Keynote talk - Hybrid and causal machine learning in the Earth sciences

Most problems in Earth sciences aim to do inferences about the system, where accurate predictions are just a tiny part of the whole problem. Inferences mean understanding variables relations, deriving models that are physically plausible, that are simple parsimonious, and mathematically tractable. While machine learning models excel as approximators, they often disregard fundamental physics laws, compromising consistency and confidence. We propose exploring the interplay between domain knowledge and machine learning with hybrid and causal machine learning models as necessary steps toward understanding the data-generating process. I will discuss recent approaches in the field to attain consistent and explainable results. This work outlines a collective, long-term AI agenda for developing algorithms that can discover knowledge in the Earth system.

Camps-Valls, G., Svendsen, D. H., Cortés-Andrés, J., Moreno-Martínez, Á., Pérez-Suay, A.,  Adsuara, J., Martín, I., Piles, M., Muñoz-Marí, J., Martino, Luca & Martino, L. (2020). Living in the physics and machine learning interplay for earth observation. arXiv preprint arXiv:2010.09031.

‍

Tuia, D., Roscher, R., Wegner, J. D., Jacobs, N., Zhu, X., & Camps-Valls, G. (2021). Toward a collective agenda on ai for earth science data analysis. IEEE Geoscience and Remote Sensing Magazine, 9 (2), 88-104.

‍

Camps-Valls, G., Gerhardus,A., Ninad, U., Varando, G., Martius, G., Balaguer-Ballester, E., Vinuesa, R., Diaz, E., Zanna, L., Runge, J. (2023), Discovering causal relations and equations from data, Physics Reports, Volume 1044, pp. 1-68.