Redouane Lguensat

Dr. Redouane Lguensat is a MOPGA Postdoc on Project Hermès from March 2020.

His research interests are Data assimilation, physically constrained deep learning and methods for parameter estimation. His PhD thesis consisted in the use of analog methods (K-Nearest Neighbor Regression) to develop a new data-driven approach to tackle data assimilation: the Analog Data Assimilation (AnDA). In parallel he also investigated the use of deep learning methods for the segmentation of oceanic eddies from satellite-derived sea related maps and developed EddyNet. In 2018, Redouane Lguensat was a recipient of a postdoctoral grant from CNES (French Space Agency), where he investigated the use of deep neural networks inside ocean numerical codes (e.g. 1-layer quasi geostrophic model).

orcid.org/0000-0003-0226-9057

Select bibliography

• Redouane Lguensat, Julien Le Sommer, Ronan Fablet, Sammy Metref, Emmanuel Cosme. Learning Generalized Quasi-Geostrophic Models Using Deep Neural Numerical Models Machine Learning and the Physical Sciences workshop, NeurIPS 2019. Vancouver, Canada. Arxiv, code
• Redouane Lguensat, Pierre Tandeo, Pierre Ailliot, Manuel Pulido and Ronan Fablet. The Analog Data Assimilation. Monthly Weather Review, 2017, vol. 145, no 10, p. 4093-4107. journal webpage
• Redouane Lguensat, Miao Sun, Ronan Fablet, Evan Mason, Pierre Tandeo, Ge Chen. EddyNet: A Deep Neural Network For Pixel-Wise Classification of Oceanic Eddies. IEEE Geoscience and Remote Sensing Symposium (IGARSS) 2018, Valencia, Spain. IEEExplore link. pdf
• Redouane Lguensat, Phi Huynh Viet, Miao Sun, Ge Chen, Tian Fenglin, Bertrand Chapron, Ronan Fablet Data-driven Interpolation of Sea Level Anomalies using Analog Data Assimilations. Remote Sensing, 2019. pdf

• Ronan Fablet, Phi Huynh Viet, and Redouane Lguensat. (2017) Data-Driven Models for the Spatio-Temporal Interpolation of Satellite-Derived SST Fields. IEEE Transactions on Computational Imaging 3:4, 647-657. journal webpage, pdf

• Google Scholar
• Personal Website