Machine Learning Approaches for Determining Causality in Coupled Earth System Data

Determining causality in earth system data is challenging because of strong coupling between different variables. Study of state-of-art statistical approaches that are designed to infer causality between variables that are strongly coupled on different time scales--for example, ocean-atmospheric coupling and land-vegetation-atmospheric coupling, and for nonlinear coupling. Methods include but not limited to Granger causality, generalized equilibrium feedback assessment, step-wise generalized equilibrium feedback assessment, empirical dynamic modeling, and area weighted connectivity. Offers stimulating group learning experience through reading papers and discussion, and if possible, application of some of methods to earth system data.

S/U grading.