© Crown copyright Met Office Impacts of the ocean mesoscale on mean state and coupled phenomena Malcolm Roberts, Met Office Hadley Centre.

© Crown copyright Met Office

Impacts of the ocean mesoscale on mean state and coupled phenomena

Malcolm Roberts, Met Office Hadley Centre

Thanks

• UK-Japan Climate Collaboration team: Pier Luigi Vidale, Marie-Estelle Demory, Adam Clayton, John Donners

• Dave Stevens and HiGEM team

• Akira Kuwano-Yoshida and group

• Yevgeny Aksenov and NOCS group

• Arne Biastoch

• Eiji Watanabe and Hiroyasu Hasumi

• Frank Bryan, Wonsun Park, Tony Rosati

Motivation

• When building a coupled model, there is demand for resources from all sides (atmosphere, ocean, sea-ice, land surface, biogeochemistry etc etc) in many areas

• Resolution (horizontal+vertical)

• Complexity, either through new parameterisations, or from extra components (e.g. Earth system components and feedbacks)

• Ensembles (to address uncertainty, signal to noise ratio)

• Run length – 100’s years for coupled models

• When arguing for resources, need evidence that making any component (much) more expensive is necessary to produce an improved simulation

• Cost implications (e.g. sea-ice can be multi-category, multi-level)

• Ocean biogeochemistry (many tracers required)

• Need longer timescale for ocean testing

Challenge

• Can we make a case for eddy-permitting ocean models being essential for climate simulations?

• Or at least are there questions we cannot/should not address without such a model?

• And if we choose an eddy-permitting ocean model, what resolution atmosphere model do we need?

• We also need to think about vertical resolution

• Diurnal cycle, intraseasonal coupled air-sea variability, and near-surface vertical resolution

Mesoscale ocean and coupled modelling

• Many studies using high resolution (eddy-permitting) ocean models that show an improved simulation of the ocean (mean state and variability)

• More complex to show how increased ocean resolution directly impacts on simulation of climate (e.g. atmosphere/land where people live) – may well need sufficient atmospheric resolution to be able to respond (e.g. HadCEM showed little change in climate, Roberts et al, 2004)

• Various studies using recent higher resolution observed datasets to force atmosphere models are beginning to show how smaller scales can have important effect on larger scales (e.g. TIWs, boundary currents)

• Just beginning to be enough studies, using coupled climate models, to show how enhanced ocean resolution does have important effects on the mean state of the coupled model, and its variability

• This is important, both for long-term climate modelling (centuries), as well as seasonal-decadal forecasting. Would like this to be a seamless transition when modelling.

• It may also be very important for processes such as ocean biogeochemistry, including air/sea flux exchanges and uptake of CO2

Selection of groups with “high” eddy-permitting resolution coupled models (ready or planned)

Model/

Centre

Model name

Atmos Ocean Comments

UK - MOHC/NERC

HiGEM/NUGEM

HadGEM3-H (MOAtm/NEMO)

1.25°x0.833°/ 0.833°x0.55°

0.833°x0.55°

0.33°x0.33°

0.25°x0.25°

100’s years + IPCC AR5 S2D

Later this year, hierarchy

GFDL CM2.4 1° 0.25° Soon to be 0.5° atmos, working on 0.1° ocean

NCAR, DOE (LLNL, Los Alamos)

CAM3.5 0.5°

0.25°

0.1°

0.1°

10-15 year development runs

CCSR/FRCGC/NIES

MIROC 0.85° (T213) 0.25°x0.166° IPCC AR4

LOCEAN, IPSL, FRCGC

ECHAM/ NEMO 1.7° (T106) 0.25°x0.25° Testing – 5-10 years so far

IFM-Geomar ECHAM5/ NEMO

2.8° (T63) 0.5°, 0.25° and 0.083°

Later this year, hierarchy

Overview of talk

• Ocean mesoscale and effects on atmosphere

• Ocean SST and wind-stress curl/divergence

• TIWs and Southern Ocean

• Boundary currents and precipitation

• Ocean mesoscale and sea-ice

• Impact of ocean mesoscale on ocean and coupled climate mean state and variability

• TIWs

• Coastal processes

• Hawaiian countercurrent

• Agulhas

• Climate change and Southern Ocean

SST gradients and atmosphere forcing• Small-scale ocean SST gradients affect the atmospheric

winds

• This can cause a large scale mean response

• Examples are tropical instability waves (TIWs), Southern Ocean features

• Following work by Chelton and others

(The following slides courtesy of Dave Stevens, UEA, model output from the HiGEM coupled climate model, ocean resolution 1/3x1/3°, atmosphere 5/4x5/6°)

