Stochastic Transport in Upper Ocean Dynamics��

A non exhaustive rapid overview of the efforts from the Ifremer team and close collaborators

​

​

​

WP1 : Multi-modal ocean data (numerical, in situ and satellite observations) compilation, analysis and interpretation: Model-data driven methods to decompose and calibrate slow-fast upper ocean motions

�

• A Golden age for multi-modal oceanic observations (satellite+in situ), along operational models and high-resolution numerical simulation
• Fundamental progress in the upper ocean observation: Multi-medium-res satellite data (altimeters, next SWOT, microwave sea surface temperature and salinity, color, surface winds and waves), geostationary high frequency data (SST, Sea Color), buoys, drifters, Argo interior ocean profilers ...

​

Altimetry Ocean Kinetic Energy

• A Golden age for multi-modal oceanic observations (satellite+in situ), along operational models and high-resolution numerical simulation
• Fundamental progress in the upper ocean observation: Multi-medium-res satellite data (altimeters, next SWOT, microwave sea surface temperature and salinity, color, surface winds and waves), geostationary high frequency data (SST, Sea Color), buoys, drifters, Argo interior ocean profilers ...

​

Daily 10 km Sea Surface Temperature

Year

Day

Hour

Internal Waves

Inertial

motions

​

Mesoscale

circulation

​

​

Basin

circulation

Minute

Thermo-Haline

Circulation

1000km

10km

1m

100km

1km

Horizontal scales

Tides

​

Sub-mesoscale

circulation

​

1-10 m

100 m

1000 m

Vertical

scales

Typical Ocean and Atmosphere Space-TimeScales

Ocean-Atmosphere Interactions

Differing Time-Space correlations

(some) Processes at the AIR-SEA interface

Surface stress (impacted by ocean velocity and by air velocity, which is affected by SST) 

Boundary layer thickness (which varies by 2 orders of magnitude in different stability conditions)

The slow ocean is becoming a spatial noise for the fast atmosphere forcings/motions (revisiting the Hasselmann’s climate stochastic model)

��

Long‐term impacts of ocean wave‐dependent roughness on global climate systems

Earth system complexity: Ocean-Atmosphere, local and non-local multi-scale interactions

WRCP : 2028, an International Earth System Year, with intensive observational and modeling activities to investigate the complexity of planetary dynamics

Long‐term impacts of ocean wave‐dependent roughness on global climate systems

Earth system complexity: Ocean-Atmosphere, local and non-local multi-scale interactions

WRCP : 2028, an International Earth System Year, with intensive observational and modeling activities to investigate the complexity of planetary dynamics

MABL rolls induce

Quasi-periodic perturbations in wind speed

Instantaneous change in the sea surface roughness and surface stress

MC

WS

NA

Unstable

• more negative Ri
• favorable for MC

Near-neutral

• weakly negative Ri
• favorable for WS

Stable

• positive Ri
• only turbulence left

GRIDDED MODELING PRODUCT

(i.e., model analysis and forecast)

SATELLITE ESTIMATES

(HARMONY-like SVW measurements)

02-02-2005

6:00hr

02-02-2005

6:00/9:00hr

QSCAT

ECMWF

Statistical Characterisation of Observed vs Modelled Wind Properties

• Surface winds statistics from ECMWF analysis with various satellite estimates along “long” lines in Mediterranean Sea

Distance [km]

Distance [km]

Distance [km]

Random Samples

Example of sampled lines

ASCAT

ECMWF analysis, Mediterranean Sea

ASCAT (METOP-A), 2013, Mediterranean Sea

QuikScat, 2005, Mediterranean Sea

Med Sea, averages of PSD computed over 240 km lines

WRF Winds Time window: 2014-10-06 to 2014-10-10.

Region of interest: Ligurian Sea (domain-03).

Grid resolution: 1.4 km.

Statistical Characterisation of Observed vs Modelled Wind Properties

For profoundly complex systems, a systematic misrepresentation of small-scale processes can lead to a systematic misrepresention of large-scale processes

Not only the Stokes drift depends on time, its characteristic vertical scale is also a function of time and space (fetch laws)

Methodology-assumption: �Observations bear the full probabilistic signatures of all these complex effects

• Infer spatio-temporal statistical properties from data
• Linking spatio-temporal properties with consistently-derived rigorous theoretical models and/or data analysis and Machine Learning methods
• Quantify upscale effects of unresolved sub-mesoscales processes on the large-scale circulation

Globcurrent currents, floaters, temperature

500 m numerical vorticity field

Coupled-Ocean-Atmosphere Dynamics: Sensor-Model synergies for stochastic descriptions

From macroscopic point of view, O-A interactions are nonlinear and subject to random fluctuations. The nonlinearity introduces subtleties into how noise influences a predictable horizon

Data-Physics Constrained Model: blend-approaches

Data-driven:

• Use multi-modal observations
• Sparse observations with gaps and noise
• Use analog forecasting from catalogue (data-driven and/or numerical simulations)

Model-driven:

• Use a dynamical/statistical model or a background
• Model affected by initial condition and errors

From Carrassi et al. 2018 WIREs Climate Change

Observations

Hidden state

Time Series Analysis: Koopman (Yicun Zhen et al., PRE 21),

Analog method (Paul), NN methodology (Said)

Data assimilation 4DvarNN (Ronan Fablet)

IR very High-resolution O(1km) vs Microwave Passive Medium-resolution O(25km)

• Catalogues of millions of co-located observations (learning multi-scale interactions and energy distributions, analogs and DL techniques, joint SSH-SST analysis)
• Comparisons with numerical simulations
• Extended O-W analysis (Valentin, JPO21) towards Stochastic O-W …

Snapshots Analysis

Internal and Surface wave-current interactions

Ray-paths in a fluctuating random environment; stochastic vs deterministic settings (Valentin+ODL team)

CMEMS Model vs. Observed surface current

Sentinel 1

Observations

Mercator Model 1/12°

Geostrophic + random small scales

Surface wave-current interactions

Ray-paths in a fluctuating random environment; stochastic vs deterministic settings (Valentin+ODL team)

Preparing experiments: C-SWOT-2023

• Unique context:
• Fast sampling-phase for SWOT -> Daily revisit (March -> May 2023)

​

19

CN

BF

North Gyre

Shcherbina, A. Y. et al. Statistics of vertical vorticity, divergence, and strain in a developed submesoscale turbulence field. Geophysical Research Letters 40, 4706–4711 (2013).

20

+

SHOM

CNFC

Two 7-day legs

​

- North branch

​

- Balearic front

OK

OK

Stochastic Transport in Upper Ocean Dynamics��

Let’s discuss our next steps and plans this week !

​

​

​

​

​

WP1 : Multi-modal ocean data (numerical, in situ and satellite observations) compilation, analysis and interpretation: Model-data driven methods to decompose and calibrate slow-fast upper ocean motions

