Characterizing turbulent heat transport in the ocean using machine learning

Miles Couchman (University of Cambridge) Cambridge Fluids Network - fluids-related seminars 14 October 2021 11:30am Open Plan Area, BP Institute, Madingley Rise CB3 0EZ To book an in-person seat at the seminar this Thursday please use the link below - please note that once the number is reached you will not be able to register but will be able to watch via the Zoom link at the bottom of this email. There will also be a selection of individually wrapped sandwich lunches for everyone who is attending in-person. https://stokes.cceif.group.cam.ac.uk:1763/seminarreg/163045 Face coverings are expected to be worn on arrival at the BPI Institute AND during the seminar and in all communal areas. Face coverings should only be removed once you have collected lunch and are back at your seat. Please also note that for ventilation windows in the Open Plan Area must remain open at all times. BPI Seminar - Thursday 14 October 2021 11.30 Miles Couchman, University of Cambridge Characterizing turbulent heat transport in the ocean using machine learning Turbulence enhances the mixing of heat between different layers of the ocean, playing a critical role in driving global currents that influence the Earth’s climate. Measurements of centimeter-scale velocity and temperature fluctuations, termed microstructure, currently provide the best observational means of probing such mixing and characterizing its variability in space and time. We present a new, data-driven technique for analyzing microstructure data that uses an unsupervised machine learning clustering algorithm to identify fluid regions with similar measured characteristics automatically, yielding insight into the underlying turbulent mechanisms driving mixing. Our results highlight the intermittent nature of the turbulence, demonstrating that relatively rare, but extreme events have the potential to dominate the bulk mixing statistics. As an extension to this work, we discuss a recent analysis of direct numerical simulations of stratified turbulence, revealing insight into mixing dynamics on length scales that are not yet resolvable in oceanographic measurements. To Attend Via Zoom: Time: Oct 14, 2021 11:15 AM London Join Zoom Meeting https://zoom.us/j/93848803121 Meeting ID: 938 4880 3121 To attend the meeting, just open the meeting link in e.g., a web browser, which, if you have not already installed Zoom, will start a download and the quick installation of a small client will be necessary. If you already have Zoom, you already know what to do - and the link can be entered as the message ID. The link will be live from 11.15am. By default you will be muted and not emitting video, so remember to unmute yourself before asking questions. This talk is part of the Seminars for the Centre for Environmental and Industrial Flows series.