Wind curl obsObservations

Mechanism

Colder Warmer

SST

Uniform wind above the boundary layer

Mechanism

Colder Warmer

SST


StableBoundaryLayer

TurbulentBoundary

Layer

Mechanism

Colder Warmer

SST

Weaker surface wind


StableBoundaryLayer

TurbulentBoundary

Layer

Stronger surface wind

Mechanism

Colder Warmer

SST

Weaker surface wind


StableBoundaryLayer

TurbulentBoundary

Layer

Stronger surface wind

Divergence

Colder

WarmerWind above the boundary layer

South easterly wind blowing across a perturbed oceanic front

Colder

Warmer

. 0

0

. 0

. 0

0

Surface wind stress

Surface wind stress

Similarly wind stress curl linearly related to the cross wind SST gradient

Wind stress divergence linearly related to the down wind SST gradient

where is the angle between the wind stress and SST gradient

. T. |T || | cos

Mean tausst

Mean divsst

Wind divergence and sst

Wind stress divergence (shaded) and SST (contoured) - High resolution









4 wind curl

Low resolution ocean - High resolution atmosphere

Low resolution ocean - Low resolution atmosphere

High resolution ocean - High resolution atmosphere

High resolution ocean - Low resolution atmosphere

30 year mean Wind stress curl coloured and SST contoured

Summary• The atmosphere can respond to the ocean on oceanic length scales

• The persistent small scale features in Southern Ocean wind stress curl and divergence arise from mesoscale ocean features

• The overlying wind field is strongly coupled to the Tropical Instability Waves

• A high resolution model can capture this behaviour

• There are implications for coupling strength between atmosphere and ocean, which may depend on such small-scale processes

Atmospheric response to the Gulf Stream in an AGCM

Akira Kuwano-Yoshida1, Shoshiro Minobe2, and Shang-Ping Xie3

1. Earth Simulator Center, JAMSTEC

2. Hokkaido University3. IPRC, University of Hawaii

To be submitted to J. Climate

Model and experiment design

ModelAGCM for the Earth Simulator (AFES, Ohfuchi et al. 2004) ver.2 (Enomoto et al. 2008)

Resolution T239 (~ 50km) L48

SST NCEP RTG SST (0.5 degree, daily)

Initial condition ERA40

Integration11. Feb. 2001 ~ 28. Feb. 2006

The data from 1. Mar. 2001 are analyzed.

Sensitivity experiment for SST gradient

• CNTL: Original RTG SST data

• SMTH: 2 dimensional 1-2-1 filtering was conducted to 25N - 55N, 100W - 30W region (blue square) 100 times for daily RTG SST data.

Seasonal variation : precipitation• The precipitation band

appears over the Gulf Stream through the year in TRMM3B43 and CNTL run, while the band disappears in SMTH run.

ANN

MAM

JJA

SON

DJF

TRMM3B43 CNTL SMTH

Precipitation form

• Convective precipitation is more sensitive to the Gulf Stream SST in AFES and observation.

CNTL

SMTH

TRMMConvective precipitation Large-scale precipitation

Composite for Convective precipitation in DJF

• CNTL: Cumulus precipitation (color in left panels) associated with large CAPE (bold contour in left panels) elongates along surface moisture convergence (bold contour in right panels) induced by -SST Laplacian maximum (color in right panels).

• SMTH: Cumulus precipitation passes away eastward because smoothed SST front does not induce surface convergence.

CNTL SMTH CNTL SMTH

Summary and conclusion

• Precipitation is trapped by the Gulf Stream through the year, while the strength and width have seasonal variation.

• Cumulus precipitation is more sensitive to the Gulf Stream SST front.

• Surface convergence and evaporation induced by sharp SST gradient form favorable environment for cumulus convection occurrence and maintenance.

Influence of small scales on mean climate and variability• TIWs and impact on mean climate in tropical Pacific (may

be partly model specific here)

• Example here from Met Office work, but many other authors have investigated this area in much greater detail

• Work by Jochum and co-workers has shown:

• how TIWs increase the atmosphere-ocean heat flux

• how intraseasonal eddies such as TIWs cause interseasonal-interannual variability in the climate model (through mixed layer changes)

• Influence of improved mean climate on atmosphere (including precipitation)

• Teleconnections can spread this impact globally

Eddies, SSTs and ENSO

• Rôle of resolved versus parametrised ocean mixing, for example:

– Tropical Instability Waves emerge in 90km-1/3o model (HiGEM), performing meridional mixing near Equator;

• ENSO is poorly represented in the standard Hadley Centre climate model (HadGEM1); a more realistic ENSO is simulated by the high-resolution HiGEM model;

– is this because the HiGEM mean state in the tropical Pacific is so much closer to reality ?

