Estimating the contribution of wind-waves in the coupled climate system WEALTH FROM OCEANS NATIONAL RESEARCH FLAGSHIP Mark Hemer 1, Baylor Fox-Kemper 2,

Estimating the contribution of wind-waves in the coupled climate system

WEALTH FROM OCEANS NATIONAL RESEARCH FLAGSHIP

Mark Hemer1, Baylor Fox-Kemper2, Ramsey Harcourt3 and Elodie Charles1

1CAWCR, CSIRO Marine and Atmospheric Research, Hobart, TAS. 2 Brown University, RI (formerly CIRES, University of Colorado, Boulder), USA. 3 Applied Physical Laboratory, University of Washington, USA.

Deep-water mooring workshop, Hobart, May 5-6, 2014.

Presentation Outline• Introduction/Motivation• Methodology

• Mixing models– PWP with Langmuir– SMC with Langmuir

• Model Forcing– Atmospheric– Wave– Model initialisation and integration

• Results• Ongoing Work/Concluding Remarks• If time, consideration of contribution wave drag effects on global

flux estimates

2 |

The contribution of:Wave-driven mixing of the Surface Mixed LayerIntroduction

3 |

Air-Sea flux biases in AOGCMs (e.g., CCSM3/4)

Qa

Total flux

CCSM3 Bias

CCSM4 Bias

Bates et al., 2012.

SW Latent

Ensemble mean CMIP5 MLD Bias (Sallee et al., 2013)

5 |

6 |

Cavaleri et al (2012) BAMS

Wind and waves not in equilibrium• Swell dominates global wave field. (Fan et al., 2014)• Global distribution of fraction of wave energy which is swell

7 |

The contribution of:Wave-driven mixing of the Surface Mixed LayerMethodology: Mixing Models

8 |

Wave driven surface ocean mixing: 3 possible processes

9 |

Injection of turbulence by breaking waves. Mixes to depth approximately equal to the wave height(Craig and Banner, 1994)

Langmuir mixing mixes to a depths of order 100m. (Langmuir, 1938)

Non-breaking wave mixing. It has been proposed turbulence generated by wave orbital motion can mix to depths of order 100m. (Babanin, 2006)

Wave dependent mixingDong et al. Observations

CCSM3.5 with Langmuir

CCSM3.5 Control without Langmuir

August Southern Ocean mixed layer bias(Webb et al., 2010)

Langmuir Parameterisation in mixing models• Bulk Model

• Price-Weller-Pinkel model with amended Li and Garrett (1995) Langmuir parameterisation

• Profile Models• Second-moment-closure (following Mellor-Yamada) with Langmuir

Parameterisation (Harcourt, 2012)– Alternate tunings

– E6=7 (tuning to minimise error in prediction of TKE levels)– E6=5 (tuning to predict vertical TKE to compare with vertical float

m’ments)– GV=GS=0 (Reduction to Kantha and Clayson, 2004)

• K-Profile Parameterisation of Langmuir turbulence (McWilliams, 2000; Smyth, 2002; McWilliams, 2012; Fan and Griffies, 2014)

11 |

Langmuir Parameterisation in mixing models• Bulk Model


• Profile Models• Second-moment-closure (following Mellor-Yamada) with Langmuir

Parameterisation (Harcourt, 2012)– Alternate tunings

– E6=7 (tuning to minimise error in prediction of TKE levels)– E6=5 (tuning to predict vertical TKE to compare with vertical float

m’ments)– GV=GS=0 (Reduction to Kantha and Clayson, 2004)

• K-Profile Parameterisation of Langmuir turbulence (McWilliams, 2000; Smyth, 2002; McWilliams, 2012; Fan and Griffies, 2014)

12 |

Shear instability mixingStatic instability mixing Bulk Richardson instability mixing

Price-Weller-Pinkel mixed layer model (Price et al., 1986)

Wind Stress and Heat Flux

Redistributions of T, S, r and U

1) Mixed layer depth2) Vertical profile of horizontal velocity3) Vertical profile of T, S and r

Incorporation of LC into the PWP model (Li et al.,1995)

Langmuir cells penetration depth depends on competition between vertical motion and stratification, represented by the Froude number.

