Climate change is one of the grand challenges facing humanity in the 21st century. Humankind can mainly act on two axes: prevention and adaptation. Adaptation is unavoidable: climate change is already having tangible impacts. Prevention is necessary to avoid the most catastrophic consequences. One way to adapt is through better weather forcasting. One way to prevent is through a better understanding of the climate system. Note that there are many other ways to act, for example in energy, transportation, agriculture, forest management, etc.

The climate and weather are modeled by running computer simulations. In a data-driven approach, scientists tailor the simulation to resemble reality (partly through an understanding of the physical processes, partly through their parameterization). With the availability of large data and compute, we can hope for machines to help in the automatic discovery of principles from data. The integration of machine learning (deep learning) within simulations opens up the exploration of new tradeoffs between accuracy and computational cost.

Project goal. Explore the use of geometric deep learning for climate and weather modeling. Depending on the interests of the student and the advice of a climate expert, we will focus on one particular aspect. Examples of problems include but are not limited to:

  • speeding up simulations, for example by replacing select model components with NN approximators
  • understanding the link between variables (e.g., ozone concentration and temperature)
  • multi-resolution (super-resolution, varying spatial resolution, prediction at intermediate time scales)
  • making good ensemble predictions (from 20+ climate models)
  • the detection of extreme events (e.g., cyclones, storms) from forecasts

Prerequisites. Highly motivated, ambitious, and independent student. Experience with Python programming. Experience with (Deep) Machine Learning and Data Science is desirable. Experience in climate or weather modeling would be great.

What to expect. The student will interact on a weekly basis with the supervisors and receive ample directions and advices. The student will produce midterm and final presentations. The student will be evaluated on his work during the semester, a report, the final presentation, and the produced code. This project can lead to a publication, and will count as an internship.

Supervisors & Contact

References

  1. Rolnick, David, et al. Tackling Climate Change with Machine Learning. arXiv:1906.05433 (2019).
  2. Monteleoni, Claire, Gavin A. Schmidt, and Scott McQuade. Climate informatics: accelerating discovering in climate science with machine learning. Computing in Science & Engineering (2013).
  3. Karpatne, Anuj, and Vipin Kumar. Big data in climate: Opportunities and challenges for machine learning. Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM, 2017.
  4. Reichstein, Markus, et al. Deep learning and process understanding for data-driven Earth system science. Nature (2019).
  5. Hourdin, Frédéric, et al. The art and science of climate model tuning. Bulletin of the American Meteorological Society (2017).
  6. Gentine, Pierre, et al. Could machine learning break the convection parameterization deadlock?. Geophysical Research Letters (2018).
  7. Rasp, Stephan, Michael S. Pritchard, and Pierre Gentine. Deep learning to represent subgrid processes in climate models. Proceedings of the National Academy of Sciences (2018).