abstract book for oral presentations · identifying wave processes associated with predictability...

ABSTRACTBOOKFORORALPRESENTATIONS

SESSION:(A1)MechanismsofS2Spredictability

(A1-01)

Identifyingwaveprocessesassociatedwithpredictabilityacrosssubseasonaltoseasonaltimescales

GilbertBrunet(1),JohnMethven,

(1)MeteorologicalResearchDivision,EnvironmentandClimateChangeCanada,(2)DepartmentofMeteorology,UniversityofReading,UK

Thekey tobetterpredictionofS2Svariabilityandweather regimes ina changingclimate lieswithimproved understanding of the fundamental nature of S2S phase space structure and associatedpredictability and dynamical processes. The S2S variability can be partitioned with the ModifiedLagrangianMean(MLM)approachintermsofslowdiabaticprocesses,suchasradiativeforcing,andadiabaticdynamicalprocesses.Thelattercanbedecomposedintoafinitenumberofrelativelylarge-scale discrete-like Rossby waves with coherent space-time characteristics using Empirical NormalMode(ENM)analysis.ENManalysisisbasedonprincipalcomponentanalysis,conservationlawsandnormal mode theories. These modes evolve in a complex manner through nonlinear interactionswiththemselvesandtransienteddiesandweakdissipativeprocesses.Thefoundationsandpotentialvalue of the ENM approach are presented but novel research is required to understandthepredictability and dynamical processes of thesemodes, including their excitation by resonantmechanisms.

Keywords

S2S; Predictability; Rossby Waves; Modified Lagrangian Mean; Empirical Normal Modes;ConservationLaws;PrincipalComponentAnalysis.


(A1-02)

Predictivesignalandnoiseinsub-seasonaltodecadalforecasts

ZoltanToth(1),JingZhang,JieFeng(2),RobertoBuizza(3),andMalaquiasPeña(4)

NOAA/OAR/ESRL/GSD,USA(1),UniversityofOklahoma,USA(2),ECMWF(3),UniversityofConnecticut,USA(4)

The prediction of the state of dynamical systems involve the estimation, and through temporalrelationships, the projection of the current state into the future. In non-periodic systems like thecoupledatmosphere-ocean-land-ice,ormoregenerally,theEarthsystem,informationabouttheexact state of the system is lost due to the chaotic growth of errors originating from the initialconditionandthenumericalmodelofthesystem.Wedistinguishbetween"traceable"predictabilitywherethephaseandamplitudeofindividualeventscanbestilltracked,and"climatic"predictabilitywhereonlythestatisticsofselectedevents(suchastheirfrequency),butnottheirtimingorposition,canbepredicted.

Though some high impact events such as heat waves, cold spells, or certain types of floods areassociated with low frequency, large scale events that are more predictable, many high impactweather events such as severe storms, tornados or flash floods are associatedwith smaller scaleeventsthatlosetraceablepredictabilityrelativelyquickly.Evenaftertheirtraceablepredictabilityislost, the frequency or other statistics of such smaller scale and potentially high impact weatherevents may still exhibit deviations from climatology, contingent on the traceable predictability oflargerscaleprocesses(i.e.,climaticpredictabilityofthunderstormsatagivenlocation,asafunctionof the phase of ENSO). In this study, the effectiveness of algorithms designed to separate thepredictable (traceable or climatic) signal from noise in S2S and S2D predictions such as spatial,temporal,orensemble-basedaveragingwillbecriticallyreviewed.


(A1-03)

SourcesoftropicalsubseasonalpredictabilitybeyondtheMJO

Newman,Matthew(1,2),Sardeshmukh,Prashant(1,2),Wang,Yan(1,2)

UniversityofColorado/CIRES,USA(1),NOAA/ESRL/PSD,USA,(2)

TherelativeimportanceofENSOandtheMJOtotropicalsubseasonalpredictabilityisassessedwitha coupled linear inversemodel (LIM) derived from the simultaneous and 5-day lag covariances ofobserved 5-day running mean atmospheric and oceanic anomalies for the years 1982-2009. TheoceanicportionoftheLIMstatevectorismadeupofseasurfacetemperature(SST)andseasurfaceheight(SSH)anomalies,whereastheatmosphericportionismadeupofoutgoinglongwaveradiation(OLR)and200and850hPazonalandmeridionalwindanomalies.Acomparisonof theLIM(cross-validated)hindcastskillwiththatoftheoperationalversionsoftheNCEPCVSv2andECMWFmodelsfor the years 1999-2010 is alsomade. LIM S2S skill (Weeks 3-6) is comparablewith both forecastmodels. The ECMWF hindcasts generally have highest skill forWeeks 3-4, with both the LIM andECMWFhindcastshavingaboutthesameskillinWeeks5-6.

GiventhatitsforecastskilliscomparablewithoperationalcoupledGCMsanditreproducesobservedspatio-temporalstatistics, themuchsimplerLIMisuseful fordiagnosisofpredictability,whichmaybedeterminedfromitsforecastsignal-to-noiseratios.Thestate-dependenceofpotentialLIMskillisassessedandshowntocomparewellbothwiththeLIMandoperationalmodelrealizedskill.Furtheranalysis is performed based on the fact that the eigenvectors of the LIM dynamical evolutionoperatorseparateintotwodistinct,butnonorthogonal,subspaces:an“internal”spacegoverningthenearly uncoupled subseasonal dynamics, and a “coupled” space governing the strongly coupledlonger-termdynamics.ThesesubspacesarisenaturallyfromtheLIManalysis;nobandpassfrequencyfilteringneedbeapplied.Theinternalspaceeigenmodestypicallyhavemuchshorterperiodsande-folding time scales than the coupled space eigenmodes. Additionally, theMJOmostly lies in thisinternal space, whereas ENSO mostly lies in the coupled space. Anomalies that project onto thecoupledspaceareshowntomorepredictablethanthoseprojectingontotheinternalspace,evenforrelatively “short”Weeks3-4 leads. That is, ENSO is theprimary contributor tooverall tropical skillevenontheWeeks3-4timescale,andprovidesalmostallofitonlongerS2Sforecastleads.

1


(A1-04)

CharacteristicsoftheQBO-StratosphericPolarVortexConnectiononMulti-decadalTimeScales

JudithPerlwitz(1),LantaoSun(1),JadwigaH.Richter(2),JohnR.Albers(1)

CIRES/UniversityofColoradoandNOAA/ESRLPhysicalSciencesDivision,USA(1),NCAR,USA(2)

The strength of the relationship between the quasi-biennial oscillation (QBO) and the NorthernHemispherestratosphericpolarvortex,orHolton-Tan(H-T)relationship,onmulti-decadaltimescalesisinvestigatedusinga10-memberensembleofhistoricalsimulationsfortheperiod1957-2015.Theexperimentswereconductedwiththehigher-topCommunityAtmosphereModelVersion5(CAM5)that is capable of internally generating theQBO. Consistentwith reanalysis, themodel ensemble-meanshowsa strengthening (weakening)of thepolarvortexwithwesterly (easterly)phaseof theQBO.However,substantialvariationsacrossindividualensembleswithrespecttothestrengthoftheH-T relationship and subsequent tropospheric impacts on ~60-year timescales are found that arecloselylinkedtovariationsinthefrequencyofoccurrenceofmajorstratosphericsuddenwarmingsintheQBOeastphase.It isshownthatthissensitivity isconsistentwiththeQBO’smodulationofthezero-wind line in the lower stratosphere affecting planetary wave propagation. Implications forevaluating the H-T relationship in atmospheric circulation model and QBO’s role as source ofpredictiveskillonsubseasonaltoseasonaltimescalesarediscussed.


(A1-05)

Impactofstatisticallyforecastedsea-iceboundaryconditiononthesub-seasonalpredictionusingatmosphericgeneralcirculationmodel

Kim,Ha-Rim(1),(2),Kim,Baek-Min(1),Jun,Sang-Yoon(1),Jeong,Jee-Hoon(3),Shin,Dong-Wook(4),Cocke,Steven(4),Yoon,Jin-Ho(5)

KoreaPolarResearchInstitute,KOR(1),EwhaWomansUniversity,KOR(2),ChonnamNationalUniversity,KOR(3),FloridaStateUniversity,USA(4),

GwangjuInstituteofScienceandTechnology,KOR(5)

Sub-seasonalandseasonalpredictionmodelhasbeenextensivelydevelopedandtestedespeciallyinrecentseveralyears,butthepredictabilityforone-monthaheadorlongerremainsstillchallenging.Recently,anenhancedyear-to-yearvariabilityof theArcticsea-icehasbeenemphasized in its rolefor controlling mid-latitude climate variability. In this study, we implemented a sub-seasonalprediction model named Korea Polar Prediction System (KPOPS) based on the CommunityAtmosphere Model version 4 (CAM4) by incorporating statistically predicted sea-ice boundarycondition.Sea-iceconcentration(SIC)inthemodelispredictedduringtheforecastperiodusingthestatisticalmethodbasedontheSeasonal-reliantEmpiricalOrthogonalFunction(S-EOF)method.Thestatistical prediction of SIC using the S-EOF generally outperforms dynamic sea-ice predictionprovidingagoodmotivationforthisstudy.Seasurfacetemperature(SST)outoftheArcticregionisprescribed using the forecasted SST from the climate forecast system (CFS) model. One-monthprediction skill of the KPOPS and CFS-Reforecast are compared for the early winter (November –December)andlatewinter(January–February)from2001/2002to2014/2015.OverallforecastskillofKPOPSoutperformed theCFS-Reforecast.Especially, in theearlywinter, the forecast skillof theKPOPSwas superior in theArctic region,apparently. In the latewinter, the forecast skilloverEastAsia and North Americawas significantly higher than the CFS-Reforecast. Therefore, we concludethat thesub-seasonalpredictionusingstatistically forecastedsea-iceboundaryconditionexhibitsanew possibility of better prediction given the situation that current state-of-the-art fully coupledmodelsdonotprovideareliablepredictionproductofSIC.


