The new GFDL global atmosphere and land model AM2/LM2:

Evaluation with prescribed SST simulations

The GFDL Global Atmospheric Model Development Team1

Revised for theJournal of Climate, February 2004

_____________________________________________________________________________

1The members of the GFDL Global Atmospheric Model Development Team include Jeffrey L.Anderson+, V. Balaji$, Anthony J. Broccoli+^, William F. Cooke%, Thomas L. Delworth+, KeithW. Dixon+, Leo J. Donner+, Krista A. Dunne#, Stuart M. Freidenreich+, Stephen T. Garner+,Richard G. Gudgel+, C. T. Gordon+, Isaac M. Held+, Richard S. Hemler+, Larry W. Horowitz+,Stephen A. Klein+*£§, Thomas R. Knutson+, Paul J. Kushner+*¥, Amy R. Langenhorst%, Ngar-Cheung Lau+, Zhi Liang%, Sergey L. Malyshev@, P. C. D. Milly#, Mary J. Nath+, Jeffrey J.Ploshay+, V. Ramaswamy+, M. Daniel Schwarzkopf+, Elena Shevliakova@, Joseph J. Sirutis+,Brian J. Soden+, William F. Stern+, Lori A. Thompson%, R. John Wilson+, Andrew T. Witten-berg$, and Bruce L. Wyman+.

+GeophysicalFluid DynamicsLaboratory(GFDL), NationalOceanicandAtmosphericAdminis-tration, Princeton, New Jersey

$Atmospheric and Oceanic Sciences Program, Princeton University, Princeton, New Jersey

%RS Information Systems, McLean, Virginia, presently at GFDL

#U. S. Geological Survey, Princeton, New Jersey

@Department of Ecology and Evolutionary Biology, Princeton University, Princeton, New Jersey

^Now at the Department of Environmental Sciences, Rutgers University, New Brunswick, NewJersey

§Now at the Atmospheric Science Division, Lawrence Livermore National Laboratory, Liver-more, California

¥Now at the Department of Physics, University of Toronto, Toronto, Ontario, Canada

*Global Atmospheric Model Development Team leaders (2001-2003)

£Correspondingauthor:StephenA. Klein, GeophysicalFluid DynamicsLaboratory/NOAA, Prin-ceton University Forrestal Campus/ US Route 1, P.O.Box 308, Princeton, New Jersey 08542, e-mail: Stephen.Klein@noaa.gov

Abstract

The configurationandperformanceof a new global atmosphereand land model for cli-

materesearchdevelopedat the Geophysical Fluid DynamicsLaboratory(GFDL) is presented.

The atmospheremodel,known asAM2, includesa new gridpoint dynamicalcore,a prognostic

cloud scheme,anda multi-speciesaerosolclimatology, andcomponentsfrom previous models

usedat GFDL. The land model,known asLM2, includessoil sensibleand latentheatstorage,

groundwaterstorage,andstomatalresistance.Theperformanceof thecoupledmodelAM2/LM2

is evaluatedwith a seriesof prescribedsea-surface temperature(SST) simulations.Particular

focusis givento themodel’s climatologyandthecharacteristicsof interannualvariability related

to El Niño/Southern Oscillation (ENSO).

OneAM2/LM2 integrationwasperformedaccordingto the prescriptionsof the second

AtmosphericModel IntercomparisonProject(AMIP II) anddataweresubmittedto theProgram

for ClimateModel DiagnosisandIntercomparison(PCMDI). Particularstrengthsof AM2/LM2,

asjudgedby comparisonto othermodelsparticipatingin AMIP II, includeits circulationanddis-

tributionsof precipitation.Prominentproblemsof AM2/LM2 includea cold biasto surfaceand

tropospherictemperatures,weaktropicalcycloneactivity, andweaktropicalintraseasonalactivity

associated with the Madden-Julian oscillation.

An ensembleof 10 AM2/LM2 integrationswith observedSSTsfor thesecondhalf of the

twentiethcenturypermitsa statisticallyreliableassessmentof themodel’s responseto ENSO.In

general,AM2/LM2 producesa realisticsimulationof theanomaliesin tropicalprecipitationand

extratropical circulation that are associated with ENSO.

GFDL GAMDT

1

1. Introduction

In this report, an overview is presentedof the new GFDL global atmosphereand land

modelknown as"AM2/LM2". AM2 andLM2 are,respectively, the atmosphericandterrestrial

componentsof the earth-systemmodelthat is underdevelopmentat GFDL for climateresearch

andclimatepredictionapplications.In developingAM2/LM2, the focushasbeenon consolidat-

ing andimproving thevariousversionsof suchmodelsthathavebeenusedin GFDL’spast(Ham-

ilton etal. 1995,SternandMiyakoda1995,Delworthetal. 2002).Theprincipalaim is to createa

model that realistically representsthe dynamic,thermodynamic,andradiative characteristicsof

the climatesystemandis suitablefor couplingto oceanandsea-icemodelswithout flux adjust-

ment.Balancedagainstthis aim is theneedto have a modelcomputationallyfastenoughso that

ensemble multi-century integrations may be performed.

Although AM2/LM2 incorporatesmany componentsof previous models used within

GFDL, it doesrepresenta substantialbreakfrom thepast.AM2 includesa new gridpointatmos-

phericdynamicalcore,a multi-speciesthreedimensionalaerosolclimatology, a fully prognostic

cloudschemeandamoistturbulencescheme.LM2 incorporatessoil sensibleandlatentheatstor-

age,groundwaterstorage,stomatalcontrolof transpiration,andsoil- andplant-dependentparam-

eters.Thesenew componentshave requiredmodificationandretuningof componentsthat were

carriedover from previousmodels.This hasled to a modelwith morecapabilitiesandpotential

for growth as well as a model with simulationcharacteristicsgenerallysuperiorto that of the

older GFDL models.

Our modeldevelopmenteffort is team-basedandinvolvesa broadcrosssectionof exper-

GFDL GAMDT

2

tise from within andoutsideof GFDL; this hasrequireda challengingdegreeof coordination.A

simultaneouschallengehasbeenGFDL’s transitionfrom vector to parallelcomputingarchitec-

tures.To addressthesechallenges,anin-housesoftwareframework known asthe"Flexible Mod-

eling System"(FMS, http://www.gfdl.noaa.gov/fms) hasbeendeveloped.FMS-basedcodesare

modular, useFortran 90, and are basedon standardizedinterfacesbetweencomponentmodels

(i.e. land,atmosphere,ocean,sea-ice).Thesoftwareconservatively exchangesthefluxesof heat,

moisture,andmomentumbetweencomponentmodelswhich mayhave differenthorizontalgrids.

The FMS codeorganizationisolatesthoseaspectsof the coderelatedto parallelcomputingto a

relatively simple messagepassing interface (http://www.gfdl.noaa.gov/~vb/mpp.html). As a

result,scientistsdevelopingnew codefor themodelneednot learntheintricaciesof parallelcom-

puting.Using theFMS, it hasbeenpossibleto rapidly testa varietyof modelconfigurationsand

follow parallel developmentpathsfor the atmosphere,ocean,land, and sea-icemodels.FMS

modelshave beentestedsimultaneouslyon vectorandparallelplatforms.As a consequence,the

transition to a new parallel computing environment was made with relative ease.

Section2 of thereportdocumentsthecomponentsof AM2/LM2 aswell astheboundary

conditionsfor theexperimentsperformed.Section3 providesa discussionof AM2/LM2’ s clima-

tologicalcirculation,hydrologyandradiationbudget,aswell asits variability. A brief comparison

of thequality of AM2/LM2’ s climatologyto thatof othermodelsis givenin section4 andfuture

plans are discussed in section 5.

GFDL GAMDT

3

2. Model components and boundary conditions for model integrations

Thecomponentsof AM2/LM2 aredescribedin the following threesubsections.For ease

of reference, a summary of model components is given in Table 1.

a. Grid-point dynamical core

The hydrostatic, finite difference dynamical core has been developed from models

describedin Mesingeret al. (1988)andWyman(1996).TheAM2 dynamicalcoreusesthesame

setof prognosticvariablesasin thesereferences,but hasa differenthorizontalandverticalgrid.

Thelatitude-longitudehorizontalgrid is thestaggeredArakawaB-grid (ArakawaandLamb1977)

with a resolutionof 2.5˚ longitudeby 2˚ latitude.In thevertical,a hybrid coordinategrid is used;

sigmasurfacesnearthegroundcontinuouslytransformto pressuresurfacesabove250hPa (Table

2). The model hastwenty-four vertical levels with the lowest model level about thirty meters

above thesurface.Thereareninefull levelsin thelowest1.5km above thesurface;this relatively

fine resolutionis neededby the boundarylayer turbulencescheme.Aloft the resolutionis more

coarsewith approximatelytwo km resolutionin theuppertroposphere.Five levelsarein thestrat-

osphere,with thetop level at about3 hPa. Theprognosticvariablesarethezonalandmeridional

wind components,surface pressure,temperature,and tracers.The tracersinclude the specific

humidity of water vapor and three prognostic cloud variables (section 2b3).

The model utilizes a two-level time differencingscheme.Gravity waves are integrated

usingtheforward-backwardscheme(Mesinger1977)andasplit timedifferencingschemeis used

for longeradvective andphysicstime steps(Gadd1978).Theadvective termsareintegratedwith

a modifiedEuler backward schemethat haslessdampingthanthe full backward scheme(Kuri-

GFDL GAMDT

4

haraandTripoli 1976).Thegravity wave,advectiveandphysicstimestepsare200,600,and1800

seconds, respectively.

Thevertical finite differenceschemeusedis from SimmonsandBurridge(1981),except

thatthepressuregradientformulationis replacedwith thefinite-volumemethodfrom Lin (1997).

Improvementsto theflow in thevicinity of steepmountainsresultfrom its use.Horizontaladvec-

tion usescenteredspatialdifferencing.Momentumadvection is fourth-order;temperatureand

traceradvectionaresecond-order. The vertical advectionof tracersusea finite-volumescheme

(Lin et al. 1994) with the piecewise parabolicmethodof Colella and Woodward (1984).Grid

point noiseandthe2∆x computationalmodeof theB-grid arecontrolledwith linearfourth-order

horizontaldiffusion.To preventspuriousdiffusionalongslopingcoordinatesurfaces,thediffusive

fluxesof heatandmoistureareadjustedwith a linearcorrectiontowardpressuresurfaces.A sec-

ondorderShapiro(1970)filter is appliedto thedeparturesfrom thezonalmeanof thezonalwind

componentandto thetotalmeridionalwind componentat thetopmodellevel to reducethereflec-

tion of waves.Fourierfiltering is appliedpolewardof 60˚ latitudeto damptheshortestresolvable

wavessothata longertimestepcanbetaken.Thefilter is appliedto themassdivergence,thehor-

izontal omega-alpha term, the horizontal advective tendencies, and the momentum components.

Although the numericalschemesaredesignedto conserve total energy, someaspectsof

thedynamicalcoredonot.Theseincludethehorizontaldiffusion,theShapirofilter usedat thetop

level, thetimedifferencing,andthepressuregradientpartof theenergy conversionterm.To guar-

anteeenergy conservationfor long climateruns,a globalenergy correctionis appliedto tempera-

ture.

Onemayaskwhy thisB-grid dynamicalcorewasselectedwhenastandardEulerianspec-

GFDL GAMDT

5

tral dynamicalcoreis alsoavailablewithin FMS. Throughoutthe developmentprocess,integra-

tionshave beenperformedwith this spectralcore.Most of thebiasesthatarepresentsimulations

with theB-grid corearealsopresentin simulationswith a T42 spectralcore,includinganexag-

gerateddoubleIntertropicalConvergenceZone(ITCZ) structurein the Pacific, the equatorward

biasin thepositionof theNorth Atlantic westerlies,andthepositive biasin Arctic sealevel pres-

sure.Overall figuresof merit of the sort discussedbelow are somewhat superiorin the B-grid

model,which partly reflectsthat the tuning processfocusedon integrationswith the gridpoint

model.Onenoticeabledifferenceis thatSouthAmericanrainfall is superiorin theB-grid model,

owing to thedifficulty in representingtheAndesin a spectralmodelof this resolution.Thedefi-

cient spectralmodel solution occurs despite the use of a sophisticatedspectral topography

smoothingalgorithm(Lindberg andBroccoli 1996).Apart from theseconsiderations,differences

in computationalefficiency wheninterchangingdynamicalcoresaremodestat this resolution.For

these reasons, the B-grid core has been chosen for the default model.

b. Atmospheric physics

1) RADIATION AND PRESCRIBEDOZONE AND AEROSOL CLIMATOLOGIES

TheshortwaveradiationalgorithmfollowsFreidenreichandRamaswamy(1999;hereinaf-

ter FR99).When this radiationcodewas first employed in AM2, it was deemednecessaryto

increaseits computationalefficiency. As a result,thebandstructureandthenumberof exponen-

tial-sumfit termswithin somebandshavebeenaltered,resultingin fewerpseudo-monochromatic

columnarcalculations.Specifically, the bandfrom 0 to 2500 cm-1 now hasone insteadof six

terms,owing to considerationof CO2 astheonly absorberfor this interval; thereis onebandfrom

2500to 4200cm-1 insteadof three,andthetotal numberof termsis reducedfrom twelve to eight;

GFDL GAMDT

6

thereis onebandfrom 4200to 8200cm-1 insteadof four, andthetotalnumberof termsis reduced

from twenty-fourto nine; thenumberof termsfor the8200to 11500cm-1 bandis reducedfrom

sevento five while that for the11500to 14600cm-1 bandis reducedfrom eight to two; thereare

now threebandsbetween27500and34500cm-1 insteadof five, eachwith oneterm.Altogether,

thenumberof bandsin thesolarspectrumis reducedfrom twenty-five to eighteen,while thetotal

numberof pseudo-monochromaticcolumn calculationsrequiredper grid-box is reducedfrom

seventy-two to thirty-eight. This new bandstructureand the revised exponential-sumfits have

beendevelopedandtestedwith benchmarkcalculationsusingthe HITRAN 2000line catalogue

(Rothmanet al. 2003).Despitethe reducedbandstructure,the maximumerror in the clear-sky

heatingratesremainsat the lessthan 10% as was obtainedwith the seventy-two term fit. The

errorsin the shortwave overcastsky heatingratesfor the watercloud modelconsidered(Slingo

1989)arenow about15%,increasedfrom about10%for theseventy-two termfit; for ice clouds,

the errors tend to be larger (FR99) and for the present parameterization could reach 25%.

The interactionsconsideredby this shortwave parameterizationinclude absorptionby

H2O, CO2, O3, O2, molecularscattering,andabsorptionandscatteringby aerosolsandclouds.For

waterclouds,the single-scatteringpropertiesin the solarspectrumfollow Slingo (1989);for ice

clouds,theformulationfollows Fu andLiou (1993).To accountfor theradiationbiasthatresults

from usinghorizontallyhomogeneousclouds(Cahalanetal. 1994),thecloudliquid andicewater

contentsaremultiplied by 0.85beforecalculatingboth shortandlongwave radiative properties.

Three-dimensional,monthly-meanprofilesof aerosolmassconcentrationsandtheir opticalprop-

ertiesfollow Haywood et al. (1999)andHaywood (personalcommunication).The prescription

accountsfor sea-salt(low windspeedcase)andthenaturalandanthropogeniccomponentsof dust,

carbonaceous (black and organic carbon), and sulfate aerosols.

GFDL GAMDT

7

Ozoneprofilesfollow FortuinandKelder(1998)andarebasedonobservationsfrom 1989

to 1991.Thisclimatologyhasbeenshown to yield resultsthatrepresentsubstantialimprovements

over thoseobtainedwith previousolderclimatologiesusedin theGFDL globalmodels(Ramas-

wamy and Schwarzkopf 2002).

The oceansurfaceis assumedto be Lambertian,with the albedoa function of the solar

zenith angle following the formulation of Taylor et al. (1996).

Theband-averagingof thesingle-scatteringparametersin theshortwave parameterization

is performedusingthethick-averagingtechnique(EdwardsandSlingo1996).Thedelta-Edding-

ton techniqueis employed to computethe layer reflectionandtransmissionbasedon the single-

scatteringpropertiesof that layer (FR99).The diffuseincidentbeamis assumedto be isotropic

andits reflectionandtransmissionarecomputedusinganeffective angleof 53˚, in contrastto the

4-point quadratureschemeusedin FR99.The net direct anddiffusequantitiesin eachlayer are

givenby theweightedsumof theclearandovercastsky fractionspresentin that layer. The total

shortwave fluxes and heatingratesare computedusing an adding scheme(Ramaswamy and

Bowen 1994).

