Abstract

Title: “Living in the machine learning and physics interplay for the Earth science” Abstract: Most Earth science problems involve making inferences about the system, where accurate predictions are just a tiny part of the problem. Inferences mean understanding variables relations and deriving physically interpretable models that are simple, parsimonious, and mathematically tractable. Machine learning models alone are excellent approximators but often do not respect the most elementary laws of physics, like mass or energy conservation, so consistency and confidence are compromised. I will review the main challenges ahead in the field and introduce several ways to live in the Physics and machine learning interplay that allow us constrain data-driven models with physics-priors and causal dependence constraints, improve parameterizations, emulate radiative transfer models, and capture causal modes of variability of the Earth system. This is a collective long-term agenda towards developing pragmatic AI methods capable of discovering knowledge in the Earth system.