(A1-06)

ThecharacteristicsofKelvinwavesintheatmosphere-oceancoupledsystem

Flatau,Maria(1)Baranowski,Dariusz(2),Barton,Neil(1),ChenSue(1)1,FlatauPiotr(3),JanigaMathew(1),Reynolds,Carolyn(1)andRidoutJames(1)SchmidtJerome(1)

NRL,USA(1),IGFPANPoland(2),SIO,USA(3)

We use the technique developed in Baranowski at al (2017) that allows to identify the individualKelvinwavesintheobservedTRMMprecipitationtoevaluatepredictionskilloftheindividualKelvinwavesandKelvinwave climatological characteristics in the setof forecastsusing thenewcoupledNavyglobalpredictionmodel.WhilethevariabilityinthecoupledmodelintheKelvinwaveportionofthespectrumappearstobewellrepresented,thefeaturesoftheindividualwavesareofinterestas well. In particular, our earlier work shows that the interaction of Kelvin waves with the localcirculationsover theMaritimecontinentcouldhavean impactonpropagationof theKelvinwavesand possiblyMJO through theMaritime Continent, depending on the time of the day the wavesreach Sumatra and Borneo. In this work we examine Kelvin wave characteristics in the coupledmodelforecaststoestablishhowtheirintensity,thelocationoftheoriginanddecayandinteractionwiththeMaritimecontinentcomparestoobservationsandhowthesecharacteristicsarelinkedwithMJOpropagationandMJOpredictionskill.


(A1-07)

Diagnosingsourcesofoperationalforecastmodelerrorsintropical-extratropicalinteractions

Dias,Juliana(1,2)andKiladis,George(2)

(1)CIRES,UniversityofColoradoBoulder,(2)NOAA/ESRLPhysicalSciencesDivision

The atmospheric response to variations in latent heating in the tropics is known to extend wellbeyond its source region and therefore it is thought that a reduction of tropical forecast errors isbeneficialforforecastskilloverremoteregionssuchasNorthAmerica.

In this presentation, the impact of the quality of tropical forecasts in extra-tropical week 1 to 4forecast skill is evaluated, using reforecasts from the S2Sproject. It is shown that inmostmodelsthereisapositivecorrelationbetweenperformanceoftropicalforecastsandextra-tropicalforecastatlaterleadtimes.Thestrengthofthesecorrelationsvarieswithleadtime,variableandmodel.Oneinterpretation is that when tropical heat sources arewell (poorly) predicted, extra-tropical skill isgained (lost) due to properly (improperly) triggeredRossby-likewavetrains. Thequality of the S2Stropical forecasts is shown to be related to eachmodel’s ability to simulate theMJO, as well ashigher frequency convectively coupled equatorial waves. This analysis also suggests that modelphysics,ratherthandynamics,controlthelinkbetweentropicalandextra-tropicalskill.


(A1-08)

ENSOmodulationofMJOteleconnectiontotheNorthAtlantic&Europeandimplicationsforsubseasonalpredictability

Lee,RobertW.,(1)(2),Woolnough,Steve(1)(2),Charlton-Perez,AndrewJ.(2),Vitart,Frédéric(3)

NationalCentreforAtmosphericScience,UK(1),UniversityofReading,UK(2),ECMWF,UK(3)

The Madden-Julian Oscillation (MJO), an organized eastward propagating source of enhancedconvection,actsasaRossbywavesourceandasaglobaltropicaldriverofsubseasonalpredictability.An example pathway leads this source of predictability out of the tropics to the North Atlantic /European(NAE)region,knownasateleconnection,viathejetstreamandstratosphere.ThroughthisteleconnectiontheMJOcan influenceregimesof large-scaleweatherpatterns, includingtheNorthAtlanticOscillation(NAO).TheaimofthisstudyistoinvestigatethedependenceoftheMJO–NAEteleconnections on the interannual variations in the background state associatedwith the ElNiñoSouthernOscillation(ENSO).

WeusetheCassou(2008)frameworktoshowthattheseteleconnectionsfromtheMJOtotheNAEweatherandjetregimesarestronglydependentonthephaseofENSO.Forexample,duringElNiñoyears the MJO to NAO+ teleconnection is strongly enhanced and persists throughout more MJOphases,dominatingtheclimatologicalmeanpicture,whilstduringLaNiñathisteleconnectiontotheNAO+isweakandshort-lived.FurtherNAEregimetransitionsandinsitudevelopmentalsobecomeclearerviathisperspectiveseparatedbyENSObackgroundstate.WealsodiscusstheseasonalmeanresponsetoENSOintheNAEregionthroughrectificationofthesesubseasonalteleconnectionsontotheseasonalmean.

The dependence on the background state has strong implications for subseasonal predictability,includingimplicationsforinterannualvariationsinsubseasonalpredictiveskillandalsotheneedformodelstogetthebackgroundstatecorrectinordertocorrectlyrepresenttheseteleconnections.WeanalyzetherepresentationoftheseteleconnectionsandtheirdependenceonthebackgroundstateinmodelscontributingtotheS2Sdatabase.

*Thiswork is submitted as part of the InterDecproject, fundedunder the 2015 joint JPI Climate-BelmontForumcall.


(A1-09)

TheimpactofNorthernHemispheremid-latitudevariabilityontropicalteleconnections

CristianaStan(1),ErikSwenson(2)

GeorgeMasonUniversity

TheroleofNorthernHemispheremid-latitudevariabilityonatmosphericrivers(AR)withlandfallonthewestcoastoftheU.S.isevaluatedintwosetsofensemblere-forecastsforthewinter2014/2015.The first set of re-forecasts defines the control experiment. In the second set of re-forecasts thevariabilityof theNorthernHemispheremid-latitudes is relaxed to theclimatologicalmean stateofthemodel.TheAReventsoccurringinobservationsduringthiswinterarefirstdividedbasedonthemoisture source location and trajectory,which ultimately determines the landfall location. Duringthiswinter,inobservationsallAReventsprimarilyoriginateinthetropics,whereasthere-forecastsshow some eventswithmoisture descending from the extra-tropics. Secondly, the re-forecasts ofthese categoryof events are comparedbetween the control experiment and the experimentwithsmoothedmid-latitudevariability.By removing thesynopticscalevariabilityofwinds, temperatureandmoisture, theWest Pacific trough becomes anchored and starts to deepen at 40N after oneweek.TheprobabilityofoccurrenceofARwithtropicalorigin increasesatall leadtimesuptofourweeks,whereas theprobabilityofoccurrenceofextra-tropicaleventsdecreases.After twoweeks,AReventswithmoisture ascending from the tropics tend tobecomemorepersistent and zonally-orientedandbyweeks3-4theycanextendacrossthewholePacificbasin.


(A1-10)

PredictingthedominantpatternsofsubseasonalvariabilityofwintertimesurfaceairtemperatureinextratropicalNorthernHemisphere

Lin,Hai

ECCC,Canada

Skillfully predicting persistent extreme temperature anomalies more than 10 days in advanceremains a challenge although it is of great value to the society. Here the two leading modes ofsubseasonal variability of surface air temperature over the extratropical Northern Hemisphere inboreal winter are identified with pentad (5-days) averaged data. They are well separatedgeographically,dominatingtemperaturevariabilityinNorthAmericaandEurasia,respectively.Thereexistsatwo-pentadlaggedcorrelationbetweenthesetwomodes,implyinganinter-continentallinkoftemperaturevariability.Forecastskillofthesetwomodesisevaluatedbasedonthreeoperationalsubseasonalpredictionmodels.TheresultsshowthatusefulforecastsoftheEurasianmode(EOF2)can be achieved four pentads in advance, which is more skillful than the North American mode(EOF1). The influence of the Madden-Julian Oscillation (MJO) on the forecast skill of these twotemperaturemodesisalsoanalyzed.


(A1-11)

TheroleofclouddiabaticprocessesinthelifecycleofAtlantic-Europeanweatherregimes

Grams,ChristianM.(1)

InstituteofMeteorologyandClimateResearch(IMK-TRO),KarlsruheInstituteofTechnology(KIT),Germany(1)

The large-scale midlatitude flow is dominated by Rossby wave activity along the upper-levelmidlatitude wave guide and jet stream. This activity often occurs in preferred quasi-stationary,persistent, and recurrent states, so-called weather regimes (e.g. Vautard, 1990). Many of theseregimesaredominatedbyablockinganticyclone.IntheAtlantic-Europeanregion,weatherregimesexplain most of the atmospheric variability on sub-seasonal time scales. From a forecastingperspective,theonset,persistence,andtransitionofweatherregimespresentaseverechallengeincurrentnumericalweatherpredictionmodels(Ferrantietal.,2015).

Recently,Pfahletal.(2015)showedthattransportofairmassintotheuppertropospheredrivenbylatent heat release in ascending air streams is a first-order process in blocking onset andmaintenance.ThisstudysystematicallyinvestigatestheroleofsuchdiabaticoutflowinthelifecycleofallEuropeanweatherregimes.