The longwave radiation code follows the modified form of the Simplified Exchange

Approximationand is also developedand testedusing benchmarkcomputations(Schwarzkopf

andRamaswamy1999).It accountsfor theabsorptionandemissionby theprincipalgasesin the

atmosphere,includingH2O, CO2, O3, N2O, CH4, andthehalocarbonsCFC-11,CFC-12,CFC-113

and HCFC-22.Aerosolsand clouds are treatedas absorbersin the longwave, with non-grey

absorptioncoefficientsspecifiedin theeightspectralbandsof the transferscheme,following the

methodologyadoptedin Ramachandranet al. (2000). For water clouds,the absorptioncoeffi-

GFDL GAMDT

8

cientsfollow thoseemployed in Held et al. (1993); for ice clouds,the Fu andLiou (1993)pre-

scription is used.

In both the shortwave and longwave parameterizations,the water vapor continuumis

parameterizedaccordingto theCKD 2.1 formulationof Cloughet al. (1992).Additionally, short-

waveandlongwavebandandcontinuumparametersarederivedusingtheHITRAN 2000line cat-

alogue (Rothman et al. 2003).

2) CUMULUS PARAMETERIZATION AND CONVECTIVE MOMENTUM TRANSPORT

Moist convectionis representedby theRelaxedArakawa-Schubert(RAS) formulationof

Moorthi and Suarez(1992). In this parameterization,convection is representby a spectrumof

entrainingplumeswhich produceprecipitation.Closureis determinedby relaxingthecloudwork

functionfor eachcloudin thespectrumbackto a critical valueover a fixedtime scale.A number

of local modifications have been made; these are enumerated below.

• (Thefractionof watercondensedin thecumulusupdraftswhichbecomesprecipitation

(known astheprecipitationefficiency) is specifiedto be0.975for deepconvectionand

0.5 for shallow convection.Deepconvection is definedasupdraftswhich detrainat

pressurelevels above 500 hPa whereasshallow convection is definedas updrafts

which detrainbeneath800hPa.For pressuresbetween500and800hPa, theprecipita-

tion efficiency is linearly interpolatedin pressurebetweenthe valuesfor deepand

shallow convection.This versionof RAS lackscumulusupdraftmicrophysicssuchas

that developed by Sud and Walker (1999).

GFDL GAMDT

9

• The non-precipitatedfraction of condensedwater, 0.025for deepconvectionand0.5

for shallow convection, is a source of condensate for the prognostic cloud scheme.

• Re-evaporationof convective precipitationis allowed to occur. This versionof RAS

doesnot include the effects of convective downdrafts developedin a later version

(Moorthi and Suarez 1999).

• The time scaleover which the cloud work function is relaxed to a cloud type depen-

dentvalueis modifiedsothatdeepupdraftsrelaxover a time scaleof about12 hours

but shallow updrafts relax over a time scale of only 2 hours.

• The cloud type dependentcloud work function is taken from Lord and Arakawa

(1980)exceptthat it is reducedto zerofor shallow updraftswhich detrainbelow 600

hPa.Thischangewasmadeto reduceundesirablelow cloudbehavior onENSOtimes-

calesover theEquatorialPacific Cold Tongueanda relatedinstability which occurred

when an earlier version of AM2/LM2 was coupled with a mixed-layer ocean.

In additionto thesechanges,deepconvectionis preventedfrom occurringin updraftswith

a lateral entrainmentrate lower than a critical value determinedby the depthof the sub-cloud

layer (Tokiokaet al. 1988).This modificationresultsin generalimprovementsto thedistribution

of tropicalprecipitationandanincreasein tropicaleddyandstormactivity. A deleteriouseffect is

a cooling of the uppertropical troposphereby 2K. This constraintis appliedonly to convective

updraftsthat detrainabove 500 hPa. This constrainthasnot beenappliedto shallower updrafts

becausethe resultingdecreasein the intensityof shallow convection leadsto large increasesin

tropical low cloudcover to thepoint thatadjustingothermodelcomponentsto achieve radiative

balanceis too difficult. It is surprisinghow importantweakly entrainingupdraftsin RAS areto

GFDL GAMDT

10

the distribution of low level cloud cover.

The impact of cumulusconvection on the horizontalmomentumfields hasbeenrepre-

sented by adding to the vertical diffusion coefficient for momentum a term of the form:

(1)

where is the contribution to the momentumdiffusion coefficient from cumulusconvection,

is thetotal cumulusmassflux predictedby RAS with unitskg m-2 s-1, is thedepthof con-

vectionin meters, is thedensityof air, and is a dimensionlessconstantwith value0.2.The

chosenvalueof is roughlyconsistentwith thecloudresolvingmodelresultsof MapesandWu

(2001)whoestimatethat10mmof convectiveprecipitationdampsout40-80%of themeanbaro-

clinic kineticenergy. If oneassumesthatthehorizontalflow hastheverticalstructureof a full sine

wave over a depthof 10 km, thenthis rateof decaycorrespondsto thechoiceof between0.1

and0.2. In AM2/LM2, this convective momentumtransportmutesthetendency of AM2/LM2 to

producea doubleITCZ in the tropical Pacific andresultsin a morerealisticregressionof zonal

surfacewind stressin the equatorialPacific on NINO3 SSTs.A deleteriousimpact is the sharp

reductionof tropical transienteddy activity which has in part motivated the inclusion of the

Tokioka modificationdescribedabove. The marked consequencesof including this convective

momentumtransporton the ENSO spectrumfrom a coupledmodel using AM2/LM2 will be

described elsewhere.

Thechoiceof adown-gradientdiffusive formulationof convectivemomentumtransportin

placeof themoreconventionalmass-fluxformations(e.g.Gregory et al. 1997),wasbasedin part

onconcernsregardingnumericalstability (Kershaw etal. 2000).Giventheuncertaintiesasto how

Kcu γ Mcd ρ⁄=

Kcu

Mc d

ρ γ

γ

γ

GFDL GAMDT

11

convective organizationmodifiesverticalmomentumtransportsandthe inability of a large-scale

model to addressthe questionof convective organization,it was felt that this simpler scheme

might beadequate.Onecanmimic thetendenciesproducedby theGregory et al. (1997)scheme

with diffusion if the vertical structureto the meanflow is simpleenough(linear in pressure,for

example).However, thediffusioncoefficientsfrom (1) arelargerat low levelsthanthosegivenby

anequivalentmass-fluxformulation.Thelargermomentumtendenciesat low levelsgeneratedby

(1) appearto beimportantto theadvantagesobtainedfrom thisdiffusive formulationin acoupled

model.

3) CLOUD SCHEMEAND RADIATION BALANCE TUNING

Large-scaleclouds are parameterizedwith separateprognostic variables for specific

humidity of cloud liquid and ice. Cloud microphysicsareparameterizedaccordingto Rotstayn

(1997)with anupdatedtreatmentof mixedphaseclouds(Rotstaynet al. 2000).Fluxesof large-

scalerain and snow are diagnosedand the amountof precipitationflux inside and outsideof

cloudsis trackedseparately(Jakob andKlein 2000).Theparticlesizeof liquid cloudsneededfor

radiationcalculationsis diagnosedfrom theprognosedliquid watercontentandanassumedcloud

droplet numberconcentrationwhich is specifiedto be 300 cm-3 over land and 100 cm-3 over

ocean.For ice clouds,the particlesize is specifiedasa function of temperaturebaseduponan

analysisof aircraft observations(Donneret al. 1997).Cloudsareassumedto randomlyoverlap.

Becauseof the coarsevertical resolutionin the uppertroposphere(Table2), this assumptionis

acceptablethere,but for cloudsin the lower tropospherethis assumptionis poor (HoganandIll-

ingworth 2000).

Cloudfractionis alsotreatedasa prognosticvariableof themodelfollowing theparame-

GFDL GAMDT

12

terizationof Tiedtke (1993)with two importantchanges.Thefirst changeinvolvesthe treatment

of super-saturatedconditionsin grid-cells.In theseconditions,it is judgedthat theparameteriza-

tion hasomittedsomemissingcondensationprocess.In Tiedtke (1993),any vapor in excessof

supersaturationwascondenseddirectly into precipitationwithout makingcloud water. In AM2/

LM2, this excessvaporis condensedinto cloud insteadof precipitation.This is justifiedbecause

the AM2/LM2 implementationomits somekey condensationterms,suchasthe boundarylayer

condensation source term from Tiedtke (1993).

Thesecondchangeinvolvestheerosionconstant,akey unknown parameterin theTiedtke

parameterization,that governsthe rateat which sub-gridscalemixing dissipatescloudsin sub-

saturatedgrid cells.Ratherthanusea singleglobally constantvalue(Tiedtke 1993),the erosion

constantis madea functionof thestateof thegrid cell in AM2/LM2. If verticaldiffusionis acting

in a grid cell, the erosionconstantis set to the large value of 5 x 10-5 s-1 which ensuresrapid

dissolutionof cloudsin sub-saturatedcells. If convectionis occurringwithout verticaldiffusion,

theerosionconstantis setto thesmallervalueof 4.7 x 10-6 s-1. If neitherconvectionnor vertical

diffusionis occurringin a grid cell, thentheerosionconstantis setto theevensmallervalueof 1

x 10-6 s-1, the original valueusedin Tiedtke (1993).The relative valuesof the erosionconstant

reflect the degree of sub-grid scale turbulenceand mixing occurring in a grid cell. In fully

turbulentlayers,themixing is rapidsothatpartially cloudyregimesshouldbemoretransitory(in

theabsenceof sourcesof partialcloudiness)thanin quiescentconditionsfor which partly cloudy

conditionscouldexist for a long time.Theerosionconstantin thepresenceof convectionis very

influential in controllingthebrightnessof tradecumulusregions.However, thevalueof 4.7x 10-6

s-1 usedin the presenceof convection is about40 times smallerthan the value for the erosion

constantsuggestedby analysisof large-eddysimulationsof trade cumuli from the Barbados

GFDL GAMDT

13

Oceanographicand MeteorologicalExperiment(BOMEX) (Siebesmaet al. 2003). This may

partly explain why the shortwave reflectionfrom tradecumulusregionsis too large (Figure10,

below).

Themodel’s radiationbudgetis tunedsothatthelong-termglobalandannualmeanoutgo-

ing longwaveandabsorbedsolarradiationarecloseto observedandthatthenetradiative balance

is between0 and1 W m-2. This is accomplishedprimarily throughadjustmentsto theclouddrop

radiusthresholdvaluefor theonsetof raindropformation(avalueof 10.6µm is used),theerosion

constantin thepresenceof convection,andto thespecifiedprecipitationefficiency for deepcon-

vectionin RAS.Althoughacritical radiusof 10.6µm is smallerthancanbejustified,it is consid-

erably larger thanvaluesusedpreviously in other large-scalemodels.The valueusedin AM2/

LM2 is perhapscloseenoughto realisticvaluesthatthelack of sub-gridscalevariability to cloud

waterin microphysicalcalculationsmaybethereasonthatlarge-scalemodelstunethisparameter

(Rotstayn 2000, Pincus and Klein 2000).

4) SURFACE FLUXES

Surfacefluxes are computedusing Monin-Obukhov similarity theory, given the atmos-

phericmodel’s lowestlevel wind, temperature,andhumidity andthe surfaceroughnesslengths,

temperature,and humidity. To recognizethe contribution to surfacefluxes from sub-gridscale

wind fluctuations,a ‘gustiness’componentproportionalto thesurfacebuoyancy flux is addedto

the wind speedusedin the flux calculations(Beljaars1995). Oceanicroughnesslengthsfor

momentum,heat,andmoistureareprescribedaccordingto Beljaars(1995).As a result of this

prescriptionfor roughnesslengths,the exchangecoefficientsfor momentumincreasewith wind

speedwhereasthe heatand scalarexchangecoefficients remain fairly constantacrossa wide

GFDL GAMDT

14

range of wind speed.

The treatmentof surfacefluxes in highly stableconditionsrequiresspecialattentionas

with traditionalformulations,thetemperatureof thesurfacewill decouplefrom thatof theatmos-

phereleadingto excessive coolingof thewinter landsurface(Derbyshire1999).In orderto pre-

ventthisdecoupling,thestability functionsaremodifiedsothatmixing will occurfor Richardson

numbersgreaterthan0.2.This pragmaticfix for a problemcommonto many modelswill hope-

fully bereplacedwith a morephysically basedtreatmentbasedon theactive researchin this area

(Holtslag 2003).

Recognizingthatflow over hills with horizontallengthscalessmallerthanthosethatgen-

erategravity waves inducessubstantialdrag on the atmosphere,a parameterizationfor ‘oro-

graphicroughness’hasbeenintroduced(Wood and Mason1993). In this parameterization,an

‘effective roughness’lengthproportionalto thestandarddeviationof orography atsub-gridscales

is usedto enhancethe exchangecoefficient for momentum.The exchangecoefficients for heat

and scalarsare unaltered.In the absenceof this parameterization,anomalouslow level jets

occurred in the vicinity of steep orography gradients.

5) TURBULENCE

Verticaldiffusioncoefficientsarepredictedaccordingto their physical context. A K-pro-

file schemebaseduponLock etal. (2000)is usedfor convectiveboundarylayersandnear-surface

convective layersdrivenby stronglongwave coolingfrom cloudtops(i.e. stratocumulusconvec-

tion). Thetopof theconvectiveboundarylayeris determinedby lifting anearsurfaceparcelto its

level of neutralbuoyancy. Likewise,thebottomof thestratocumuluslayer is determinedby low-

GFDL GAMDT

15

eringa radiatively cooledparcelto its level of neutralbuoyancy. For bothtypesof convection,the

mixing acrossthetop of theselayersis prescribedwith anentrainmentparameterizationwhich is

basedupona combinationof observation and large eddysimulationresults.For the convective

boundarylayer, the entrainmentparameterizationfollows that of Lock et al. 2000for which the

entrainmentrateis proportionalbothto thesurfacebuoyancy flux andthesurfacewind stressand

is inverselyproportionalto thestrengthof theinversionat thetopof theconvective layer. For stra-

tocumuluslayers,a parameterizationfor the entrainmentrate is usedwhich approximately

reduces to:

(2)

where is thelongwaveflux divergenceacrosscloudtop in W m-2, is theheatcapacityof

air at constantpressure,and is thejump in liquid watervirtual potentialtemperatureacross

the entrainmentinterface.This parameterizationdiffers from Lock et al. 2000in that the buoy-

ancy reversal term hasbeenomitted which is justified as follows. To accuratelycalculatethe

buoyancy reversaltermrequiresa goodpredictionof the liquid watercontentat cloudtop. How-

ever, confidencein themodel’s predictionof cloudtop liquid wateris low becauseno provisions

have beenmadeto accountfor thesub-gridvertical structureof the inversionlayerasis donein

Lock (2001)andGrenierandBretherton(2001).In theabsenceof thebuoyancy reversalterm,the

radiatively driven entrainmentrate hasbeenenhancedby increasingthe constantin (2) to 0.5,

approximately double the value used in Lock et al. (2000).

For layersof theatmospherewhich arenot partof eithera convective planetaryboundary

layer or a stratocumuluslayer, a local mixing parameterizationis used.For unstablelayers,the

we

we

0.5∆FLW

ρcp∆θvl----------------------=

∆FLW cp

∆θvl

GFDL GAMDT

16

mixing coefficientsof Louis (1979)areused.For stableturbulent layers,conventionalstability

functionsfor whichmixing ceaseswhentheRichardsonnumberexceeds0.2areusedexceptnear

the surface.If the surfacelayer is understableconditions,the stability functionswhich provide

enhancedsurfacefluxesfor Richardsonnumbersin excessof 0.2 (section2b4)areblendedwith

theconventionalstability functionsin thelowestkilometerof theatmosphere.This is doneto pro-

vide a smooth transition from enhanced mixing near the surface to conventional mixing aloft.

Finally, the vertical diffusion coefficients are given ‘memory’ by making the diffusion

coefficientsprognosticvariablesanddampingtheirvaluesto thosediagnosedfrom theinstantane-

ousstatewith adampingtimescaleof onehour. This treatmentpreventsa2∆t oscillationin stable

turbulent layers.

6) GRAVITY WAVE DRAG

Orographicgravity wavesareparameterizedaccordingto Pierrehumbert(1986)andStern

andPierrehumbert(1988).Themomentumflux is a functionof its surfacevalue andtheverti-

calprofileof thesaturationflux requiredfor wavebreaking.Thesurfaceflux is specifiedas:

(for ) (3)

where is the low level horizontalwind velocity, is the low level Brunt-Vaisalafrequency,

and is aneffective horizontalmountainwavelengthwith a fixedvalueof 100km. is a func-

tion of the Froudenumber ( , where is the sub-gridscalemountainheight)andis

specified as:

τs

τ∗ τs

τsρU

3

Nλ----------G F( )–= N

20>

U N

λ G

F Nh U⁄= h

GFDL GAMDT

17

. (4)

In AM2, theconstants and have bothbeentunedto a valueof 1 to optimizethesimulation

of zonalmeanwindsandsealevel pressuregradients.Theheightdependentvalueof thesatura-

tion flux is given by:

(5)

where is theverticalwavelengthof thegravity wavesdeterminedfrom WKB theory. Theflux

at a given level is equal to the flux in the level immediately below or , whichever is smaller.

