1 of 49

Development of the GEOS-MITgcm atmosphere-ocean model for coupled data assimilation

​

Ehud STROBACH1,2, Andrea MOLOD2, Atanas TRAYANOV2,3, William PUTMAN2, Gael FORGET4, Jean-Michel CAMPIN4, Chris HILL4, Dimitris MENEMENLIS5, Patrick HEIMBACH6

�1University of Maryland, United States, 2NASA / GMAO, United States, 3Science Systems and Applications, Inc., 4Massachusetts Institute of Technology, United States, 5Jet Propulsion Laboratory, California Institute of Technology, United States, 6University of Texas at Austin, United States

​

​

1

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

2 of 49

Overall motivation of the research program

    • Couple the models underlying the MERRA-2 atmospheric reanalysis (GEOS) and the ECCO-v4 ocean state estimate (MITgcm).
    • Develop a prototype ocean-ice-atmosphere coupled data assimilation system.
    • Work toward closed budget global data assimilation system.

​

2

Jun 2018

Applications

    • Recent sea ice and ice sheet changes.
    • Sub-seasonal to decadal climate predictions.
    • Observation System Simulation Experiments (OSSEs).

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

3 of 49

Overview

  • Part I: Towards a closed budget planetary assimilation system GEOS- MIT model.
  • Part II: Air sea interactions in the high resolution GEOS-MIT.
  • Part III: Consequences of different air-sea feedbacks on the ocean using MITgcm with MERRA-2 forcing.

3

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

4 of 49

Part I: �Towards a closed budget planetary assimilation system�  GEOS-MIT model�

4

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

5 of 49

GEOS GCM main relevant features

  • Dynamical core: finite-volume (Lin, 2004)
  • Physics:
    • Moist processes: Based on the Relaxed Arakawa–Schubert (RAS) scheme (Moorthi and Suarez,1992).
    • Turbulent mixing: Non-local scheme (Lock et al., 2000).
    • Surface layer: Monin–Obukhov similarity theory (Helfand and Schubert,1995).
    • Radiation: long wave (Chou and Suarez, 1994), short wave (Chou and Suarez, 1999).
    • Gravity wave drag: Orographic (McFarlane, 1987) and non-orographic (Garcia and Boville, 1994).
    • Land surface model: Koster et al. (2000).
    • Chemistry: Goddard Chemistry, Aerosol, Radiation, and Transport (GOCART, Chin et al. 2002).
    • Glacial thermodynamic: Cullather et al. (2014)

5

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

6 of 49

MITgcm main relevant features

  • Dynamical core: finite-volume (Adcroft et al., 1997).
  • Nonlinear free-surface & real freshwater flux.
  • Physics:
    • Sub-grid scale eddy parameterization: (Gent and Mcwilliams, 1990; Redi,1982).
    • Ocean vertical mixing:
      • KPP - The nonlocal K-profile parameterization scheme (Large et al., 1994).
      • GGL90 - TKE vertical mixing scheme (Gaspar et al., 1990).

6

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

7 of 49

GEOS-MIT air-sea coupling

Exchange grid

  • The exchange grid is a new grid composed of all cells enclosed by the two grids intersections.
  • Exchange of properties between the ocean and the atmosphere is done on the exchange grid.
  • Conservative exchange of water heat and momentum.

Sea-ice

  • Thermodynamics – Los Alamos Sea Ice model (CICE4) (Hunke and Lipscomb 2010).
  • Advection – viscous-plastic (VP) model (Hilber, 1979; Hilber, 1980; Losch et al. ,2010).

7

Jun 2018

Atmospheric

grid

Oceanic

grid

Exchange grid

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

8 of 49

Experimental setup

  • 10 year run (2000/04-2010/04)
  • Atmosphere – GEOS:
    • Atmospheric initial conditions – MERRA-2
    • Horizontal grid type – Cubed sphere, 1◦ X1◦ .
    • Vertical grid type – hybrid sigma-pressure, 72 levels
  • Ocean – MITgcm
    • Oceanic initial conditions – ECCO-v4
    • Horizontal grid type – Lat-Lon-Cap, 1◦ X1◦
    • Vertical grid type – z* rescaled height vertical coordinate, 50 levels