An extendeddefinition of 7 year-roundAtlantic-Europeanweather regimes from37 years of ERA-Interim reanalysis data is used (Grams et al. 2017). This is based on an EOF analysis and k-meansclustering of normalized low-pass-filtered 500hPa geopotential height anomalies. The weatherregime indexofMichelandRivière (2011) isused to furtherobjectivelydefine important lifecyclestages such as the regime onset, mature stage, decay, or transition. The role of cloud-diabaticprocessesinEuropeanweatherregimesisassessedbasedontimelaganalysisofwarmconveyorbelt(WCB)activityattheselifecyclestages.

Results indicate that the period prior to regime onset is characterized by important changes inlocationandfrequencyofWCBoccurrence.Thesechangespersistduringtheearlystageofaregimelife cycleandweaken thereafter.Most importantly,prior to theonsetof regimescharacterizedbyblocking,astatisticallysignificantincreaseinWCBactivityoccursupstreamoftheevolvingblockinganticycloneevenbeforeblocking isdetectable.Thissuggeststhatthe injectionofairmass intotheupper troposphere by diabatic WCB outflow helps to establish the blocking anticyclone.Furthermore,diabaticWCBoutflowhelpsmaintaining theblockinganticyclone laterduring the lifecycle.Finally,arecentforecastbustdemonstratesthechallengesinpredictabilityofweatherregimelifecyclesimposedbythemulti-scaleinteractionsestablishedthroughdiabaticoutflow.

This study corroborates the importance of correctly representing cloud-diabatic processes innumericalweatherprediction(NWP)modelsacrossmultiplescalesinordertopredictthelarge-scalecirculationaccurately.AseamlessNWPapproachthat isabletopredictEuropeanweatherregimesonsubseasonaltimescaleswouldallowformanifoldapplicationssuchasintheenergysector.

Ferranti, L., S. Corti, andM. Janousek, 2015: Flow-dependent verificationof the ECMWFensemble over theEuro-Atlanticsector.Q.J.R.Meteorol.Soc.,141,916–924,doi:10.1002/qj.2411.

Grams,C.M.,R.Beerli,S.Pfenninger,I.Staffell,andH.Wernli,2017:BalancingEurope/’swind-poweroutputthrough spatial deployment informed by weather regimes. Nature Climate Change, 7, 557–562,doi:10.1038/nclimate3338.

Michel,C.,andG.Rivière,2011:TheLinkbetweenRossbyWaveBreakingsandWeatherRegimeTransitions.JournaloftheAtmosphericSciences,68,1730–1748,doi:10.1175/2011JAS3635.1.

Pfahl,S.,C.Schwierz,M.Croci-Maspoli,C.M.Grams,andH.Wernli,2015:Importanceoflatentheatreleaseinascendingairstreamsforatmosphericblocking.NatureGeosci,8,610–614,doi:10.1038/ngeo2487.

Vautard,R., 1990:Multipleweather regimesover theNorthAtlantic: analysis of precursors and successors.Mon. Wea. Rev., 118, 2056–2081, doi:10.1175/1520-0493(1990)1182056:MWROTN2.0.CO;2. 2 t 2 h

SESSION:(A2)ModellingissuesinS2Sprediction

(A2-01)

Theartandscienceinsub-seasonalforecastsystemdesign

YuheiTakaya

MeteorologicalResearchInstitute,JapanMeteorologicalAgency

Anoperationalforecastsystemrequiresahighlevelofforecastquality,timelinessandcost-efficiencyunder various competing resource requirements. As such, it is vitally important for operationalmodelling centres to design and configure forecast systems, including the sub-seasonal forecastsystem,soastomaximizetheaccuracy,forecastskill,andbenefitsofforecastinformation.

Apart from modelling issues in representing physical and dynamics processes, there are manyunresolved issues in the forecast systemdesign and configuration: initialization,model resolution,ensemble size, ensemble generation, hindcast and forecast configurations. The choice of theseconfigurationsdemandscarefulconsiderationsinceeachchoiceimpactscostandquality.Undertheconstraintoffixedresources,thesechoicescompeteagainstoneotherinthetradespace.

This presentation reviews current practices in sub-seasonal forecast system design at operationalcentresanddiscussestheprosandconsofseveralapproaches.Existingissuesregardingtheoptimalforecastsystemdesignforthesub-seasonalpredictionsinclude(1)benefitsanddeficienciesofburstand Lagged Average Forecasting (LAF) ensemble approaches for operational forecasts, (2) optimalconfigurationsofreal-timeforecastssuchasfrequencyandensemblesize,(3)optimalconfigurationsof reforecasts in terms of ensemble size, length and frequency, (4) techniques of ensemblegenerationanddataassimilation.Thesub-project“ensemblegeneration”intheS2SPhase2focusesonsomeoftheseissues.

Initialshocksandmodeldriftsareother importantaspects insub-seasonalforecastsystems. Inthebig picture of the sub-seasonal forecast systems, global observations and analyses of past andpresent are essential in maintaining the operational sub-seasonal forecast service. Oceanicobservations are necessary for systems employing atmosphere-ocean coupled models. Withincreasesinthecomplexityofforecastsystems,itisbecomingdifficulttomakethesystemconsistentinhindcastsandforecasts.Inthesub-seasonaltime-range,theinitialshocksandgradualincreaseofmodelbiases(modeldrifts)appearduetotheimbalanceanddifferencebetweenmodelandanalysisstates.The inconsistencyofanalysisandmodelstatescauses initial shocks indifferent timescales,forexample,initialshocksoftheatmospheremaylastaboutoneweekandthoseoftheoceanandlandmay last severalweeksandmore.Therefore reducing theseshocksanddrifts isoneofmajorfociofthesub-seasonalsystemdevelopment.

Alloftheaboveaspectsaffectthequalityofforecastsandneedtobeproperlyconfiguredtocreateanoptimizedsystem,a“masterpiece”ofasub-seasonalforecastsystem.


(A2-02)

InitialisationandmodelerrorsimulationintheECMWFcoupledensembles

Buizza,Roberto(1),Bonavita,Massimo(1),Holms,Elias(1),Laloyaux,Patrick(1),Lang,Simon(1),Leutbecher,Martin(1),Lock,Sarah-Jane(1),Vitart,Frederic(1)

(1)EuropeanCentreforMedium-RangeWeatherForecasts(ECMWF)

OnekeyelementoftheECMWFensemblesisthemethodologyusedtosimulateinitialuncertainties,which is based on a combination of singular vectors and perturbations defined by the ECMWFensembleofdataassimilations(EDA).AsecondkeyelementoftheECMWFensemblesistheuseofstochasticschemestosimulatemodeluncertainties.Inthistalk,IwillbereviewingthesetwoaspectsoftheECMWFensembles.


(A2-03)

DevelopmentofaUnifiedForecastSystematNCEPforS2SPrediction

Saha,Suranjana(1),Wu,Xingren(2),Wang,Jiande(3),Tripp,Patrick(4),Woollen,Jack(5),Lee,Hyun-Chul(6),Bhattacharjee,Partha(7),vandenDool,Huug(8),Pena,Malaquias(9)

EMC/NCEP,USA(1),EMC/IMSG,USA(2),EMC/IMSG,USA(3),EMC/IMSG,USA(4),EMC/IMSG,USA(5),EMC/IMSG,USA(6),EMC/IMSG,USA(7),CPC/Innovim,USA(8),EMC/IMSG,USA(9)

NCEP’smissionforS2SpredictionrequiresdevelopingasuccessormodeltothepresentoperationalCFSv2. For several years, a concerted effort has taken place, both internal and external toEMC/NCEP, to create the infrastructureof amulti-component global coupled system in theNEMSframework.Atthepresenttime,theatmosphericspectralGSMiscoupledwiththeMOM5.1oceanmodel and the CICE5 seaice model. The land surface model is still internal to the GSM. ThisconfigurationwillchangewhenthespectraldynamicswillbereplacedbytheFV3dynamiccoreandtheoceanmodelwillbeupgradedtoMOM6.AverificationmodulehasbeendevelopedtovalidatetheUFSasitevolvesandconvergestoitsfinalconfiguration.Thismoduleconsistsofmaking14435-day forecasts from the 1st and the 15th of eachmonth, over a 7-year period from April 2011 toMarch2018.Calibrationclimatologiesare firstprepared forall variables thatare studiedby fittingfourharmonics and themean to themodel time series, aswell as to thematchingobserved timeseriesusedforverification.Someofthevariablesstudiedarez500,SST,T2mandPRATE,aswellasU850,U250andOLRtostudyMJOprediction.ForecastsofthevariousconfigurationsandthecontroloperationalCFSv2arecomparedintermsofRMSEandAC,bothwithandwithoutsystematicerrorcorrection(SEC).SpecialemphasisisgiventoNHZ500,US-landT2mandPRATEverifiedagainstCPC-‘daily’observations,andSSTandPRATEfortheNino3.4area.Preliminaryresultsshowthatthenewsystem with MOM6 and CICE5, without any type of tuning, is equal or better than the controloperationalCFSv2overthelast7years.

Inthenearfuture,thefinalconfiguration,whichmayincludetheWavewatch-IIIforwaves,GOCARTforaerosolsandNoah-MPforthelandsurface,willbefine-tunedforoptimalperformance,especiallywithregardtothephysicsparameterizationsofconvection,radiation,microphysicsandcloudsandthecouplingbetweenthedifferentcomponents.Theverificationmoduledescribedabovewillhelpguide the rapiddevelopmentof theUFSas it canbeexecutedquickly toget interimassessments.Beyond such first steps, many more integrations to build an ensemble, covering longer forecastranges,usingmorefrequent initialstatesandmanymoreyearswillberequired.AweaklycoupledDA system of all the components is also being developed to provide initial states for the UFS. AcompleteReanalysisfrom1979tothepresentandsubsequentReforecastsofthesystemwillalsobecarried out to provide stable calibration and skill estimates of the system before it is madeoperationalatNCEP.