Notethat this parameterizationomitsenhancedlow level dragfor high andanisotropic

effects(the stressat all levels is oppositethe directionof the low level wind). The omissionof

enhancedlow level drag is partly compensatedby enhanceddragdue to orographicroughness

(Section 2b4).

c. Land model LM2

ThelandmodelLM2 is basedon theLandDynamics(LaD) modeldescribedin detailby

Milly andShmakin(2002;hereinafterMS02).At unglaciatedlandpoints,watermaybestoredin

threelumpedreservoirs: snow pack, soil water (representingthe plant root zone),and ground

water. Energy is storedassensibleheatin 18 soil layersandaslatentheatof fusionin snow pack

andall soil layersexceptthetop layer. For simplicity thesoil latentheat,which wasneglectedby

MS02, is treatedin an idealizedfashion;every soil grid cell except the top layer is assumedto

have 300 kg m-3 “freezablewater” that is hydraulically isolatedfrom the watercycle. For water

G F( ) G∗ F2

F2

a2

+------------------

=

G∗ a

τ∗

τ∗ ρU2DG∗ λ⁄–=

D

τ∗

F

GFDL GAMDT

18

massbalance,soil waterandgroundwaterarenot allowed to freeze,regardlessof temperature.

Evapotranspirationfrom soil is limited by a non-water-stressedbulk stomatalresistanceand a

soil-water-stressfunction.Drainageof soil waterto groundwateroccurswhenthewatercapacity

of therootzoneis exceeded.Groundwaterdischargeto surfacewateris proportionalto groundwa-

ter storage.Model parametersvary spatiallyasfunctionsof mappedvegetationandsoil typesbut

aretemporallyinvariant.CertainLaD-modelparametervaluesweremodifiedfrom thoseassigned

by MS02 for coupling with AM2; these are described below.

Parametersaffecting surfacealbedo(snow-free surfacealbedo,snow albedo,andsnow-

maskingdepth)weretunedon the basisof a comparisonof modeloutputwith NASA Langley

SurfaceRadiationBudgetdataanalyses(Darnellet al. 1988,Guptaet al. 1992).Additionally, to

improve albedofields,threesparse-vegetationclassesof Matthews (1983)werere-assignedrela-

tive to MS02sothatonly theMatthews’ ‘desert’classremainedasdesertin LM2; theotherthree

werere-definedasgrassland.In anotherdeparturefrom MS02,thegeographicvariationof snow-

freealbedoof desertwasprescribedon thebasisof annualmeanalbedofrom theEarthRadiation

BudgetExperiment(ERBE,Barkstromet al. 1989).This wasdonebecausealbedoof desertshas

largeregionalvariations;to representall desertswith asinglealbedo,asis donefor theotherveg-

etation classes, was judged to produce unacceptably large errors.

WhentheLaD modelwasfirst run asLM2 coupledto AM2, computedvaluesof evapora-

tion from land weregenerallysmallerthanexpectedfor the AM2 precipitationandsurfacenet

radiation.To remedythis bias, the non-water-stressedvaluesof bulk stomatalresistancewere

reducedglobally in LM2 by a factorof 5 from the valuespreviously determinedby stand-alone

tuningof theLaD model(MS02).Themagnitudeof this reductionwaschosento produceratesof

GFDL GAMDT

19

evaporationhaving relationsto precipitationandsurfacenet radiationconsistentwith the semi-

empirical relationof Budyko (1974).The necessityfor sucha large parameteradjustmentwas

unexpectedandis underinvestigation.Discrepanciesbetweenstand-aloneandcoupledtuningof

theLaD modelmayberelatedto fundamentalproblemsin thestand-alonetuningstrategy, which

does not permit atmospheric feedbacks.

Theheatcapacityof soil for soil depthslessthan0.3m wasreducedgloballyby afactorof

4 so that the diurnal temperaturerangeof near-surfaceair simulatedby AM2/LM2 is generally

consistentwith theClimateResearchUnit (CRU) observationsof New et al. (1999).Theneedto

adjust the heat capacityin order to increasethe diurnal temperaturerangeis understandable,

becauseit compensatesfor systematicerrorsin theoriginal model.In humid regions,the model

assumptionof anisothermalsurface,in whichthevegetationcanopy andsoil surfaceareatacom-

mon temperature,promotesexcessive sensibleheat flux into the ground. In arid regions, the

modeluseof a global averagesoil wetnessleadsto overestimationof the soil heatcapacityand

thermalconductivity. Theneedfor thisadjustmentis not surprisingbecauseMS02focusedon the

long-termmeanwater and energy balances,quantitiesthat are very insensitive to the soil heat

capacity.

Theverticalstructureof thesoil levelswaschangedfrom theMS02valuessothatthetotal

soil depthis 6 m with the thicknessof soil levels changingfrom 0.02m at the top to 1 m at the

bottom.Relative to MS02,thethickernear-surfacelevelssuppressnumericalproblemsintroduced

when the near-surfaceheatcapacitywas reduced,and the deepersoil domainpermitsthe full

effect of seasonal heat storage to be realized.

GFDL GAMDT

20

d. Boundary conditions and integrations performed

The standard integration described in this study uses the observationally-based AMIP II

SST and sea-ice prescriptions (Gates et al. 1999). The period of integration is from 1 January

1979 to 1 March 1996. The integration was initialized from another spun-up integration of the

model with slightly different boundary conditions and forcing from the AMIP prescription. The

model output from this integration was submitted to the PCMDI in February 2004. A monthly cli-

matology was formed from this integration for the years 1979 through 1995 and was compared to

observations in sections 3a and 4.

A second set of integrations discussed below is a 10-member ensemble of 50-year integra-

tions, from January 1951 to December 2000, that uses another SST and sea-ice data prescription

developed by J. Hurrell at NCAR (personal communication). The data from these integrations are

used in the analysis of variability related to ENSO (sections 3b1 and 3b2) and the Northern Annu-

lar Mode (section 3b3).

3. Simulation characteristics

a. Model climatology

1) GENERAL CIRCULATION

Figure 1 shows the difference in annual- and zonal-mean temperature between the long-

term mean of the AMIP II integration of AM2/LM2 and a 50-year climatology from the National

Center for Environmental Prediction (NCEP) reanalysis (Kalnay et al. 1996). The model exhibits

a cold troposphere and warm stratosphere bias throughout the year. Typical errors in seasonal

GFDL GAMDT

21

meantemperaturesare2 K and4 K for troposphericandstratospherictemperatures,respectively.

Thelargestmodelbiasoccursat thehigh latitudesof theSouthernHemispherecoldbiasfrom 100

to 500hPa.This biasis commonto many climatemodels;however, themagnitudeof theerror in

AM2/LM2 is smallerthanin othermodels.Analysesfrom bothNCEPandtheEuropeanCentre

for MediumRangeWeatherForecasts(ECMWF, Gibsonetal. 1997)indicatezonalmean200hPa

December-January-February(DJF) temperaturesof about225-227K at 60˚S to 90˚S,whereas

AM2/LM2 givestemperaturesof 219 K for this region. In contrast,the medianAMIP II model

hastemperaturesof about211K (P. Gleckler, personalcommunication).Reasonsfor this reduced

cold bias of 200 hPa temperature are under investigation.

Thetroposphericcold biasevident in Figure1 extendsto the landsurface,ascanbeseen

in the annual-mean2 m air temperaturebias with respectto the CRU climatology (Fig. 2).

Although the annual-meanis fairly representative of the full seasonalcycle, warm biasesdo

appearin someseasonsin specificregions.Suchbiasesareapparent,for example,in borealwinter

over centraland northwesternNorth America,and in borealsummerover the southernUnited

States (Fig. 3). The latter bias corresponds to a mean temperature of 303 K or 30˚C.

Figure4 displaystheannualandzonal-meanzonalwindsfrom AM2/LM2, NCEPreanal-

ysis,andthedifference.As for temperature,thelargestwind errorsareconcentratedin thetop lev-

els of the model. Typical error amplitudesthroughoutthe seasonalcycle are 1-2 m s-1 in the

troposphereand5-10 m s-1 in the stratosphere.Biasesthat persistthroughoutthe seasonalcycle

includeawesterlybiasin thetropicalmiddletroposphereanda tendency for theextratropicaljets

to have 1-2 m s-1 errorsthatarewesterlynear40˚ latitudeandeasterlynear60-80˚latitude.These

dipolar error patternscorrespondto negative annular-modetype signaturesin eachhemisphere

GFDL GAMDT

22

(Thompson and Wallace 1998, Thompson and Wallace 2000).

Figure4 suggestsannularmodetypeerrorscontinueto surface.Figure5 displaysthelong-

termannual-andzonal-meanzonalwind stressover theoceanfor AM2/LM2 andthreeobserva-

tional baseddatasets(seecaptionfor details).Although the spreadis large amongthe observa-

tional datasets,robust biasesare apparent:the AM2/LM2 surface wind stressamplitude is

approximately30%too largein thesubtropicsandtheextratropicalpatterndisplaysdistinctly the

annular-mode type equatorward shift, particularly in the Northern Hemisphere.

Figure 6 illustrates the long-term meanNorthern HemisphereDJF sea level pressure

(SLP) for AM2/LM2, the NCEP reanalysis,and the difference.The equatorward shift of the

NorthernHemispheresurfacecirculationevident in the annual-meanwind stress(Fig. 5) corre-

spondsto abiastowardstrongerthanobservedSLPgradientsequatorwardof 60˚N,particularlyin

theNorth Atlantic. Otherbiasesincludea slightly strongerthanobservedIcelandicandAleutian

lows,anda high pressurebiasof 8 to 10 hPa over theEasternHemisphereArctic. This errorpat-

ternis accompaniedby anomalouseasterliesin northwestRussia,which appearthroughtempera-

ture advectionby the meanflow, to contribute to the enhancedcold biasin that region (Fig. 2).

This temperatureadvectionsignal,theArctic high-pressurebiasandthelow pressureerrorpattern

at lower latitudesarealso signaturesof annular-modetype anomalies(Thompsonand Wallace

2000;section3b4andFig. 17). Note that this biaspatternis commonto many models(Fig. 2 of

Walsh et al. 2002).

Figure7 displaysthedeparturefrom zonalmeanof the500hPaDJFgeopotentialheight,a

usefulmetric of the model’s ability to producea realisticplanetarywave pattern.Typical errors

areon theorderof 20 m, with themostprominenterrorsbeingananomalouslystrongridgecen-

GFDL GAMDT

23

teredover the North AmericanPacific coastand a weaker than observed negative-to-positive

dipoleover theHudson-Bay-to-North-Atlanticsector. Consistentwith thelattererror, the200hPa

zonalwind over theNorth Atlantic displaysa jet axisthathasinsufficient southwest-northeasttilt

(not shown).

2) PRECIPITATION, RADIATION, CLOUDS, AND WATER VAPOR

Figure 8 comparesthe annualmeanclimatological precipitationfor the model to the

observationalclimatologyof Xie andArkin (1997),alsoknown asCMAP. Althoughthecorrela-

tion coefficient is high, 0.9, the root meansquareerror, 0.85mm d-1, is about40%of thespatial

standarddeviationof thefield, 2 mmd-1. Themostprominenterrorsaredeficitsof precipitationto

thewestof theMaritime continent,in theSouthandtropicalAtlantic convergencezones,andin

the easternPacific ITCZ. Precipitationexcessesoccur in tropical Africa, the westernIndian

ocean,andthenorthwesttropicalPacific oceans.In theannualmean,therearefaint signaturesof

a doubleITCZ marked by excessive precipitationnear5˚S in the easternPacific and Atlantic.

Although this error is larger during the March-April-May season,we highlight it herebecause

coupled ocean-atmosphere models using AM2/LM2 exhibit a much more severe double ITCZ.

AM2/LM2 simulatestoomuchsummertimeprecipitationin Siberia,Alaska,andNorthern

Canadawith themodelproducingdoubletheCMAP precipitation.Thepositive biasin summer-

timehigh latitudeprecipitationis alsopresentin theannualmeanandis commonto many models

(Fig. 13 of Walshet al. 2002).However, on an annualmeanbasistheredoesnot appearto be a

biasin precipitationminusevaporation;outflows from riversfeedingtheArctic oceanarenotsys-

tematicallyoverestimated(notshown). Theglobalmeanprecipitationis ~0.15mmd-1 higherthan

the CMAP mean of 2.68 mm d-1 (Table 3).

GFDL GAMDT

24

Figures9 and10comparethelong-termannualmeanoutgoinglongwave radiation(OLR)

and net shortwave absorbed(SWAbs) from AM2/LM2 to the ERBE observations.Root mean

squareerrorsareabout8 W m-2 for OLR and13 W m-2 for SWAbs.Over thetropicaloceans,the

error patterns,particularly for OLR, resemblesthoseof the precipitationerrors,suggestingthat

improvementsin the simulationof precipitationwould be accompaniedby improvementsin the

radiationfields.An interestingexceptionto this is that OLR is overestimatedover tropical land

areaswherethereis not a systematicunderestimateof precipitation(e.g.tropicalAfrica). For the

shortwave radiation budget, the most prominenterror is the overestimationof SWAbs in the

coastalzonesof theeasternsubtropicaloceans.Althoughthemodelis effective in creatingstrato-

cumulusclouds further offshore, there is a severe deficit of coastalstratocumulus.This may

reflectthefact thatnot enoughcarehasbeentakenwith therepresentationof entrainmentacross

thestronginversionsat the top of theboundarylayer (Lock 2001).Away from thecoasts,in the

tradecumulusregionsof thesubtropics,thereis anoverestimationof thereflectedshortwave.This

may partly indicatethat the erosionconstantin the presenceof convectionis too small (section

2b3)and/orthattheuseof randomcloudoverlapassumptionis poorfor theseregions.Altogether

thepatternof “dim-stratocumulus-bright-trades”is endemicto atmosphericmodels(Siebesmaet

al. 2004).

In the extratropics,a prominentoverestimateof SWAbs of about10-20W m-2 occursat

nearlyall longitudesof thesouthernoceanatabout60˚S.Theerroroccursin theopenoceanareas

adjacentto theseaice margin andhasleadto anomalouslywarmSSTsin a coupledmodelbuilt

with AM2/LM2. Throughcomparisonto datafrom theInternationalSatelliteCloudClimatology

Project(ISCCP, Rossow andSchiffer 1999),this error appearsto be dueto an underestimateof

midlevel topped clouds.

GFDL GAMDT

25

Although themodel’s radiationhasbeentunedto resemblethe top of atmosphere(TOA)

radiationbudget,thesurfaceradiationbudgetis somewhatindependent.Both theestimatesof the

GlobalEnergy andWaterExperiment(GEWEX) (Stackhouseet al. 2004)andtheGoddardInsti-

tutefor SpaceStudies(GISS)(Zhanget al. 2003)indicatethattheshortwave absorbedat thesur-

faceis about5 W m-2 too low (Table3). Thelow biasin netshortwave absorbedresultsfrom the

excessshortwave cloud forcing, which is the differencebetweenclear sky and all-sky or total

shortwave fluxes.Fromearlierintegrationsof AM2/LM2 without thespecifiedthreedimensional

monthly climatologyof aerosols,the SWAbs at the SFCis reducedby 5 W m-2 while the long-

wave coolingof thesurfaceis reducedby lessthan1 W m-2 dueto thepresenceof aerosols.With

regard to the surface longwave budget, it appearsthat AM2/LM2 overestimatesthe longwave

cooling by about 10 W m-2, although under clear skies there is less bias.

With regard to the turbulent surfacefluxes,the modeloverestimatesthe Kiehl andTren-

berth(1997)estimateof evaporationby about5 W m-2 andunderestimatesthesensibleheatflux

by a similar amount.Note that the sumof the Kiehl andTrenberth(1997)turbulent heatfluxes,

102 W m-2, is lower thanthe eitherthe GEWEX or GISSestimatesof the surfacenet radiation,

about115W m-2, by 10 to 15 W m-2. Giventhesignificantremaininguncertaintiesin thesurface

energy budget,the biasesin the model’s global meanturbulent heatfluxesarenot well defined.

Indeedthe model’s valueslie within the rangeof observationalestimatesquotedin Table1 of

Kiehl and Trenberth (1997).

Figure 11 comparesAM2/LM2’ s annualmeantotal cloud amountto the satellitesesti-

matesof ISCCP. Thedatausedis theD2 adjustedmonthlymeantotalcloudamounts(Rossow and

Schiffer 1999). (A more thoroughcomparisonof AM2/LM2 clouds to ISCCP data using an

GFDL GAMDT

26

‘ISCCPSimulator’ (Klein andJakob 1999,Webbet al. 2001)will bereportedelsewhere).AM2/

LM2 doesnot produceenoughcloudsover oceansbetween20˚ and40˚ latitude,particularly in

thecoastalstratocumuluszone.Quantitatively, themodelhasaroot-meansquareerrorof 0.1rela-

tive to both ISCCP and the surface observer climatology of Warren et al. (1986, 1988) (not

shown). Theglobally averagedcloudcover of AM2/LM2 of 0.66lies in betweenthe ISCCPD2

valueof 0.69andthesurfaceobservers’valueof 0.62.Anothernoticeableproblemof themodelis

the excessive wintertime cloudinessin northernEurasiaand North America; surfaceobservers

indicateabout0.5 cloudcover in theseregionswhereasAM2/LM2 hascloudcover in excessof

0.8. Much of this differenceoccursin low cloudinesswherethe modelhasover 0.7 low cloudi-

nessbut the surfaceobservers report low cloudinessunder0.3 (not shown). Averagedover the

oceans,AM2/LM2’ s liquid waterpathof 77 g m-2 is comparableto the two satelliteestimates

(Table3) (Greenwaldetal. 1993;Wengetal. 1997);however, this is achievedby anexcessof liq-

uid waterpathovermidlatitudestormtracksandadeficitover tropicalandsubtropicaloceans(not

shown). Themodel’s simulationof ice waterpathcannotbeassesseddueto thelack of a reliable

observational product with global coverage.

