Atmospheres) Foundation Model, a novel model that combines masked autoencoders (MAE) and self-DIstillation with NO labels (DINO) for analyzing global atmospheric patterns in satellite imagery. By integrating these complementary self-supervised learning approaches, our model simultaneously captures both local features and global dependencies. We address two critical challenges in satellite data analysis: reconstructing missing regions and estimating precipitation patterns as our first downstream tasks. The model demonstrates superior temporal pattern capture compared to standard MAE approaches, while maintaining robust performance in downstream tasks. Our experimental results show strong gap-filling capabilities across varying mask ratios and accurate precipitation estimation with limited training data, achieving a false alarm ratio of 0.088 and structural similarity of 0.881. This work represents an advancement in self-supervised learning for atmospheric science, providing a foundation for improved weather monitoring and climate analysis. The trained model weights and accompanying code are publicly available as open-source on Hugging Face here:this https URL.
View on arXiv@article{asanjan2025_2505.18179,
title={ GAIA: A Foundation Model for Operational Atmospheric Dynamics },
author={ Ata Akbari Asanjan and Olivia Alexander and Tom Berg and Clara Zhang and Matt Yang and Jad Makki and Disha Shidham and Srija Chakraborty and William Bender and Stephen Peng and Arun Ravindran and Olivier Raiman and David Potere and David Bell },
journal={arXiv preprint arXiv:2505.18179},
year={ 2025 }
}