(A2-04)

SufficientresolutionforS2Spredictions

Sardeshmukh,Prashant(1,2).Magnusson,Linus(3),Wang,Aaron(1,2),Bacmeister,Julio(4)

CIRES,CU,USA(1),PSD/ESRL/NOAA,USA(2),ECMWF,UK(3),NCAR,USA(4)

SubseasonaltoSeasonal(S2S)predictionsareinherentlyprobabilistic.Theyareaboutpredictingtheprobabilitydistributionsoffuturestatesgiveninitialconditionsandexternalforcing.Thequestioniswhetheroneneedsultra-high-resolutionmodels to represent suchdistributions,orwhether lowerresolutionmodelswithasuitablecombinationofdeterministicplusstochasticparameterizationsofsmall-scale feedbacks are sufficient for this purpose. From a computational viewpoint, lowerresolutionmodelsareobviouslymoreattractive,especiallyforgenerating largeforecastensemblestopindowntheforecastprobabilitydistribution.Ontheotherhandonecouldarguethatultra-highresolutionisnecessary,giventheexistenceofcoherentstructuresateveryscaleandtheabsenceofa spectral energy gap, to adequately capturemulti-scale interactions and the statistics of extremevaluesi.e.thetailsofthegenerallynon-Gaussianprobabilitydistributions.

ComparisonsoftheprobabilitydistributionsofdailyanomaliesinlongglobalatmosphericGCMrunsmadeatECMWFandNCARatresolutionsrangingfromT95(about130km)toT2047(about6.5km)are,however,revealinginthisregard.Thedistributionsareindeednon-Gaussian,buttoanexcellentapproximation differ only in their widths but not their shapes at the different resolutions. Thisremarkableresult isarguedtobeconsistentwiththestochasticallygeneratedskewed(SGS)natureof the distributions, and that beyond T511 (about 25 km) themain impact of higher resolution ismerely toenhance theeffectively stochastic forcingof the large-scaleeddiesby small-scale fluxes.This suggests thata resolutionofaboutT511,utilizinga suitablecombinationofdeterministicandstochastic parameterizations to accurately represent variances and energy spectra, should besufficient for S2S predictions. Indeed we have recently confirmed in a large ensemble forecastexperimentperformedwith theNCEP/GFSmodel (80-member15-day forecasts at T254 resolutionfor100forecastcases)thataddingstochasticparameterizationstothemodelnotonlyincreasestheensemble spread but also unambiguously reduces the error of the ensemblemean forecast. Thisincreases both the deterministic and probabilistic skill of the forecasts at Day 15, and makes itindistinguishablefromthatofthemuchhigherresolutionoperationalGFSforecasts.Theincreaseofensemble spread through the additional stochastic forcing is not surprising, but the unambiguousreduction of the ensemblemean error is. It results from amultiplicative (state-dependent) noise-inducedmodificationofthedeterministicforecastevolutionoperator,aswillbedescribedinthetalk


(A2-05)

Seasonalpredictionexperimentsinaglobalcoupledsystembasedonanon-hydrostaticglobalatmosphericmodel

Jung-EunEstherKim,Myung-SeoKoo,andSong-YouHong

KIAPS

TheKoreaInstituteofAtmosphericPredictionSystems(KIAPS)begananationalprojectindevelopinganewglobalatmosphericmodelsystemin2011.AsofFebruary2018,the12-kmKoreanIntegratedModel(KIM)system,whichconsistsofanewspectral-elementnon-hydrostaticdynamicalcoreonacubed sphere and the state-of-the-art physics parameterization package, has been launched in areal-time forecast framework,with the initial conditions obtained from the advanced hybrid four-dimensionalensemblevariationaldataassimilation(4DEnVar)over itsnativegrid.Further,KIMhasbeencoupledwiththeoceanmodel,HYCOM,forexpansionoftheapplicationtoseasonalpredictionandclimatestudies.Predictabilityexperimentswiththiscouplesystemarebeingconductedandtheresultswill be presented at the conference,with a focus on the stochastic coupling of ocean andatmosphere.


(A2-06)

TowardReducingCloud-RadiationErrorsfromDay1toWeek4Prediction

Benjamin,Stan(1),Grell,Georg(2)

NOAA/ESRL,USA

Cloud-radiationrepresentationinmodelsforsubgrid-scalecloudsisaknowngapfromsubseasonal-to-seasonalmodelsdowntostorm-scalemodelsapplied for forecastdurationofonlya fewhours.NOAA/ESRL is confirming this issue from3-kmmodel (HRRR) for short-range forecasting includingsub-grid-scale cloud representation up to a 60-km subseasonal model testing a common suite ofscale-aware physical parameterizations. Resulting refinements to this common suite of physicalparameterizations for scale-aware deep/shallow convection and boundary-layer mixing over thiswiderangeoftimeandspatialscaleswillbereportedinthispresentation.Evaluationofcomponentsofthissuiteisbeingevaluatedforcloud/radiation(usingSURFRAD,CERES,METARceiling)andnear-surface(METAR,mesonet,aircraft,rawinsonde).

ESRL/GSD has developed a parameterization suite (turbulent mixing, deep/shallow convection, 9-layer land/snow/vegetation model) to improve PBL biases (temperature and moisture) includingbetterrepresentationofcloudsandprecipitation.Thisparameterizationsuitedevelopmenthasbeenaccompaniedby aneffort for improveddata assimilationof clouds, near-surfaceobservations andradarfortheatmosphere-landsystem.

The Grell-Freitas scheme (2014, Atmos. Chem. Phys.) and MYNN boundary-layer EDMF scheme(Olson/Benjaminetal.2016Mon.Wea.Rev.),andRUC land-surfacemodel (Smirnovaetal.2016Mon.Wea.Rev.)havebeenappliedandtestedextensivelyfortheNOAAhourlyupdated3-kmHigh-Resolution Rapid Refresh (HRRR) and 13-km Rapid Refresh model/assimilation systems over theUnitedStatesandNorthAmerica,with targeting toward improvement toboundary-layerevolutionand cloud-radiation representation in all seasons. This representation is critical for both warm-season severe convective storm forecasting and for winter-storm prediction of snow and mixedprecipitation.

At the same time the Grell-Freitas scheme has been applied also as an option for subseasonalforecastingtowardimprovedUSweek3-4predictionwiththeFIM-HYCOMcoupledmodel(Greenetal2017,MWR,andSunetal2018a,b,MWR,accepted).


(A2-07)

Meanstatebias,cloud-radiationfeedbacks,andMJOpredictionskillintheS2Smodels

Kim,Daehyun(1),Lim,Yuna(2),Son,Seok-Woo(2)

UniversityofWashington,USA(1),SeoulNationalUniversity,SouthKorea(2)

The Madden-Julian Oscillation (MJO), the dominant mode of tropical intraseasonal variability,provides amajor source of tropical and extratropical predictability on subseasonal timescale.Weexamined the relationshipofMJOpredictionskillwithbiases in thecolumnwatermeanstateandlongwave cloud-radiation feedbacks in the models participating in subseasonal- to-seasonal (S2S)predictionproject.InmostS2Smodels,anotabledrybiasdevelopswithinafewdaysofforecastleadtime in the deep tropics, especially across the Maritime Continent. The dry bias weakens thehorizontal moisture gradient over the Indian Ocean and western Pacific, likely dampening theorganizationandpropagationoftheMJO.MostS2Smodelsalsounderestimatethelongwavecloud-radiation feedbacks in the tropics, which may affect the maintenance of the MJO convectiveenvelope.Themodelswithsmallerbiasinthemeanhorizontalmoisturegradientandthelongwavecloud-radiationfeedbacksshowhigherMJOpredictionskills,suggestingthatimprovingthosebiaseswouldenhanceMJOpredictionskilloftheoperationalmodels.


(A2-08)

Process-basedMJOhindcastevaluationinSubX

HyemiKim(1),MatthewA.Janiga(2),DeepthiAchuthavarier(3,4),KathyPegion(5)

StonyBrookUniv.,USA(1),NRL,USA(2),USRA,USA(3),GMAO,USA(4)GeorgeMasonUniv.,USA(5)

There is a growing interest in forecastingweather and climatewithin the subseasonal time range.TheMadden-JulianOscillation(MJO),anorganizedenvelopeoftropicalconvection,isrecognizedasoneoftheleadingsourcesofsubseasonalpredictability.Theoreticalmodelsandfieldcampaignssuchas the Dynamics of the MJO (DYNAMO) experiment have led to a better understanding of theprocesses which are critical to simulating the MJO in numerical models. This has guidedimprovements to numerical models, particularly cumulus parametrizations, leading to significantadvances in MJO prediction. Nevertheless, estimates of the potential predictability of the MJOsuggestthatthereisstillconsiderableroomforimprovement.

The goal of this study is to examine how the ability of operational models to represent criticalphysicalprocesseswhichareimportanttotheMJOisrelatedtoMJOpredictionskill.Recentstudieshaveidentifiedseveralprocess-basedmetricswhicharerelatedtotheabilityofnumericalmodelstorepresent MJO behavior in multi-year climate runs and MJO hindcast performance during case-studies.Upuntil now,however, evaluationsofMJO fidelity in subseasonal forecastsovermultipleyearshavefocusedonperformance-basedratherprocess-basedmetrics.Internationalcollaborativeefforts such as the NOAA/MAPP SubX and WWRP/WCRP S2S projects, which make availablenumerous fields from the current generation of subseasonal prediction systems, present anunprecedentedopportunity torelateMJOperformancetoprocess-basedmetrics. In thisstudy,werelateMJOprediction skill in SubXhindcasts to four process-basedmetrics: (i) errors in themeanstateofmoistureandtherepresentationof(ii)moisture-convection,(iii)radiation-convection(iv)air-sea interactionprocess.Resultsfromthiscomparisonshedlightonwhichphysicalprocesses intheatmosphere-ocean system need to be better represented in numerical models to produce betterMJOpredictions.