– if so, this is a good example of a smaller scale phenomenon affecting the large scale mean state and, through it, a major element of climate variability.

Eddy heat flux convergence

Ocean temperature profile along Equator

L

H

Obs

Roberts et al, 2009, in press

SST errors compared to HadISST climatology for HadGEM1.1 and HiGEM1.1

HadGEM1.1 HiGEM1.1

CMAP observations

El Niño DJF precipitation anomalies (mm/day)

Ocean and sea-ice interaction

• Models with different ocean resolutions

• in Arctic, clear differences in the inflow routes

• Important impact on sea-ice extent

• definite improvement in the Norwegian Atlantic Current,

• the Arctic Circumpolar Current,

• the inflow of the Pacific Water through the Chukchi Sea,

• in southward re-circulation of the Atlantic water in the Nordic Seas and in the West Greenland Current in the Baffin bay

Potential Temperature at ca. 100 m depth in September in NEMO and OCCAM

OCCAM 1/12°98 mSep 2004

NEMO 1/4°LIM295 mSep 2001

NEMO 1°CICE495 mSep 2001

NEMO 1°LIM295 mSep 2001

Courtesy Y. Aksenov, NOCS

Pacific inflow – Watanabe and Hasumi, 2009, JPO (submitted)• This study deals with how the water entering the Arctic Ocean

from the Pacific behaves.

• It first flows along the Alaskan coast in the clockwise direction, and then is transported toward the centre of the Beaufort Gyre by baroclinic eddies in summer.

• The eddy activity depends on the summer-time sea ice margin.

• When the Alaskan coastal area is open, the coastal current tends to be strong, which favours baroclinic eddy generation.

• In such a situation, eddies also transport solar heat absorbed in the open coastal region toward the central Canada Basin.

• This heat transport induces further retreat of the sea-ice margin or delays re-freezing of the offshore region.

• Could be important feedback mechanism, both for simulation of sea-ice extent for future climate, as well as improving present-day simulated mean state.

• Uses regional ocean-sea-ice model with 2.5 km resolution

Model bathymetry in the spherical coordinate system which is rotated so that the singular points are on the equator. B.S. = Bering Strait, H.C. = Herald Canyon, C.C. = Central Channel, N.R. = Northwind Ridge, C.R. = Colville River mouth,

Coastal processes

• Coastal upwelling regions are a strongly coupled system of clouds, winds and SST

• Important for ocean primary productivity (hence sink for CO2), fisheries (20% of global catch), and coastal communities

• Global models often have large warm biases (poor stratocumulus/SST simulation) off western coasts of America/Africa

• Work done by John Donners (UJCC/NCAS-Climate, Univ. of Reading)

04/21/23

Change in summer SSTs at 4 locations when increasing atmosphere res. only from N96 (150km)

to N144 (90km)

California

Africa

S. America

Africa

04/21/23

Ocean upwelling

summerwinter

04/21/23

Deep upwelling

Observations: tens to hundreds m/month

~ 25m/s ==> 65 m/month

High resolution ocean (1/3°)

~ 5m/s ==> 13 m/month

Low resolution ocean (1°)

Warning: different color scales!

Seasonal cycle of SST – need high resolution in both components


Summary

• Global coupled models lacked resolution to represent upwelling adequately; regional studies used atmosphere- or ocean-only models.

• Both higher atmosphere and ocean resolution improve the seasonal cycle of SSTs: either due to cloud or upwelling effects.

• Upwelling at higher ocean resolution is concentrated near the coastline, drawing colder water from deeper down the water column.

• Cloud feedback enhances SST cooling due to upwelling.

Wind stress curl/ocean current interaction• Scale interactions between atmosphere and

ocean can cause changes to the mean state – often involve wind stress curl

• Changes to both atmosphere and ocean resolutions both play a role – an important question as to how one should balance their relative resolutions

• Sasaki and Nonaka (2006) showed how different forcing fields (NCEP vs QuikSCAT winds) can also cause differences to the Hawaiian Lee Countercurrent – due to wind stress curl propagated from Hawaiian Islands

• The following uses the coupled HiGEM model

Wind stress curl (colours, N/m2) and ocean current at 35m (contours at 5cm/s intervals) for a matrix of coupled atmos/ocean model resolutions (150 or 90km Atm, 1 or 1/3° Ocn models)

