The meridional overturning circulation controls the fluxes of heat and carbon in the ocean on long time scales. Dense waters are sinking to the abyss at high latitudes, and are uplifted on their way south to come back to the surface and close the circulation. The classical view is that upwelling of abyssal waters is driven by turbulent mixing associated with breaking internal waves in the ocean interior. But this has been contradicted by observations, and a new paradigm has emerged in the recent years. Instead, the deep branch of the overturning circulation is shaped by highly localized turbulent processes that generate mixing close to the bottom of the ocean and drive upwelling of water masses over sloping topography. However, these small-scale turbulent processes are not well understood nor parameterized, which limits the accuracy of ocean models and the predictive skills of climate models. Theoretical and process studies are now beginning to highlight the role played by submesoscale (< 30 km) processes in the ocean bottom layer. But we still miss a clear picture of their phenomenology and impacts on the large scale circulation.
The goals of the DEEPER project are (1) to quantify the impacts of deep-sea submesoscale processes and internal waves on mixing and water mass transformations, (2) to explore ways of parameterizing these impacts using the latest advances in machine learning, i.e. applying deep learning to the deep ocean.
A hierarchy of numerical simulations will be built to characterise submesoscale processes and their interaction with the internal wave field in the deep ocean using cutting-edge realistic modelling with the ROMS/CROCO model. This problem requires simulations that resolve submesoscale processes (1-30 km), and a large enough domain to generate realistic levels of internal waves and allow to evaluate impacts on the large scale circulation. This will be the first time a terrain-following model — particularly advantageous to study flow-topography interactions — will be used at sub-kilometer resolution over a domain as large as the full Atlantic Ocean.
We will quantify how submesoscale processes and internal waves modify the large-scale budgets of energy and buoyancy, and impact mixing and water-mass transformations. We will characterize the impact of boundary potential vorticity (PV) fluxes related to diabatic processes on the mean large-scale PV budget and associated circulation. We will then characterise processes responsible for mixing in the different regions of the Atlantic Ocean: breaking of internal tides, near-inertial waves, or lee waves; formation of hydraulic jumps; or small scale instability such as gravitational, centrifugal or symmetrical instability. We will also investigate processes driving the restratification of the bottom layer such as submesoscale baroclinic instability, or deep frontogenesis. We will select some critical locations, where we observe a significant impact on water-mass transformations, and investigate the processes at play in more detail by means of very-high resolution simulations, using the non-hydrostatic version of CROCO, and idealized process studies. This downscaling will allow to check the sensitivity of the processes to the model resolution and the validity of the hydrostatic hypothesis.
Finally, we will take advantage of machine learning methods to parameterise these processes for coarser simulations. We will train a deep neural network using submesoscale resolving simulations from this project to predict mixing and eddy buoyancy fluxes for mesoscale-resolving and coarser simulations. This parameterization will then be tested over other geographical regions where similar outputs are available, and applied to different numerical models (in particular the eNATL60 simulation using the NEMO model). The long-term objective will be to be able to implement it in a global climate simulation to assess impacts on the overturning circulation over long time scales.
Monsieur Jonathan Gula (Université de Bretagne Occidentale (UBO, Laboratoire d'Océanographie Physique et Spatiale (LOPS))
The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.
UBO-LOPS Université de Bretagne Occidentale (UBO, Laboratoire d'Océanographie Physique et Spatiale (LOPS)
Help of the ANR 256,932 euros
Beginning and duration of the scientific project: March 2020 - 48 Months