us CLIVARU.S. CLIVAR

U.S. CLIMATE VARIABILITY AND PREDICTABILITY (CLIVAR)

1717 Pennsylvania Avenue, NW, Suite 250, Washington, DC 20006• www.usclivar.org

Summer 2011, Vol. 9, No. 2

Atlantic Basin ResearchChallenges

by Michael Patterson, Director

This issue of Variations presentsfour articles highlighting mod-eling research presented at the

CLIVAR-sponsored workshop onCoupled Ocean-Atmosphere-LandProcesses in the Atlantic held inMiami, March 23-25, 2011. Theworkshop reviewed model biases insimulating the Atlantic (notablyocean circulation, sea surface tem-peratures, surface winds, clouds,and precipitation) and identifiedhypotheses on mechanisms respon-sible for the biases (see summaryarticle on back page).

Kirtman et al. summarizesresults of experiments usingincreased ocean model resolution toassesses the impact of resolvedocean fronts and eddies on the sim-ulated mean large-scale climate.Brian Medeiros presents a compari-son of simulated surface and atmos-pheric state of the eastern tropicalAtlantic and Pacific stratocumulusregions, identifying differences inmodeled characteristics of the tworegions. Christina Patricola et al.investigate contributions of biases inwesterly trade winds as well asAmazonian and tropical Africanprecipitation to warm summer SSTbiases in the eastern equatorialAtlantic. Sang-Ki Lee et al. exam-

Continued on Page Two

VARIATIONS

Impact of Ocean Model Resolution..........1Southern Hemipshere Stratocumulus Decksin CAM ....................................................5Calendar ....................................................8Tropical Atlantic Bias Problem ................9Secular and Multidecadal Warming of theAtlantic Ocean ........................................13Tropical Atlantic Workshop ....................16

IN THIS ISSUE

Impact of Ocean Model Resolution onCCSM Climate Simulations

Ben P. Kirtman1, Cecilia Bitz2, Frank Bryan3, William Collins4, John Dennis3 ,Nathan Hearn3, James L. Kinter5 III, Richard Loft3, Clement Rousset1, Leo

Siqueira1, Cristiana Stan5, Robert Tomas3, and Mariana Vertenstein3

1University of Miami, Miami, Florida2University of Washington, Seattle, Washington

3National Center for Atmospheric Research, Boulder, Colorado4University of California, Berkeley, California

5Center for Ocean-Land-Atmosphere Studies, Calverton, Maryland

There is a growing demand forenvironmental predictionsthat include a broader rangeof space and time scales and

that include a more complete representa-tion of physical processes. Meeting thisdemand necessitates a unified approachthat will challenge the traditional bound-aries between weather and climate sci-ence, and will require a more integratedapproach to the underlying Earth systemscience and the supporting computationalscience. One of the consequences of thisunified or seamless approach is the needto explore much higher spatial resolutionin weather and climate models. Increasedmodel resolution has the potential to bet-ter resolve relevant features, and, moreimportantly, to accurately represent theinteractions and feedbacks among thevarious physical and dynamical process-es (Randall et al. 2003; Hurrell et al.2009; Shukla et al. 2008; Brunet et al.2010). It is also recognized that interac-tions across time and space scales arefundamental to the climate system itself.The large-scale climate, for instance,determines the environment formicroscale (order 1 km) and mesoscale(order 10 km) variability which thenfeedback onto the large-scale climate. Inthe simplest terms, the statistics ofmicroscale and mesoscale variability sig-

nificantly impact the simulation of cli-mate. In typical climate models at, say,200 km horizontal resolution1, thesevariations occur on unresolved scales,and the microscale and mesoscaleprocesses are parameterized in terms ofthe resolved variables.

Several recent studies havefocused on the importance of atmospher-ic model resolution in the simulation ofclimate (May and Roeckner 2001;Brankovic and Gregory 2001; Pope andStratton 2002; Kobayashi and Sugi 2004;Hack et al. 2006; Navarra et al. 2008;Gent et al. 2009; Kinter et al. 2011). Thereported results range from little or nochange in the mean and variable climate(i.e., Hack et al. 2006) to significant dif-ferences in the cycle of El Niño andtheSouthern Oscillation (ENSO; Navarraet al. 2008) and in sea surface tempera-ture (SST) biases in the upwellingregions (i.e., Gent et al. 2009 – hereafterG09).

McClean et al. (2010) andBryan et al. (2010) were the first toexamine the question of the resolutiondependence of simulations with CCSM(Community Climate System Model)that incorporated an eddy-resolvingocean component . In particular,McClean et al. (2010) used the sameversion of the CCSM component modelsas used in this study and G09, but with

1Throughout this paper, model “resolution” refers to the spacing of model grid elements.

V9N2 8/8/11 10:47 AM Page 1

U.S. CLIVARine the multi-decadal variability ofthe Atlantic Meridional OverturningCirculation (AMOC) and associatedmeridional ocean heat transport inthe 20th Century.

Atlantic meridional heat trans-port was a focus of the 2011 USAMOC annual science meeting, heldjointly with the UK Rapid ClimateChange Programme in Bristol, UK,12-15 July. Understanding Atlanticvariability on a range of time scaleswas explored, with a main focus ofthe role of the AMOC. Presentationabstracts are available athttp://www.noc.soton.ac.uk/rapid/ic2011/. Aditionally, the recentlyreleased Special Issue of Deep-SeaResearch II entitled Climate and theAtlantic Overturning MeridionalCirculation presents a series ofpapers on observing and monitoringthe characteristics of AMOC, meth-ods for tracking overturning trans-ports, predictability and prediction,and the potential for abrupt climatechange (see http://www.elsevier.com/wps/find/journaldescription.cws_home/116/description).

Finally, I wish to take this oppor-tunity to announce my acceptance ofthe permanent position of the U.S.CLIVAR Office Director, effective May31, 2011. I look forward to engag-ing the US climate research commu-nity and the US climate researchfunding agencies in the months andyears ahead to advance climate sci-ence planning and implementation inthe US and to foster fruitful interna-tional collaborations.

Page 2

Variations

U.S. CLIVAR Office1717 Pennsylvania Ave., NWSuite 250, Washington, DC 20006usco@usclivar.org

Staff: Michael Patterson, EditorCathy Stephens, Assistant Editor and Staff Writer

© 2011 U.S. CLIVAR Permission to use any scientific material (text and fig -ures) published in this Newsletter should be obtainedfrom the respective authors. Suggestion citation ofnewsletter articles: Authors, year: Title. U.S. CLIVARNewsletter, Vol.(No.), pp. (unpublished manuscript).

Continued from Page One

horizontal atmospheric model resolutionof approximately 0.25º coupled to the0.1º ocean component. The coupledmodel was run for 20 years, and pro-duced simulated SST that is too cold inthe sub-polar and mid-latitude NorthernHemisphere, but more realistic Aghulaseddy pathways compared to ocean-onlysimulations at comparable resolution.Bryan et al. (2010) examined theMcClean et al. (2010) simulation as wellas two additional experiments separatelyprobing the ocean and atmosphericmodel resolution. As in McClean et al.(2010), the Bryan et al. (2010) simula-tions were run for approximately 20years. Bryan et al. (2010) focused prima-rily on the coupling between the loweratmosphere and the SST, and found amore realistic pattern of positive correla-tion between high-pass filtered surfacewind speed and SST when t he oceancomponent model is eddy-resolving.Both of these earlier studies are viewedas predecessors for the present work. Thefundamental difference is that the focusof this work is on climate variability,which requires simulations of muchlonger duration than 20 years. Additionaldetails and results from the work pre-sented here can be found in Kirtman etal. (2011).2. Model Configure and ExperimentalDesigna. CCSM3.5

The model used for this study is theNCAR Community Climate SystemModel version 3.5 (CCSM3.5) (Neale etal. 2008; G09). The atmospheric compo-nent model, the Community AtmosphericModel (CAM), is based on a finite vol-ume discretization rather than the spec-tral discretization of the governing equa-tions used in earlier versions of CAM,and has extensive changes in the parame-terization of sub-grid-scale processes thathave resulted in a significant improve-ment in the simulation of tropical vari-ability relative to CCSM3.0 (Neale et al.2008). Changes in the other componentmodels, while less extensive, have alsocontributed to a reduction in systematicbiases (Jochum et al. 2008; G09).

b. Increasing the Ocean ModelResolution

Two experiments are reported here.The first experiment (i.e., control,referred to as LRC) is a 155-year pres-ent-day climate simulation of the 0.5ºatmosphere (zonal resolution 0.625º,meridional resolution 0.5º) coupled toocean and sea-ice components withzonal resolution of 1.2º and meridionalresolution varying from 0.27º at theequator to 0.54º in the mid-latitudes on adipole grid (Murray, 1996) with 60 verti-cal levels. This control experiment isidentical to the “high-resolution” experi-ment in G09 in terms of the model con-figuration, but differs in its initial stateand climate forcing. The G09 experimentwas a transient climate simulation initial-ized with a state extracted at year 1980from a 20th century integration of themodel at coarser resolution. The initialcondition for our experiments was takenfrom the end of a previously completedpresent day control simulation carriedout with an earlier version of CCSM, sothat the ocean state is fully “spun-up”and the initialization shock ought to beminimized; however, as will be shown,some climate drift remains. The secondsimulation uses the same atmosphericmodel coupled to 0.1º ocean and sea-icecomponent models and will be referredto as HRC06. The ocean model configu-ration in this case is identical to the cou-pled climate simulation of McClean etal. (2011). In addition to the change inhorizontal resolution from the controlexperiment, there are commensuratechanges in the parameterization of hori-zontal sub-grid scale dissipation, and useof a different, 42-level vertical grid.3. Results

One of the key motivating factors forthis study is to assess how resolvedocean fronts and eddies impact the simu-lated mean large-scale climate.Specifically, Fig. 1a shows the NorthAtlantic SST climatology from HRC06(shaded) and LRC superimposed (blackcontours). Fig. 1b is in the same format,except the figure shows the climatologi-cal rainfall. In terms of the surface tem-perature, HRC06 produces much sharper

2 In common parlance, eddy-resolving models have horizontal resolution of less than 1/6º, in contrast tocoarse-resolution models with 1º or greater grid spacing or eddy-permitting ocean models whose resolutionlies between 1/6º and 1º.