At the top of atmosphere,the magnitudeof the global and annualaveragedshortwave

cloudforcing is overestimatedby about5 W m-2 but thelongwavecloudforcing is underestimated

by about5 W m-2 (Table3). Becausethetotal OLR hasbeentunedto matchERBEobservations,

theunderestimateof the longwave cloudforcing indicatesa similar significanterror in theclear-

sky OLR. Although the clear-sky samplingbiasmay contribute a few W m-2 to this difference

(HartmannandDoelling 1991),themodel’s clear-sky OLR is probablytoo low for two reasons.

First, the tropospherehasa cold biasrelative to re-analyses(Fig. 1). Second,asshown in Figure

12, themodelhasa moistbiasin theuppertropospherein comparisonto estimatesof uppertrop-

GFDL GAMDT

27

ospheric(~ 200-500hPa) relative humidities from the TIROS OperationalVertical Sounder

(TOVS) (SodenandBretherton1993).This moist bias is in excessof that due to the clear-sky

samplingbias of the observations.This moist bias deducedfrom satelliteobservationsis con-

firmed by both ECMWF andNCEPreanalyseswhich indicatea moist bias in both relative and

absolutehumidity in the middle tropical troposphere(not shown). The model’s column water

vapor(Table3), a measureprimarily of lower troposphericwatervapor, is slightly low, partially

reflecting the model’s cold bias.

b. Model variability

1) TROPICAL PRECIPITATION PATTERN ASSOCIATED WITH ENSO

The El Niño/SouthernOscillation (ENSO) is one of the most importantcontributors to

atmosphericvariability on interannualtimescales.TheENSOrelatedtropicalprecipitationanom-

aliesrepresenta redistributionof diabaticheatsourcesandsinksthatstronglyinfluencetheglobal

atmosphericcirculation.Theresponseof AM2/LM2 to theprescribedENSO-relatedSSTanoma-

liesaredepictedin Figure13,whichshows thedistributionof regressioncoefficientsof precipita-

tion rateon thestandardizedNINO3 index. TheNINO3 index is definedastheareallyaveraged

SSTanomalyin theregion5˚S-5˚N,150˚W-90˚W, andis acommonlyusedindicatorof theampli-

tudeandpolarityof ENSOevents.Thesecoefficientshavebeenmultipliedby onestandarddevia-

tion of the NINO3 index; thus they representtypical precipitationanomaliesthat accompany a

onestandarddeviation increasein theNINO3 index. Theupperpanelof Figure13 is basedon the

ensembleaverageof the10 AM2/LM2 runsfor theDJFseasonin the1951-2000period(section

2d),andthelowerpanelis computedusingGlobalPrecipitationClimatologyProject(GPCP)data

(Huffman et al. 1997) for 1979-2000.

GFDL GAMDT

28

The simulatedprecipitationsignalsduring ENSOaregenerallyin goodagreementwith

theobservations,asinferredfrom theGPCPdatasetandfrom stationmeasurements(Ropelewski

andHalpert1987).Both panelsof Figure13 indicatethatwarmENSOeventsareassociatedwith

positive precipitationanomaliesacrossmuch of the equatorialPacific, and negative anomalies

over equatorialSouthAmericaandthe neighboringAtlantic waters,aswell asthe northwestern

and southwesternsubtropicalPacific. One discrepancy betweenmodel and observation is seen

along the equatorover the easternIndonesianArchipelago.The GPCPpatternshows negative

rainfall anomaliesin thatregion duringwarmevents,whereasthemodelresultportraysnear-nor-

mal conditions.

2) EXTRATROPICAL TELECONNECTIONSTO ENSO

The impactof ENSO-relatedSSTanomalieson theextratropicalcirculationis illustrated

in Figure 14, which displaysthe regressioncoefficients of 200 hPa height on the standardized

NINO3 SSTindex for theDJFseason.ThesechartshavebeenconstructedusingNCEPreanalysis

data(lower panels)andtheensembleaverageof the10 AM2/LM2 integrations(upperpanels).In

analogywith Figure13, the regressionstatisticsin Figure14 portraythe typical 200 hPa height

anomaliesin responseto a one-standarddeviation SSTforcing from the tropicalPacific. Similar

chartshave beenpresentedby Horel andWallace(1981),amongmany others,to illustrate the

relationship between ENSO and the extratropical flow pattern.

The comparisonbetweenthe upperand lower panelsin this figure revealsconsiderable

spatialsimilaritiesbetweenthe simulatedandobserved wavetrainsemanatingfrom the NINO3

region to theeasternNorth Pacific/NorthAmericansectorandtheSouthernOceans.Theoverall

resemblancebetweenthe teleconnectionpatternsderived from the modeloutputandre-analysis

GFDL GAMDT

29

datais attributableto therealisticsimulationof the tropicalprecipitationforcing associatedwith

ENSO(Fig. 13).A noticeableerroris thatAM2/LM2 underestimatesthemagnitudeof themodel

anomaly over Canada.

Thecapabilityof themodelto reproducetheNorthernHemispherecirculationanomalies

observed in individual El Niño andLa Niña eventsis now examined.For eachof theprominent

warmandcold eventsin the1951-2000period(e.g.,seelisting in Trenberth1997),theanomaly

patternsof DJF200hPaheightwerecomputedusingtheNCEPreanalysisandtheensemblemean

of the 10 model integrations.The spatialcorrelationcoefficient, root meansquare(rms) differ-

ence,andratiobetweenthespatialvariancesof themodelandobservedfieldsin theNorthPacific/

North Americansector(20˚-70˚N,60˚-180˚W)aredisplayedusinga ‘Taylor’ diagram(Gateset

al. 1999andTaylor 2001) in Figure15. In this diagram,eachevent is indicatedby a dot anda

labelcorrespondingto thelasttwo digitsof theyear;for instance,thestatisticsfor the1982/83El

Niño eventareindicatedusingthelabel ‘82’. Thespatialcorrelationcoefficient betweenthesim-

ulatedandobservedanomaliesis at the0.5 level or greaterfor four (1969,1982,1991and1997)

outof theeightwarmevents,andall sevencoldevents.While thespatialvarianceof theensemble

meanmodelpatternis noticeablylower thanthatof theobservations,inspectionof theTaylordia-

gramfor thetenindividual membersof theensemble(not shown) revealsthatthespatialvariance

of thesemembersis in betteragreementwith theobservations.TheTaylor diagramfor individual

samplesfurther illustratesthat,for thoseeventswith high spatialcorrelationbetweentheensem-

ble-meanand observations(e.g. 1982and1997), the agreementbetweenmany model samples

and the observations is also high.

GFDL GAMDT

30

3) ENSO- MONSOONRELATIONSHIPS

Ample observational and model evidenceexists for the impact of ENSO on the Asian-

Australianmonsoons;e.g., seethe brief review of pertinentstudiesby Lau and Nath (2000).

Warm ENSOepisodesaregenerallyaccompaniedby below normalprecipitationduring the wet

summermonsoonsover the Indian subcontinent(IND) andnorthernAustralia(AUS). Addition-

ally, the dry winter monsoonover southeastAsia (SEA) weakensin El Niño eventsresultingin

above averagerainfall amounts.The polarity of theseanomaliestendsto reverseduring cold

events.

The simulationof theseENSO-monsoonrelationshipsby AM2/LM2 hasbeenevaluated

by examiningthemodel’s 10-memberensemblemeanprecipitationanomaliesin theabove-men-

tionedregionsfor eachmonsoonseasonin the 1951-2000period.The relationshipbetweenthe

model’sprecipitationanomaliesin thesemonsoonregionsandtheNINO3 SSTanomaliesis illus-

tratedin the upperpanelsof Figure16. The simulatedprecipitationanomaliesin IND andAUS

during the local summerseasonarenegatively correlatedwith NINO3 SSTanomalies,and the

wintertimerainfall in SEA exhibits a positive correlationwith the ENSOforcing. Many of the

outstandingENSO episodes(coloreddots and squares)are accompaniedby notablesimulated

rainfall perturbationssimulatedin the regions considered.The correlationcoefficients between

monsoonprecipitationamountsin theAM2/LM2 modelrunsandtheNINO3 index, asindicated

in theupperright cornerof individual panels,maybecomparedwith thosededucedfrom GPCP

observationalestimates(lowerpanelsof Fig. 16).Thenoticeablyweakercorrelationsbetweenthe

observed Indian rainfall and the NINO3 index (lower left panel) reflect the much diminished

Indian monsoon- ENSOrelationshipsduring the recentdecadescoveredby the GPCPdataset

(Kumaret al. 1999).Thecorrelationcoefficientsbetweentheobservedrainfall anomaliesandthe

GFDL GAMDT

31

NINO3 index for theoutstandingENSOepisodes(shown above the lower panelswithout paren-

thesis) are based on only five events, and are hence subject to considerable sampling fluctuations.

4) NORTHERN ANNULAR MODE

Apart from ENSO,the dominantpatternof interannualclimate variability is associated

with theannularmodesof theextratropicalatmosphericcirculationfield. Shown in Figure17 are

distributionsof sealevel pressure(SLP) andsurfacetemperatureanomaliesassociatedwith the

NorthernAnnularMode(NAM, alsoreferredto astheArctic Oscillation;seeThompsonandWal-

lace(1998,2000))for theobservationsandAM2/LM2. TheNAM is definedasthefirst empirical

orthogonalfunction(EOF)of sealevel pressureover thedomainfrom 20˚Nto 90˚N.Thecontours

indicatethe sealevel pressurechangesassociatedwith a 1 hPa increaseof a NAM index. The

NAM index is definedasthedifferencein sealevel pressurebetweentheArctic andmidlatitude

extremaof theEOFpattern,multipliedby theEOFtimeseries,therebygiving anindex with units

of hPa.

Themodelhasa highly realisticsimulationof thespatialpatternof theNAM. Thecolor

shadingindicatesthenear-surfaceair temperatureanomaliesassociatedwith a 1 hPa increasein

theNAM index. TheAM2/LM2 NAM patternshown is themeanof NAM patternscomputedsep-

aratelyfor eachof the10 ensemblemembersintegratedwith observedSSTsfrom 1951to 2000.

Examinationof theNAM patternfrom eachof the10 membersof theensemblerevealsrelatively

small intra-ensemblevariationsin the spatialpatternof the NAM. Consistentwith the observa-

tions,a positive phaseof thesimulatedNAM is associatedwith a quadrupolefield of temperature

anomalies:warmanomaliesover southeasternNorth AmericaandnorthernEurasiaandnegative

anomaliesover northeasternNorth AmericaandnorthernAfrica throughthe Middle East.The

GFDL GAMDT

32

primary discrepancy betweenthe simulatedand observed temperatureanomaliesoccursover

northwesternNorth America, with larger negative temperatureanomaliesin AM2/LM2 than

observed. This is consistent with Northeast Pacific SLP gradients that are stronger than observed.

5) TROPICAL TRANSIENT ACTIVITY

Transientactivity in the tropics is evaluatedby examinationof 2 phenomena:tropical

cyclones and the Madden-Julian Oscillation (MJO, Madden and Julian 1972).

Tropicalcyclonesin AM2/LM2 aredetectedusingthealgorithmof Vitart etal. (1997)and

comparedto theNationalClimateDataCenter’s global tropicalcyclonedataset(Neumannet al.

1999). Figure 18 displays genesislocation frequenciesfor the years 1979-1995.AM2/LM2

underestimatesthenumberof stormsquitesignificantly, particularlyin theNorthAtlantic andthe

EasternPacific, whereno stormsoccur. The seasonalcycle in the NorthernHemisphere(not

shown) is alsoquitepoor, with themodellaggingobservationsby severalmonths.Overall AM2/

LM2’s simulationof tropical cyclonesis inferior to that of someothermodels(Bengtssonet al.

1995; Vitart and Stockdale 2001).

An assessmentof theMJOis madeby examiningthestructureandbehaviour of intrase-

asonalvariability (ISV), definedhereasvariability with timescalesbetween30 and90 days.Fig-

ure19 displaysthewave frequency spectra,with theannualcycle removed,for deviationsof the

200hPa zonalwind from its zonalmean.Pentaddatafrom JanuarythroughDecember, averaged

between5˚Sand5˚N, from 1979through1995wereusedto computespectrafor eachyear. These

spectrawerethenaveragedover the17yearsandsmoothedfurtherby theapplicationof a3-point

Hanningwindow. A broadpeakin the intraseasonalrangeis evident in thespectrafor theNCEP

GFDL GAMDT

33

reanalyses,with the observed maximaprimarily in the 40 to 60 day range.AM2/LM2 shows

weaker peaksin thevicinity of 35 and50 dayswith additionalpower at periodsaround90 days,

implying a somewhat slower propagation speed.In addition,a strongerpreferencefor eastward

propagation is seenin the NCEPreanalysesthanin AM2/LM2. MJO structureandpropagation

characteristicsmay be studied by applying extendedempirical orthogonal function analysis

(EEOF)to precipitationin theregion 30˚S- 30˚N and30˚E- 90˚W. To focuson MJO timescales

thedatais band-passed(30 to 90 day)andtheEEOFanalysisis performedusinglagsfrom -7 to

+7 pentads(-35 to + 35 days).CompositeMJO life cycles are obtainedby using peaksin the

EEOFmode-1time seriesto identify centersof events.Theneachevent is takento be-7 pentads

to +7 pentadsrelative to thesemid-pointsandall thusidentifiedeventsareaveragedtogether. Fig-

ure 20 shows a comparisonof MJO compositelife cycles,during Novemberto April, from the

AM2/LM2 on the left and the CMAP observed precipitationon the right. The monthsfrom

Novemberto April wereselectedastheMJO is mostactive then.Eachpanelin thefigurerepre-

sentsanaverageof 3 adjacenttime lags.TheCMAP observationsdisplaycoherentintraseasonal

activity in the centralandeasternIndian Oceanwhich propagateseastward acrossthe maritime

continentinto thewesternPacific. (Thegreendashedlines in Figure20 indicatethepropagation

of theanomalies)AM2/LM2 showsweaker, lesscoherentactivity with perhapssomeslowereast-

wardpropagationfrom themaritimecontinentinto thewesternPacific. (Notethat theAM2/LM2

anomalieshave beenmultiplied by 2 for displayin Figure20) AM2/LM2 is particularlydeficient

in the Indian Oceansouthof the Bay of Bengal whencomparedto CMAP. Waliseret al. 2003,

indicatethat this is a commondeficiency of large-scalemodels.Overall, AM2/LM2’ s simulation

of the MJO is fairly poor.

GFDL GAMDT

34

4. Comparison of AM2/LM2 climatology to other models

It is of generalinterestto comparethecapabilityof AM2/LM2 to reproduceobservedcli-

matewith thatof othermodels.To do so,Taylor diagrams(seelegendof Figure15 for a detailed

explanationof thesediagrams)havebeencalculatedfor eightvariablesusingAM2/LM2, two pre-

vious GFDL models,andfour non-GFDLmodels(Fig. 21). The first row of Figure21 displays

variablesassociatedwith surfaceclimate,includingborealwinterocean-onlySLP, borealsummer

NorthernHemisphereland-onlysurfaceair temperatures,andannualmeanocean-onlyzonalwind

stress.Thesecondrow displaysvariablesrelatedto hydrology:annualmeantropicalprecipitation,

shortwave cloudforcing,andtotal cloudamount.Thelastrow displaysvariablesrelatedto upper

troposphericcirculation:theborealwinter200hPaeddygeopotentialin theNorthernHemisphere

and the 200 hPa zonal wind.

ThepreviousGFDL modelsincludetheGFDL climatemodelrecodedinto FMS software

which is known locally as the ManabeClimate Model (MCM) (Delworth et al. 2002) and the

modeldevelopedby theGFDL’s formerexperimentalpredictiongroup(DERF)(SternandMiya-

koda1995).Thedatafrom modelsoutsideof GFDL wereacquiredfrom thearchivemaintainedat

thePCMDI andrepresentstheir official submissionto AMIP II. Theoutsidemodelsincludethe

CCM3.5of the NationalCenterfor AtmosphericResearch,ECHAM4 of the Max PlanckInsti-

tute, the ECMWF model CY18R5, and HadAM3 from the United Kingdom’s Meteorological

Office. Theexperimentaldataproducedby thenon-GFDLmodelsweresubmittedto PCMDI in

either1998or 1999(seehttp://www-pcmdi.llnl.gov/amip/STATUS/incoming.htmlfor documen-

tation).