(A2-09)

Theocean-atmospheredialogintheMJO:Physicalprocessesvs.systematicbiasesinforecastmodels

DeMott,CharlotteA.andKlingaman,NicholasP.

ColoradoStateUniversity(CAD)andUniversityofReading(NPK)

Convection and circulation anomalies driven by the Madden Julian oscillation regulate fluxes ofenergy, momentum, and fresh water to the upper ocean throughout theWarm Pool. The oceanrespondstothis forcing,andfeedsbacktotheatmospherethroughtheSST-modulationofsurfacefluxes on intraseasonal and shorter timescales, and through column moisture variations on low-frequencytimescales.

ManystudieshavedemonstratedimprovedMJOpredictionskillwhenoceanfeedbacksareincludedin dynamic forecast models, indicating an important role for ocean-atmosphere communicationwithintheMJO.Understandingtheessentialprocessesofthisocean-atmospherecommunication,ordialog,hasbeenhinderedbythecomplexityofprocessesthatregulatetheoceanresponsetoMJOforcing, the substantial event-to-event variability in apparent ocean feedbacks toMJO convection,and the broad diversity of changes toMJO prediction skill among atmosphere-onlymodels whencoupledtodynamicoceanmodels.

In Part I of this study, we first assess the “nature of the conversation” between ocean andatmosphere in coupled prediction models (CPMs) by applying a newly developed set of air-seainteractiondiagnostics toCPMmembersof theS2Sdatabase.We find thatmanyCPMsareoverlyreliant on SST-modulated surface fluxes for themaintenance and propagation ofMJO convection.The larger-than-observed SST variability in thesemodels and its contribution to theMJOmoisturebudget is rooted in dry biases in the atmospheric boundary layer that over-amplify intraseasonalsurfacelatentheatfluxes.

InPartIIofthisstudy,weassesssourcesofoceanpredictabilityforMJOforecastsbyCPMsintheS2Sdatabase.We focus on periodswhen the equatorial Indian Ocean exhibits enhanced eastward orenhanced westward surface currents, which have been linked to the regulation of MJO intensity(Moumetal.2016);andonperiodswhentheupperoceaniscalm,favoringtheformationofdiurnalwarmlayersandadiurnalcycleofshallowconvection(RuppertandJohnson2015).

Moum,J.N.,K.Pujiana,R.-C.Lien,andW.D.Smyth,2016:OceanfeedbacktopulsesoftheMadden-JulianoscillationintheequatorialIndianOcean.NatureComm.,DOI:10.1038/ncomms13203.

Ruppert, J. H., and R. H. Johnson, 2015: Diurnallymodulated cumulusmoistening in the preonsetstage of the Madden–Julian oscillation during DYNAMO. J. Atmos. Sci., 72, 1622–1647

SESSION:(A3)S2Sensemblepredictionsandforecastinformation

(A3-01)

Howfarinadvancecanwepredictchangesinlarge-scaleflowleadingtoseverecoldconditionsoverEurope?

FerrantiLaura,MagnussonLinus,VitartFrederic,RichardsonDavid

ECMWF

ThepotentialofearlywarningforseverecoldconditionsisexploredusingtheS2Sdataarchive.Weuseofa2dimensionalphasespacetostudythetimeevolutionofflowpatternsassociatedwithhigh-impact temperatureanomalies.AnumberofS2S systemshave someskill in thepredictionof coldspellsoverEurope,wellbeyondthemediumrange.WefindthattheMJOimpactonthepredictiveskilloflarge-scaleflowoverEuropeisasymmetric.


(A3-02)

AverificationframeworkforSouthAmericansub-seasonalprecipitationpredictions

Coelho,Caio(1),Firpo,Mári(1),Andrade,Felipe(1)

CenterforWeatherForecastandClimateStudies(CPTEC),NationalInstituteforSpaceResearch(INPE),Brazil(1)

We propose a verification framework for South American sub-seasonal (weekly accumulated)precipitationpredictionsproducedonetofourweeksinadvance.Theproposedframeworkassessesbothhindcastandnear real time forecastquality focusingona selectionof themost fundamentalattributes (association, discrimination, reliability and resolution). These attributes are measuredusingvariousdeterministicandprobabilisticverificationscores.Suchanattribute-basedframeworkallowstheproductionofverificationinformationinthreelevelsaccordingtotheavailabilityofsub-seasonalhindcastsandnearrealtimeforecastssamples.Theframeworkisalsousefulforsupportingfuture routine sub-seasonal prediction practice by helping forecasters to indentifymodel forecastmeritsanddeficienciesandregionswheretobesttrustthemodelguidanceinformation.Thethreeinformation levels of the proposed framework are defined according to the verification samplingstrategyandarereferredtoastargetweekhindcastverification,allseasonhindcastverification,allseasonnearrealtimeforecastverification.Theframeworkis illustratedusingECMWFsub-seasonalprecipitationpredictions.For the investigatedperiod (australautumn), reasonableaccordancewasindentified between hindcasts and near real time forecast quality across the three levels ofverification informationproduced.Overall, sub-seasonal precipitationpredictions producedone totwoweeks in advancepresentedbetter performance than thoseproduced three to fourweeks inadvance.ThenortheastregionofBrazilconsistentlypresentedfavorablesub-seasonalprecipitationprediction performance through the computed verification scores, particularly in terms ofassociationanddiscriminationattributes.Thisregionwasthereforeidentifiedasaregionwheresub-seasonalpredictionsproducedonetofourweeksinadvancewiththeECMWFmodelaremostlikelyto be successful in South America. When aggregating all predictions over the South Americancontinenttheprobabilisticassessmentshowedmodestdiscriminationability,withpredictionsclearlyrequiringcalibrationforimprovingreliabilityandpossiblycombinationwithpredictionsproducedbyothermodelsforimprovingresolution.Theproposedframeworkisalsousefulforprovidingfeedbackto model developers in identifying strengths and weaknesses for future sub-seasonal predictionssystemsimprovements.


(A3-03)

HowmuchcanModelOutputStatisticsimprovesub-seasonalpredictiveskill?

A.G.Munoz(1),C.A.S.Coelho(2),A.W.Robertson(1)andS.J.Mason(1)

IRI,USA(1),CPTEC,Brazil(2)

Recent research has highlighted the potential for improving predictive skill at the sub-seasonaltimescale,whichcouldbethebasisforenhanced,actionableforecastsforclimateservicesinvolvingwateranddisastermanagement,health,energyandfoodsecurity.TheWMO'sWorldWeatherandWorld Climate Research Programme’s Subseasonal-to-Seasonal Prediction Project (S2S) has madeavailableanextensivedatabasewithbothhindcastsandalmost-realtimeforecastatthistimescale.Leadtimesarelongenoughthatmuchoftheinformationintheatmosphericinitialconditionsislost,but at the same time are too short for other sources of predictability (e.g., ocean boundaryconditions) to have a strong influence in skill. Presently, sub-seasonal skill is still limited, and ingeneral raw uncalibrated forecasts cannot be used to develop climate services. An obviousalternativeistomakeuseofavarietyofrobustbias-correctionandcalibrationmethods--alsoknownas Model Output Statistics, MOS-- available for other timescales, such as the seasonal one.Nonetheless,sometechnical issuescanhinderthisapproach.WediscussproblemsandadvantagesofapplyingMOStosub-seasonalforecasts,analyzingthespatio-temporalvariabilityofskillinseveralmodelsandmethods.


(A3-04)

Regime-dependentpredictabilityandforecasterrorspectraofinitializedforecasts

Berner,Judith

NCAR,USA

Recentworkhasdemonstrated that forecasts initiated fromstates thatprojectontocertain large-scalepatternstypicallyassociatedwithlow-frequencyvariability,canexhibitextendedforecastskill.Soaree.g. forecasts initialized in thenegativephaseof theNorthAtlanticOscillationmore skillfulovertheEuro-Atlanticsectorthanaverage.

Here we will use the S2S database to investigate if these findings carry over to other modes ofvariability,suchase.g.thePacificNorthAmericanpattern(PNA).Wewilldetermine if initializationfrom a positive or negative PNA phase in the S2S database are associated with changes in sub-seasonalpredictability.

Finally we examine the forecast error growth of these initialized forecasts in an attempt to linkextended predictability to the classical predictability theory proposed by Lorenz (1969).


(A3-05)

AdvancingAtmosphericRiverandBlockingForecastsintoSubseasonal-to-SeasonalTimescales

Barnes,Elizabeth(1),Baggett,Cory(1),Maloney,Eric(1),Mayer,Kirsten(1),Mundhenk,Bryan(1),Nardi,Kyle(1),Tseng,Kai-Chih(1)

(1)ColoradoStateUniversity

Atmosphericrivers(ARs)arenarrowcorridorsofhighatmosphericwatervaportransportthatcancausefloodingandhighwindswheretheyoccur.WedemonstrateherethatthepotentialexiststoadvanceforecastleadtimesofatmosphericriversintoS2Stimescalesthroughknowledgeoftwooftheatmosphere’smostprominentoscillations;theMadden-Julianoscillation(MJO)andtheQuasi-biennialoscillation(QBO).