V9N2 8/8/11 10:47 AM Page 2

VARIATIONS

Page 3

gradients than LRC along the coast ofNorth America and in the Gulf Streamseparation region. Figure 1b indicatessignificant structural changes in the sim-ulated rainfall associated with oceanmodel resolution. For example, the axisof maximum rainfall in HRC06 followsthe maximum SST gradient so that therainfall hugs the US coast and extendsout into the open Atlantic as part of theGulf Stream extension. The LRC simula-tion captures some aspects of the rainfallmaximum along the US coast, but per-haps as expected, fails to capture theeast-west oriented maximum along theGulf Stream extension.

With respect to how higher oceanmodel resolution impacts the seasonal-to-interannual variability, we concentrate onthe monthly mean SST anomalies(SSTA). Figure 2a, for example, showsthe ratio of the SSTA monthly standarddeviation for HRC06 compared to LRC.The standard deviation is calculated onthe atmospheric model grid. The SST

uniform reduction along the equator of0.2ºC in the ENSO-related standarddeviation. We hypothesize that thereduction of ENSO variance is an indi-rect effect of the ubiquitous ocean sur-face warming (see Fig. 2b) and modestincreases in ocean stratification in thePacific that ultimately lead to a smallreduction in the ENSO variance (similarto the CCSM response to doubled CO2levels – see Collins et al. 2010). Thisreduced ENSO variability leads toreduced variance in the tropical IndianOcean and the north tropical Atlantic viawell-known teleconnections.

The second aspect of interest in thereduction of variance in Fig. 2b appearsas a subtle feature; namely, the standarddeviation in HRC06 in the westernPacific between 140ºE and 160ºE has apronounced plateau compared to LRC.This is a distinctly positive aspect of theHRC06 simulation compared to theobserved variance shown in Fig. 2b. Forinstance, most current coupled models

produce ENSO variance that extends toofar into the western Pacific leading toENSO teleconnections that have thewrong sign in the western Pacific.Kirtman and Vecchi (2010), Wu andKirtman (2007) and Wu et al. (2007)argue heuristically that the excessivevariance and teleconnection errors aredue to “excessive coupling” between theatmosphere and the ocean. We hypothe-size that the enhanced ocean resolutionin HRC06 has the effect of adding rela-tively high frequency noise (see Fig. 2c)to the coupled system and destructivelyinterfering with the excessive coupling.

Figure 2c shows the high-pass fil-tered SSTA standard deviation along theequator for 10 years of daily data fromLRC and HRC06 respectively. Thehigh-pass filtered standard deviation islarger for HRC06 compared to LRCindicating that the eddies enhance thevariance but primarily at sub-monthlytime scales.

In keeping with the objective toexamine how resolved ocean fronts andeddies impact the large-scale climate,the results obtained suggest somenotable climatic impacts. For example:(i)The SST front associated with theGulf Stream is better resolved in thehigh-resolution simulation. This leads tolarge structural changes in the meanrainfall. Similar changes in the currentsare seen in the vicinity of the Kuroshio,but the increased ocean resolution isapparently not as important for main-taining the SST gradient, as the temper-ature and rainfall differences are smallcompared to those near the Gulf Stream.(ii)As expected, the variability ofmonthly mean SSTA increases withincreasing ocean resolution throughoutthe extra-tropics. This increase is mostnotable in the western boundary currentregions and the Southern Ocean.Contrary to expectation, ENSO variabil-ity decreases and some of the associatedtropical SSTA teleconnections are weak-er. (iii)The structure of the SSTA vari-ability along the equator in the westernPacific also is quite different in HRC06and LRC. There is a distinct plateau inthe variability in HRC06 that is consis-tent with reduced coupling and ENSO

Figure 1. (a) Annual mean SST in the North Atlantic for

HRC06 (shaded) and for LRC (contour) in degrees

Celsius. (b) Annual mean surface current speeds in cm

sec-1 for HRC (shaded) and LRC (contour).

3Only 10 years of daily data was available for this calculation.

variability applied to theatmosphere is clearlyenhanced in HRC06throughout most of themid-latitudes and the sub-tropics. The core regionsof substantially enhancedvariance include theNorthern Hemispherewestern boundary currentzones and the SouthernOcean from the Atlanticcoast of South Americaextending through to thePacific side of theAustralian continent.

In contrast to theextra-tropics, throughoutmost of the tropical oceanthere is a reduction ofvariance in the eddy-resolving simulation(HRC06). Figure 2b indi-cates that there is a robustreduction of variability inthe tropical Pacific. Thereare two aspects to thisreduction of variance thatare of particular interest.First, there is a nearly

V9N2 8/8/11 10:47 AM Page 3

U.S. CLIVAR

Page 4

events that do not extend as far to thewest as in the case of LRC. Acknowledgements:We acknowledge the support of theNational Science Foundation (J. Kinterand C Stan through AGS 0830068 andOCI 0749290; B. Kirtman through OCI0749165, AGS 0754341 and AGS0850897; C. Bitz through ARC 0938204;F. Bryan, J. Dennis, N. Hearn, R. Loft, R.Tomas and M. Vertenstein through itssupport of NCAR). B. Kirtman alsoacknowledges support from NOAANA08OAR3420889. Computingresources were provided by the NationalInstitute of Computational Sciences at theUniversity of Tennessee through an awardmade by the TeraGrid ResourceAllocations Committee.References

Brankovic C. and Gregory D., 2001:Impact of horizontal resolution on seasonalintegrations. Clim. Dyn., 18:123–143.doi:10.1007/s003820100165.

Brunet, G., and Coauthors, 2010:Collaboration of the Weather and ClimateCommunities to Advance Subseasonal-to-

Kirtman, B. P., C. Bitz, F. Bryan, W.Collins, J. Dennis, N. Hern, J. L. Kinter III,R. Loft, C. Rousset, L. Siqueira, C. Stan, R.Tomas, and M. Vertenstein, 2011: Impact ofocean model resolution on CCSM climatesimulations. Clim. Dyn. (submitted).

Kobayashi C. and Sugi, M., 2004:Impact of horizontal resolution on the simula -tion of the Asian summer monsoon and tropi -cal cyclones in the JMA global model. Clim.Dyn., 23, 165–176. doi:10.1007/s00382-004-0427-8.

May W, and Roeckner, E., 2001: Atime-slice experiment with the ECHAM4AGCM at high resolution: the impact of hori -zontal resolution on annual mean climatechange. Clim. Dyn., 17, 407–420,doi:10.1007/s003820000112.

McClean, J.L., D.C. Bader, F.O.Bryan, M.E. Maltrud, J.M. Dennis, A.A.Mirin, P.W. Jones, M. Vertenstein, D.P.Ivanova, Y.-Y. Kim, J.S. Boyle, R.L. Jacob, N.Norton, A. Craig, and P.H. Worley, 2010:Aprototype two decade fully coupled fine reso -lution CCSM simulation. J. Climate (submit -ted).

Murray, R., 1996: Explicit genera -tion of orthogonal grids for ocean models. J.Comp. Phys., 126, 251-273.

Navarra A et al., 2008:Atmospheric horizontal resolution affectstropical climate variability in coupled mod -els. J. Climate, 21,730– 750,doi:10.1175/2007JCLI1406.1.

Neale, R.B., J.H. Richter, and M.Jochum, 2008: The impact of convection onENSO: from a delayed oscillator to a seriesof events. J. Climate, 21, 5904-5924.

Pope V, and Stratton R, 2002: Theprocesses governing horizontal resolutionsensitivity in a climate model. Clim. Dyn., 19,211–236. doi:10.1007/s00382-001-0222-8.

Randall, D. A., M. Khairoutdinov,A. Arakawa , and W. Grabowski, 2003:Breaking the cloud-parameterization dead-lock. Bull. Amer. Meteor. Soc., 84,1547–1564.

Shukla, J., R. Hagedorn, M. Miller,T. N. Palmer, B. Hoskins, J. Kinter, J.Marotzke, J. Slingo, 2009: Strategies:Revolution in climate prediction is both nec -essary and possible: A declaration at theWorld Modelling Summit for ClimatePrediction. Bull. Amer. Meteor. Soc., 90,175–178.

Wu, R., and B. P. Kirtman, 2007:Regimes of local air-sea interactions andimplications for performance of forced simu -lations. Climate Dynamics, 29, 393-410.

Wu, R., B. P. Kirtman, and K.Pegion, 2007: Surface latent heat flux and itsrelationship with sea surface temperature inthe National Centers for EnvironmentalPrediction Climate Forecast System simula -tions and retrospective forecasts. Geophys.Res. Lett., 34, L17712,doi:10.1029/2007GL030751.

Danabasoglu G, Oleson KW, Bitz C,Truesdale JE (2006) CCSM–CAM3 climatesimulation sensitivity to changes in horizon -tal reslution. J. Climate, 19, 2267–2289, doi:10.1175/JCLI3764.1.

Hurrell, J., G. A. Meehl, D. Bader,T. L. Delworth, B. Kirtman, B. Wielicki,2009: A unified modeling approach to cli -mate system prediction. Bull. Amer. Meteor.Soc., 90, 1819–1832.

Jochum M, Danabasoglu G,Holland MM, Kwon YO, Large WG (2008)Ocean viscosity and climate. J. Geophys.Res., 113, doi:10.1029/2007JC004515.

