Atmosphere Ocean Science Friday Seminar

---

Deep Learning of Ocean Turbulence

Speaker: Prof. Laure Zanna, Associate Professor

Location: Warren Weaver Hall 1314

Date: Friday, September 20, 2019, 3:30 p.m.

Synopsis:

Ocean eddies play an important role in the ocean circulation. Mesoscale eddies impact processes such as the mixing of climatic tracers, the modulation of large-scale ocean currents and the transfer of energy. However, mesoscale eddies are not fully resolved in current climate models, creating the need for parameterisations of the missing turbulent processes associated with these eddies.  I will discuss briefly the “traditional” approach to turbulence parametrisation, and contrast it to a more “modern” approach using the latest developments in machine learning, specifically convolutional neural networks and relevance vector machine. I will show some results from both approaches in idealised ocean models, and discuss how we can best merge our knowledge of mathematics and physics of ocean turbulence with fast emerging tools from deep learning.