The dynamical relationship between atmospheric rivers, theMJO and theQBO is hypothesized tooccurthroughmodulationofNorthPacificblocking.Wepresentanempiricalpredictionschemeforanomalous atmospheric river activity based solely on theMJO and QBO and demonstrate skillfulsubseasonal“forecastsofopportunity”5+weeksahead.Wethengoontodiscusstheabilityofstate-of-the-artNumericalWeather Predictionmodels (from the S2SDatabase) to forecast atmosphericrivercharacteristicsandtheMJO-inducedblockingpatternsonS2Stimescales.

1


(A3-06)

Identifyingthecapacityofdynamicalmodelstoforecastsubseasonalextremes:Multi-modelensembles

Collins,Dan(1),LaJoie,Emerson(1),Jia,Liwei(1),Strazzo,Sarah(1),Becker,Emily(1)

NOAA,USA(1)

Forecastingtemperatureandprecipitationonsubseasonaltimescalesbeyondtwoweekslead-timeisat the limitsofpredictabilityandmodelingcapabilities.TheNOAAClimatePredictionCenter (CPC)relies on both dynamical and statisticalmodels tomakeoperational and experimental, above andbelowmedian,temperatureandprecipitationforecastsforweeks3-4.Bothstatisticalanddynamicalforecastmodelsattempttoutilizetheenhancedpredictabilityduringactiveclimateevents,relatedto modes of climate variability, such as ENSO and theMadden-Julian Oscillation. Other than thepredictabilityduetodecadaltimescalechangesinclimate,muchoftheskillofsubseasonalforecastsis related to these drivers of climate variability. Furthermore, much of the utility of subseasonalforecastsliesintheforecastofextremesintemperatureandprecipitation,whichbytheirnatureareintermittent and likely associated with high amplitude climate drivers. While a calibrated multi-modelensemble(MME)ofdynamicalmodelforecastshasproventobeoneofthemostskillfultoolsinCPCoperational,subseasonalforecasts,skillremainslowforprecipitationforecastsandattimes,nearzeroforallforecasts.Therefore,identificationofforecastsofopportunity,whenpredictabilityisenhanced,couldgreatlyimprovetheutilityofforecastsonthistimescale.Toanalyzethepotentialtoidentifyforecastsofopportunity,weexaminetheskillofforecastsofextremes(identifiedasthe15thand 85th percentiles from the past climatological distributions) when they are predicted, usinghindcastsfromtheSubXMME.Inthisway,weexamineifintermittentforecastsoflargermagnitudesignalsrepresentthebestopportunitytoobtaininformationforextended-leadsubseasonalforecasts(weeks 3-4) including extremes, and determine appropriatemetrics of forecasts of opportunity. Ithasbeenfoundthatamulti-modelensemblesignificantlyimprovesthecapacitytoidentifyforecastsofopportunityforextremes.


(A3-07)

TheSubseasonalExperiment(SubX)

Pegion,Kathy(1),Kirtman,Ben(2),Collins,Dan(3),Burgman,Rob(4),Lajoie,Emerson(5,3)

GeorgeMasonUniversity(1),RSMAS/U.Miami(2),FloridaInternationalUniversity(3),NOAA/CPC(4),Innovim,Inc(5)

The Subseasonal Experiment (SubX) is a NOAA/Climate Testbed project focused on subseasonalpredictability and predictions. Seven global models have produced seventeen years of ensembleretrospective forecasts initialized weekly to investigate subseasonal prediction and predictability.Additionally,thisprojectbeganproducingreal-timepredictionsinsupportoftheNOAA/NWSClimatePredictionCenterasguidancefortheirweek-3/4outlooksinJuly2017.

Thispresentationwillprovideanoverviewoftheprojectandthepublicallyavailabledatasets.Wewillalsoshowacomprehensiveevaluationofdeterministicandprobabilistic skill forweeks1-4,aswellasmodelbiases,fortheindividualmodelsandmulti-modelensemble,demonstratingthevalueofthemulti-modelensemblepredictionsystem.


(A3-08)

SubseasonalpredictionofwintertimeEastAsiantemperaturebasedonatmosphericteleconnections

Yoo,Changhyun(1),Johnson,Nathaniel(2),Chang,Chueh-Hsin(3),Feldstein,Steven(4),Kim,Young-Ha(1)

EwhaWomansUniv.,Korea(1),PrincetonUniv,.USA(2),NationalTaiwanUniv.,Taiwan(3),PennStateUniv.,USA(4)

AstatisticalmodelisconstructedbasedonthelaggedcompositefieldsassociatedwiththeNorthernHemisphereteleconnectionpatternstopredicttheEastAsianwintertimesurfaceairtemperatureforlead times out to 6 weeks. The prediction skill of the statistical model is compared with that ofhindcast simulations of the Global Seasonal forecasting model version 5. We found that fourteleconnections,i.e.,theWestPacific,EastAtlantic,Scandinavian,andEastAtlantic/WesternRussianteleconnectionpatterns, provide skillful predictionsover EastAsia for lead timesbeyond4weeks.When combinations of the teleconnections is used, the statisticalmodel outperforms the climatemodelforleadtimesbeyond3weeks,especiallyduringthosetimeswhentheteleconnectionsareintheiractivephases.


(A3-09)

ForecastingspringtimeSahelianheatwavesatseasonalandsub-seasonaltimescales

Batté,Lauriane(1),Ardilouze,Constantin(1),Déqué,Michel(1)

CNRMUMR3589(Météo-France/CNRS),France(1)

This study evaluates the skill of Météo-France seasonal and sub-seasonal forecasting systems inanticipatingextremeheatconditionsover theWestAfrican region.Bothsystemsarebasedon theCNRM-CMcoupledglobalclimatemodel.

The skill in capturing interannual variability of minimum, maximum and humidity-correctedapparenttemperatureheatwave indices isanalyzedfortheMarchtoJune(MAMJ)peakseason inre-forecastsinitializedendofJanuaryovera22-yearperiod.Althoughcorrelationislimitedatagrid-point level, significant skill is found for regional-scale indiceswhencomparing tobothERA-InterimandBESTreferencedatasets.

Atthesub-seasonalscale,correlationdropssharplybeyondthedeterministicrange.

Real-time forecasts for the 2016 season are then studiedboth in termsof anomalies andusing aweathertypeapproach.Thisanalysisshedsmorelightonthepotentialinformationandlimitationsofearly-warningsystemsattheseextendedtimescalesandtheconditionalskill relatedto large-scalephenomenasuchasENSO.


(A3-10)

AdvancesinoperationalsubseasonalpredictionofheatandcoldwavesforU.S.cities

Toma,Violeta(1),Curry,Judith(2),Marks,Jesse(3),Webster,Peter(4)

ClimateForecastApplicationsNetwork(1),(3),UniversityofWashington(2)

CFANhasbeenmakingoperationalprobabilisticforecastsofsurfaceairtemperature,heatandcoldwavesforselectedU.S.citiessince2010toaddressneedsoftheenergysectorinanticipatingnaturalgas demand. An innovativemulti-model prediction systemusing theCFSv2 andECMWF forecastshasbeendeveloped to exploit the advantagesof eachmodel. Several categories of heat and coldwavesarepredictedforeachcity,basedonboththeseverityanddurationoftheevent.TheECMWFandCFSv2forecaststreamsarecalibratedagainsthindcastsusingaRankAnalogapproach,togetherwithaPDFmappingtechnique.Additionally,ensembleclusteringisusedseparatelyfortheECMWFandCFSv2forecaststreams,basedonselfclusteringanddominantweatherregimes,thattakesintoaccountcombinationsofwell-knownNHteleconnectionpatterns.Anobjectiveweightingtechniquebased on past performance of each model is used to provide CFAN’s final forecast. A forecastconfidence assessment is made for each forecast based on identification of forecast ‘windows ofopportunity’ determined from the current and predicted strength of teleconnections. Forecastevaluationstatisticsarepresentedasafunctionofregion,seasonandcirculationregime.

SESSION:(A4)S2Sforecastsfordecisionmaking

(A4-01)

Applicationsofsubseasonaltoseasonal(S2S)predictions

White,ChristopherJ.(1,2);Carlsen,Henrik(3);Robertson,AndrewW.(4);Klein,Richard(5);Lazo,JeffreyK.(6);Kumar,Arun(7);Vitart,Frederic(8);CoughlandePerez,Erin(4,9);Ray,AndreaJ.(10);

Brown,TimothyJ.(11)

(1)DepartmentofCivilandEnvironmentalEngineering,UniversityofStrathclyde,Glasgow,UK;(2)AntarcticClimateandEcosystemsCooperativeResearchCentre,Hobart,Australia;(3)Stockholm

EnvironmentInstitute,Stockholm,Sweden

Whilelong-rangeseasonaloutlookshavebeenoperationalformanyyears,untilrecentlythe2weeksto a season timescale – referred to as ’subseasonal-to-seasonal’ (S2S) andwhich sits between themedium- to long-range forecasting timescales – has received relatively little attention. The S2Stimescalehaslongbeenseenasa‘predictabilitydesert’,yetanewgenerationofS2Spredictionsarestarting to bridge the gap betweenweather forecasts and longer-range prediction. Decisions in arangeofsectorsaremadeinthisextended-rangeleadtime,thereforethereisastrongdemandforthisnewgenerationofpredictions.

At least ten international weather centres now have some capability for issuing experimental oroperationalS2Spredictions,includingtheEuropeanCentreforMedium-RangeWeatherForecasting(ECMWF) and the National Oceanic and Atmospheric Administration (NOAA). International effortsarenowunderwaytoidentifykeysourcesofpredictability,improveforecastskillandoperationaliseaspectsofS2Sforecasts,howeverchallengesremaininadvancingthisnewfrontier.IfS2Spredictionsare to be used effectively, it is important that along with scientific advances, we learn how todevelop,communicateandapplytheseforecasts.

