top of page

Research

I am interested in understanding of ocean's role in climate, particularly the role of ocean meso-scale processes in large-scale circulation and its representation in climate models.  My current research focuses on the development of high resolution ocean model at 1/12° and 1/36° that can be coupled with high resolution atmosphere model:  
- spin-up strategy;
- improvement of eddy parameterisation.

IMG_0114.PNG

Machine Learning for reconstruction of ocean surface pCO2

The global ocean is a major sink of excess carbon dioxide CO2 that has been emitted to the atmosphere since the beginning of the industrial revolution. The ocean carbon sink is estimated by surface ocean partial pressures of carbon dioxide pCO2. However, in-situ measurements of pCO2 are sparse in space and time. 

​

We developed a Machine Learning method based on Feed-Forward Neural Network and SOCAT database to reconstruct ocean surface pCO2 over the global ocean. The detailed description of ML data-driven method is published in LSCE-FFNN-v1: a two-step neural network model for the reconstruction of surface ocean pCO2 over the global ocean.

The data product is available at Copernicus Marine Environment Monitoring Service.

​

In frame of AtlantOS project we provided observation system simulation experiences (OSSE) to obtain an optimal observation system for surface ocean pCO2 in the Atlantic Ocean and the Atlantic sector of the Southern Ocean. The results are published in Observation system simulation experiments in the Atlantic Ocean for enhanced surface ocean pCO2 reconstructions.

IMG_0113.PNG

Improvment of meso-scale parameterisation in ocean global models using Machine Learning

The role of mesoscale eddies is crucial for ocean circulation and its energy budget. At scales of 10 to 300 km, the mesoscale eddies transfer hydrographic properties and energy at different spatial and temporal scales, hence contributing to equilibrating large scale ocean dynamics and thermodynamics, which is paramount for long-term climate modelling [Olbers et al., 2012]. They also affect biogeochemical tracers, which in return influence ocean thermodynamics (through light penetration), climate and ecosystems, hence representing correctly their effect in ocean models is of greatest importance.

​

In the framework of the project Hermès, founded under the Make Our Planet Great Again call, we worked on the reconstruction of turbulent part of buoyancy flux vertical component using Machine Learning and filtered and averaged over coarse grid (large-scale) outputs of high-resolution regional ocean model eNATL60.

 

First results were presented at the AGU 2019: Reconstruction of Sub-grid-scale Buoyancy Fluxes from Large-Scale ocean Variables.

Full report can be found here.

IMG_0096.PNG

Machine Learning and parametrisation in global biogeochemical model PlankTOM

The transfer of ecosystem dynamics into carbon export to the deep ocean is the subject of further development in global carbon cycle models. Models build the CO2 flux from the atmosphere to the ocean by absorbing carbon into the ocean interior. When carbon is transported from the surface to intermediate and deep ocean, more CO2 can be absorbed at the surface. Even small variability in sinking organic carbon fluxes can have a large impact on air-sea CO2 fluxes, and on the amount of CO2 emissions remained in the atmosphere.

​

In models the concentration of sinking materials is presented by small and large particulate organic carbon concentration. Small particulate organic carbon sinks with the constant speed 3 m/d. The sinking speed of large organic carbon is presented by the function of particle density and ballast by CaCO3 and SiO2. The representation of sinking velocity for small and large particulate organic carbon needs further improvement.  

At the first step we started from the reconstruction of small and large particulate organic carbon concentration over the global ocean based on the observations and using Machine Learning.

​

First results were presented at the EGU2021: Using new observations and Machine Learning to improve organic sinking processes in the PlankTOM global ocean biogeochemical model.

bottom of page