Vertical penetration is inhibited when Fr reaches a critical value Frc = 0.6 (LG97).

This was parameterised by Li and Garret using:

(after LG93)

This gives an additional stability criteria (i.e., that deepening of the SML will not occur if)

Under the assumption of a fully developed sea.

Nh

wFr dn

cFrFr

*31

31

*

72.0 uLau

uw s

dn

h

ub

265.0

Amended Fr scaling of Langmuir Mixing in PWPFlor et al. (2010) suggest wdn = 5.2wrms, so Froude based stability criteria becomes:

wrms obtained from LES model scaling (Van Roekel et al., 2012):

, where c1 = 1.5 and c2=5.4.

For the case where wind and waves are non-aligned,

a = angle b/w wind and langmuir cell direction, qww is angle b/w stokes drift and wind.

422

12*

2 0.16.0 ttrms LacLacuw

wwMLsSLprojt u

uLaLa

cos

cos

2.0,

*22

ww

S

ww

z

h

uu

coslogsin

arctan1

0

*

6.02.52

crmsdn Frhg

w

Nh

wFr

Shear instability mixingStatic instability mixing Bulk Richardson instability mixing

PWP mixed layer model with LC parameterisation

Wind Stress and Heat Flux

Redistributions of T, S, r and U

1) Mixed layer depth2) Vertical profile of horizontal velocity3) Vertical profile of T, S and r

Langmuir mixing

6.02.52

hg

wFr rms

Second moment closure model with langmuir paramn

(Harcourt, 2013)

Turbulent Kinetic Energy Equation

17 |

lB

qPP

x

qqlS

zDt

DqbS

kq

1

322

2

j

iji x

Uuu

''PS = Shear Production Pb = Buoyancy Flux '

jjug

k

jSkk

L

x

uuu

tDt

D

)(

Harcourt (2013) modifies for impact of Craik-Leibovich vortex forcing using

Harcourt SMC model with langmuir turbulence (2013, JPO)

andx

uuu

t

u

Dt

uDwhereuuuufugug

x

uuu

x

uuu

x

uuu

x

uuu

x

u

x

uuu

x

pu

x

pu

x

uuu

Dt

uuD

k

jSkk

jjL

ljjkllijklkjiij

i

Sk

kjj

Sk

kik

ikj

k

jki

k

j

k

iji

ij

ji

k

jikjiL

,)(,

22

CL vortex force (Craik and Leibovich, 1976) included in momentum equation after McWilliams et al (1997):

uj is surface-wave phase-averaged Eulerian velocity, uj

S is the surface wave Stokes drift, P is non-hydrostatic pressure, r0 is reference density, fk is Coriolis components, gk is gravitational acceleration, n is viscosity q is a thermodynamic scalar with expansion coefficient a and diffusivity kq.

Deriving Reynolds stress and flux equations for fluctuations uj’, q’ and a slowly- or non-fluctuating Stokes drift gives:

Kantha and Clayson (2004) included this production

term

The contribution of:Wave-driven mixing of the Surface Mixed LayerMethodology: Model Forcing

19 |

Large and Yeager (2004) CORE normal year forcing Standard air-sea flux dataset of WGOMD Coupled Ocean-Ice Reference Experiment

Atmospheric Fields• ISCCE: Daily varying shortwave and longwave radiative fluxes (no diurnal cycle)• NCEP/NCAR: Six-hourly varying meteorological fields (1948-2006)

10m TA, humidity, zonal/ meridional winds, SLP• GCGCS: Annual mean river runoff• GCGCS: Monthly varying precipitation (12 time steps per year)