8

Jun 2018

Cubed sphere grid (left) and Lat-Lon-Cap (right)

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

9 of 49

Net heat flux

9

Jun 2018

GEOS-MIT

ECCO-v4

ECCO-v4 – GEOS-MIT

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

10 of 49

Global ocean temperature drift

10

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

11 of 49

Net fresh water flux

11

Jun 2018

GEOS-MIT

ECCO-v4

ECCO-v4 – GEOS-MIT

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

12 of 49

Global sea level and salt

12

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

13 of 49

Conclusions (part I)

  • GEOS-MIT model is now operational.
  • Last pieces of the sea-ice advection scheme are now being integrated into the model.
  • Tuning is about to commence using Green’s function method (Menemenlis et al., 2005).

13

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

14 of 49

Part II:�Air sea interactions in the high resolution GEOS-MIT

14

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

15 of 49

Current objectives of this study

    • Develop a high resolution coupled ocean-atmosphere run for studying air sea interactions and simulating an observation system.
    • Investigate the ability of the coupled model to capture the strong observed positive correlations between SST and wind stress/speed.
    • Compare near-surface diagnostics of the fully coupled ocean-atmosphere set-up to equivalent atmosphere-only simulations.

15

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

16 of 49

Background: observed SST/wind speed anomaly correlations

16

Jun 2018

“Most often negative correlations between SST and surface wind speed variability are observed in the extra-tropics for seasonal means and on the basin scale”

Xie et al (2004)

SST-wind relation in the North Pacific and Atlantic Oceans, (left) COADS SST (color shade), surface wind vectors, and SLP regressed upon the Pacific decadal oscillation index (Mantua et al. 1997). (right) COADS SST (color in ◦C ) and NCEP surface wind (m s-1) composites in Jan-Mar based on a cross-equatorial SST gradient index (Okumura et al. 2001).

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

17 of 49

Background: observed SST/wind stress anomaly correlations

17

Jun 2018

Two-month averages (January–February 2008) of spatially high-pass-filtered sea surface temperature (SST) overlaid as contours on spatially high-pass-filtered wind stress.

Agulhas Return Current

Gulf Stream

“Satellite observations have revealed a remarkably strong positive correlation between sea surface temperature (SST) and surface winds on oceanic mesoscales of 10–1000 km.”

Chelton et al., Oceanography (2010)

​

1 DegC contours

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

18 of 49

Background: modeled SST/wind speed correlation

18

Jun 2018

Temporal correlation of high-pass filtered surface wind speed with SST. (a) 1.0° ocean and 0.5° atmosphere (b) 0.1° ocean and 0.5° atmosphere (c) 0.1° ocean and 0.25° atmosphere. (d) Satellite observations.

“… the output of a suite of Community Climate System Model (CCSM) experiments indicates that … correlation between SST and surface wind stress, is realistically captured only when the ocean component is eddy resolving.”

Bryan et al., J. Clim. (2010)

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

19 of 49

Methods - models

  • Atmosphere – GEOS:
    • Horizontal grid type – Cubed sphere, 1/8⚬ X1/8⚬
    • Vertical grid type – hybrid sigma-pressure, 72 levels
  • Ocean – MITgcm
    • Horizontal grid type – Lat-Lon-Cap, 1/12 ◦ X1/12 ◦
    • Vertical grid type – z* rescaled height vertical coordinate, 90 levels

19

Jun 2018

Cubed sphere grid (left) and Lat-Lon-Cap (right)

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

20 of 49

Methods - experimental setup

  1. Ocean only – MITgcm (OGCM):
    • Jan, 1 – Jun 15, 2012
    • Forcing: 0.14◦, 6 hourly ECMWF
  2. Atmosphere Only – GEOS (AGCM)
    • Feb, 9 – Apr 9, 2012
    • Forcing: SST and ice fraction from run 1
    • Initial conditions: MERRA-2
  3. Coupled – GEOS-MITgcm (AOGCM)
    • Feb, 9 – Apr 9, 2012
    • Ocean initial conditions: from run 1
    • Atmospheric initial conditions: MERRA-2 (same as the run 2)

​

​

20

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

21 of 49

Ocean surface current

21

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

22 of 49

Precipitation

22

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

23 of 49

Wind stress (shading) and SST (contours)

23

Jun 2018

Wind Stress [N m-2]

GEOS-MITgcm: Agulhas Return Current

GEOS : Agulhas Return Current

GEOS-MITgcm: Gulf Stream

GEOS: Gulf Stream

Solid Black – positive anomaly

White – zero

Dashed black – negative anomaly

Both GEOS and GEOS-MITgcm show positive correlation between wind stress and SST consistent with previous results

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

24 of 49

24

Jun 2018

The linear relation between the stress and the SST in our coupled model is closer to the observed values compared to the previous modeling study.

