Offre en lien avec l’Action/le Réseau : – — –/Doctorants
Laboratoire/Entreprise : Institut des Géosciences de l’Environnement (IGE)
Durée : 12 mois (+ext.)
Contact : Julien.Lesommer@univ-grenoble-alpes.fr
Date limite de publication : 2021-07-31
Contexte :
The Institute for Environmental Geosciences (IGE) is looking for a post-doctoral researcher to develop subgrid scale parameterizations for ocean models based on machine learning in the frame of the M2LINES project (https://m2lines.github.io). He/she will focus on improving the representation of submesoscale processes in the NEMO-OCE ocean model (https://www.nemo-ocean.eu) with Deep Neural Networks.
Sujet :
The postdoctoral researcher will use a coarse graining approach applied to high resolution ocean numerical simulations and Deep Neural Networks in order to formulate parameterizations of the impact of unresolved turbulent processes in coarser resolution ocean models. The work will involve analysing several kilometric resolution ocean model simulations through a cloud-based system and formulating inverse problems for estimating the contribution of unresolved processes on ocean dynamics from coarse grained information. The inverse problems will then be solved with physics-aware machine learning algorithms based on Deep Neural Networks. The learned subgrid closures will eventually be tested within the NEMO-OCE ocean model (https://www.nemo-ocean.eu) using SmartSim (https://github.com/CrayLabs/SmartSim) in order to assess their a posteriori skills in realistic simulations. The subgrid closures will focus in priority on the representation of the impact of submesoscale variability on air-sea interactions and ocean surface boundary layer dynamics in ocean climate models. The research will be conducted as part of the M2LINES project (https://m2lines.github.io) and will involve several international collaborators.
Profil du candidat :
The candidates research track record should demonstrate their ability to carry out cutting edge research in one of the following fields : ocean fine scale processes, ocean modelling, turbulent closures, data-driven large eddy simulation via machine learning. Their research background should demonstrate their strong interest in approaches bridging physical science and computational science. Computational skills should include Python, FORTRAN and some experience with one of the prominent software libraries in machine learning (in particular PyTorch or TensorFlow). They should speak and write English fluently and be able to interact in a multicultural environment. They will need to demonstrate curiosity, autonomy and initiative
Formation et compétences requises :
The expected candidates should hold a PhD in physical oceanography, atmospheric science or computational fluid dynamics.
Adresse d’emploi :
https://bit.ly/3AyOiD5