GFDL GAMDT

35

Broadly speaking,Figure 21 indicatesthat AM2/LM2 producesa model climate better

thanthoseof the previous GFDL models.The quality of AM2/LM2’ s climate is comparableto

thatproducedby thenon-GFDLmodels.In somevariables(SLP, wind stress,200hPa circulation

and precipitation),the AM2/LM2 model is at the front rank, but for shortwave cloud forcing

AM2/LM2 is slightly worse.It is importantto statethreecaveatsof this modelcomparison:these

Taylor diagramscompareonly modelclimatologieswith no resultsshown for differentaspectsof

modelvariability; the performanceof non-GFDLmodelsmay have improved in the yearssince

theirsubmissionof datato AMIP II; andtheTaylordiagramsarebasedon largescalepatternsand

do not assess important regional biases.

5. Future work

A new global atmosphereandland modelAM2/LM2 developedat GFDL hasbeenpre-

sentedand the model evaluatedusing simulationsin which the model is forced with observed

SSTsandseaice. In this final section,the suitability of AM2/LM2 for couplingwith an ocean

model and future plans for global atmosphere and land modeling at GFDL are discussed.

An importantgoal for this work is to coupleAM2/LM2 to an oceanmodelwithout flux

adjustments.This hasbeenaccomplishedand will be reportedelsewhere.Here,a preliminary

indicationof theability of AM2/LM2 to couplewith anoceanmodelis givenby estimatesof the

implied poleward oceanicheattransportfor the Atlantic, Indo-Pacific, and world oceanbasins

(Fig. 22). For comparison,observation basedestimatesof oceanicheattransportderived from

atmosphericdata(TrenberthandCaron2001)andoceanicdata(GanachaudandWunsch2003)

arealsoshown. AM2/LM2’ s implied oceanicheattransportis in reasonableagreementwith the

GFDL GAMDT

36

observedestimatesin theAtlantic basin,althoughit is typically nearthelow endof theconfidence

intervalsreportedfor theobservedestimates,andin a few cases,suchastheGanachaudandWun-

schestimatesfor 24˚N and19˚S,themodellies outsidetheconfidenceintervalson the low side.

In theNorth Atlantic, AM2/LM2’ s simulationof about1 petawatt (1015 W) at 10˚N-30˚Nrepre-

sentsa significant improvementover that implied by the atmosphericcomponentof the older

GFDL R30 climate model (Delworth et al. 2002),which had too small implied poleward heat

transport(0.7petawattsat 15˚N,not shown). In theIndo-Pacific basin,themodel’s implied oce-

anic heattransporthasa positive bias relative to both setsof observed estimates,and exceeds

TrenberthandCaron’s 1 standarderror limit over all latitudes.This positive bias for the Indo-

Pacific is reflectedin a similar but lesspronouncedbiasfor theworld oceanbasin.Theseresults

indicatethat in theNorth Pacific AM2/LM2 removesheatfrom theoceanat a greaterratethanis

supportedby eithersetof observations.This differencefrom observationslikely contributesto a

largecold biasin theNorth Pacific whenAM2/LM2 is coupledto anoceanmodel(detailsto be

reported elsewhere).

Thedevelopmentof thenext versionof theatmosphericmodel,AM3, is well underway

with a numberof changesbeingexploredandevaluatedthroughthemodeldevelopmentprocess.

These include:

• Replacementof the B-grid dynamicalcorewith a finite volumedynamicalcore(Lin

2004).

• Replacementof RASwith aconvectionschemewhich includesrepresentationsof ver-

tical velocitiesandmicrophysicsin cumulusupdraftsanddowndrafts,andparameter-

ized mesoscale circulations (Donner et al. 2001).

GFDL GAMDT

37

• Replacementof randomcloudoverlapwith by a resolutioninvariantoverlapscheme.

This will be accomplishedby a stochastictreatmentof clouds(Pincuset al. 2003).

Also underconsiderationis the replacementof the prognosticcloud fraction scheme

(Tiedtke 1993)with a statisticalcloudschemewith prognostichigherordermoments

similar to Tompkins (2002).

• Addition of morevertical levels at the top of the modelto bettersimulatethe strato-

sphereand its couplingwith the troposphere.Considerationis beinggiven to a new

anisotropicorographicgravity wave scheme(Garner2003)andto a convectively gen-

erated gravity wave scheme (Alexander and Dunkerton 1999).

• Addition of prognosticchemistryandaerosolmodulesbasedon thechemistryscheme

developed for use in the MOZART-2 chemical transport model (Horowitz et al. 2003).

• Replacementof LM2 with anew dynamiclandsurfacemodelwith carbonandvegeta-

tion dynamics.This new landmodel,LM3, includesthevariousprocessesthatdeter-

mine the amountof carbonstoredin the soil and the vegetation.Theseprocesses

includechangesin CO2 concentrationsandotherenvironmentalfactors,naturaldistur-

bances(e.g.fire), andanthropogeniclanduse(e.g.deforestationandagriculturalcrop-

land abandonment).

Acknowledgments. FormerGFDL directorJerryMahlmanis thanked for his encourage-

mentandsupportof the Flexible Modeling SystemandcurrentGFDL directorAnts Leetmaais

thankedfor his charteringof theGlobalAtmosphericModel DevelopmentTeam.Reviews of the

manuscriptby Olivier Pauluis, Mike Winton and two anonymous reviewers are appreciated.

Assistanceregardingsurfaceradiationbudgetdataprovidedby Paul StackhouseandYuanchong

Zhang is appreciated.

GFDL GAMDT

38

References

Alexander, M.J. andT.J.Dunkerton,1999:A spectralparameterizationof mean-flow forcing due

to breaking gravity waves.J. Atmos. Sci., 56, 4167-4182.

Arakawa, A. andV. R. Lamb,1977:Computationaldesignof the basicdynamicalprocessesof

the UCLA generalcirculation model. Methods in Computational Physics, Vol. 17, J.

Chang, Ed., Academic Press, 173-265.

Barkstrom,B. R., E. Harrison,G. Smith, R. Green,J. Kibler, R. Cess,and the ERBE Science

Team,1989:EarthRadiationBudgetExperiment(ERBE)archival andApril 1985results.

Bull. Amer. Met. Soc., 70, 1254-1262.

Beljaars,A. C. M., 1995:Theparameterizationof surface-fluxesin large-scalemodelsunderfree

convection.Quart. J. Roy. Meteor. Soc., 121, 255-270.

Beljaars,A. C. M., 1998:Roleof theboundarylayerin a numericalweatherpredictionmodel,in

ClearandCloudyBoundaryLayers,A. A. M. Holtslag,P. G. Duynkerke,andP. J.Jonker,

Eds., Royal Netherlands Academy of Arts and Sciences, 287-304.

Bengtsson,L., M. Botzet and M. Esch,1995: Hurricane-typevorticesin a generalcirculation

model.Tellus, 47A,175-196.

Budyko, M. I., 1974: Climate and Life. Academic, 508 pp.

Cahalan,R. F., W. Ridgway, W. J. Wiscombe,T. L. Bell, andJ. B. Snider, 1994:The albedoof

fractal stratocumulus clouds.J. Atmos. Sci., 51, 2434-2460.

GFDL GAMDT

39

CERSAT-IFREMER,2002:MeanWind Fields(MWF product),Usermanual.Volume1: ERS-1,

ERS-2,andNSCAT. C2-MUT-W-05-IF. Brest,France,72 pp. (availableat ftp://ftp.ifre-

mer.fr/ifremer/cersat/documentation/gridded/mwf-ers/mwf_vol1.pdf)

Clough, S. A., M. J. Iaconoand J-L. Moncet, 1992: Line-by-line calculationsof atmospheric

fluxes and cooling rates: Application to water vapor. J. Geophys. Res., 97, 15761-15785

Colella,P. andP. R. Woodward,1984:Thepiecewiseparabolicmethod(PPM)for gas-dynamical

simulations.J. Comput. Phys., 54, 174-201.

Darnell,W. L., W. F. Staylor, S.K. Gupta,andF. M. Denn,1988:Estimationof surfaceinsolation

using sun-synchronous satellite data.J. Clim., 1, 820-836.

A. daSilva, A. C. Young,and S. Levitus, 1994.Atlas of surfacemarinedata1994.Volume 1:

Algorithmsandprocedures.TechnicalReport#6,U.S.Departmentof Commerce,NOAA,

NESDIS, 1994.

Delworth, T. L., R. J. Stouffer, K. W. Dixon, M. J. Spelman,T. R. Knutson,A. J. Broccoli, P. J.

Kushner, and R. T. Wetherald,2002: Review of simulationsof climate variability and

change with the GFDL R30 coupled climate model.Clim. Dyn., 19, 555-574.

Derbyshire,S. H., 1999:Boundary-layerdecouplingover cold surfacesasa physical boundary

instability. Bound.-Layer Meteor., 90, 297-325.

Donner, L. J., C. J. Seman,B. J. Soden,R. S. Hemler, J. C. Warren,J. Strom,andK.-N. Liou,

1997:Large-scaleice cloudsin theGFDL SKYHI generalcirculationmodel.J. Geophys.

Res., 102, 21745-21768.

GFDL GAMDT

40

Donner, L. J.,C. J.Seman,R. S.Hemler, andS.Fan,2001:A cumulusparameterizationincluding

massfluxes, convective vertical velocities,and mesoscaleeffects: thermodynamicand

hydrological aspects in a general circulation model.J. Clim., 14, 3444-3463.

Edwards,J.M. andandA. Slingo,1996:Studieswith aflexible new radiationcode.I: Choosinga

configuration for a large-scale model.Quart. J. Roy. Meteor. Soc., 122, 689-719.

Fortuin, P. andH. Kelder, 1998:An ozoneclimatologybasedon ozonesondeandsatellitemeas-

urements.J. Geophys. Res., 103, 31709-31734.

Freidenreich,S. M., andV. Ramaswamy, 1999:A new multiple-bandsolarradiative parameteri-

zation for general circulation models.J. Geophys. Res., 104, 31389-31409.

Fu, Q. and K.N. Liou, 1993: Parameterizationof the radiative propertiesof cirrus clouds.J.

Atmos. Sci., 50, 2008-2025.

Gadd,A. J.,1978:A split explicit integrationschemefor numericalweatherprediction.Quart. J.

Roy. Meteor. Soc., 104, 569-582.

Ganachaud,A. andC. Wunsch,2003:Large-scaleoceanheatandfreshwatertransportsduringthe

World Ocean Circulation Experiment.J. Clim., 16, 696-705.

Garner, S. T., 2003:A topographicdragclosurewith analyticalbaseflux. Preprintsto the 14th

annualconferenceon AtmosphericandOceanicFluid Dynamics,SanAntonio,American

Meteorological Society, 120-124.

Gates,W.L., andCoauthors,1999:An overview of the resultsof the AtmosphericModel Inter-

GFDL GAMDT

41

comparison Project (AMIP I). Bull. Amer. Met. Soc., 80, 29-55.

Gibson, J. K., P. Kallberg, S. Uppala, A. Hernandez, A. Nomura, and E. Serrano, 1997: ERA

description. ECMWF Reanalysis Project Rep. Series 1, European Centre for Medium-

Range Weather Forecasts, Reading, United Kingdom, 66 pp.

Gregory, D., R. Kershaw, and P. M. Innes, 1997: Parameterization of momentum transport by

convection. II: Tests in single-column and general circulation models. Quart. J. Roy.

Meteor. Soc., 123, 1153-1183.

Grenier, H. and C. S. Bretherton, 2001: A moist PBL parameterization for large-scale models and

its application to subtropical cloud-topped marine boundary layers. Mon. Wea. Rev., 129,

357-377.

Greenwald, T. J., G. L. Stephens, and T. H. Vonder Haar, 1993: A physical retrieval of cloud liq-

uid water over the global oceans using Special Sensor Microwave/Imager (SSM/I) obser-

vations. J. Geophys. Res., 98, 18,471-18,488.

Gupta, S. K., W. L. Darnell, and A. C. Wilber, 1992: A parameterization for longwave surface

radiation from satellite data: Recent improvements. J. Appl. Met., 31, 1361-1367.

Gupta, S. K., N. A. Ritchey, A. C. Wilber, C. H. Whitlock, G. G. Gibson, and P. W. Stackhouse

Jr., 1999: A climatology of surface radiation budget derived from satellite data. J. Clim.,

12, 2691-2710.

Hamilton, K. P., R. J. Wilson, J. D. Mahlman, and L. Umscheid, 1995: Climatology of the SKYHI

troposphere-stratosphere-mesosphere general circulation model. J. Atmos. Sci., 52, 5-43.

GFDL GAMDT

42

Harrison, E. F., P. Minnis, B. R. Barkstrom, V. Ramanathan, R. C. Cess, and G. G. Gibson, 1990:

Seasonal variation of cloud radiative forcing derived from the Earth Radiation Budget

Experiment. J. Geophys. Res., 95, 18,687-18,703.

Hartmann, D. L. and D. Doelling, 1991: On the net radiative effect of clouds. J. Geophys. Res.,

96, 869-891.

Haywood, J. M., V. Ramaswamy, and B. J. Soden, 1999: Tropospheric aerosol climate forcing in

clear-sky satellite observations over the oceans. Science, Science, 283, 1299-1303.

Held, I., R. S. Hemler and V. Ramaswamy, 1993: Radiative-convective equilibrium with explicit

two-dimensional moist convection. J. Atmos. Sci., 50, 3909-3927.

Hogan, R. J. and A. J. Illingworth, 2000: Deriving cloud overlap statistics from radar. Quart. J.

Roy. Meteor. Soc., 126, 2903-2909.

Horel, J. D. and J. M. Wallace, 1981: Planetary-scale atmospheric phenomena associated with the

Southern Oscillation. Mon. Wea. Rev., 109, 813-829.

Holtslag, A. A. M., 2003: GABLS initiates intercomparison for stable boundary layer case.

GEWEX News, 13 (2), 7-8.

Horowitz, L. W., S. Walters, D. L. Mauzerall, L. K. Emmons, P. J. Rasch, C. Granier, X. X. Tie,

J.-F. Lamarque, M. G. Schultz, G. S. Tyndall, J. J. Orlando, and G. P. Brasseur, 2003: A

global simulation of tropospheric ozone and related tracers: Description and evaluation of

MOZART, version 2. J. Geophys. Res., 108, 4784, doi:10.1029/2002JD002853.

GFDL GAMDT

43

Huffman,G. J.,R. F. Adler, P. Arkin, A. Chang,R. Ferraro,A. Gruber, J.Janowiak, A. McNab,B.

Rudolf, andU. Schneider, 1997:The Global PrecipitationClimatologyProject(GPCP)

combined precipitation dataset.Bull. Amer. Met. Soc., 78, 5-20.

Jakob,C. andS.A. Klein, 2000:A parameterizationof theeffectsof cloudandprecipitationover-

lap for use in general circulation models.Quart. J. Roy. Meteor. Soc., 126, 2525-2544.

Jones,P. D., 1994:Hemisphericsurfaceair temperaturevariations:a reanalysisandanupdateto

1993.J. Clim., 7, 1794-1802.

Kalnay, E., andCoauthors,1996:TheNCEP/NCAR40-yearReanalysisProject.Bull. Amer. Met.

Soc., 77, 437-471.

Kershaw, R., A. L. M. Grant,S.H. Derbyshire,andS.Cusack,2000:Thenumericalstability of a

parameterizationof convective momentumtransport.Quart. J. Roy. Meteor. Soc., 126,

2981-2984.

Kiehl, J. T. andK. E. Trenberth,1997:Earth’s annualglobal meanenergy budget.Bull. Amer.

Met. Soc., 78, 197-208.

Klein, S. A., andC. Jakob, 1999:Validationandsensitivities of frontal cloudssimulatedby the

ECMWF model.Mon. Wea. Rev., 127, 2514-2531.

Kumar, K. K., B. Rajagopalan,andM. A. Cane,1999:On the weakeningrelationshipbetween

Indian monsoon and ENSO.Science, 284, 2156-2159.

Kurihara,Y. andG. J. Tripoli, 1976:An iterative time integrationschemedesignedto preserve a

GFDL GAMDT

44

low-frequency wave.Mon. Wea. Rev., 104, 761-764.

Lau,N.-C.,andM.J.Nath,2000:Impactof ENSOonthevariability of theAsian-Australianmon-

soons as simulated in GCM experiments.J. Clim., 13, 4287-4309.

Lin, S.-J.,W. C. Chao,Y. C. Sud,andG. K. Walker, 1994:A classof thevanLeer-typetransport

schemesandits applicationto themoisturetransportin a generalcirculationmodel.Mon.

Wea. Rev., 122, 1575-1593.