Linear relation between wind stress and SST

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

25 of 49

 

25

Jun 2018

Agulhas Return Current

Wind speed is lagging the SST by ~1day

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

26 of 49

 

26

Jun 2018

Gulf Stream

Wind speed is lagging the SST by ~1day

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

27 of 49

Possible mechanism

27

Jun 2018

Positive SST anomaly

Positive wind speed anomaly

Negative SST anomaly

Negative wind speed anomaly

Increase instability

and draw

horizontal

momentum

from upper

levels

Increase

upward latent and sensible heat fluxes

Increase stability

Reduce upward latent and sensible

heat flux

Days

Days

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

28 of 49

 

28

Jun 2018

MERRA-2

(2006-2015)

An observational based product (MERRA-2) demonstrates cycles of several days both in surface wind and SST

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

29 of 49

 

29

Jun 2018

GEOS-MIT

MERRA-2

The GEOS-MIT model is able to reproduce the MERRA-2 spectral density but with higher SST amplitudes. May indicate strong air-sea interactions.

GEOS

2 months

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

30 of 49

Conclusions (part II)

  • First analysis of the ~10km coupled GEOS-MITgcm model reproduces realistic synoptic and mesoscale patterns.
  • The coupled model shows positive correlations between SST and wind speed/stress, and the relation is slightly closer to observational estimates compared to previous simulations.
  • The fact that the atmosphere-only experiment can reproduce the positive correlation suggests that the atmosphere responds to the ocean and not the opposite.
  • Daily time series suggest a three-four-day cycle induced by air-sea feedbacks.

30

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

31 of 49

Part III:�Consequences of different air-sea feedbacks on the ocean using MITgcm with MERRA-2 forcing

31

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

32 of 49

Background

  • Atmospheric and oceanic reanalyses (or state estimation) optimize the simulated states of atmosphere and ocean based on minimization of model-data difference.
  • No feedbacks between the ocean and the atmosphere.
  • Results in errors in the estimations of air-sea fluxes and can have implications on the circulation of a forced ocean model.

32

Jun 2018

  • Investigate the differences between MERRA-2 and ECCO-v4 in terms of air-sea heat flux.
  • Investigate the possible effect of MERRA-2 atmosphere on an ECCO-v4 ocean using different forcing methods (turning on and off different feedbacks).
  • Document the agreement of the MITgcm sensitivity simulations with available observations.

Current motivation

Strobach et al (2018), under review

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

33 of 49

Ocean models thermal forcing methods

Three main methods:

  • Relaxing ocean surface to prescribed SST.
  • Providing air-sea fluxes from observations/reanalysis.
  • Providing atmospheric surface state variables from observations/reanalysis and calculating air-sea fluxes interactively using bulk formulae and black body radiation.

33

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

34 of 49

Relaxation to prescribed SST

  • The heat flux is proportional to the difference between the surface temperature and the first model level.
  • Does not require any atmospheric information.
  • Does not constrain air-sea fluxes themselves to be realistic in any way.
  • Probably the oldest and the simplest way to force ocean models.

34

Jun 2018

Haney, JPO (1971)

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

35 of 49

Forcing with fluxes

  • Realistic air-sea fluxes from the observations/reanalysis.
  • No feedback between the ocean and the atmosphere.
  • A link between the thermal and hydrological forcing is provided by the atmospheric observations/reanalysis.
  • Considered as a strongly constrained atmosphere.

35

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

36 of 49

Forcing with state variables

  • Air-sea flux is calculated from surface state variables and the changing SST.
  • Include feedbacks between the ocean and the atmosphere.
  • A link between the thermal and hydrological forcing is provided by the bulk formulae.
  • A useful compromise and is the most commonly used today.
  • Considered as a less constrained atmosphere (more active feedbacks).

