Deep learning to infer eddy heat fluxes from sea surface height patterns of mesoscale turbulence

This is a Preprint and has not been peer reviewed. The published version of this Preprint is available: This is version 2 of this Preprint.

Add a Comment

You must log in to post a comment.


There are no comments or no comments have been made public for this article.


Download Preprint


Tom George, Georgy Manucharyan, Andrew Thompson


Oceans play a major role in Earths climate by storing and transporting heat via turbulent currents called mesoscale eddies. However, direct monitoring of eddy-driven heat fluxes is currently impossible because it requires simultaneous surface and subsurface observations of velocity and heat content, while only surface properties of mesoscale eddies can be comprehensively measured by satellites in the form of sea surface height (SSH) anomalies. Nonetheless, surface and subsurface expressions of eddies are dynamically linked, suggesting that surface observations may contain at least partial information about subsurface flows and thus heat transport. Here we used deep convolutional neural networks (CNNs) to demonstrate that SSH expressions of mesoscale turbulence contain sufficient information to predict about 64\% of eddy heat flux variance, with CNNs significantly outperforming other conventional data-driven techniques. Our results suggest that deep CNNs could provide an effective pathway towards an operational monitoring of eddy heat fluxes using satellite altimetry.



Earth Sciences, Oceanography and Atmospheric Sciences and Meteorology, Physical Sciences and Mathematics, Physics, Planetary Sciences


Deep learning, Satellite altimetry, Convolutional Neural Networks, Heat Fluxes, Mesoscale Turbulence


Published: 2019-11-14 18:26

Older Versions

GNU Lesser General Public License (LGPL) 2.1