Lin, S.-J.,1997: A finite-volume integration methodfor computingpressure-gradientforce in

general vertical coordinates.Quart. J. Roy. Meteor. Soc., 123, 1749-1762.

Lin, S.-J.,2004:A “vertically Lagrangian”finite-volumedynamicalcorefor globalmodels.sub-

mitted toMon. Wea. Rev.

Lindberg, C. andA. J. Broccoli, 1996:Representationof topography in spectralclimatemodels

and its effect on simulated precipitation.J. Clim., 9, 2641-2659.

Lock, A. P., A. R. Brown, M. R. Bush,G. M. Martin, andR. N. B. Smith,2000:A new boundary

layer mixing scheme.Part I: Schemedescriptionand single-columnmodel tests.Mon.

Wea. Rev., 128, 3187-3199.

Lock, A. P., 2001:Thenumericalrepresentationof entrainmentin parameterizaionsof boundary

layer turbulent mixing.Mon. Wea. Rev., 129, 1148-1163.

Lord, S. J. andA. Arakawa, 1980:Interactionof a cumuluscloudensemblewith the large-scale

environment. Part II. J. Atmos. Sci., 37, 2677-2692.

GFDL GAMDT

45

Louis,J.-F., 1979:A parametericmodelof verticaleddyfluxesin theatmosphere.Bound.-Layer

Meteor., 17, 187-202.

Madden,R. A. andP. R. Julian,1972:Descriptionof global-scalecirculationcells in the tropics

with a 40-50 day period.J. Atmos. Sci., 29, 1109-1123.

Mapes,B. E. andX. Wu, 2001:Convective eddymomentumtendenciesin long cloud-resolving

model simulations.J. Atmos. Sci., 58, 517-526.

Matthews, E., 1983:Global vegetationandland use:new high-resolutiondatabasesfor climate

studies.J. Appl. Met., 22, 474-487.

Mellor, G.L. andT. Yamada,1982:Developmentof a turbulent closuremodel for geophysical

fluid problems.Rev. Geophys. Space Phys., 20, 851-875.

Mesinger, F., 1977: Forward-backward scheme,and its use in a limited areamodel. Contrib.

Atmos. Phys., 50, 186-199.

Mesinger, F., Z. I. Janjic,S. Nickovic, D. Gavrilov andD. G. Deaven,1988:Thestep-mountain

coordinate:Model descriptionandperformancefor casesof Alpine leecyclogenesisand

for a case of an Appalachian redevelopment.Mon. Wea. Rev., 116, 1493-1518.

Milly , P. C. D., andA. B. Shmakin,2002:Global modelingof land waterandenergy balances.

Part I: The land dynamics (LaD) model.J. Hydrometeor., 3, 283-299.

Moorthi, S.andM. J.Suarez,1992:RelaxedArakawa-Schubert:aparameterizationof moistcon-

vection for general circulation models.Mon. Wea. Rev., 120, 978-1002.

GFDL GAMDT

46

Moorthi, S. andM. J. Suarez,1999:Documentationof version2 of Relaxed Arakawa-Schubert

cumulusparameterizationwith convective downdrafts.NOAA TechnicalReportNWS/

NCEP 99-01, U.S. Dept. of Commerce, Camp Springs, MD.

Neumann,C. J., B. R. Jarvinen,C. J. McAdie andJ. D. Elms, 1999:Tropical cyclonesof the

North Atlantic ocean,1871-1998. NCDC Historical Climatology Series,6-2, 190 pp.

(downloaded from http://dss.ucar.edu/datasets/ds824.1/)

New, M., M. Hulme,andP. Jones,1999:Representingtwentieth-centuryspace-timeclimatevari-

ability. Part I: Developmentof a 1961-90meanmonthly terrestrialclimatology. J. Clim.,

12, 829-856.

Pierrehumbert,R. T., 1986.An essayon the parameterizationof orographicgravity wave drag.

Proceedings from ECMWF 1986 Seminar, Vol. I, 251-282.

Pincus,R. andS. A. Klein, 2000:Unresolvedspatialvariability andmicrophysicalprocessrates

in large-scale models.J. Geophys. Res., 105, 27059-27065.

Pincus,R., H. W. Barker, andJ.-J.Morcrette,2003:A fast,flexible, approximatetechniquefor

computingradiative transferin inhomogeneousclouds.J. Geophys. Res., 108, 4376-4380,

doi:10.1029/2002JD003322.

Ramachandran,S.,V. Ramaswamy, G. L. Stenchikov, andA. Robock,2000:Radiative impactof

the Mount Pinatubovolcanic eruption:Lower stratosphericresponse.J. Geophys. Res.,

105, 24409-24429.

Ramaswamy, V., andM. M. Bowen,1994:Effectof changesin radiatively activespeciesuponthe

GFDL GAMDT

47

lower stratospheric temperatures.J. Geophys. Res., 99, 18909-18921.

Ramaswamy, V., and M. D. Schwarzkopf, 2002: Effects of ozoneand well-mixed gaseson

annual-meanstratospherictemperaturetrends.Geophys. Res. Let., 29, 2064, 10.1029/

2002GL015141.

Randel,D. L. andco-authors,1996:A new globalwatervapordataset.Bull. Amer. Met. Soc., 77,

1233-1246.

Ropelewski, C. F. andM. S.Halpert,1987:Globalandregionalscaleprecipitationpatternsasso-

ciated with the El Niño/Southern Oscillation.Mon. Wea. Rev., 115, 1606-1626.

Rossow, W. B. andR. A. Schiffer, 1999:Advancesin UnderstandingCloudsfrom ISCCP. Bull.

Amer. Met. Soc., 80, 2261-2288.

Rothman,L.S. and30 co-authors,2003:TheHITRAN molecularspectroscopicdatabase:edition

of 2000 including updates through 2001.J. Quant. Spectr. Rad. Transfer, 82, 5-44.

Rotstayn,L. D., 1997:A physically basedschemefor thetreatmentof stratiformcloudsandpre-

cipitationin large-scalemodels.I: Descriptionandevaluationof microphysicalprocesses.

Quart. J. Roy. Meteor. Soc., 123, 1227-1282.

Rotstayn,L. D., 2000:On the“tuning” of autoconversionparameterizationsin climatemodels.J.

Geophys. Res., 105, 15495-15507.

Rotstayn,L. D., B. F. Ryan,andJ.Katzfey, 2000:A schemefor calculationof theliquid fraction

in mixed-phase clouds in large-scale models.Mon. Wea. Rev., 128, 1070-1088.

GFDL GAMDT

48

Schwarzkopf, M. D. andV. Ramaswamy, 1999:Radiative effectsof CH4, N2O, halocarbonsand

theforeign-broadenedH2O continuum:A GCM experiment.J. Geophys. Res., 104, 9467-

9488.

Shapiro,R., 1970:Smoothing,filtering andboundaryeffects.Rev. Geophys. Space Phys., 8, 359-

387.

Siebesma,A. P., C. S.Bretherton,A. Brown, A. Chlond,J.Cuxart,P. G. Duynkerke,H. Jiang,M.

Khairoutdinov, D. Lewellen, C.-H. Moeng,E. Sanchez,B. Stevens,and D. E. Stevens,

2003:A large-eddysimulationintercomparisonstudyof shallow cumulusconvection.J.

Atmos. Sci., 60, 1201-1219.

Siebesma,A. P., C. Jakob,G. Lenderink,R. Neggers,J.Teixeira,J.Calvo, A. Chlond,H. Grenier,

C. Jones,M. Koehler, H. Kitagawa, P. Marquet,A. Lock, F. Mueller, D. Olmeda,C.

Severijn, 2004: Cloud representationin generalcirculation modelsover the Northern

Pacific Ocean:A EUROCS intercomparisonstudy. submittedto Quart. J. Roy. Meteor.

Soc.

Simmons,A. J.andD. M. Burridge,1981:An energy andangular-momentumconservingvertical

finite-difference scheme and hybrid vertical coordinates.Mon. Wea. Rev., 109, 758-766.

Slingo,A., 1989:A GCM parameterizationfor theshortwaveradiativepropertiesof waterclouds.

J. Atmos. Sci., 46, 1419-1427.

Soden,B. J.andF. P. Bretherton,1993:Uppertroposphererelative humidity from theGOES6.7

µm channel: Method and climatology for July 1987.J. Geophys. Res., 98, 16,669-16,688.

GFDL GAMDT

49

Stackhouse,P. W., Jr., S. J. Cox, S. K. Gupta,J. C. Mikovitz, and M. Chiacchio,2003: The

WCRP/GEWEXSurfaceRadiationBudgetDataSetRelease2: A 1 degreeresolution,12

year flux climatology. in preparation.

Stern,W. F. and R. T. Pierrehumbert,1988: The impact of an orographicgravity wave drag

parameterizationon extendedrangepredictionswith a GCM. Preprintsto the8th annual

conferenceon NumericalWeatherPrediction,Baltimore,AmericanMeteorologicalSoci-

ety, 745-750.

Stern,W. F. and K. Miyakoda, 1995: Feasibility of seasonalforecastsinferred from multiple

GCM Simulations.J. Clim., 8, 1071-1085.

Sud,Y. C. andG. K. Walker, 1999:Microphysicsof cloudswith the relaxed Arakawa-Schubert

scheme(McRAS).Part I: Designandevaluationwith GATE PhaseIII Data.J. Atmos. Sci.,

56, 3196-3220.

Taylor, J.P., J.M. Edwards,M. D. Glew, P. Hignett,andA. Slingo,1996:Studieswith a flexible

new radiationcode.Part II: Comparisonswith aircraftshort-wave observations.Quart. J.

Roy. Meteor. Soc., 122, 839-861.

Taylor, K. E., 2001:Summarizingmultiple aspectsof modelperformancein a singlediagram.J.

Geophys. Res., 106, 7183-7192.

Thompson,D. W. andJ.M. Wallace,1998:Thearcticoscillationsignaturein thewinter timegeo-

potential height and temperature fields.Geophys. Res. Lett., 25, 1297-1300.

Thompson,D. andJ. M. Wallace,2000Annular modesin the extratropicalcirculation.Part I:

GFDL GAMDT

50

Month-to-month variability. J. Clim., 13, 1000-1016.

Tiedtke, M., 1993:Representationof cloudsin large-scalemodels.Mon. Wea. Rev., 121, 3040-

3061.

Tokioka,T., K. Yamazaki,A. Kitoh, andT. Ose,1988:Theequatorial30-60dayoscillationand

the Arakawa-Schubertpenetrative cumulusparameterization.J Meteor. Soc. Japan, 66,

883-901.

Tompkins,A. M., 2002:A prognosticparameterizationfor thesubgrid-scalevariability of water

vaporandcloudsin large-scalemodelsandits useto diagnosecloudcover. J. Atmos. Sci.,

59, 1917-1942.

Trenberth, K. E., 1997: The definition of El Niño.Bull. Amer. Met. Soc., 78, 2771-2777.

Trenberth,K. E., andJ. M. Caron,2001: Estimatesof meridionalatmosphereandoceanheat

transports.J. Clim., 14, 3433-3443.

Vitart, F., J. L. Anderson,andW. F. Stern,1997:Simulationof interannualvariability of tropical

storm frequency in an ensemble of GCM integrations.J. Clim., 10, 745-760.

Vitart, F., andT. N. Stockdale,2001:Seasonalforecastingof tropicalstormsusingcoupledGCM

integrations. Mon. Wea. Rev., 129, 2521-2537.

Waliser, D., W. Lau, W. Stern,andC. Jones,2003:Potentialpredictabilityof theMadden-Julian

oscillation.Bull. Amer. Met. Soc., 84, 33-50.

Walsh,J.E.,V. M. Kattsov, W. L. Chapman,V. Govorkova,andT. Pavlova,2002:Comparisonof

GFDL GAMDT

51

Arctic climatesimulationsby uncoupledandcoupledglobal models.J. Clim., 15, 1429-

1446.

Warren,S.G.,C. J.Hahn,J.London,R. M. Chevin, andR. L. Jenne,1986:Globaldistributionof

totalcloudcoverandcloudtypeamountsover land.NCAR TechnicalNoteTN-273+STR,

Boulder, CO, 29 pp. + 200maps.[Availablefrom CarbonDioxide InformationAnalysis

Center, Oak Ridge, Tennessee.]

Warren,S.G.,C. J.Hahn,J.London,R. M. Chevin, andR. L. Jenne,1988:Globaldistributionof

total cloud cover and cloud type amountsover ocean.NCAR Technical Note TN-

317+STR,Boulder, CO,44pp.+ 170maps.[Availablefrom CarbonDioxide Information

Analysis Center, Oak Ridge, Tennessee.]

Webb,M., C. Senior, S. Bony, andJ.-J.Morcrette,2001:CombiningERBE andISCCPdatato

assessclouds in the Hadley Centre,ECMWF, and LMD atmosphericclimate models.

Clim. Dyn., 17, 905-922.

Weng,F., N. C. Grody, R. Ferraro,A. Basist,andD. Forsyth,1997:Cloudliquid waterclimatol-

ogy from the Special Sensor Microwave/Imager. J. Clim., 10, 1086-1098.

Wood,N. andP. Mason,1993:The pressureforce inducedby neutralturbulent flow over hills.

Quart. J. Roy. Meteor. Soc., 119, 1233-1267.

Woodruff, S. D., R. J. Slutz, R. L. Jenneand P. M. Steurer, 1987: A Comprehensive Ocean-

Atmosphere Data Set.Bull. Amer. Met. Soc., 68, 1239-1239.

Wyman,B. L., 1996:A step-mountaincoordinategeneralcirculationmodel:Descriptionandval-

GFDL GAMDT

52

idation of medium-range forecasts.Mon. Wea. Rev., 124, 102-121.

Xie, P. andP. A. Arkin, 1997:Globalprecipitation:A 17-yearmonthlyanalysisbasedon gauge

observations,satelliteestimates,andnumericalmodeloutputs.Bull. Amer. Met. Soc., 78,

2539-2558.

Zhang,Y.-C., W. B. Rossow, A. A. Lacis,V. Oinas,andM. I. Mishchenko, 2003:Calculationof

radiative flux profiles from the surfaceto top-of-atmospherebasedon ISCCPandother

globaldatasets:Refinementsof theradiative transfermodelandtheinputdata.in prepara-

tion for J. Geophys. Res.

GFDL GAMDT

53

Table captions

Table 1. Brief description of AM2/LM2 components.

Table2. Coefficientsak andbk for calculationof interfacelevel values.Thecoefficientsareused

in theSimmons-Burridge(1981)formulap = ak + bk * ps , wherep is pressureandps is surface

pressure.Thepressuresp andgeopotentialheightsz of interfacelevelsusinga scaleheightof 7.5

km andps = 1013.25 hPa are also shown.

Table3. Selectedglobal annualmeanradiationbudgetandhydrologic quantities.Observational

datasourcesareERBE:Harrisonet al. (1990);GEWEX: Stackhouseet al. (2004);GISS:Zhang

et al. (2003);KT: Kiehl andTrenberth(1997);NVAP: Randelet al. (1996),GR: Greenwald et al.

(1993); WG: Weng et al. (1997); ISCCP: Rossow and Schiffer (1999); SFC: Warren et al.

(1986,1988); CMAP: Xie and Arkin (1997); GPCP: Huffman et al. (1997).

GFDL GAMDT

54

Component Description

Dynamics B-grid model, 2.5˚ longitude by 2.0˚ latitude

24 vertical levels with the effective model top at about 40 km

Radiation Diurnal cycle with full radiation calculation every 3 hours

Effects of H2O, CO2, O3, O2, N2O, CH4, and 4 halocarbons included

Longwave: Simplified Exchange Approximation (Schwarzkopf and Ramaswamy

1999)

Clough et al. (1992) CKD 2.1 H2O continuum parameterization

Shortwave: Exponential Sum Fit with 18 bands (Freidenreich and Ramaswamy

1999)

Liquid cloud radiative properties from Slingo (1989)

Ice cloud radiative properties from Fu and Liou (1993)

Aerosols: Prescribed monthly three-dimensional climatology from chemical

transport models

Species represented include sulfate, hydrophilic and hydrophobic car-

bon, dust, and sea salt

Clouds 3 prognostic tracers: cloud liquid, cloud ice, and cloud fraction

Cloud microphysics from Rotstayn (1997)

Cloud macrophysics from Tiedtke (1993)

Table 1. Brief description of AM2/LM2 components.

GFDL GAMDT

55

Table 1 (cont.). Brief description of AM2/LM2 components.