Kinter III, J. L., B. Cash, D.Achuthavarier, J. Adams, E. Altshuler, P.Dirmeyer, B. Doty, B. Huang, L. Marx, J.Manganello, C. Stan, T. Wakefield, E. Jin, T.Palmer, M. Hamrud, T. Jung, M. Miller, P.Towers, N. Wedi, M. Satoh, H. Tomita, C.Kodama, T. Nasuno, K. Oouchi, Y. Yamada,H. Taniguchi, P. Andrews, T. Baer, M. Ezell,C. Halloy, D. John, B. Loftis, R. Mohr, andK. Wong, 2011: Revolutionizing climate mod -eling – Project Athena: A multi-institutional,international collaboration. Bull. Amer.Meteor. Soc. (submitted).

Kirtman, B.P., and G. Vecchi,2010: Why Climate Modelers Should WorryAbout the Weather. WMO Monograph: TheGlobal Monsoon System: Research andForecast, 2nd Ed.

Figure 2.(a) SSTA monthly mean standard deviation ratio

(HRC06/LRC) - dimensionless. Values greater than 1.0 indicate

more variance in HRC06. (b) SSTA monthly mean standard

deviation near the equator (1S-1N) in the Pacific for LRC (red)

and HRC06 (blue) in degrees Celsius. Observational estimates

are given in green. (c) SSTA high pass filtered SSTA standard

deviation near the equator (1S-1N) in the Pacific for LRC (red)

and HRC06 (blue).

Seasonal Prediction.Bull. Amer. Meteor.Soc., 91, 1397–1406.

Bryan, F.O., R. Tomas, J. M.Dennis, D. B.Chelton, N. G. Loeb,J. L. McClean, 2010:Frontal scale air–seainteraction in high-resolution coupled cli -mate models. J.Climate, 23,6277–6291. doi:10.1175/2010JCLI3665.1

Collins, M.,and co-authors, 2010:The impact of globalwarming on the tropi -cal Pacific Ocean andEl Nino. NatureGeoscience, 3, 391-397.

Gent, P.R.,S.G. Yeager, R.B.Neale, S. Levis, andD.A. Bailey, 2009:Improvements in ahalf degree atmos -phere/land version ofthe CCSM. Clim.Dynamics, 34, 819-833.

Hack JJ,Caron JM,

V9N2 8/8/11 10:47 AM Page 4

VARIATIONS

Page 5

Viewed from space, the sub-tropical stratocumulus decksappear as bright sheets ofcloud anchored to the west

coasts of continents sweeping towardthe deeper tropics. Closer inspectionshows that these large climatologicalfeatures consist of thin clouds lyingatop a well-mixed atmospheric bound-ary layer just a few hundred metersabove the surface. These characteristics– large areal coverage, low altitude, andhigh reflectivity – make the subtropicalstratocumulus decks conspicuous in theglobal energy budget (e.g., Ramanathanet al., 1989). Because of their connec-tion to the energy budget, stratocumu-lus decks have been the target of manyobservational campaigns; 20th Centuryinvestigations focused mainly on stra-tocumulus in the northern hemisphere.The last decade witnessed increasedinterest in, and relatively sustainedobservations of the southeast Pacificstratocumulus deck (n.b., EPIC(Bretherton et al., 2004) and VOCALS-REx (Wood et al., 2011)).

Compared to the southeast Pacific,the southeast Atlantic stratocumulusdeck has remained obscure. While itsexistence has long been recognized(e.g., Schubert et al., 1979), includingits large extent and seasonality(Hanson, 1991; Klein and Hartmann,1993), there are limited in situ observa-tions, few satellite-based studies, and ahandful of modeling studies of this fea-ture. This apparent lack of interest issurprising because the tropical Atlantic,especially near the coast of southwest-ern Africa, has been noted for large,persistent sea-surface temperature bias-es in climate models. Cloud-SST feed-back has been a favored explanation forthe biases (Wahl et al., 2011), so under-

standing the characteristics of the clouddeck and representing these clouds inclimate models should be a priority forthe climate modeling community.

The latest version of the NCAR cli-mate model, the Community EarthSystem Model (CESM), containsimprovements across the componentmodels. Among these are major changesin the parameterized physical processesin the Community Atmosphere Model,Version 5 (CAM5), many of whichdirectly affect the representation ofclouds. Despite the changes, coupledsimulations continue to exhibit tempera-ture biases near eastern boundary cur-rents, particularly in the southeastAtlantic.

Inspired by recent evaluations ofsoutheast Pacific stratocumulus in mod-els (Wyant et al., 2010; Hannay et al.,2009), we revisited the southeast Pacificwith two versions of CAM recentlyreleased: CAM41 which closely resem-bles previous versions, and CAM52

which incorporates many changes intro-duced with the CESM. The results showthat the updated physics produce morerealistic stratocumulus, but some impor-tant deviations from observationsremain that could impact climate simu-lations (Medeiros et al., 2011). As anextension of that work, we performedsimulations with CAM5, describedbelow, of the southeast Atlantic stra-tocumulus deck. Because observationsare more limited for the southeastAtlantic, a validation of the model isdifficult, but it is interesting to compareand contrast the two major southernhemisphere stratocumulus decks as rep-resented by CAM5.

To focus on the parameterizedphysics associated with the subtropicalstratocumulus decks, we use a short-term forecast framework that mimicsthe forecast/analysis cycle of numerical

weather prediction (Phillips et al., 2004;Boyle and Klein, 2010). The principle issimple: perform short forecasts with aclimate model by starting from anobserved atmospheric state. The benefitis that the large-scale circulation adjustsaway from the realistic conditions slow-ly, while the fast processes (i.e., para-meterized physics) act quickly, so fore-cast errors are tied to parameterizationerrors. This allows a close inspection ofparameterized processes within a realis-tic large-scale setting, but precludesinvestigating the model’s climatologicalbehavior (e.g., seasonality or interannu-al variability). Here we apply thisframework to produce 5-day forecastsfrom CAM5 with 3-hourly output savedfor both the southeast Pacific and thesoutheast Atlantic. The forecasts coverOctober 2006, as October is climatolog-ically the month during which the stra-tocumulus decks reach their maximumspatial extent. A new forecast beginseach day at midnight UTC, startingfrom the atmospheric state derived fromECMWF operational analyses (fordetails, see Medeiros et al., 2011).

Figure 1 summarizes the environ-mental conditions simulated for October2006. The figure composites the fore-casts by averaging all forecasts overdays 2-5, which avoids any initializa-tion artifacts that might occur in thefirst hours. Both regions are stronglyinfluenced by subtropical anticyclones,with southerly and southeasterly windsblowing along the coast as the circula-tion rounds the eastern flank of the highpressure centers. These winds induceEkman transport and coastal upwellingthat help to keep the sea-surface cool,though in these forecasts surface tem-perature is set to the observed values.The streamlines cross isolines of sea-surface temperature traveling towardthe trades, signaling low-level cold

Comparing the Southern Hemisphere Stratocumulus Decks in the Community Atmosphere Model

Brian Medeiros, National Center for Atmospheric Research, Boulder, Colorado

1 http://www.cesm.ucar.edu/models/ccsm4.0/cam/2 http://www.cesm.ucar.edu/models/cesm1.0/cam/

V9N2 8/8/11 10:47 AM Page 5

U.S. CLIVAR

Page 6

advection. Above the atmosphericboundary layer, large-scale sinkingmotion delivers warm, dry air to thelower troposphere. Where this warm,dry air meets the cool, moist boundarylayer air the trade wind inversion isfound. The inversion is sharp in thestratocumulus regions, occurring in lessthan 100 m (e.g., Fedorovich et al.,2004), and cloud amount is correlatedwith inversion strength on seasonaltimescales (Klein and Hartmann, 1993).Simulating the inversion is a challengefor climate models because of theircoarse vertical resolution; the CAM5 istypical of these models with 30 verticallevels, 8 of which are below 800 hPa.Even so, the model captures some sem-blance of the inversion. A crude proxyfor the inversion strength is the so-called lower-tropospheric stability (thedifference between the 700 hPa and sur-face potential temperatures), shown bythe colored field in Figure 1.

The simulated cloud fraction andliquid water path are shown in Figure 2.Comparing Figures 1 and 2, the corre-spondence between lower-troposphericstability and cloud amount is evident. Inboth regions, the maximum cloud coveris displaced downstream from the maxi-mum stability. This displacement maybe due local effects near the coast, but

warrants further investigation.Compared to satellite observations ofOctober 2006, the cloud distributions forboth regions appear realistic, though lesscloudy than some estimates. This lowbias has been typical for climate modelsfor many years. Related to this bias, inthe southeast Pacific the cloud deck dis-sipates too close to the continent.

The large-scale conditions appearfairly similar for the two regions, andthe cloud patterns are roughly commen-surate with those conditions and resem-ble observations from the same time.

The vertical structure of the clouds,however, contains some subtle differ-ences. An example is shown by Figure3, which shows the composite forecast(averaging all forecasts provides anestimate of the monthly mean for eachforecast time) of cloud fraction andcloud liquid water for a single gridpoint in each region. The southeastPacific point shows the location nearesta moored buoy that has been the focusof substantial study (e.g., de Szoeke etal., 2010). Without a similar target forthe southeast Atlantic, a point was cho-sen with a similar mean (vertically inte-grated) low-cloud fraction (printed inthe figure). The figure illustrates twodifferences between the regions. First,the southeast Pacific has a larger diur-nal cycle of cloud cover. Second, thecloud-base, and to a lesser extent cloud-top, is less well defined (i.e., gradientsare less sharp) in the southeast Atlantic.While the figure shows averages forjust two individual points, these charac-teristics appear to be robust within thestratocumulus decks.