36

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

37 of 49

Experimental setup

  • MITgcm in its ECCO-v4-r2 configuration (1992-2011).
  • MERRA-2 surface heat fluxes/state variables.
  • Runoff from Fekete et al. (2002)
  • Sea-ice, as implemented in MITgcm can only be forced by computing a sea-ice-specific set of bulk-formulae, so all solutions use the same forcing method over sea-ice.

37

Jun 2018

For the agreement with observations:

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

38 of 49

MERRA-2-flux: MITgcm forced with MERRA-2 fluxes

Negative MERRA-2 net heat flux to the ocean resulted in a large SST reduction compared with observationally based products.

38

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

39 of 49

MERRA-2-state: MITgcm forced with MERRA-2 state variables

SST restored but errors propagated to the water cycle and resulted in a global mean sea level increase of 2.7m

39

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

40 of 49

MERRA-2-state: MITgcm forced with MERRA-2 state variables

40

Jun 2018

  • Latent heat flux feedback acted to increase SST.
  • Sensible heat and long wave radiation Feedbacks are secondary to Latent heat.

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

41 of 49

MERRA-2 state: importance of different budget terms

 

41

Jun 2018

SST [◦C]

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

42 of 49

MERRA-2 state: importance of different budget terms

 

42

Jun 2018

SST [◦C]

 

 

Forcing the ocean using bulk formulae can be interpreted as a strong relaxation to observed SST

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

43 of 49

Forcing with turbulent fluxes

  • Latent heat and sensible heat fluxes from observations/reanalysis.
  • Emitted long wave feedback between the ocean and the atmosphere.
  • A link between the thermal and hydrological forcing is provided the atmospheric observations/reanalysis.
  • May reflect a “compromise” state between the ocean and the atmosphere in a coupled DA system.

​

43

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

44 of 49

MERRA-2-turb: MITgcm forced with MERRA-2 turbulent fluxes

  • More realistic SST compared with MERR-2-flux (but still cold).
  • No change in the water cycle compared with MERRA-2 reanalysis.

44

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

45 of 49

MERRA-2-turb: MITgcm forced with MERRA-2 turbulent fluxes

45

Jun 2018

  • Long wave radiation acted to increase SST.
  • No change to the water cycle compared to MERRA-2.

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

46 of 49

Time Evolution: Comparison of Different Forcing Methods

46

Jun 2018

SST drift in MERRA-2-flux and MERRA-2-turb. SSS drift in MERRA-2-state

MERRA-2-flux

MERRA-2-turb

MERRA-2-state

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

47 of 49

Model-Data Misfit in the different experiments

47

Jun 2018

  • MERRA-2-state has the smallest cost function for temperature.

​

  • MERRA-2-turb has the smallest cost function for salinity but not far from MERRA-2-state.

Cost function (Forget et al., 2015)

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

48 of 49

Conclusions (part III)

  • We performed three experiments using different forcing methods
    • MERRA-2-flux: the negative heat flux in MERRA-2 resulted in a near-steady sea surface temperature decline of ~0.25 ◦C/year.
    • MERRA-2-state: reduced this drift and effectively restored SST to those used in MERRA-2, but, it resulted in a large increase in sea level and an SSS drift.
    • MERRA-2-turb: resulted in larger temperature errors than in MERRA-2-state, but it has a better physical justification in terms of the water budget.
  • For some ocean modeling applications, the traditional surface forcing with atmospheric state variables and bulk formulae may not be optimal.
  • Our results strongly and unambiguously argue for next-generation data assimilation climate studies to involve fully coupled systems.

48

Jun 2018

MERRA-2-flux

MERRA-2-turb

MERRA-2-state

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration

49 of 49

Next steps/future work

  • Finalize sea-ice advection.
  • Model tuning using green’s function method.
  • Increasing horizontal resolution (~1km).
  • Initialized sub-seasonal to decadal prediction system.
    • Observation System Simulation Experiments (OSSE).

​

49

Jun 2018

Global Modeling and Assimilation Office

gmao.gsfc.nasa.gov

GMAO

National Aeronautics and Space Administration