Large and Yeager (2004, 2009) Bulk Formula• Sfc BC determines fluxes of heat, freshwater and momentum• Net heat flux = Sensible + Latent + Shortwave + Longwave• Net freshwater flux = Precip – Evap + River Runoff• Net momentum exchange is driven by windstress

Large and Yeager (2004) Solution set• Solves bulk formula under CORE forcing using Hadley OI-SST

20 |

Large and Yeager (2004) CORE normal year forcing Standard air-sea flux dataset of WGOMD Coupled Ocean-Ice Reference Experiment

Atmospheric Fields• ISCCE: Daily varying shortwave and longwave radiative fluxes (no diurnal cycle)• NCEP/NCAR: Six-hourly varying meteorological fields (1948-2006)

10m TA, humidity, zonal/ meridional winds, SLP• GCGCS: Annual mean river runoff• GCGCS: Monthly varying precipitation (12 time steps per year)

Large and Yeager (2004, 2009) Bulk Formula• Sfc BC determines fluxes of heat, freshwater and momentum• Net heat flux = Sensible + Latent + Shortwave + Longwave• Net freshwater flux = Precip – Evap + River Runoff• Net momentum exchange is driven by windstress

Large and Yeager (2004) Solution set• Solves bulk formula under CORE forcing using Hadley OI-SST

21 |

CORE normal year consistent wave field

22 |

WaveWatch III (v4.08)CORE normal year winds forcing.1° spatial resolution 13 month integration: (dec) spinup + 1 whole CORE normal yr. Full directional spectra archived 6-hourly at 4 ° resolution.

Stokes’ drift

23 |

25 |

Model Initialisation:

Real ARGO profile. Most representative summer solstice (shallow MLD) profile (taken within 1.5° radius of wave archive location, +/- 20 days from summer solstice date.

Temperature (°C)

Dep

th (

m)

26 |

Model Forcing:

CORE Normal Year (Large and Yeager, 2004).Bulk fluxes, using evolving mixing model SST

•Daily varying shortwave and longwave radiative fluxes (no diurnal cycle)•Six-hourly varying meteorological fields (1948-2006) 10m TA, humidity, zonal/ meridional winds, SLP• Annual mean river runoff•Monthly varying precipitation (12 time steps per year)• Six-hourly varying stokes velocity profile from CORE forced wave model

Summer Solstice

1-D Mixing Model:Harcourt (2012): SMC with Langmuir

Temperature (°C)

Dep

th (

m)

Model applied with zero waves (solid) and COREwave (dashed) fields

27 |

Model Forcing:



Temperature (°C)

Dep

th (

m)



28 |

Model Forcing:



Temperature (°C)

Dep

th (

m)



29 |

Model Forcing:


Summer


Temperature (°C)

Dep

th (

m)



30 |

Model Forcing:


Summer


Temperature (°C)

Dep

th (

m)



31 |

Model Forcing:


Summer


Temperature (°C)

Dep

th (

m)



32 |

Model Forcing:


Summer

Autumn


Temperature (°C)

Dep

th (

m)



33 |

Model Forcing:


Summer

Autumn


Temperature (°C)

Dep

th (

m)



34 |

Model Forcing:


Summer

Autumn


Temperature (°C)

Dep

th (

m)



35 |

Model Forcing:


Summer

Autumn

Winter


Temperature (°C)

Dep

th (

m)



36 |

Model Forcing:


Summer

Autumn

Winter


Temperature (°C)

Dep

th (

m)



37 |

Model Forcing:


Summer

Autumn

Winter


Temperature (°C)

Dep

th (

m)



38 |

Model Forcing:


Summer

Autumn

Winter

Spring


Temperature (°C)

Dep

th (

m)



NO WAVES WAVES ARGO PROFILES50N, 145WHarcourt SMC

Experiments carried out

• Carry out annual integration at each location we have spectral wave data archived (globally on 4° grid)

• Repeat computations with alternative mixing schemes• Harcourt (2012) SMC with langmuir parameterisation