Both stratocumulus regions show aclear diurnal cycle, with maximumcloud cover during night and earlymorning, consistent with observations.Capturing the correct diurnal behavioris crucial for climate simulationsbecause these clouds impact the globalenergy balance through their shortwaveeffects. Underestimating the cloud frac-

Figure 1. Streamlines at 850 hPa for CAM5 forecasts of the (left) southeast Pacific

and (right) southeast Atlantic for October 2006. The color contours show average

lower-tropospheric stability. Blue contour lines show the prescribed sea-surface tem-

perature (K).

Figure 2. Average low-cloud fraction (contour lines, %) and liquid water path

(color, kg m2) simulated by CAM5 for October 2006.

V9N2 8/8/11 10:47 AM Page 6

VARIATIONS

Page 7

tion during the daytime undercuts theclimatic impact of the clouds, and couldlead to positive cloud-sea surface tem-perature feedbacks. Medeiros et al.(2011) suggest that the daytime breakupof the southeast Pacific cloud deck isexaggerated by CAM5. As shortwaveheating in the cloud layer compensateslongwave cooling, the cloud layer andsubcloud layer become decoupled, cut-ting off the cloud layer from its moisturesource. Along with the decoupling, shal-low cumulus convection rises throughthe cloud layer and entrains warm, dryair that enhances cloud evaporation.Such decoupling is common in thesoutheast Pacific, but the cloud dissipa-tion in CAM5 is excessive, weakens thecloud radiative effect, and could lead toclimate biases.

The southeast Atlantic clouds exhibita smaller amplitude diurnal cycle thanthe southeast Pacific, and the differenceis in the daytime minimum. The cloud-top appears lower for the Atlantic stra-tocumulus, which is consistent withsatellite observations (Zuidema et al.,2009). Cloud-base also appears lower.The boundary layer depth is similarbetween the regions, though the liftingcondensation level (LCL) is lower andmore variable in the southeast Atlantic.The variability in LCL explains why thecloud-base level is not distinct in theregion. The difference between theboundary layer depth and the LCLapproximates the thickness of the cloudlayer or (when the difference is small)decoupling. Using this diagnostic,decoupling is less frequent in the south-east Atlantic than southeast Pacific.Since the cloud layer breaks up moreeasily once decoupled, this differencebetween the regions probably accountsfor the difference in daytime cloud frac-tion and might explain the difference inthe size of the stratocumulus decks dur-ing this month. The less frequent decou-pling in the Atlantic might imply somesoutheast Pacific biases are regionallyspecific.

This comparison of the southernhemisphere stratocumulus decks raisessome interesting questions. First amongthese is: given the limited temporal

the climatic relevance of stratocumulusdecks.

AcknowledgementThe author thanks ECMWF for provid-

ing the analysis data, and D. L. Williamson,J. E. Kay, and Y. Zhang for their thoughtfulcomments. This work was supported by theOffice of Science (BER), U.S. Departmentof Energy, Cooperative Agreement DE-FC02-97ER62402. NCAR is sponsored bythe National Science Foundation.

ReferencesBoyle, J. and S. Klein, 2010: Impact of

horizontal resolution on climate model fore -casts of tropical precipitation and diabaticheating for the TWP-ICE period. J.Geophys. Res., 115 (D23), D23113, doi:10.1029/2010JD014262.

Bretherton, C. S., et al., 2004: The EPIC2001 stratocumulus study. Bull. Amer.Meteor. Soc., 85 (7), 967–977,doi:10.1175/BAMS-85-7-967.

de Szoeke, S. P., C. W. Fairall, D. E.Wolfe, L. Bariteau, and P. Zuidema, 2010:Surface flux observations on the southeast -ern tropical Pacific Ocean and attributionof SST errors in coupled ocean–atmospheremodels. J. Climate, 23 (15), 4152–4174,doi:10.1175/2010JCLI3411.1.

Fedorovich, E., R. Rotunno, B. Stevens,and D. Lilly, (Eds.), 2004: Atmospheric tur -bulence and mesoscale meteorology: scien -tific research inspired by Doug Lilly.Cambridge University Press.

Hannay, C., D. L. Williamson, J. J.Hack, J. T. Kiehl, J. G. Olson, S. A. Klein,C. S. Bretherton, and M. Köhler, 2009:Evaluation of forecasted southeast Pacificstratocumulus in the NCAR, GFDL, and

Figure 3. Composite forecasts at two grid points, given

at top of panels for the southeast Pacific (upper) and

southeast Atlantic (lower). The colored contours show

the grid-box average liquid water content, and the black

contours show the cloud fraction. Time is in UTC, intro-

ducing a temporal shift between the panels.

scope of the forecasts,how do these stratocumu-lus decks vary on longertimescales, both in themodel and in reality?That answer could help toaddress a second ques-tion: is the less frequentdecoupling, and smallerdiurnal cycle, in thesoutheast Atlantic a per-sistent difference? Theseissues can be examined inthe model, but it is notclear if suitable observa-tions exist to compare theregions. If these differ-ences are typical, what isthe cause, given the simi-larity in the large-scaleenvironment? Are precip-itation or aerosol process-es important, or doregional influences (e.g.,nearby topography) intro-duce idiosyncrasies to thesubtropical stratocumulusdecks? Answering suchquestions will requiredetailed in situ observa-tions, comprehensivesatellite data sets, andfocused modeling studies,and the answers may pro-vide new insight about

V9N2 8/8/11 10:47 AM Page 7

U.S. CLIVAR

Page 8

ECMWF models. J. Climate, 22 (11),2871–2889, doi:10.11752F2008JCLI2479.1.

Hanson, H. P., 1991: Marine stratocu -mulus climatologies. International Journalof Climatology, 11 (2), 147–164,doi:10.1002/joc.3370110204.

Klein, S. A. and D. L. Hartmann, 1993:The seasonal cycle of low stratiform clouds.J. Climate, 6, 1587–1606.

Medeiros, B., D. L. Williamson, C.Hannay, and J. G. Olson, 2011: SoutheastPacific stratocumulus in the CommunityAtmosphere Model (in preparation).

Phillips, T. J., et al., 2004: Evaluatingparameterizations in general circulationmodels: Climate simulation meets weatherprediction. Bull. Amer. Meteor. Soc., 85 (12),1903–1915, doi:10.1175/BAMS-85-12-1903.

Ramanathan, V., B. R. Barkstrom, and E.F. Harrison, 1989: Climate and the Earth’sradiation budget. Physics Today, 42 (5),22–32, doi:10.1063/1.881167.

Schubert, W. H., J. S. Wakefield, E. J.Steiner, and S. K. Cox, 1979: Marine stra -tocumulus convection. Part I: Governingequations and horizontally homogeneoussolutions. J. Atmos. Sci., 36, 1286–1307.

Wahl, S., M. Latif, W. Park, and N.Keenlyside, 2011: On the tropical AtlanticSST warm bias in the Kiel climate model.Climate Dynamics, 36, 891–906,doi:10.1007/s00382-009-0690-9.

Wood, R., et al., 2011: The VAMOSOcean-Cloud-Atmosphere-Land StudyRegional Experiment (VOCALS- REx):goals, platforms, and field operations.Atmos. Chem. Phys., 11 (2), 627–654,doi:10.5194/acp-11-627-2011.

Wyant, M. C., et al., 2010: ThePreVOCA experiment: modeling the lowertroposphere in the Southeast Pacific. Atmos.Chem. Phys., 10 (10), 4757–4774,doi:10.5194/acp-10-4757-2010.

Zuidema, P., D. Painemal, S. de Szoeke,and C. Fairall, 2009: Stratocumulus cloud-top height estimates and their climatic impli -cations. J. Climate, 22 (17), 4652–4666,doi:10.1175/2009JCLI2708.1.

CLIVAR Working Group on Seasonal toInterannual Prediction12-14 September 2011Trieste, ItalyAttendance: Invitedhttp://www.clivar.org/organization/wgsip/wgsip14/wgsip14.php

Land-Ocean Interactions in the CoastalZone (LOICZ) Conference12-15 September 2011Yantai, ChinaAttendance: Openhttp://www.loicz-osc2011.org/

ICES Annual Science Conference19-23 September 2011Gdansk, PolandAttendance: Openhttp://www.ices.dk/iceswork/asc/2011/index.asp

3rd Symposium on the Ocean in aHigh-CO2 World24-27 September 2011Monterey, CaliforniaAttendance: Openhttp://www.highco2_iii.org/main.cfm?cid=2259&nid=14766

SAMOC4 Meeting27-29 September 2011Simons Town, South AfricaAttendance: Openhttp://www.aoml.noaa.gov/phod/SAMOC

CLIVAR Working Group on CoupledModeling19-21 October 2011Boulder, COAttendance: Invitedhttp://www.clivar.org/wgcm/wgcm-15/wgcm15.php

CLIVAR/CliC/SCAR Southern OceanPanel Meeting19-21 October 2011Boulder, COAttendance: Invitedhttp://www.clivar.org/organization/south-ern/SOP7_meet.php

Interdisciplinary Climate ChangeResearch Symposium22-29 October 2011Colorado Springs, COAttendance: Limitedhttp://disccrs.org/

WCRP Open Science Conference24-28 October 2011Denver, COAttendance: Openhttp://www.wcrp-climate.org/confer-ence2011/

International Conference on Energyand Meteorology8-11 November 2011Gold Coast, AustraliaAttendance: Openhttp://www.icem2011.org/

Earth Observation for Ocean-Atmosphere Interactions Science29 November - 2 December 2011Frascati, ItalyAttendance: Openhttp://www.eo4oceanatmosphere.info/

AGU Fall Meeting5-9 December 2011San Francisco, CAAttendance: Openhttp://sites.agu.org/fallmeeting/

Calendar of CLIVAR and CLIVAR-relatedmeetings

Further details are available on the U.S. CLIVAR and International CLIVAR websites: www.usclivar.org and www.clivar.org

V9N2 8/8/11 10:47 AM Page 8

VARIATIONS

Page 9

Coupled atmosphere-oceangeneral circulation models(AOGCMs) have long beenplagued by biases in the tropi-

cal Atlantic. Observed SSTs in the east-ern equatorial Atlantic cool by ~5°Cfrom May to August as the Atlantic coldtongue forms, however, AOGCMs with-out flux correction fail to capture this(Davey et al. 2002; Breugem et al.2006; Richter and Xie 2008). The coldtongue development is sensitive to thecross-equatorial southerly flow of theWest African monsoon (Philander andPacanowski 1981; Mitchell and Wallace1992), and the monsoon is in turn influ-enced by regional SSTs (Lamb 1978;Ward 1998; Vizy and Cook 2002).Thus, the inability of coupled models torealistically represent the tropicalAtlantic compounds the uncertainty offuture climate change simulations.