Convection Relaxed Arakawa Schubert (Moorthi and Suarez 1992)

Detrainmentof cloudliquid, iceandfractionfrom convectiveupdrafts

into stratiform clouds

A lower bound imposed on lateral entrainment rates for deep convec-

tive updrafts (Tokioka et al. 1988)

Convective momentum transport represented by vertical diffusion pro-

portional to the cumulus mass flux

Vertical diffusion Surface and stratocumulus convective layers represented by a K-pro-

file scheme with prescribed entrainment rates (Lock et al. 2000)

Surface fluxes from Monin-Obukhov similarity theory

Gustinessenhancementto wind speedusedin surfaceflux calculations

(Beljaars 1995)

Enhanced near-surface mixing in stable conditions

Orographic roughness effects included

Gravity Wave Drag Orographic drag from Stern and Pierrehumbert (1988)

Land Model Isothermal surface (soil-snow-vegetation)

3 water stores: snow, root zone and ground water

18 soil-temperature levels to 6 m total depth

Stomatal control of evapotranspiration

Latent heat storage in soil

Surface parameters dependent on 8 soil and 8 vegetation types

Component Description

GFDL GAMDT

56

Table2. Coefficientsak andbk for calculationof interfacelevel values.Thecoefficientsareusedin theSimmons-Burridge(1981)formulap = ak + bk * ps , wherep is pressureandps is surfacepressure.Thepressuresp andgeopotentialheightsz of interfacelevelsusinga scaleheightof 7.5km andps = 1013.25 hPa are also shown.

k ak (Pa) bk p (hPa) z (km)

1 0 0 0 -

2 903.45 0 9 35.40

3 3474.8 0 35 25.30

4 7505.6 0 75 19.52

5 12787 0 128 15.52

6 19111 0 191 12.51

7 21855 0.043568 263 10.12

8 22884 0.11023 341 8.18

9 22776 0.19223 423 6.56

10 21716 0.28177 503 5.26

11 20073 0.36950 575 4.25

12 18110 0.45324 640 3.44

13 16005 0.53163 699 2.79

14 13878 0.60387 751 2.25

15 11813 0.66956 797 1.80

16 9865.9 0.72852 837 1.43

17 8074.0 0.78080 872 1.13

18 6458.1 0.82660 902 0.87

19 5028.0 0.86621 928 0.66

20 3784.6 0.90004 950 0.48

21 2722.0 0.92854 968 0.34

22 1829.0 0.95221 983 0.23

23 1090.2 0.97163 995 0.13

24 487.56 0.98735 1005 0.06

25 0 1 1013 0

GFDL GAMDT

57

Table3. Selectedglobal annualmeanradiationbudgetandhydrologic quantities.ObservationaldatasourcesareERBE:Harrisonet al. (1990);GEWEX: Stackhouseet al. (2004);GISS:Zhanget al. (2003);KT: Kiehl andTrenberth(1997);NVAP: Randelet al. (1996),GR: Greenwald et al.(1993); WG: Weng et al. (1997); ISCCP: Rossow and Schiffer (1999); SFC: Warren et al.(1986,1988); CMAP: Xie and Arkin (1997); GPCP: Huffman et al. (1997).

Measure Source Observation AM2/LM2

Top of Atmosphere Radiation Budget (W m-2)

Shortwave absorbed ERBE 240.2 235.7

Outgoing longwave radiation ERBE 235.3 235.3

Clear-sky shortwave absorbed ERBE 288.4 289.1

Clear-sky outgoing longwave radiation ERBE 264.8 260.0

Shortwave cloud forcing ERBE -48.2 -53.4

Longwave cloud forcing ERBE 29.5 24.7

Surface Energy Budget (W m-2)

Shortwave absorbed GEWEX/GISS 164.6/165.2 159.8

Net longwave GEWEX/GISS -47.1/-50.9 -57.8

Clear-sky shortwave absorbed GEWEX/GISS 214.7/218.4 216.9

Clear-sky downward longwave GEWEX/GISS 309.6/313.5 313.9

Shortwave cloud forcing GEWEX/GISS -50.1/-53.3 -57.2

Longwave down cloud forcing GEWEX/GISS 35.6/31.1 24.5

Sensible heat flux KT 24 18.7

Latent heat flux KT 78 82.2

Hydrologic Quantities

Column integrated water vapor (kg m-2) NVAP 24.5 23.4

Columnintegratedoceaniccloudliquid (g m-2) GR/WG 76.2/63.4 77.1

Column integrated cloud ice (g m-2) - - 37.6

Total cloud amount (fraction) ISCCP/SFC 0.69/0.62 0.66

Surface Precipitation (mm d-1) CMAP/GPCP 2.68/2.65 2.84

GFDL GAMDT

58

Figure captions

Figure 1. Long-term annual and zonal mean temperature difference between NCEP/NCAR rean-

alysis climatology and AM2/LM2 (AM2/LM2 minus NCEP). Contour interval is 0.5 K.

Figure 2. Long-term annual mean 2 m temperature difference between CRU climatology and

AM2/LM2 (AM2/LM2 minus CRU). Contour interval is 2 K.

Figure 3. Long-term mean 2 m temperature difference for North America between CRU climatol-

ogy and AM2/LM2 (AM2/LM2 minus CRU) for a) December-January-February (DJF) and b)

June-July-August (JJA). Contour interval is 2 K.

Figure 4. Long-term annual and zonal mean zonal wind in m s-1 for (a) NCEP/NCAR reanalysis,

(b) AM2/LM2, and (c) AM2/LM2 minus NCEP. Contour interval is 5 m s-1 in (a) and (b) and 1 m

s-1 in (c).

Figure 5. Long-term annual and zonal mean zonal wind stress in Pa over the ocean for the ship-

based climatology of COADS (blue) (daSilva et al. 1994; Woodruff et al. 1987), ECMWF reanal-

ysis (red) (Gibson et al. 1997), the ERS satellite scatterometer (green) (CERSAT-IFREMER

2002), and AM2/LM2 (black). The sign convention is such that a positive stress indicates an east-

erly stress on the atmosphere and a westerly stress on the ocean.

Figure 6. Long-term Northern Hemisphere DJF mean SLP minus 1013.25 hPa for (a) AM2/LM2,

(b) NCEP/NCAR reanalysis, and (c) AM2/LM2 minus NCEP. Contour interval is 3 hPa for (a)

and (b) and 1 hPa for (c); the zero contour is not plotted in (c). Note that regions with mean sur-

face pressure below 950 hPa have been masked.

GFDL GAMDT

59

Figure7. Long-termDJFmeandepartureof 500hPa geopotentialheightfrom its zonalmeanfor

(a) AM2/LM2, (b) NCEP/NCARreanalysisclimatology, and(c) AM2/LM2 minusNCEP. Con-

tour interval is 25 m in (a) and(b) and10 m in (c). Statisticsat thebottomof (a) and(b) include

theNorthernHemispheremeanandstandarddeviation.Statisticsat thebottomof (c) includethe

differencein NorthernHemispheremeans,theroot meansquareerror, andthecorrelationcoeffi-

cient.

Figure8. Annual long-termmeanprecipitationin mm d-1 for (a) AM2/LM2, (b) CMAP observa-

tions,and(c) AM2/LM2 minusCMAP. Statisticsat thebottomof (a) and(b) includetheglobal

meanandstandarddeviation.Statisticsat thebottomof (c) includethedifferencein globalmeans,

the correlation coefficient, and the root mean square error.

Figure 9. Annual long-termmeanoutgoinglongwave radiation(OLR) in W m-2 for (a) AM2/

LM2, (b) ERBEobservations,and(c) AM2/LM2 minusERBE.Statisticsat thebottomof (a) and

(b) includetheglobalmeanandstandarddeviation.Statisticsat thebottomof (c) includethedif-

ference in global means, the correlation coefficient, and the root mean square error.

Figure10.Annuallong-termmeanabsorbedsolarradiation(SWAbs) in W m-2 for (a)AM2/LM2,

(b) ERBE observations,and(c) AM2/LM2 minusERBE.Statisticsat the bottomof (a) and(b)

includetheglobalmeanandstandarddeviation. Statisticsat thebottomof (c) includethediffer-

ence in global means, the correlation coefficient, and the root mean square error.

Figure11. Annual long-termmeantotal cloud amount(fraction) for (a) AM2/LM2, (b) ISCCP

observations,and(c) AM2/LM2 minusISCCP. Statisticsat thebottomof (a) and(b) includethe

global mean and standard deviation. Statistics at the bottom of (c) include the difference in global

GFDL GAMDT

60

means,thecorrelationcoefficient, andthe root meansquareerror. Note that thesestatisticshave

been computed only over the domain 60˚S-60˚N, as high latitude ISCCP data is unreliable.

Figure12.Annuallong-termmeanuppertropospherichumidity in percentfor (a) AM2/LM2, (b)

TOVS observations,and(c) TOVS minusAM2/LM2. Statisticsat thebottomof (a) and(b) indi-

catethe global mean.Statisticsat the bottomof (c) includethe differencein global means,the

correlation coefficient, and the root mean square error.

Figure13.Distributionsof theregressioncoefficientsof precipitationrateversusthestandardized

NINO3 SSTindex, ascomputedusingtheensemblemeanof the10-memberAMIP-style integra-

tionswith theAM2/LM2 for 1951-2000(upperpanel)andtheGPCPdatasetfor 1979-2000,both

for theDecember-January-Februaryseason.Contourinterval: 1 mm d-1. Zerocontouris omitted.

Contours for -0.5 and +0.5 mm d-1 are inserted.

Figure14. Distributionsof the regressioncoefficientsof 200 hPa heightversusthe standardized

NINO3 SSTindex, ascomputedusingtheensemblemeanof the10-memberAMIP-style integra-

tionswith theAM2/LM2 (upperpanels)andNCEPreanalyses(lower panels)for theDecember-

January-Februaryseasonof the 1951-2000period.Resultsfor the northernandsouthernextrat-

ropicsareshown in theleft andright panels,respectively. Contourinterval is 5 m. Thezero-con-

tour is not plotted.

Figure15. Taylor diagramdepictingthe relationshipsbetweentheobservedDJF200hPa height

anomaliesin the North Pacific / North Americansector(20˚-70˚N,60˚-180˚W)during selected

ENSO events and the correspondingensemble-meanpatternsas simulatedin the 10-member

AMIP-style runswith theAM2/LM2. Resultsfor individualwarmandcoldENSOeventsarepre-

GFDL GAMDT

61

sentedusingredandbluedots,respectively. Thetwo-digit labelfor eachdot indicatestheyearof

the event in question.The spatial correlationcoefficient is given by the cosineof the angle

betweentheabscissaandthestraightline joining theorigin andthedot representingtheeventof

interest;correlationvaluesaregiven alongthe outersolid arc. The ratio betweenthe simulated

andobservedspatialvariancesis givenby thedistancebetweenthedotandorigin; innerandouter

solid arcs indicate ratios of 1 and 1.5, respectively. The root mean square(rms) difference

betweenthesimulatedandobservedpattern,asnormalizedby thespatialvarianceof theobserved

field, is givenby thedistancebetweenthedotandthepointwith coordinates(1,0) in thediagram;

inner and outer dotted arcs indicate normalized rms differences of 0.5 and 1, respectively.

Figure16. Scatterplotsof the precipitationanomaliesin threemonsoonregions [IND - Indian;

AUS - northernAustralia;SEA - southeastAsia; boundariesof theseregionsaredepictedin Fig-

ure3 of LauandNath(2000)]versustheNINO3 SSTanomalies.Theabscissain all panelsrepre-

sentsthe standardizedSST anomaly in the NINO3 region. The ordinateaxis representsthe

standardizedprecipitationanomalyin IND duringtheJJA (left panels),andin SEA (middlepan-

els)andAUS(right panels)duringtheDJFseason.TheupperpanelsarebasedonAM2/LM2 out-

put for the 1951-2000period.The lower panelsshow the observational estimatesprovided by

GPCPfor the shorterperiod of 1979-2000.In eachpanel, the anomaliesof precipitationand

NINO3 index for agivenyeararejointly depictedby asmalldotor square.Theoutstandingwarm

andcold ENSOeventsarehighlightedusingcoloreddotsandsquares,respectively. Thedatafor

all remainingyearsareplottedusingblackdots.Thecorrelationcoefficient for thedataentriesin

eachpanelis shown in theupperright cornerof thatpanel.Correlationvaluesbasedon all avail-

able yearsare given in parentheses.Correlationvaluesbasedon the available warm and cold

ENSO events only are given without parentheses.

GFDL GAMDT

62

Figure17.Spatialpatternof anomaliesin sealevel pressure(SLP, contours)andsurfacetempera-

ture (color shading)associatedwith a 1 hPa increasein an index of theNorthernAnnularMode

(NAM), alsoreferredto asthe Arctic Oscillation.The anomaliesshown arefrom the monthsof

NovemberthroughApril only. TheSLPanomaliesarecomputedby multiplying thelinearregres-

sioncoefficientsateachgrid pointby a1 hPa increasein aNAM index. Theshadingindicatesthe

surfaceair temperatureanomaliesin ˚C associatedwith a 1 hPa increaseof a NAM index andis

computedin a similar manner. TheNAM is definedby computinganempiricalorthogonalfunc-

tion (EOF)of SLPfor all pointsnorthof 20˚N.A NAM index is thencalculatedasthedifference

betweentheminimumandmaximumof thespatialpatternof thefirst EOFmultipliedby its asso-

ciatedtime series,therebyyielding an index with units hPa. (a) SpatialpatternNAM anomalies

for AM2/LM2. A 10memberensembleof experimentswasconductedusingobservedSSTvaria-

tionsfrom 1951to 2000.For eachensemblemembera NAM patternwascomputedasdescribed

above. The spatialpatternshown is the 10 memberensemblemeanof the NAM regressionpat-

terns.The temperatureshown is the 2 m surfaceair temperatureover both land andocean.(b)

Similar to (a) but for observationaldata.The EOF of SLP is adaptedfrom ThompsonandWal-

lace,(1998); the surfacetemperaturedatais from Jones(1994).Surfaceair temperatureis used

over land, while SST is used over oceanic regions including ice-covered areas.

Figure18. Frequency of tropical cyclonegenesisfor (a) AM2/LM2 and(b) observations.Units

are number of storms per year in a box of size 4˚ latitude by 5˚ longitude.

Figure 19. Wave frequency spectraof 5˚N to 5˚S 200 hPa zonal wind variancefor AM2/LM2

(top) and NCEP reanalyses (bottom). Contour interval is 40 m2s-2.

Figure20.CompositeNorthernHemispherewinter (November-April) Madden-JulianOscillation

GFDL GAMDT

63

from 30-90dayfilteredprecipitationin mm d-1. Mapsbasedon AM2/LM2 areshown in the left

column and thosebasedon CMAP observationsin right column.Sequentialmapsare 10 day

meanscenteredon lagsof -30 days,-15 days,0 days,+15 days,and+30days.Thesuperimposed

greendashedlines indicatepropagation of the disturbance.Note that the valuesfor AM2/LM2

values have been enhanced by a factor of 2.

Figure21.Taylor diagramsfor selectedvariablescomparingtheskill of AM2/LM2 (redcircle) in

reproducingtheobservedclimatologyto thatof olderGFDL models(MCM andDERF, greenand

orangecircles respectively) and other modelsparticipatingin AMIP II (CCM3.5, ECHAM4,

ECMWF, andHADAM3). Note that all non-GFDLmodelsareplottedwith a blue diamondto

preventuniqueidentification.Theselectedvariablesincludethoseassociatedwith surfaceclimate

(SLP, landsurfaceair temperature,andoceanicwind stress,toprow), watercycleandclouds(pre-

cipitation,shortwave cloudforcing,andtotal cloudamount,middlerow), anduppertropospheric

circulation(200 hPa eddygeopotentialandzonalwind, bottomrow). The observationalsources

for thesedatainclude the NCEPreanalysesfor SLP and200 hPa eddygeopotentialandzonal

wind, ECMWF reanalysesfor oceanicwind stress,ERBEfor cloudradiative forcing, ISCCPfor

total cloud amount,CMAP for precipitationandCRU for land surfaceair temperature.Beneath

eachvariablenameis anindicationof thegeographicaldomainandseasonusedin thecalculation.

Figure22. Poleward oceanicheattransportin petawatts(1015 W) from observationalbasedesti-

matesand implied by AM2/LM2 (dark black line). The observed estimatesare derived from

atmosphericdata(TrenberthandCaron2001;redlineswith dashedlinesindicatingplusor minus

onestandarderror; basedon NCEP-derived products)or oceanicdata(GanachaudandWunsch

2003). Results are shown for the a) Atlantic, b) Indo-Pacific, and c) world ocean basins.

GFDL GAMDT

64

80S 60S 40S 20S Eq 20N 40N 60N 80N

100

200

300

400

500

600

700

800

900

1000 −5

−4.5

−4

−3.5

−3

−2.5

−2

−1.5

−1

−0.5

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

Figure 1. Long-term annual and zonal mean temperature difference between NCEP/NCAR rean-alysis climatology and AM2/LM2 (AM2/LM2 minus NCEP). Contour interval is 0.5 K.

GFDL GAMDT

65

Figure 2. Long-term annual mean 2 m temperature difference between CRU climatology andAM2/LM2 (AM2/LM2 minus CRU). Contour interval is 2 K.

GFDL GAMDT

66

��� DJF

b) JJA

Figure 3. Long-term mean 2 m temperature difference for North America between CRU climatol-ogy and AM2/LM2 (AM2/LM2 minus CRU) for a) December-January-February (DJF) and b)June-July-August (JJA). Contour interval is 2 K.