Inthispresentation,wepresentthepotentialoftheemergingoperationalS2Sforecaststothewiderweatherandclimateapplicationscommunity.WepresentthefirstcomprehensivereviewofsectoralapplicationsofS2Spredictions, includingpublichealth,disasterpreparedness,watermanagement,energy and agriculture, which was recently published in Meteorological Applications:https://rmets.onlinelibrary.wiley.com/doi/full/10.1002/met.1654. We explore the value ofapplications-relevant S2S predictions, and highlight the opportunities and challenges facing theiruptake. We show how social sciences can be integrated with S2S development – fromcommunicationtodecision-makingandvaluationofforecasts–toenhancethebenefitsof ‘climateservices’ approaches for extended-range forecasting. We highlight the availability of datarepositories of near real-time S2S forecasts and hindcasts, including the WWRP-WCRP(http://apps.ecmwf.int/datasets/data/s2s) and North American Multimodel Ensemble (NMME;http://www.cpc.ncep.noaa.gov/products/NMME/data.html) repositories, and discuss how they arepromotingtheuse(andaidingthedevelopment)ofS2Spredictions.

While S2S forecasting is at a relativelyearly stageofdevelopment,we conclude that it presents asignificant newwindowof opportunity that canbeexplored for application-ready capabilities thatcould allow many sectors the opportunity to systematically plan on a new time horizon.


(A4-02)

Developingnewwatershed-basedclimateforecastproductsforhydrologistsandwatermanagers

SarahBaker(1,2),AndrewWood(3),BalajiRajagopalan(1),FlavioLehner(3),PeitaoPeng(4),andKevinWerner(5)

(1)UniversityofColoradoatBoulder,Boulder,CO,USA;(2)BureauofReclamation,Boulder,CO,USA;(3)NCAR,Boulder,CO,USA;(4)CPC,CollegePark,MD,USA;(5)NOAA,Seattle,WA,USA

Operationalsub-seasonaltoseasonal(S2S)climatepredictionshaveadvancedinskillinrecentyearsbutarenotyetbroadlyutilizedbystakeholdersinthewatermanagementsector.Whilesomeofthechallenges that relate to fundamental predictability are difficult or impossible to surmount, otherhurdles related to forecast product formulation, translation, and accessibility can be directlyaddressed. These include products being misaligned with users’ space-time needs, productsdisseminatedinformatsuserscannoteasilyprocess,andproductsbasedonrawmodeloutputsthatarebiasedrelativetouserclimatologies.Ineachoftheseareas,morecanbedonetobridgethegapbyenhancing theusability,quality,and relevanceofwater-orientedpredictions. Inaddition,waterstakeholderimpactscanbenefitfromshort-rangeextremespredictions(suchas2-3daystormsor1-weekheatwaves)atS2Stime-scales,forwhichfewproductsexist.

Wepresentresultsfromaresearchtooperations(R2O)effortsponsoredbytheNOAAMAPPClimateTestbed to (1) create post-processed climate prediction products to reduce hurdles in waterstakeholderadoption,and to (2) createnewproductsdepicting theprobabilityofextremesatS2Slead times. The project is currently using NCEP’s Climate Forecast System version 2 (CFSv2) andNationalMulti-modelEnsemble(NMME)reforecastsandforecaststodevelopreal-timewatershed-based climate forecastproducts, and to trainpost-processingapproaches toenhance the skill andreliability of raw real-time S2S forecasts. Prototype S2S climate data products (forecasts andassociatedskillanalyses)arenowbeingoperationallydisseminatedfromNCARonapublicwebsitetofacilitatefurtherproductdevelopmentthroughinteractionswithwatermanagers.Products includeCFSv2-basedbi-weeklyclimateforecasts (weeks1-2,2-3,and3-4) forsub-regionalscalehydrologicunits,andNMME-basedmonthlyandseasonalpredictionproducts.

Website:http://hydro.rap.ucar.edu/s2s/


(A4-03)

ImprovingthepredictabilityofstreamflowforhydropowerproductioninCanadausingS2Sensemblemeteorologicalforecasts

Bazile,Rachel(1),Boucher,Marie-Amélie(2),Perreault,Luc(3),Leconte,Robert(4)

UniversitédeSherbrooke,Canada(1),UniversitédeSherbrooke,Canada(2),IREQ,Canada(3),UniversitédeSherbrooke,Canada(4)

Currently,long-termhydrologicalforecastsatHydro-Quebec(Quebec’smainhydro-powerproducer)areExtendedStreamflowPredictions(ESP),derivedfromclimatology.Thisprocessrestsonastrongassumptionofstationaritywhichmaynotholdinachangingclimate.Ourmainworkinghypothesisisthat hydrological forecasts based on dynamicmeteorological forecasts have better predictive skillthanESP.By«better»,wemeantheyaremoreskillfulthanESPforlead-timeslongerthan10days.Inmanyoperationalcontexts,ESPareusedinsteadofdynamicalforecastsbeyondthisperiod.

To verify our hypothesis,we rely on a testbed of ten catchments exploited byHydro-Quebec forhydropower.HSAMI,theconceptualglobalhydrologicalmodelusedoperationallybyHydro-Québec,was successively fed by three different types of ensemblemeteorological forecasts: 1- the 30-dayaheadforecastsproducedbytheEuropeanCenterforMediumrangeWeatherForecasts(ECMWF),retrieved from the global S2S database (Vitart et al. 2017); 2- the 7-month ahead forecasts fromECMWF’s System4 (Molteni et al. 2011); 3- the ECMWF’s SEAS5 forecasts (released in November2017).Inallcases,theforecastedprecipitationandtemperature(minandmax)wereusedasinputstoHSAMI,withdaily time steps.Theperiod from1995 to2014 is the same for the three typesofforecasts,althoughtheyhavedifferentissuedates.

Biasesinrawmeteorologicalforecastswerefirstquantified.Insomecases,biasesweresolargethatrawforecastsdidnotleadtoanygaininpredictabilityandonemightaswelluseclimatologyinsteadof forecasts as inputs to HSAMI. Consequently, biases were removed from precipitation andtemperatureforecastsusingasimplebutefficientlinearscaling.

Hydrological forecasts skillwasassessedusing theContinuousRankedProbabilityScore (CRPS),aswell as the reliability diagram and the rank histogram. ESP were used as a benchmark, as theyrepresent«thesystemtobeat».

Forstreamflowforecasts,performancemetricsanddiagramswerecomputedusing(1)theobservedstreamflow;and (2)asimulationrun forwhichHSAMIwas fedwithprecipitationandtemperatureobservations. The latter allows for an assessment of forecasts that is free from model andobservationuncertainties.However,weconsider important toverify forecasts skill against real-lifeobservations,despitetheassociatedflawsanduncertainties,asthegoalofforecastingispreciselytoanticipatethoseobservations.

The relative performance of the three different types of ensemble forecasts is discussed andcompared according to watersheds and seasons. For most watersheds, streamflow and volumeforecastscomputedfromdynamicalforecastswerefoundtobemoreskillfulthanESPforthefirst30days,which confirms our hypothesis. Surprisingly however, streamflow forecasts based on 30-daymeteorological forecasts from the S2S database did not in general outperform those based onSystem4andSEAS5.

Issues related to the fair comparison of different types of forecastswhich do not share commonissue dates still need to be addressed in future studies, as well as possibilities for a seamlesscombination of medium and long term meteorological forecasts into the same hydrologicalforecastingframework.


(A4-04)

ExperimentalsubseasonalforecastingofatmosphericrivervariationsforthewesternU.S.duringWinters2017-2018and2018-2019

DuaneWaliser(1),MikeDeFlorio(1),AlexGoodman(1),AneeshSubramanian(2),MartyRalph(2),BinGuan(3),FredericVitart(4),JeanineJones(5)

JPL/Caltech/NASAUSA(1),CW3E/UCSDUSA(2),UCLAUSA(3),ECMWFUK(4),CaliforniaDept.WaterResourcesUSA(5)

WeutilizetheGuanandWaliser(2015)atmosphericriver(AR)detectionalgorithmandDeFlorioetal.(2018)ARactivityforecastapproachonthreeoperationalsubseasonalforecastsystems(ECMWF,NCEP,andECCC)topredictARactivityovertheeasternPacificOceanandwesternU.S.,withafocuson subseasonal variations, and in particular the week-3 lead time. The predictand for thesesubseasonalforecastsisthelikelihoodofanARoccurringatanytimeduringagivenweek,withtheprimary focus being the week-3 window. Forecast verification statistics for these subseasonal ARforecasts will be presented, as well as case studies demonstrating periods where the AR activityappeared toexhibit enhanced/diminishedpredictability andhowwell and consistently themodelsperformed during these periods. This is a collaborative activity between the Jet PropulsionLaboratory/NASA, the Center for Western Weather and Water Extremes of the University ofCalifornia at SanDiego, and the Joint Institute forRegional Earth SystemScience (JIFRESSE) at theUniversity of California, Los Angeles, with sponsorship from the California Department of WaterResources. This activity leverages the ECMWF, NCEP and ECCC hindcasts of the Subseasonal toSeasonal (S2S) Prediction Project, along with their real-time data stream counterparts, andrepresentsoneoftheS2SPredictionProject'sPilotProjectsforapplicationsuse.Wewillalsodiscussthe"operational" framework for thecollaboration, the interferences fromthe2017-18winter,andchangesandplansforthe2018-19winter.