The typical tropical Atlantic warmSST bias covers two regions – the east-ern equatorial Atlantic (EEA) and south-eastern tropical Atlantic (SETA) – andseveral distinct bias mechanisms havebeen proposed, including interactionsbetween the two. Modeling studies sug-gest that an under-representation of low-level clouds and the associated exces-sive surface shortwave radiation (Huanget al. 2007; Wahl et al. 2011), an under-representation of coastal upwelling(Large and Danabasoglu 2006), and sen-sitivity to entrainment efficiency at thebase of the ocean mixed layer(Hazeleger and Haarsma 2005) con-tribute to SST bias in the SETA. Xu etal. (2011) recently find that warm sub-surface temperature bias in the EEA canbe advected south by sub-surface cur-rents and upwelled along the Benguelacoast, contributing to the warm SSTbias in the coastal upwelling zone.

SST bias by unrealistically deepeningthe thermocline within the newly devel-oped Texas A&M University high-reso-lution coupled regional climate model(TAMU_CRCM). We also discuss therole of atmosphere-ocean feedbacks inamplifying the bias.

The atmospheric component of theTAMU_CRCM is the Weather Researchand Forecasting model (WRF;Skamarock et al. 2008) Version 3.1.1,which is configured with 27 km hori-zontal resolution and 28 vertical levelson a domain covering 110°W – 27°Eand 46°S – 61°N. The ocean is repre-sented by the Regional Ocean ModelingSystem (ROMS; http://www.myroms.org/), configured with 9 km horizontalresolution and 30 vertical levels over98°W – 22°E and 33°S – 52°N. SST,precipitation minus evaporation, surfacewind, and latent heat, sensible heat, andradiative fluxes are exchanged everyhour. Both models use alignedArakawa C grids so variables are notinterpolated during coupling. Boundaryconditions are based on the 0.5° resolu-tion 6-hourly 2001 – 2008 climatologyof the National Center forEnvironmental Prediction ClimateForecast System Reanalysis (NCEPCFSR; Saha et al. 2010), except forSSTs in the uncoupled simulations,which are prescribed from the 0.25° res-olution NOAA Optimum InterpolationSST V2 (Reynolds et al. 2007) daily2001 – 2008 climatology. Initializingthe coupled model with the CFSR SST,which is damped to the Reynolds SST,introduces relatively little bias.

Extensive numerical experimentswere carried out to test the sensitivity ofthe model mean climate to physicalparameterizations. Here we highlighttwo uncoupled atmosphere-only and

Several studies demonstrate that theEEA SST bias is connected to a spring-time westerly equatorial trade wind bias(DeWitt 2005) that erroneously deepensthe thermocline, inhibiting the summercold tongue development (Chang et al.2007; Richter and Xie 2008; Richter etal. 2011; Wahl et al. 2011). UncoupledAGCM simulations with observed SSTsalso simulate the trade wind bias, butrelatively weakly, suggesting that itoriginates in atmospheric models and isamplified by the Bjerknes feedback(Bjerknes 1969) in coupled models.These studies link the trade wind bias toa zonal pressure gradient driven by defi-cient (excessive) rainfall over theAmazon (tropical Africa), althoughChang et al. (2008) find the Amazon isthe primary driver.

The formation of spurious oceanicbarrier layers (BLs), which reduceentrainment and turbulent mixing ofcold water at depth into the mixed layer,may also contribute to the warm EEASST bias (Breugem et al. 2008). Asouthward bias of the intertropical con-vergence zone (ITZC) is common inatmospheric GCMs (Biasutti et al.2006), and in coupled models it may ini-tiate a positive ITCZ – BL – SST feed-back, in which the freshwater flux of thedisplaced ITCZ reduces the ocean sur-face salinity, supporting BL formationand warmer SSTs, further anchoring theITCZ. However, unrealistic BLs are notin all coupled simulations (Wahl et al.2011; Richter et al. 2011), suggestingthey may be a secondary error source(Balaguru et al. 2011).

Here we test the hypothesis that aspringtime westerly trade wind bias,which is linked to deficient (excessive)Amazon (Congo basin) precipitation,contributes to the warm summer EEA

An Investigation of the Tropical Atlantic Bias Problem Using aHigh-Resolution Coupled Regional Climate Model

Christina M. Patricola1, Ping Chang2, R. Saravanan1, Mingkui Li3, and Jen-Shan Hsieh1

1Department of Atmospheric Sciences, Texas A&M University, College Station, Texas2Department of Oceanography, Texas A&M University, College Station, Texas

3Ocean University of China, Qingdao, China

V9N2 8/8/11 10:47 AM Page 9

U.S. CLIVAR

Page 10

western Atlantic basin in w27_dry.These biases in w27r9_dry, as well asmajor biases in the large-scale circula-tion throughout the troposphere (notshown) are related to an over-represen-tation of low-level clouds over the

Amazon that leads to a cold land-sur-face temperature bias and descendingmid-tropospheric vertical motion. Thetrade wind and Amazon precipitationbiases of w27_dry are similar to thosesimulated by most uncoupled and cou-pled GCMs discussed in Richter andXie (2008). The rainfall and wind bias-es are similar between the uncoupledand coupled regional model experiments(except for a stronger westerly windbias of 1 – 2 m/s along the west coast ofequatorial Africa in w27r9_wet (Fig. 1c)compared to w27_wet) indicating that,as in many GCMs, the precipitation andtrade wind bias in TAMU_CRCM origi-nates in the atmospheric component.

Although the westerly springtimeequatorial trade wind bias in w27r9_dry(Fig. 1d) is of a stronger magnitude andspans a greater portion of the Atlanticbasin than that of w27r9_wet (Fig. 1c),the warm summer EEA SST bias ismore severe in w27r9_wet. The JJAaveraged SST and 10-m wind bias ofw27r9_wet and w27r9_dry are shown inFigure 2a and b, respectively. Inw27r9_wet there is a cold SST bias of 1– 2°C in the western equatorial Atlanticand a warm bias of 4 – 5°C concentrat-

two coupled atmosphere-ocean experi-ments (run May 1 – October 1) conduct-ed with atmospheric parameterizationschosen to test the hypothesis that tradewind and precipitation biases drive thesummer EEA SST bias. An uncoupled(w27_wet) and a coupled (w27r9_wet)simulation are run with longwave radia-tion, shortwave radiation, and convec-tion represented by the Rapid RadiativeTransfer Model (RRTM), Goddard, andBetts-Miller-Janjic scheme, respectively.An additional uncoupled (w27_dry) andcoupled (w27r9_dry) set of simulationsis run with the CAM longwave andshortwave radiation and Kain-Fritschconvection. Figure 1 shows the precipi-tation and 10-m wind bias in the firstmonth of integration. The w27_wetsimulation (Fig. 1a) produces a wet bias– in some parts over 20 mm/day – overthe Amazon and Congo basin with aweak (1 – 2 m/s) westerly trade windbias in the EEA. The w27_dry experi-ment (Fig. 1b) simulates a rainfall deficit(over 10 mm/day) over the Amazon withweaker drying over equatorial Africa andan erroneous rainfall band along thenortheastern South American coast.There is also a considerable westerlytrade wind bias (over 10 m/s) in the

Figure 1. Bias in 10-m wind (m/s, vector) over the ocean and precipitation (mm/day,

shaded) in May from the (a) w27_wet, (b) w27_dry, (c) w27r9_wet, and (d)

w27r9_dry simulations. Precipitation bias is relative to the 0.25° Tropical Rainfall

Measuring Mission (TRMM) 3B43V6 (Huffman et al. 2007) 2001 – 2008 climatology

and the 10-m wind bias is relative to the NCEP CFSR 2001 – 2008 climatology.

Figure 2. Bias in SST (°C, shaded) relative to the 2001 – 2008 Reynolds climatol-

ogy and 10-m wind (m/s, vector) over the ocean relative to the NCEP CFSR 2001 –

2008 climatology during JJA from the (a) w27r9_wet and (b) w27r9_dry simulations.

(c) Thermocline (20° isotherm) depth (m) during JJA averaged 2°S – 2°N from the

NCEP CFSR 2001 – 2008 climatology (black), w27r9_wet (blue), and w27r9_dry

(red). (d) The zonal wind at 10-m (m/s) over the ocean during JJA averaged 2°S –

2°N from the NCEP CFSR 2001 – 2008 climatology (black), w27r9_wet (solid blue),

w27r9_dry (solid red), w27_wet (dash blue), and w27_dry (dash red).