– Alternate tunings– E6=7 (tuning to minimise error in prediction of TKE levels)– E6=5 (tuning to predict vertical TKE to compare with vertical float m’ments)– GV=GS=0 (Reduction to Kantha and Clayson, 2004)


40 |

The contribution of:Wave-driven mixing of the Surface Mixed LayerResults

41 |

Percentage increase in MLD with introduction of SMC (E6=7) langmuir mixing ~180 days after Summer Solstice

Zonal mean MLD after 180days after 365 days

Annual Cycle (PAPA)

43 |

tQQ LI

waves

nowave

)( endofyear

waves

nowavesOHC 7 Wm-2

WAVES

NOWAVES

Mean waveflux is

~0.3 Wm-2

Annual Mean Surface Heat Flux Equivalent with addition of Langmuir mixing (Wm-2)

yearend

nowavewaveequiva time

OHCOHCQ

,

E6=7

Implied Annual Mean Northward Heat Transport (Wm-2)

Latitude

Watts

NO WAVESWAVES

Note: Plot not directly comparable to Atm flux calcs. Can not assume that storage is uniformally distributed, as 1d model with/without waves is calculating spatial distribution of storage. If we remove this influence (by the above assumption applied previously to ensure convergence to zero at Northern boundary), the lines overlay one another.

yxQa

~8.6Wm-2

~11.5Wm-2

Global equivalent bias

SMC Langmuir implies ocean stores an extra 1.0PW





Reduced to Kantha&ClaysonApproximationGV=GS=0

NO WAVES WAVES ARGO PROFILES50N, 145WPWP-Lc

Percentage increase in MLD with introduction of PWP langmuir mixing ~180 days after Summer Solstice


Summarising the contribution of langmuir parameterisation in mixing models

51 |

Model Equivalent global mean Surface heat flux

Equivalent storage of global ocean

SMC_LC (E6=7) 2.9 Wm-2 1.0 PW

SMC _LC(E6=5) 2.2 Wm-2 0.77 PW

SMC _LC_KC 1.0 Wm-2 0.35 PW

PWP_LC (@180days) 3.0 Wm-2 1.0 PW

c.f., annual mean global energy flux from wind-waves = 70TW annual mean global energy flux from waves-OML= 68TW (Rascle et al., 2008)

The contribution of:Wave-driven mixing of the Surface Mixed LayerOngoing work/Concluding remarks

52 |

Ongoing work• Wave Mixing Processes

• Comparison with other mixing models with wave mixing parameterisations. – Qiao et al. Bv scheme– Ghantous/Babanin scheme– KPP with langmuir implementations (FanGriffies, McWilliams, …)

• Comparison with observations– Opportunities with SOFS or other platforms?

• Reality check - Determine swell dissipation from mixing – are values reasonable from wave perspective or are we removing too much wave energy.

• Implementation within– 3D ocean model (influence of buoyancy restoring forces)– Coupled Ocean Carbon model

53 |

Ongoing work• Other wave driven processes

• Attempted estimates for effects of wave dependent drag on bulk fluxes– 1D sense. Surface fluxes are sensitive to parameterisation of wave

dependent drag, introducing a range of uncertainty of approximately 1PW in atmosphere-ocean heat exchange (~3Wm-2)

– Coupled atmosphere-wave global climate model (Elodie’s talk tmrw)• Effects of other processes (albedo, mass fluxes - marine aerosol prodn and

bubble injection, sea-ice/wave intern, etc) – Ultimately a fully coupled atm-wave-ocean-sea-ice (&carbon?) model

54 |

Conclusions of contribution of waves to climate system.

• Quantitative estimates of the contribution of wave driven langmuir mixing on the annual heat budget have been determined

• Wave driven forcing of 1-d mixing models applied globally show an approximate 25% increase in mixed layer depth in extra-tropical storm belts, which is greater during the winter mixing season. Expressed as a surface heat flux, this is equivalent to up to 10Wm-2, or an additional heat uptake of ~1PW to the global ocean over one year.