LoA-LoO

HiA-HiOLoA-HiO

HiA-LoO

Agulhas

• Transport of properties (heat, salt) between Indian and Atlantic Oceans potentially very important for long-term Atlantic meridional overturning

• This is (almost) always wrong in climate model simulations – since their resolution is rarely sufficient to resolve the retroflection

• Impact on mean model properties, but also makes models more difficult to use and compare to observations (changing water mass properties in the Atlantic)

Agulhas region in HadGEM1.1 (top) and HiGEM1.1 (bottom)

Retroflection and eddy-shedding at high resolution, mean flow at low resolution – significant impacts on water mass properties -> climate change detection

Agulhas

• Need high resolution for retroflection (better that ½ degree?), and even higher for Agulhas ring variability

• Important implications for ocean and coupled climate:

• Water mass erosion

• Heat/freshwater transport and long-term MOC

• Understanding of the mechanisms of MOC variability, for comparison with observations

• The following courtesy of A. Biastoch, IFM-GEOMAR

DRAKKAR Hierarchy [DRAKKAR Group, 2007], based on NEMO v2.3 [Madec, 2008]

• Tripolar grid, coupled configuration with 2-way interaction between

Global coarse-resolution model (1/2° )

Regional high-resolution model (1/10° )

• 46 vertical levels, partial cell topography

• State-of-the-art physics / parameterizations

• Thermodynamic-dynamic sea-ice model

• 20-yr spinup with global model

• O(50-yr) atmospheric forcing:

NCEP/NCAR-derived (“CORE”),

applied via Bulk formulae,

6h/1d-resolution, inter-annual

variability (1958-2004)

• ~O(months) on (NEC SX-8/9)

• 5-daily output several TB ORCA05

AG01

Nested Agulhas Model using AGRIFNested Agulhas Model using AGRIF

Courtesy A. Biastoch

Agulhas-Induced MOC VariabilityAgulhas-Induced MOC Variability

±1.5 Sv decadal variability rapid communication to the North Atlantic

MOC-difference at

1000m: Exp. with

minus without

Agulhas nest

(interannually

filtered)

Difference in MOC and

North Brasil Current

at 6°S

[Biastoch, Böning, Lutjeharms; Nature, 2008]

Principle of Signal PropagationPrinciple of Signal Propagation

Topographic Shelf WavesTopographic Shelf Waves

Rossby Waves / Agulhas Rings

Rossby Waves / Agulhas Rings

Large-scale horizontal circulation and eddy kinetic energy


Agulhas Variability vs. Subpolar Deepwater FormationAgulhas Variability vs. Subpolar Deepwater Formation

Complete Forcing

Effect of Agulhas mesoscale variability

Effect of heat flux variability

Standard deviation of interannual MOC strength

Agulhas influence reaches into northern hemisphere

… in tropics of similar amplitude as those by subpolar deepwater formation


Ocean-atmosphere coupling

• Couple 0.5° atmosphere with either 1° or 0.1° ocean model (CAM3.5)

• Look at relationship between ocean SST and atmospheric winds

• Although much more variability in high resolution ocean case, the regression coefficient (coupling strength) is unchanged – and weaker than observed

• Suggests missing or misrepresented physics in the atmosphere model

• Supplied by F. Bryan, NCAR

Lower resolution ocean

Higher resolution ocean

Climate change and Southern Ocean• Look at the response of models with different horizontal

resolution to climate change (1% to 2x CO2 forcing) - 1° vs ¼° ocean resolution

• Westerly winds shift poleward under climate change in both models

• Response of ocean and hence climate is different

• Eddy transport works against mean flow, and changes compensate the increased Ekman flow

• Hence change in oceanic energy transport much reduced at higher resolution

• Atmospheric energy transport reduced in turn

• Hence response of coupled climate to forcing is different at higher resolution

• Work from R. Farneti, GFDL, using CM2.4 and CM2.1

Change in ocean energy transport (2x CO2 – control)

Blue = eddy-permitting

Change in atmosphere energy transport (2x CO2 – control)

Blue = eddy-permitting

Heat transport components of the control and 2xCO2 integrations using the higher resolution ocean model CM2.4

Summary/issues

• Increasing number of examples of how mesoscale ocean processes play important role in ocean, atmosphere and climate mean state and variability

• Are there recommendations to make with regards to ocean resolution and e.g. IPCC-class models

• Is there enough evidence that eddy-permitting models are needed in order to produce “good” climate projections?

• Does the timescale under consideration (annual/decades vs centuries) make any difference


Questions

© Crown copyright Met Office Impacts of the ocean mesoscale on mean state and coupled phenomena Malcolm Roberts, Met Office Hadley Centre.

Documents