V9N2 8/8/11 10:47 AM Page 10

VARIATIONS

Page 11

ed along the west coast of equatorialAfrica that extends west to 5°W anddecreases to 2 – 3°C. The warm biasalso extends into the SETA. The tropi-cal Atlantic SST bias in w27r9_wetresembles that of typical AOGCMs,suggesting that simply increasing modelresolution is unlikely to solve this biasproblem. Over the EEA there is a west-erly surface wind bias of 2 – 3 m/s thatis strongest along the coast. The EEAwarm SST bias peaks in July at up to5°C along the coast, while the SETAbias becomes progressively worsethroughout the integration (not shown).The warm bias in both regions is con-fined to the coast during the initialdevelopment and spreads west in thefollowing months. The w27r9_dry sim-ulation (Fig. 2b) produces a cool SSTbias in the western equatorial Atlanticsimilar to that of w27r9_wet, but unlikew27r9_wet simulates a weak bias (+/-1°C) in the EEA and a cold bias in theSETA. The westerly wind bias in thewestern Atlantic persists into the sum-mer in w27r9_dry. This simulation alsoexhibits a weak easterly bias in the EEAand a strong northerly wind biasbetween the equator and 10°N associat-ed with a southward displacement of theITCZ. The stronger warm EEA SSTbias in w27r9_wet suggests that in thisregional coupled model the bias is moresensitive to the local rather than basin-wide equatorial trade wind bias, and toexcessive rainfall in the Congo basinrather than deficient Amazon rainfall.

The response of the thermocline,represented as the 20°C isotherm depth,supports the idea that the EEA SST biasis more sensitive to local trade winds.Figure 2c shows the thermocline aver-aged from 2°S – 2°N in JJA from theCFSR reanalysis and the two coupledsimulations. There is a steep zonal gra-dient in the reanalyzed thermocline,which is deep in the west and shallow inthe east. To the west of 10°W,w27r9_wet and w27r9_dry simulate arelatively small (5 – 10 m) thermoclinebias, whereas to the east w27r9_dry(w27r9_wet) produces a thermoclinethat is too deep by 20 m (over 30 m).In w27r9_wet, the bias in thermoclinedepth is so large that the zonal gradient

is incorrect. The easterly wind bias inthe EEA in w27r9_dry may partly coun-teract the thermocline bias driven by thealmost basin-wide westerly bias. Whilethe basin-wide wind bias in w27r9_drydoes induce a thermocline bias in theEEA, it is not as severe as the bias driv-en by the local winds in w27r9_wet(Fig. 1c and 2a).

The SST and thermocline responsein w27r9_wet, together with the differ-ence in the 10-m zonal wind biasbetween w27r9_wet and w27_wet, sug-gest that the Bjerknes feedback isimportant on a local scale in supportingthe EEA SST bias. Figure 2d shows the10-m zonal wind averaged 2°S – 2°Nduring JJA from the CFSR reanalysisand the coupled and uncoupled simula-tions. In both the “wet” and “dry”experiments, the trade wind bias isamplified by coupling, similar to typicalAGCM and AOGCM simulations. Inthe EEA there is a slight (less than 1m/s) westerly bias in w27_wet, that isamplified (over 2 m/s) and strongestalong the west coast of Africa in thecorresponding coupled simulation. Inthe “dry” case, the coupling worsensthe westerly wind bias in the center ofthe Atlantic rather than EEA.

Figure 3a – d shows the precipita-tion bias for the uncoupled and coupledsimulations in JJA. The w27_dry (Fig.

3b) and w27r9_dry (Fig. 3d) simulationsboth produce a dry bias (over 10mm/day) over the Amazon, a weakerdry bias over equatorial Africa, and –similar to many GCMs – a southwardshift of the ITCZ that is amplified in thecoupled simulation. In both w27_wet(Fig. 3a) and w27r9_wet (Fig. 3c), thelatitudinal position of the ITCZ resem-bles the observed, but rainfall rates aretoo high. There is also a wet (dry) biasover the western (eastern) Amazon anda strong wet bias of up to 20 mm/dayover equatorial Africa. In w27r9_wetthere is also a wet precipitation bias ofover 20 mm/day concurrent with thewarm EEA SST bias that does not existin the corresponding uncoupled simula-tion. This would suggest a rainfall – BL– SST feedback. However, erroneousBLs are simulated in both w27r9_wet(Fig. 3e) and w27r9_dry (Fig. 3f), indi-cating that the role of barrier layers maybe secondary to the wind-driven bias.Coupled simulations with corrected pre-cipitation minus evaporation flux areplanned to better understand this prob-lem.Acknowledgements

This work is supported by DOEgrant # DE-SC0004966 and by NOAAgrant # NA09OAR4310135.Simulations were run at the Texas A&MSupercomputing Facility. The authors

Figure 3. Bias in pre-

cipitation (mm/day,

shaded) in JJA rela-

tive to the TRMM

2001 – 2008 climatol-

ogy from the (a)

w27_wet,

(b) w27_dry,

(c) w27r9_wet, and

(d) w27r9_dry simu-

lations. Barrier layer

thickness (m, com-

puted as in Breugem

et al. 2008) in JJA

from (e and f) the

CFSR 2001 – 2008

climatology (con-

toured every 10 m)

and (e) w27r9_wet

(shaded) and (f)

w27r9_dry (shaded).

V9N2 8/8/11 10:47 AM Page 11

thank Howard Seidel for his initial con-tribution to the development ofTAMU_CRCM.

ReferencesBalaguru, K., P. Chang, R. Saravanan,

and C. J. Jang, 2011: The Barrier Layer ofthe Atlantic warmpool: Formation mecha -nism and influence on the mean climate.Tellus (submitted).

Biasutti, M., A. H. Sobel, and Y. Kushnir,2006: AGCM Precipitation biases in thetropical Atlantic. J. Climate, 19, 935 – 925.

Bjerknes, J., 1969: Atmospheric telecon -nections from the Equatorial Pacific.Monthly Weather Review, 97, 163 – 172.

Breugem, W. –P., W. Hazeleger, and R. J.Haarsma, 2006: Multimodel study of tropicalAtlantic variability and change. Geophy.Res. Lett., 33, doi: 10.1029/2006GL027831.

Breugem, W. –P., P. Chang, C. J. Jang, J.Mignot, and W. Hazeleger, 2008: Barrierlayers and tropical Atlantic SST biases incoupled GCMs. Tellus, 60A, 885 – 897.

Chang, C. -Y., J. A. Carton, S. A.Grodsky, and S. Nigam, 2007: Seasonal cli -mate of the tropical Atlantic sector in theNCAR Community Climate System Model 3:Error structure and probable causes ofcrrors. J. Climate, 20, 1053 – 1070.

Chang, C. -Y., S. Nigam, and J. A.Carton, 2008: Origin of the springtime west -erly bias in Equatorial Atlantic surfacewinds in the Community Atmosphere ModelVersion 3 (CAM3) simulation. J. Climate, 21,4766 – 4778.

Davey, K. M., and Coauthors, 2002:STOIC: A study of coupled model climatol -ogy and variability in tropical ocean regions.Clim. Dyn., 18, 403 – 420.

DeWitt, D. G., 2005: Diagnosis of thetropical Atlantic near-equatorial SST bias ina directly coupled atmosphere-ocean generalcirculation model. Geophys. Res. Lett., 32,doi:10.1029/2004GL021707.

Hazeleger, W., and R. J. Harrsma, 2005:Sensitivity of tropical Atlantic climate to mix -ing in a coupled ocean–atmosphere model.Clim. Dyn., 25, 387 – 399.

Huang, B., Z. –Z. Hu, and B. Jha, 2007:Evolution of model systematic errors in thetropical Atlantic basin from coupled climatehindcasts. Clim. Dyn., 28, 661 – 682.

Huffman, G.J., R.F. Adler, D.T. Bolvin,G. Gu, E.J. Nelkin, K.P. Bowman, Y. Hong,E.F. Stocker, D.B. Wolff, 2007: The TRMMmulti-satellite precipitation analysis: Quasi-global, multi-year, combined-sensor precipi -tation estimates at fine scale. J.Hydrometeor., 8, 38-55.

Lamb. P. J., 1978: Case studies of tropi -cal Atlantic surface circulation patterns dur -ing recent sub-Saharan weather anomalies:1967 and 1968. Monthly Weather Review,106, 482 – 491.

Large, W. G., and G. Danabasoglu,2006: Attribution and impacts of upper-ocean biases in CCSM3. J. Climate, 19,2325 – 2346.

Mitchell, T.P. and J.M. Wallace, 1992:The annual cycle in equatorial convectionand sea surface temperature, J. Climate, 5,1140-1156.

Philander, S. G. H., and R. C.Pacanoswki, 1981: The oceanic response tocross-equatorial winds (with application tocoastal upwelling in low latitudes). Tellus,33, 201 – 210.

Reynolds, R. W., T. M. Smith, C. Liu, D.B. Chelton, K. S. Casey, and M. G. Schlax,2007: Daily high-resolution-blended analy -ses for sea surface temperature. J. Climate,20, 5473 – 5496.

Richter, I., and S. –P. Xie, 2008: On theorigin of equatorial Atlantic biases in cou -pled general circulation models. Clim. Dyn.,31, 587 – 598.

Richter, I., S. –P. Xie, A. T. Wittenberg,and Y. Masumoto, 2011: Tropical Atlanticbiases and their relation to surface wind

U.S. CLIVARstress and terrestrial precipitation. Clim.Dyn., doi:10.1007/s00382-011-1038-9.

Saha, S., and Coauthors, 2010: TheNCEP Climate Forecast System Reanalysis.Bull. Amer. Met. Soc., 1015 – 1057.

Skamarock, W. C., and Coauthors, 2008:A description of the Advanced ResearchWRF Version 3. NCAR Tech. Note,NCAR/TN-475+STR, 113 pp.

Vizy, E. K., and K. H. Cook, 2002:Development and application of a mesoscaleclimate model for the tropics: Influence ofsea surface temperature anomalies on theWest African monsoon. J. Geophys. Res.,107, doi: 10.1029/2001JD000686.

Wahl, S., M. Latif, W. Park, and N.Keenlyside, 2001: On the tropical AtlanticSST warm bias in the Kiel Climate Model.Clim. Dyn., 36, 891 – 906.

