The Earth and its climate form a complex system. In the last few years, research in machine learning has created new techniques for the analysis of high-dimensional non-linear systems. Many of these new techniques could improve our ability to understand and predict the Earth.
In this session, we aim to connect researchers from machine learning (ML) and computational geoscience to identify opportunities that advance the state-of-the-art in Earth and climate modeling. We invite participants to discuss (1) cutting-edge machine learning advances that are relevant to Earth and climate science problems, such as advances in the modeling and simulation of non-linear systems with generative adversarial networks; new tools for interpretable ML; or methods for placing physical constraints on ML models; (2) creative new applications of reinforcement learning techniques to Earth and climate science problems; and (3) geoscience problems that reveal needs for new research in machine learning, e.g. extreme event problems involving skewed or poorly-labeled datasets.