GFDL GAMDT

67

AM2/LM2

200

400

600

800

1000

Reanalysis

200

400

600

800

1000−25

−20

−15

−10

−5

0

5

10

15

20

25

30

35

40

45

AM2 − Reanalysis

80S 60S 40S 20S Eq 20N 40N 60N 80N

200

400

600

800

1000

−10−9−8−7−6−5−4−3−2−1012345678910

Figure 4. Long-term annual and zonal mean zonal wind in m s-1 for (a) NCEP/NCAR reanalysis,(b) AM2/LM2, and (c) AM2/LM2 minus NCEP. Contour interval is 5 m s-1 in (a) and (b) and 1 ms-1 in (c).

GFDL GAMDT

68

Figure 5. Long-term annual and zonal mean zonal wind stress in Pa over the ocean for the ship-based climatology of COADS (blue) (daSilva et al. 1994; Woodruff et al. 1987), ECMWF reanal-ysis (red) (Gibson et al. 1997), the ERS satellite scatterometer (green) (CERSAT-IFREMER2002), and AM2/LM2 (black). The sign convention is such that a positive stress indicates an east-erly stress on the atmosphere and a westerly stress on the ocean.

GFDL GAMDT

69

Figure 6. Long-term Northern Hemisphere DJF mean SLP minus 1013.25 hPa for (a) AM2/LM2,(b) NCEP/NCAR reanalysis, and (c) AM2/LM2 minus NCEP. Contour interval is 3 hPa for (a)and (b) and 1 hPa for (c); the zero contour is not plotted in (c). Note that regions with mean sur-face pressure below 950 hPa have been masked.

GFDL GAMDT

70

Figure7. Long-termDJFmeandepartureof 500hPa geopotentialheightfrom its zonalmeanfor(a) AM2/LM2, (b) NCEP/NCARreanalysisclimatology, and(c) AM2/LM2 minusNCEP. Con-tour interval is 25 m in (a) and(b) and10 m in (c). Statisticsat thebottomof (a) and(b) includetheNorthernHemispheremeanandstandarddeviation.Statisticsat thebottomof (c) includethedifferencein NorthernHemispheremeans,theroot meansquareerror, andthecorrelationcoeffi-cient.

GFDL GAMDT

71

Figure8. Annual long-termmeanprecipitationin mm d-1 for (a) AM2/LM2, (b) CMAP observa-tions,and(c) AM2/LM2 minusCMAP. Statisticsat thebottomof (a) and(b) includetheglobalmeanandstandarddeviation.Statisticsat thebottomof (c) includethedifferencein globalmeans,the correlation coefficient, and the root mean square error.

GFDL GAMDT

72

Figure 9. Annual long-termmeanoutgoinglongwave radiation(OLR) in W m-2 for (a) AM2/LM2, (b) ERBEobservations,and(c) AM2/LM2 minusERBE.Statisticsat thebottomof (a) and(b) includetheglobalmeanandstandarddeviation.Statisticsat thebottomof (c) includethedif-ference in global means, the correlation coefficient, and the root mean square error.

GFDL GAMDT

73

Figure10.Annuallong-termmeanabsorbedsolarradiation(SWAbs) in W m-2 for (a)AM2/LM2,(b) ERBE observations,and(c) AM2/LM2 minusERBE.Statisticsat the bottomof (a) and(b)includetheglobalmeanandstandarddeviation. Statisticsat thebottomof (c) includethediffer-ence in global means, the correlation coefficient, and the root mean square error.

GFDL GAMDT

74

Figure11. Annual long-termmeantotal cloud amount(fraction) for (a) AM2/LM2, (b) ISCCPobservations,and(c) AM2/LM2 minusISCCP. Statisticsat thebottomof (a) and(b) includetheglobal mean and standard deviation. Statistics at the bottom of (c) include the difference in globalmeans,thecorrelationcoefficient, andthe root meansquareerror. Note that thesestatisticshavebeen computed only over the domain 60˚S-60˚N, as high latitude ISCCP data is unreliable.

GFDL GAMDT

75

Figure12.Annuallong-termmeanuppertropospherichumidity in percentfor (a) AM2/LM2, (b)TOVS observations,and(c) TOVS minusAM2/LM2. Statisticsat thebottomof (a) and(b) indi-catethe global mean.Statisticsat the bottomof (c) includethe differencein global means,thecorrelation coefficient, and the root mean square error.

GFDL GAMDT

76

Figure13.Distributionsof theregressioncoefficientsof precipitationrateversusthestandardizedNINO3 SSTindex, ascomputedusingtheensemblemeanof the10-memberAMIP-style integra-tionswith theAM2/LM2 for 1951-2000(upperpanel)andtheGPCPdatasetfor 1979-2000,bothfor theDecember-January-Februaryseason.Contourinterval: 1 mm d-1. Zerocontouris omitted.Contours for -0.5 and +0.5 mm d-1 are inserted.

GFDL GAMDT

77

Figure14. Distributionsof the regressioncoefficientsof 200 hPa heightversusthe standardizedNINO3 SSTindex, ascomputedusingtheensemblemeanof the10-memberAMIP-style integra-tionswith theAM2/LM2 (upperpanels)andNCEPreanalyses(lower panels)for theDecember-January-Februaryseasonof the 1951-2000period.Resultsfor the northernandsouthernextrat-ropicsareshown in theleft andright panels,respectively. Contourinterval is 5 m. Thezero-con-tour is not plotted.

GFDL GAMDT

78

1.0

0.99

0.95

0.9

0.8

0.7

0.6

0.5

0.4

0.30.

20.10.0

-

-

-

-

-

-

-

-

---

5455

57 64

65

69

70

72

73

75

82

87

8891

97

0 0.5 1 1.50

0.5

1

1.5

Correlation

Figure15. Taylor diagramdepictingthe relationshipsbetweentheobservedDJF200hPa heightanomaliesin the North Pacific / North Americansector(20˚-70˚N,60˚-180˚W)during selectedENSO events and the correspondingensemble-meanpatternsas simulatedin the 10-memberAMIP-style runswith theAM2/LM2. Resultsfor individualwarmandcoldENSOeventsarepre-sentedusingredandbluedots,respectively. Thetwo-digit labelfor eachdot indicatestheyearofthe event in question.The spatial correlationcoefficient is given by the cosineof the anglebetweentheabscissaandthestraightline joining theorigin andthedot representingtheeventofinterest;correlationvaluesaregiven alongthe outersolid arc. The ratio betweenthe simulatedandobservedspatialvariancesis givenby thedistancebetweenthedotandorigin; innerandoutersolid arcs indicate ratios of 1 and 1.5, respectively. The root mean square(rms) differencebetweenthesimulatedandobservedpattern,asnormalizedby thespatialvarianceof theobservedfield, is givenby thedistancebetweenthedotandthepointwith coordinates(1,0) in thediagram;inner and outer dotted arcs indicate normalized rms differences of 0.5 and 1, respectively.

79

-4 -3 -2 -1 0 1 2 3 4-4

-3

-2

-1

0

1

2

3

4

Pre

cipi

tati

on a

nom

aly

1957

1965

1969

1972

1982

1987

1991

1997

IND

-4 -3 -2 -1 0 1 2 3 4-4

-3

-2

-1

0

1

2

3

4

Pre

cipi

tati

on a

nom

aly

IND

-4 -3 -2 -1 0 1 2 3 4Nino3 SST anomaly

-4

-3

-2

-1

0

1

2

3

4

AM2/LM2

SEA

-4 -3 -2 -1 0 1 2 3 4Nino3 SST anomaly

-4

-3

-2

-1

0

1

2

3

4

GPCP Observations

SEA

-4 -3 -2 -1 0 1 2 3 4-4

-3

-2

-1

0

1

2

3

4

1954

1955

1964

1970

1973

1975

1988

AUS

-4 -3 -2 -1 0 1 2 3 4-4

-3

-2

-1

0

1

2

3

4AUS

-.70 .80 -.68

-.43 .69 .27(-.06) (.58) (-.23)

(-.31) (.64) (-.46)

El N

ino

La

Nin

a

all yrsEn+Ln

Figure16. Scatterplotsof the precipitationanomaliesin threemonsoonregions [IND - Indian;AUS - northernAustralia;SEA - southeastAsia; boundariesof theseregionsaredepictedin Fig-ure3 of LauandNath(2000)]versustheNINO3 SSTanomalies.Theabscissain all panelsrepre-sentsthe standardizedSST anomaly in the NINO3 region. The ordinateaxis representsthestandardizedprecipitationanomalyin IND duringtheJJA season(left panels),andin SEA (mid-dle panels)andAUS (right panels)during theDJFseason.Theupperpanelsarebasedon AM2/LM2 output for the 1951-2000period.The lower panelsshow the observationalestimatespro-videdby GPCPfor theshorterperiodof 1979-2000.In eachpanel,theanomaliesof precipitationandNINO3 index for a givenyeararejointly depictedby a smalldot or square.Theoutstandingwarm andcold ENSOeventsarehighlightedusingcoloreddotsandsquares,respectively. Thedatafor all remainingyearsareplottedusingblackdots.The correlationcoefficient for the dataentriesin eachpanelis shown in theupperright cornerof thatpanel.Correlationvaluesbasedonall availableyearsaregiven in parentheses.Correlationvaluesbasedon the availablewarm andcold ENSO events only are given without parentheses.

GFDL GAMDT

80

(b) Observations

(a) AM2/LM2

-0.4 -0.35 -0.3 -0.2-0.25 -0.15 -0.1 -0.05 0 0.40.350.30.2 0.250.150.10.05

Figure17.Spatialpatternof anomaliesinsea level pressure(SLP, contours)andsurfacetemperature(colorshading)asso-ciatedwith a 1 hPa increasein an indexof the NorthernAnnular Mode (NAM),alsoreferredto astheArctic Oscillation.The anomalies shown are from themonthsof NovemberthroughApril only.The SLP anomaliesare computed bymultiplying the linear regressioncoeffi-cients at each grid point by a 1 hPaincreasein a NAM index. The shadingindicates the surface air temperatureanomaliesin ˚C associatedwith a 1 hPaincreaseof a NAM index and is com-putedin a similar manner. The NAM isdefined by computing an empiricalorthogonalfunction(EOF)of SLPfor allpoints north of 20˚N. A NAM index isthencalculatedasthedifferencebetweenthe minimum andmaximumof the spa-tial patternof thefirst EOFmultiplied byits associatedtime series,therebyyield-ing an index with units hPa. (a) SpatialpatternNAM anomaliesfor AM2/LM2.A 10 memberensembleof experimentswasconductedusingobservedSSTvari-ations from 1951 to 2000. For eachensemblemembera NAM pattern wascomputedas describedabove. The spa-tial pattern shown is the 10 memberensemblemeanof the NAM regressionpatterns.The temperatureshown is the2m surfaceair temperatureover both landand ocean. (b) Similar to (a) but forobservational data.The EOF of SLP isadaptedfrom Thompsonand Wallace,(1998); the surface temperaturedata isfrom Jones(1994).Surfaceair tempera-ture is usedover land,while SSTis usedover oceanicregions including ice-cov-ered areas.

GFDL GAMDT

81

Figure18. Frequency of tropical cyclonegenesisfor (a) AM2/LM2 and(b) observations.Unitsare number of storms per year in a box of size 4˚ latitude by 5˚ longitude.

GFDL GAMDT

82

Figure 19. Wave frequency spectraof 5˚N to 5˚S 200 hPa zonal wind variancefor AM2/LM2(top) and NCEP reanalyses (bottom). Contour interval is 40 m2s-2.

GFDL GAMDT

83

-30 days

-15 days

0 days

15 days

30 days

Figure20.CompositeNorthernHemispherewinter (November-April) Madden-JulianOscillationfrom 30-90dayfilteredprecipitationin mm d-1. Mapsbasedon AM2/LM2 areshown in the leftcolumn and thosebasedon CMAP observationsin right column.Sequentialmapsare 10 daymeanscenteredon lagsof -30 days,-15 days,0 days,+15 days,and+30days.Thesuperimposedgreendashedlines indicatepropagation of the disturbance.Note that the valuesfor AM2/LM2values have been enhanced by a factor of 2.

GFDL GAMDT

84

MCM

AM2/LM2CCM3.5ECHAM4ECMWFHADAM3{

DERF

-

-

-

-

-

--

-----

1.0

0.99

0.95

0.9

0.8

0.7

0.6

0.5

0.40.

30.20.

1

0.0

0 0.5 1 1.50

0.5

1

1.5

0 0.5 1 1.50

0.5

1

1.5

Sea Level Pressure

Correlation

Obs

0.5

1.0

global,ocean only,DJFSurface Temperature

-

-

-

-

-

--

-----

1.0

0.99

0.95

0.9

0.8

0.7

0.6

0.5

0.40.

30.20.

1

0.0

0 0.5 1 1.50

0.5

1

1.5

0 0.5 1 1.50

0.5

1

1.5

Correlation

Obs

0.5

1.0

30N-90N,land only,JJA

-

-

-

-

-

--

-----

1.0

0.99

0.95

0.9

0.8

0.7

0.6

0.5

0.40.

30.20.

1

0.0

0 0.5 1 1.50

0.5

1

1.5

0 0.5 1 1.50

0.5

1

1.5

Zonal Surface Wind Stress

Correlation

Obs

0.5

1.0

60S-60N,ocean only,ANN

-

-

-

-

-

--

-----

1.0

0.99

0.95

0.9

0.8

0.7

0.6

0.5

0.40.

30.20.

1

0.0

0 0.5 1 1.50

0.5

1

1.5

0 0.5 1 1.50

0.5

1

1.5Precipitation

Correlation

Obs

0.5

1.0

30S-30N,ANN

-

-

-

-

-

--

-----

1.0

0.99

0.95

0.9

0.8

0.7

0.6

0.5

0.40.

30.20.

1

0.0

0 0.5 1 1.50

0.5

1

1.5

0 0.5 1 1.50

0.5

1

1.5Shortwave Cloud Forcing

Correlation

Obs

0.5

1.0

60S-60N,ANN

-

-

-

-

-

--

-----

1.0

0.99

0.95

0.9

0.8

0.7

0.6

0.5

0.40.

30.20.

1

0.0

0 0.5 1 1.50

0.5

1

1.5

0 0.5 1 1.50

0.5

1

1.5Total Cloud Amount

Correlation

Obs

0.5

1.0

60S-60N,ANN

200hPa Z*

-

-

-

-

-

--

-----

1.0

0.99

0.95

0.9

0.8

0.7

0.6

0.5

0.40.

30.20.

1

0.0

0 0.5 1 1.50

0.5

1

1.5

0 0.5 1 1.50

0.5

1

1.5

Correlation

Obs

0.5

1.0

30N-90N,DJF

-

-

-

-

-

--

-----

1.0

0.99

0.95

0.9

0.8

0.7

0.6

0.5

0.40.

30.20.

1

0.0

0 0.5 1 1.50

0.5

1

1.5

0 0.5 1 1.50

0.5

1

1.5200hPa Zonal Wind

Correlation

Obs

0.5

1.0

global,DJF

Figure21.Taylor diagramsfor selectedvariablescomparingtheskill of AM2/LM2 (redcircle) inreproducingtheobservedclimatologyto thatof olderGFDL models(MCM andDERF, greenandorangecircles respectively) and other modelsparticipatingin AMIP II (CCM3.5, ECHAM4,ECMWF, andHADAM3). Note that all non-GFDLmodelsareplottedwith a blue diamondtopreventuniqueidentification.Theselectedvariablesincludethoseassociatedwith surfaceclimate(SLP, landsurfaceair temperature,andoceanicwind stress,toprow), watercycleandclouds(pre-cipitation,shortwave cloudforcing,andtotal cloudamount,middlerow), anduppertroposphericcirculation(200 hPa eddygeopotentialandzonalwind, bottomrow). The observationalsourcesfor thesedatainclude the NCEPreanalysesfor SLP and200 hPa eddygeopotentialandzonalwind, ECMWF reanalysesfor oceanicwind stress,ERBEfor cloudradiative forcing, ISCCPfortotal cloud amount,CMAP for precipitationandCRU for land surfaceair temperature.Beneatheachvariablenameis anindicationof thegeographicaldomainandseasonusedin thecalculation.

GFDL GAMDT

85

−90 −60 −30 0 30 60 90Latitude

−2

−1

0

1

2

PW

−2

−1

0

1

2

PW

−2

−1

0

1

2P

W

Trenberth & Caron Ganachaud & WunschAM2/LM2

a) Atlantic

b) Indo−Pacific

c) All basins

Figure22. Poleward oceanicheattransportin petawatts(1015 W) from observationalbasedesti-matesand implied by AM2/LM2 (dark black line). The observed estimatesare derived fromatmosphericdata(TrenberthandCaron2001;redlineswith dashedlinesindicatingplusor minusonestandarderror; basedon NCEP-derived products)or oceanicdata(GanachaudandWunsch2003). Results are shown for the a) Atlantic, b) Indo-Pacific, and c) world ocean basins.