Ward, N. M., 1998: Diagnosis and short-lead time prediction of summer rainfall intropical North Africa at interannual andmultidecadal timescales. J. Climate, 11,3167 – 3191.

Xu, Z., M. Li, P. Chang, and R.Saravanan, 2011: Oceanic origin of tropicalAtlantic biases (in preparation).

Page 12

Figure1. Observed (a)

North (0o – 60oN) and

(b) South Atlantic (30oS

– 0o) SSTs during the

instrumental period,

obtained from ERSST3.

The green line in each

plot is obtained by per-

forming a 11-year run-

ning average.

Figure associated with article

on the following page

V9N2 8/8/11 10:47 AM Page 12

VARIATIONS

As seen in the extended recon-structed sea surface tempera-ture data (ERSST3; Smith etal., 2008), the North Atlantic

sea surface temperature (SST) duringthe instrumental period (Figure 1a) con-tains both a secular increase at a rate of~ 0.4oC per 100yrs during 1901-2010and a robust multidecadal signal of simi-lar amplitude, known as the Atlanticmultidecadal oscillation (AMO; Kerr,2000). Coupled general circulation mod-els forced by external factors such asgreenhouse gases and solar variations doa poor job of reproducing the AMO inthe 20th century, suggesting that it mustarise instead from internal interactionsof the climate system (e.g., Knight,2009). It has been shown that in anunforced coupled climate model simula-tion the Atlantic meridional overturningcirculation (AMOC) exhibits decadal tomultidecadal (15 - 70 yrs) variations andis closely linked both dynamically andstatistically to the AMO (e.g., Delworthet al., 1993). However, neither the long-term variations of the AMOC nor thelink between the AMOC and AMO hasbeen demonstrated with observations inthe 20th century because historicalobservations of the AMOC are not up tothe task.

Contrary to the North Atlantic SST,the South Atlantic SST during the instru-mental period (Figure 1b - shown left)has increased almost linearly at a rate of~ 0.9oC per 100yrs during 1901-2010,clearly surpassing the global average of0.6 ~ 0.7oC per 100 yrs. Consistent withthis surface trend, recently updated andbias-corrected instrumental records indi-

cate that the heat content of the AtlanticOcean in the upper 700m has substan-tially increased during the 1970s –2000s at a rate (~ 8x1022 J per 40yrs)almost matching that of the PacificOcean (~ 6x1022 J per 40yrs) and IndianOcean (~ 2x1022 J per 40yrs) combined(Levitus et al., 2009), even though theAtlantic Ocean covers less than 20% ofthe global ocean in surface area. As acandidate mechanism for the differentialinter-ocean warming, Lee et al. (2011)pointed out the potential role played bythe global overturning circulation. Theyargued that as the upper ocean warmsglobally during the 20th century, theinter-ocean heat transport associatedwith the global overturning circulationshould increase until the deep oceanfully adjusts to the surface warming orthe global overturning circulation slowsdown. Since the Atlantic Ocean is char-acterized with advective heat conver-gence, they argued that the AtlanticOcean should therefore gain extra heatfrom other oceans. They used a surface-forced global ocean-ice coupled modelto test this hypothesis and to furthershow that the increased AMOC at 30oS,forced by the increased wind stress curlover the region from 50o to 30oS, hascontributed greatly to the increasedinter-ocean heat transport from theIndian Ocean during the latter half ofthe 20th century.

To sum up, it appears that theAMOC and the associated meridionalocean heat transport hold the key to ourunderstanding of the observed secularand multidecadal warming of the Northand South Atlantic Oceans since themid-20th century. Therefore, it is vital to

reconstruct the history of AMOC andassociated meridional ocean heat trans-port in the 20th century, covering atleast one full cycle of the AMO to beable to cleanly distinguish the seculartrend from the multidecadal variations.Here, we attempt an ocean model-basedreconstruction, which is likely to be ourbest chance for assessing the history ofAMOC in the 20th century because theobserved surface flux fields, which con-strain ocean-only (or ocean-ice coupled)models, are available on relatively longtime scales. 20th Century Reanalysis (20CR)

None of the surface-forced oceanmodel studies so far has been simulatedwith the surface forcing prior to themid-20th century because the surfaceforcing data, which are typically derivedfrom atmospheric reanalysis productssuch as NCEP-NCAR reanalysis, arelimited to the last 50 - 60 years.Recently, the newly developed NOAA-CIRES 20th Century Reanalysis (20CR)has been completed (Compo et al.,2010). The 20CR provides the first esti-mate of global surface momentum, heatand freshwater fluxes spanning the late19th century and the entire 20th century(1871-2008) at daily temporal and 2ospatial resolutions. Model Experiments

The global ocean-ice coupled modelof the NCAR Community ClimateSystem Model version 3 (CCSM3)forced with the 20CR is used as the pri-mary tool in this study. The oceanmodel is a level-coordinate model basedon the Parallel Ocean Program (POP),

Page 13

Secular and Multidecadal Warming of the Atlantic Ocean sincethe mid-20th Century

Sang-Ki Lee1,2, David B. Enfield1,2, Wonsun Park3, Erik van Sebille5, Chunzai Wang2, Steven Yeager4,Ben Kirtman5, and Molly Baringer2

1Cooperative Institute for Marine and Atmospheric Studies, University of Miami, Miami, Florida2Atlantic Oceanographic and Meteorological Laboratory, NOAA, Miami, Florida

3Leibniz Institute of Marine Sciences (IFM-GEOMAR), Kiel, Germany4National Center for Atmospheric Research, Boulder, Colorado

5Rosenstiel School for Marine and Atmospheric Science, University of Miami, Miami, Florida

V9N2 8/8/11 10:47 AM Page 13

U.S. CLIVAR

divided into 40 vertical levels. The icemodel, the NCAR Community Sea IceModel version 5, is a dynamic-thermo-dynamic ice model. Both the ocean andice models have 320 longitudes and384 latitudes on a displaced pole gridwith a longitudinal resolution of about1.0 degrees and a variable latitudinalresolution of approximately 0.3 degreesnear the equator. See Doney et al.(2007) for more detailed descriptionsabout the CCSM3 ocean-ice model(CCSM3_POP hereafter).

After the total of 900 years of spinup runs (see Lee et al. (2011) for adetailed description of the spin upruns), three model experiments are per-formed. In the control experiment(EXP_CTR), the CCSM3_POP is inte-grated for 1871-2008 using the real-time daily 20CR surface flux fields.The next two experiments are idealized

EXP_REM, respectively.Model Results

Figure 2a shows the simulated NorthAtlantic Ocean heat content change inthe upper 700m in reference to the 1871- 1900 period obtained from the threemodel experiments, along with theobserved trend of the North AtlanticOcean heat content during 1969 – 2008.The observed trend is referenced at 1969in Figure 2 for a better visual compari-son with the simulated trends. The simu-lated North Atlantic Ocean heat contentin EXP_CTR increases moderately dur-ing the 1930s - 1940s, and then decreas-es during the 1940s - 1970s, after whichit increases substantially much like theAMO from observations (Figure 1a).Between the 1970s and 2000s, itincreases by 3 ~ 4x1022 J. This largeincrease is reasonably close to theobserved North Atlantic Ocean heat con-

tent increase of ~ 5.5x1022 J during thesame period (Levitus et al., 2009), sug-gesting that the model experiment(EXP_CTR) reproduces reasonably wellthe heat budget trend of the NorthAtlantic Ocean after the 1960s.

If the northward heat transport in theSouth Atlantic at 30oS is fixed at its1871-1900 level by considering the fullytransient surface fluxes only north of30oS (EXP_ATL), the simulated NorthAtlantic Ocean heat content increasesduring the 1930s - 1940s, and thendecreases during the 1940s - 1970s,almost perfectly reproducing that ofEXP_CTR prior to the 1970s. However,the North Atlantic Ocean heat content, inthis case, increases by only ~ 1x1022 Jduring the 1970s and 2000s, and thuscannot explain the observed NorthAtlantic Ocean heat content increaseduring the same period.

On the other hand, if the northwardheat transport in the South Atlantic at30oS is allowed to vary in real time byconsidering the fully transient surfacefluxes south of 30oS while keeping thesurface fluxes over the Atlantic Ocean attheir 1871-1900 levels (EXP_REM), theNorth Atlantic Ocean heat contentincreases by ~ 2 x1022 J during the1970s – 2000s explaining a moderateportion of the observed trend. In thiscase, however, the multidecadal signalduring the 1930s – 1970s, which isclearly simulated in both EXP_CTR andEXP_ATL is completely missing. Theabsence of this multidecadal signal inEXP_REM and its presence inEXP_ATL clearly suggest that the multi-decadal swing in EXP_CTR prior to the1970s is caused by processes internal tothe Atlantic Ocean. During the 1970s –2000s, on the other hand, remoteprocesses (i.e., increased inter-oceanheat transport from the Indian Ocean;see Lee et al. (2011)) have contributedmore to the large increase in the NorthAtlantic heat content, although internalprocesses have also contributed.

Figure 2b is the same as Figure 2a,except for the South Atlantic Ocean heatcontent change. The simulated SouthAtlantic Ocean heat content inEXP_CTR remains unchanged until the

Page 14

experiments designed to under-stand the Atlantic Ocean heatcontent change with and with-out the influence of the north-ward heat transport change at30oS. The remote ocean warm-ing experiment (EXP_REM) isidentical to EXP_CTR exceptthat the surface forcing fieldsnorth of 30oS are from thedaily 20CR surface flux fieldsfor the period of 1871-1900exactly like the spin-up experi-ment, whereas those south of30oS are real time as inEXP_CTR. The Atlantic Oceanwarming experiment(EXP_ATL) is also identical toEXP_CTR except that the sur-face forcing fields south of30oS are from the daily 20CRsurface flux fields for the peri-od of 1871-1900 as in the spin-up experiment, whereas thosenorth of 30oS are real time asin EXP_CTR. Note that theAtlantic Ocean warms onlythrough anomalous surfacewarming (i.e., local processes)in EXP_ATL, and only throughanomalous northward oceanheat transport at 30oS (i.e.,remote processes) in

Figure 2. Simulated (a) North and (b) South

Atlantic Ocean heat content changes in the upper

700m in reference to 1871-1900 obtained from the

three model experiments. The thick black lines in

(a) and (b) are the observed trends of the North

and South Atlantic Ocean heat content, respec-

tively, reproduced from Levitus et al. (2009).