• The sensitivity of global heat and momentum budgets to available parameterisations of drag is non-trivial, with seastate dependent parameterisations introducing uncertainty of ~1PW in surface atmosphere-ocean heat exchange. This supposed contribution however remains within the bounds set by alternative wind-dependent parameterisations of drag.

• Estimates of the contribution of other wave processes to follow

55 |

Centre for Australian Weather and Climate Research: A Partnership between CSIRO and the Bureau of Meteorology

Mark Hemer

t +61 3 6232 5017e [email protected] www.cmar.csiro.au/sealevelCSIRO WEALTH FROM OCEANS NATIONAL RESEARCH FLAGSHIP

Thank you

Acknowledgements: Work supported through:

the CIRES Visiting Fellowship Program, the CSIRO Wealth from Oceans National Research Flagship; and the CSIRO OCE Julius fellowship fund.

57 |

Cavaleri et al (2012) BAMS

MethodologyAtmosphere

Assess sensitivity of CORE 1 Large & Yeager ocean-atmosphere surface fluxes to wave dependent parameterisations of sfc roughness.

58 |

Ocean

Annual cycle integration of 1-d mixing models with parameterisation of langmuir mixing, applied globally. Assess sensitivity to inclusion of wave driven mixing.

one year.o

ConclusionsAtmosphere

Wave dependent parameterisN of roughness leads to up to 1PW of additional heat transfer to the ocean (within range of alternate wind-dependent schemes.

59 |

OceanLangmuir mixing of the surface ocean

shows ~25% increase in MLD in extra-tropical storm belts during winter. Equivalent to an additional heat uptake of ~1 PW to the global ocean over one year.

c.f. 1.4PW is the estimated heat flux transported by the Gulf Stream

The Future

Many other processes to be considered

This ‘back of the envelope’ approach used to support decisions as to where to focus effort in a coupled model (Elodie’s talk)

60 |

The Large and Yeager CORE forcing•Annual mean river runoff•Monthly varying precipitation (12 time steps per year),•Daily varying shortwave and longwave radiative fluxes (365 time steps per year, and so no diurnal cycle and no leap years)•Six-hourly varying meteorological fields (1948-2006)10m air temperature, humidity, zonal/ meridional winds, SLP

Large and Yeager (2004, 2009) Bulk Formula

Surface boundary condition determines fluxes of heat, freshwater, and momentumSolved separately for ocean and sea-ice covered areas of each grid cellNet Heat Flux = Sensible + Latent + Shortwave + LongwaveNet freshwater flux = Precipitation -Evaporation + River Runoff (+ Glacial Calving)

Net momentum exchange is driven by windstress, accounting for ocean-ice stress, and ocean currents

Large and Yeager Solution set (mean monthly 1948-2006)Solves bulk formula under CORE forcing with Hadley OI-SSThttp://dss.ucar.edu/datasets/ds260.2/

Drag Coefficient vs Wind Speed (SOFS, 47S, 142E)

61 |

10000764.0142.07.2

1010

10

u

uCDN

*

2* 11.0

ug

uzo

*

5.4*

11.0/

2

50

uCuz ppo

*

5.4 11.0/1200

uhhz psso

21

1

w

*

2* 11.0

ug

uzo

2

2

10

10log

o

DN

z

C

62 |

Latent Heat Flux Mean Bias (Param - CORE)

Charnock Wave Age: Oost et al.

Wave Steepness: YellandTaylor Wave Stress: JanssenViterbo

,,, uSSTqqCQ satCOREEvCOREE

q

r

DNEN

zZ

CC

ln

Wm-2

0

-200

63 |

Total Heat Flux QA = QS + QE + QL + QH (Mean Bias)

Charnock Wave Age: Oost et al.

