I am a seismologist studying earthquakes and Earth structures.

Machine Learning in Earth and Planetary Sciences

Graduate course, Harvard University, Earth and Planetary Sciences, 2019

This is the seminar-based class taught by Marine Denolle and Bradley Lipovsky in the fall of year 2019. They are both currently professors at University of Washington. I was fortunately to be an invited guest lecturer for this class. I gave a lecture on the topic of “Machine Learning in Seismology”. The slides can be found lecture01 lecture02

Course Description

This graduate research seminar combines machine learning with applications in the Earth and Planetary Sciences. The class will consist of literature review augmented with significant data curation and coding. Students will decide on topics and machine learning tools may include supervised learning (e.g., for seismic tomography, detection of earthquakes, impact craters, exoplanets), physics-driven discovery (e.g., estimating simple ODE’s to represent wave propagation, ENSO cycles, the earthquake cycle), time series analysis (e.g., on glacier length time series and/or climate reanalysis information), acceleration of multiphysics numerical modeling and for emulation of sub-grid scale processes (e.g., for small-scale earthquake processes, cloud parameterization, or ice sheet hydrology).

Learning outcomes are:

  1. knowledge of the successes and challenges in using machine learning (ML) for specific problems in the Earth and Planetary Sciences; 2. knowledge in the ML tools in supervised and unsupervised learning;
  2. knowledge gained through hands-on experience in a coding project;
  3. familiarity with ethical standards and debates surrounding reproducibility, transparency, open source code development, open data availability, and open access publishing.

The seminar will build toward a group or individual project given at the end of the term. Students are expected to read and present papers and work at least 2-3 hours per week outside of their class on their projects. Computing will be performed on FAS Research Computing, a personal desktop, or other cloud computing options that could become available. Class time will be dedicated to presentations and coding activities.