V9N2 8/8/11 10:47 AM Page 14

VARIATIONS1960s, after which it increases monoton-ically, consistent with the similarincrease in the South Atlantic SST dur-ing the same period from observations(Figure 1b). Between the 1970s and2000s, it increases by ~ 2 x1022 J,slightly less than the observed SouthAtlantic Ocean heat content increase of~ 2.5x1022 J during the same period(Levitus et al., 2009). As in EXP_CTR,the South Atlantic Ocean heat content inEXP_REM is also characterized with amonotonic increase after the 1960s, butwith a smaller amplitude of 1~ 2x1022 Jduring the 1970s – 2000s. In the case ofEXP_ATL, however, there is no appar-ent change in the South Atlantic heatcontent throughout the 20th century.These results derived from Figure 2blead to a conclusion that remoteprocesses mainly drive the SouthAtlantic Ocean heat content increaseduring the 1970s - 2000s in EXP_CTR.

Figure 3a shows the time-averagedAMOC during 1979-2008 obtained fromEXP_CTR. The simulated maximumstrength of the AMOC at 35oN is only11 Sv (1Sv = 106 m3s-1), which issmaller than the observed range of 14 ~20Sv. Despite the smaller maximumstrength, the overall spatial structure ofthe simulated AMOC is quite close tothat derived from observations (e.g.,Lumpkin and Spear, 2007). Figure 3bshows the time series of the simulatedAMOC index (maximum overturningstream function) at three different lati-tudes, namely 30oS, the equator, and60oN, in reference to the 1871 – 1900period. It is clear that the AMOC at allthree latitudes increase after the 1950s,with the largest amplitude at 30oS, alesser amplitude at the equator, and thesmallest amplitude at 60oN. This sug-gests that both the North (0o – 60oN)and South (30oS – 0o) Atlantic Oceansare subject to advective heat conver-gence during the 1960s – 2000s, consis-tent with the heat content increase in theNorth and South Atlantic Oceans duringthe 1970s – 2000s (Figure 2).

Additional work is needed to under-stand how the AMOC is linked with themultidecadal variations of the NorthAtlantic Ocean heat content prior to the1960s.

AcknowledgmentsThis study was motivated and bene-

fited from the AMOC discussion groupof the research community atUM/RSMAS and NOAA/AOML. Wewish to thank Igor Kamenkovich and allthe participants who led the AMOC dis-cussion group during the past year. Weacknowledge helpful suggestions fromPing Chang. This work was supportedby grants from the National Oceanic andAtmospheric Administration’s ClimateProgram Office and by grants from theNational Science Foundation.References

Compo, G. P., and collaborators, 2011:The twentieth century reanalysis project.Quarterly J. Roy. Meteorol. Soc., 137, 1-28.doi: 10.1002/qj.776.

Delworth, T., S. Manabe, and R. Stouffer,1993: Interdecadal variations of the thermo -haline circulation in a coupled ocean-atmos -phere model. J. Climate, 6, 1993–201.

Doney, S. C., Steve Yeager, G.Danabasoglu, W. G. Large, and J. C.McWilliams, 2007: Mechanisms governinginterannual variability of upper-ocean tem -

perature in a global ocean hindcast simula -tion. J. Phys. Oceanogr., 37, 1918–1938.

Kerr, R. A., 2000: A North Atlantic cli -mate pacemaker for the centuries. Science,288, 1984-1986. doi:10.1126/science.288.5473.1984.

Knight, Jeff R., 2009: The Atlantic multi -decadal oscillation inferred from the forcedclimate response in coupled general circula -tion models. J. Climate, 22, 1610–1625.

Lee S.-K., W. Park, E. van Sebille, C.Wang, D. B. Enfield, S. Yeager, B. P.Kirtman, and M. O. Baringer, 2011: Whatcaused the significant increase in Atlanticocean heat content since the mid-20th centu -ry? Geophys. Res. Lett. (submitted).

Levitus, S., J. I. Antonov, T. P. Boyer, R.A. Locarnini, H. E. Garcia, and A. V.Mishonov, 2009: Global ocean heat content1955–2008 in light of recently revealedinstrumentation problems. Geophys. Res.Lett., 36, L07608.doi:10.1029/2008GL037155.

Lumpkin, R. and K. Speer, 2007: Globalocean meridional overturning. J. Phys.Oceanogr., 37, 2550-2562.

Page 15

Figure 3. (a) Time-aver-

aged AMOC during

1979-2008 and (b) 11-

year running averaged

time series of the simu-

lated AMOC index

(maximum overturning

stream function) at

30oS, the equator, and

60oN in reference to

1871-1900 obtained

from EXP_CTR.

V9N2 8/8/11 10:47 AM Page 15

Subscription requests, and changes of address should be sent to the attention of the U.S. CLIVAR Office (cstephens@usclivar.org)

Coupled GCMs suffer from commonbiases in the Pacific and Atlantic basins.While the newer-generation CCSM4 hasdemonstrated an overall reduced SSTbias compared to CCSM3, it neverthe-less retains a similar SST-bias spatialpattern (Gent et al., 2010). Significantly,however, model improvements have hadless impact on the simulated SSTs inthe Atlantic than for the Pacific. TheAtlantic still exhibits the most severebias problem among all the tropicaloceans in the current generation of cli-mate models. In fact, the Atlantic biasproblem is still so severe that some ofthe most fundamental features of theequatorial Atlantic Ocean – the east-west equatorial SST gradient and theeastward shoaling thermocline – cannotbe reproduced by most coupled climatemodels.

The lack of progress in the Atlanticbias problem may be attributed to twomajor factors: 1) the complex nature ofthe bias problem and 2) a lack offocused attention from the researchcommunity. Hypotheses for a complexAtlantic bias problem tend to draw onthe fundamental observation that theAtlantic basin is far smaller than thePacific basin. The smaller Atlantic basincompared to the Pacific encourages atighter and more complex land-atmos-phere-ocean interaction with not just theeast side of the ocean basin, but also itswest side. Hypotheses can be oceanic(e.g., insufficient upwelling in the east-ern basin because local processes are notresolved); atmospheric (e.g., poor repre-sentation of the winds; cloud decks that

U.S. CLIVAR OFFICE1717 Pennsylvania Avenue, NWSuite 250Washington, DC 20006

U.S. CLIVAR contributes to the CLIVAR Program and is a member ofthe World Climate Research Programme

This material was developed with federal support of NASA, NOAA andNSF through the NSF Cooperative Agreement No. AGS-0926904. Anyopinions, findings, conclusions or recommendations expressed in thismaterial are those of the author(s) and do not necessarily reflect theviews of the sponsoring agencies.

U.S. CLIVAR

are too thin); land-based (e.g., poorlyrepresented deep convection over land;biomass-burning aerosol direct radiativeimpacts that vary with whether theunderlying surface is cloud or ocean)and can be of both local and remote ori-gins (e.g., remote forcings from thePacific onto the Atlantic). As alreadysuggested by the examples given here,many hypotheses reflect truly coupledprocesses between ocean, land, andatmosphere. Several of these are detailedfurther here in other articles.

In March 23-24, 2011, a workshopwas held at the University of Miami tospecifically focus on the bias problem inthe tropical Atlantic. A goal of the work-shop was to bring together disparatecommunities working on research rele-vant to the hypotheses for the modelbiases, and to identify a network ofinterested researchers. The workshopreceived support from both US fundingagencies and from WCRP. Participationin the workshop was international and ata high level of expertise. Approximately85 people participated; the agenda andall the presentations are availablethrough http://www.clivar.org/organiza-tion/atlantic/meetings/tropical_bias/miami.php. Interest in this problem is clear-ly high.

A major workshop objective was todevelop a coherent synthesis of the state-of-the-art knowledge on the AtlanticSST model biases and their causes forthe southeast and eastern tropicalAtlantic, as well as a set of sharpenedhypotheses. This is an ambitious goal,and although many ideas were put forth,

Workshop on Coupled Ocean-Atmosphere-Land Processes in theTropical Atlantic

Paquita Zuidema, C. Roberto Mechoso, and Laurent Terray

the 2.5 days were not enough to form aconsensus view, or even to integrate allthe information. The workshop con-firmed many hypotheses, but did notrank them. The workshop also revealedthat while there are efforts underway toaddress the hypotheses using models, theinteraction with observational programsis still weak. Observational programshave collected data, but they are still inthe synthesis stage. Further work stillneeds to be done to articulate an effectiveway forward (further model analysis?new coordinated model experiments?new field programs, or modification ofexisting observational networks?), and todefine an appropriate geographical focus.

At the time of this writing, the nextsteps forward are being discussed, andare expected to lead to both formal andinformal task teams aimed at furtherarticulating, ranking, and addressingapproaches to causes for the coupledmodel SST biases. These will lead to areport highlighting similarities and dif-ferences between GCM performance inthe tropical Atlantic and Pacific; a writ-ten survey of field programs completed,in progress, or under development; betterdefinition of an appropriate geographicalfoci; development of diagnostics andmetrics for model validation; and promo-tion of international collaborationsbetween observational and modelingstudies. These efforts will help refine theultimate goal: to articulate the best wayto reduce coupled model SST biasesusing targeted process studies and modelassessments.

V9N2 8/8/11 10:47 AM Page 16