Wave Steepness: YellandTaylor Wave Stress: JanssenViterbo

coreAparamA QQ ,,

Wm-2

Global mean air-sea fluxes (Wm-2)

Corrected 2007 CORE data. C.f., Table 3 Large and Yeager, 2009, but note different masking area defined by wave model.

Flux CORE Charn Oost et al Taylor-Yelland

Janssen

Qh -12.8 -13.4 -13.1 -12.7 -13.4

Qs* 178.41 178.32 178.33 178.33 178.3

Ql -53.9 -53.9 -53.9 -53.9 -53.9

Qe -107.4 -112.0 -107.0 -105.3 -111.9

Qa 4.3 -1.0 4.3 6.4 -0.8

64 |

• 1 no consideration of whitecapping. • 2 whitecapping parameterised using wind-dependent method of Frouin et al., 2001.• 3 whitecapping parameterisation is sea-state dependent, following Zhao et al. 2003.

• This is a function of u*, thus dependent on zo parameterisation.• No wave dependent long-wave radiation flux is implemented. Note surface

emissivity has a sea-state dependent component• 4 rms (spatial) of annual mean relative to CORE calculation annual mean.

Assume bias/storage (from previous slide) is uniformly distributed across the global ocean.

Implied Northward heat transport

Latitude

Watts

Integration of this heat budget implies wave-based parameterisation can lead to an increase in heat storage of approx 1PW (Taylor and Yelland, 2001), or decreased capacity of approx 2PW, which are within the limits set by alternative wind dependent parameterisations of roughness.

Wind and waves not in equilibrium• Swell dominates global wave field. (Semedo et al., 2011)• Global distribution of fraction of wave energy which is swell for

DJF and JJA.

66 |

%

Southern Ocean GCM wind biasSwart and Fyfe, 2011

SO winds drive large components of present and future ocean uptake of heat and CO2. =>This bias has broad implications for GCM’s

Sea-state dependent drag influence

Janssen and Viterbo (1996) Sea-state dependent drag in Seasonal prediction model

Garfinkel et al. (2011)Increased ocean roughness in GEOS-5 GCM improved SO wind bias.

68 |

The Southern Ocean Wind Bias

CORE (Large and Yeager, 2004, 2009)Standard air-sea flux dataset of WGOMD

Atmospheric Fields• NCEP/NCAR

• Near surface winds, U• Near surface atmospheric temperature, q• Near surface specific humidity, q

Radiation• International Satellite Cloud Climatology

Experiment• Short wave insolation, QI

• Downwelling Long wave Radiation, QA

Precipitation• GCGCS (Merged GPCP, CMAP, S-H-Y data)

SST• Hadley Centre sea Ice and SST dataset version 1

(HadISST1)

69 |

Bulk air-sea fluxes• Bulk flux formulae

• Neutral Conditions

• Correction dependent on surface stability

70 |

XSCXSccxw xdx 2/12/1''windspeedSzXXX sea );(

)/ln(2/1

oxxn zz

c

)(1

)(2/1

2/12/1

xxn

xnx

c

cc

Zox – roughness length

Comparisons with CORE.v2

10000764.0142.07.2

1010

10

u

uCDN

Heat budget closure is dependent on parameterisation of transfer coefficient.

Roughness length• Charnock relation (constant coefficient, plus smooth flow limit)

where

• Other parameterisations suggest zo is a function of wave age, steepness or stress.

E.g., Oost et al. (2002),

and Taylor and Yelland (2001)

And Janssen and Viterbo (1996)

72 |

*

2* 11.0

ug

uzo

*

5.4*

11.0/

2

50

uCuz ppo

*

5.4 11.0/1200

uhhz psso

21

1

w

*

2* 11.0

ug

uzo

2

2

10

10log

o

DN

z

C

Documents

Estimating the contribution of wind-waves in the coupled climate system WEALTH FROM OCEANS NATIONAL RESEARCH FLAGSHIP Mark Hemer 1, Baylor Fox-Kemper 2,