(A4-05)

StratosphericinfluencesonEuropeanmonth-aheadwindpowergenerationanditspredictabilityonsubseasonaltimescales

Büeler,Dominik(1,2),Beerli,Remo(3,2),Grams,ChristianM.(1,2),Wernli,Heini(2)

IMK-TRO,KIT,Germany(1),IAC,ETHZurich,Switzerland(2),AXPOTradingAG,Switzerland(3)

Wind power is playing an increasingly important role in Europe’s electricity generation. Accurateforecastsofwind-poweroutputonvariousspatialandtemporalscalesarethereforeofhighinterestfor the energy industry. However, predictability of near-surface wind particularly on subseasonaltimescales has received relatively little attention. The stratosphere is an important source ofsubseasonal predictability in winter. We thus investigate how the lower-stratospheric circulationaffectsmonth-aheadwindelectricitygeneration inEuropeanwinterandhowthiseffect influencesthe skill of subseasonal numerical weather forecasts. In a first step, we use the ERA-Interimreanalysis and the wind-power dataset Renewables.ninja to demonstrate a strong correlationbetween the strength of the lower-stratospheric circulation and month-ahead wind electricitygeneration in different regionsof Europe. This relationship exists due to episodesof troposphere-stratospherecoupling,which lead toprolongedperiodsofeither thepositiveornegativephaseoftheNorthAtlanticOscillation(NAO).SincethesepersistentNAOperiodsareassociatedwithstrongsurfacewindanomalies,theyhaveanimportantimpactonwindelectricitygeneration,inparticularin Northern Europe. Motivated by this empirical relationship, we develop a simple statisticalforecasting approach based on the strength of the lower-stratospheric circulation,which providesskillful forecasts of month-ahead wind electricity generation in Europe. In a second step, weinvestigate the skill of different subseasonal forecastmodels from the S2S database in predictingmonth-ahead10-mwindspeedasaproxyforwindelectricitygeneration.TheskilloftheS2Smodelsisgenerallyhigherthantheskillofthesimplestatisticalforecast,particularlyforshortleadtimes.Itissubstantially drivenby the strengthof the lower-stratospheric circulationat initialization timeandthe associated stateof theNAO throughout the forecast,which reflects the empirical relationshipfromthe reanalysisdataalso in themodels.However, thereare substantialdifferences in the skillbetweendifferentEuropeanregionsaswellasmodels,withimplicationsforboththeenergyindustryand the numerical modeling community. In a future step, the study will be expanded on othermeteorologicalfieldsrelevantfortheenergyindustry.


(A4-06)

Sub-seasonaltoseasonalclimatepredictionsforenergy:theS2S4Eproject

Soret,Albert(1),Lledó,Llorenç(1),Manrique-Suñén,Andrea(1),Torralba,Verónica(1),Cortesi,Nicola(1),González-Reviriego,Nube(1),Bretonnière,Pierre-Antoine(1),Christel,Isadora(1),Doblas-

Reyes,FranciscoJ.(1,2)

BarcelonaSupercomputingCenter(BSC),Barcelona,Spain(1),InstitucióCatalanadeRecercaiEstudisAvançats(ICREA),Barcelona,Spain(2)

Despite becoming cost competitive in many settings, renewable energy diffusion remains limitedlargelydue to itsgenerationvariability.To foster renewableenergydeploymentwhilemaintainingenergysecurity,theS2S4Eprojectwillprovidesub-seasonaltoseasonal(S2S)climateforecasts.ThemainobjectiveofS2S4EistomaketheEuropeanenergysectormoreresilienttoclimatevariabilityandhighimpactevents,suchasheatwaves,byexploringthepotentialofS2Spredictionstailoredtousers’ needs. This information will enable the energy industry to assess the renewable energysourcescapacitytomeetdemandoverextendedtimehorizons(weekstomonths),focusingontheimpactofclimatevariablesonenergyoutputs.Basedonauser-centricapproachofclimateservicesdevelopment,S2S4Ewillprovideaccesstotailoredreal-timeclimatepredictionproductstooptimisedecisionmakingacrossalllevelsoftheenergysectorcommunity.ADecisionSupportTool(DST)willbe co-designed and co-developed with relevant industrial partners of the consortium whichrepresentdifferentneedsand interests in termsof regions, renewableenergysources (wind, solarand hydro) and electricity demand. TheDSTwill be developed taking into account eight historicalcasestudiespointedasthemostrelevantbyindustrialpartners,i.e.periodswithanunusualclimatebehaviouraffectingtheenergymarket.ToillustrateS2Spotential,oneofthesecasestudieswillbepresented.Inwinter2016asignificantdecreaseofwindspeedwasobservedthroughallcentralandsouthernEuropeleadingtoareductionofwindenergygeneration.Atthesametimethecoldwaveobserved inDecember had a significant impact on the power system. It created a combination oflargeincreaseinelectricitydemandandlowerthanusualrenewableenergygeneration.


(A4-07)

ExcessiveHeatEventsandHealth:BuildingResiliencebasedonGlobalScaleSubseasonal-to-SeasonalExcessiveHeatOutlookSystems

Vintzileos,Augustin

UMCP-ESSIC.USA

Excessive heat events (EHE) are the primary cause for mortality resulting from atmosphericextremes.AsthepopulationbecomesolderandEHEintensityandfrequencyisincreasingmortalityisexpectedtogrowandthusearlywarningsystemsbecomecrucial.Data fromtheS2Sdatabaseareused to demonstrate the feasibility of S2S forecasting of EHE. The paper concludes with thepresentationofanexperimentalS2Squasi-operationalexcessiveheatoutlooksystemthatfocusesonhumanhealth.


(A4-08)

Usinghydrologicpredictionskillelasticitytoquantifythebenefitsofs2sclimateinformationforhydrologicforecasting

Wood,Andrew(1),Clark,Martyn(1),Hopson,Tom(1),Arnold,Jeff(2),Brekke,Levi(3),andNowak,Kenneth (3)

(1)NationalCenterforAtmospheric Research,Boulder,CO,UnitedStates, (2)USArmyCorpsofEngineers, Seattle,WA,UnitedStates, (3)BureauofReclamation, Lakewood,CO,UnitedStates

Waterresourcesdecision-making commonly reliesonmonthly toseasonalstreamflow forecastsamong other kindsof information. The skill of suchpredictions derives from theability toestimate awatershed’sinitial moistureand energyconditionsand to forecastfutureweatherand climate.Arecent project sponsored by theUSwater agencies (theUSArmyCorpsofEngineers and theBureauof Reclamation) investigated the role of each sourceof predictabilityat S2S time scales to assesswhere and when improvements in each area can improve streamflow forecasts for use in watermanagement. The study used hydrologic simulation models for 424 US watersheds in an idealizedpredictability framework to characterize the influence of varying levels of skill in each predictabilityareaonstreamflow prediction skill.Itenabled thecalculation ofderivatives inhydrologicpredictability (ie, skill elasticities) throughout the initial conditions and future forcing skill space.Wefind that regional and seasonal variations inwatershed hydro-climatologystrongly control therelative importance of initial hydrologic conditions and S2S climate forecasts, leading to strikingdifferences between rainfall driven and snowmelt drivenwatersheds. The resulting analysis providesinsights on the relative benefits of investments toward improving watershed monitoring (throughmodeling and measurement) versus improved climate forecasting and application. A somewhatcounterintuitivebut encouragingfinding was that climate forecast skill improvementscan beamplified instreamflowprediction skill,whichmeans thatclimate forecastsmayhavegreaterbenefitforS2Sflowforecasting thanmaybeexpected fromclimate forecastskillalone.


(A5-01)

TheLandSurface“SweetSpot”BetweenWeatherandClimate

Dirmeyer,Paul

COLA/GeorgeMasonUniversity

Weatherforecastingistypicallyconsideredadeterministicinitialvalueproblemoftheatmosphere.Climate on seasonal to decadal time scales is strongly driven by ocean surface temperatures andnear-surfaceheatcontent,andforecastsattheselongertimescalesareprobabilisticinnature.Inthedifficulttransitionregionbetweendeterministicandprobabilisticforecasts,the landsurfacehas itsgreatest potential impact on numerical forecasts. Anomalies in land surface states affect theatmospherethroughanomaloussurfacefluxesthatmanifestviathediurnalcycle.Thisoccurswhereandwhentheatmosphereissensitivetoits lowerboundary,wherethelandsurfaceanomaliesarelargeenoughandpersistentenoughtohaveasignificanteffect.This terrestrial sourceofpotentialpredictability isa“low-hanging fruit”buthasbeenunder-exploited inoperational forecasting foranumber of reasons. Reliable, high-quality real-time observations of land surface states needed toinitializethiscomponentofforecastmodelsareonlynowbecomingwidelyavailable,andlanddataassimilation systems that can ingest this information are relatively new. Models have not beendeveloped, calibrated or validated with regards to the coupled processes that link land andatmosphere,partiallybecauseofthehistoricallackofnecessarysurface,near-surfaceandboundary-layer data, and partly because the culture of model development has been fragmented alongtraditional disciplinary lines. This has resulted in both land surface and atmospheric modelersspendingmuch time and effort trying to compensate for the systematic errors passed across themodelinterface,ratherthandevelopinganappropriatelycoupledmodelsystem.Theopportunityisnowtorealizethepotentialpredictabilityoffromthelandsurface,whoseinherentmemoryisatthesubseasonaltimescalesthatareofsuchgreatinteresttoday.

SESSION:(A5)Landinitializationandprocesses

(A5-02)

Theroleofthemidlatitudeoceaninsub-seasonalprediction

Saravanan,Ramalingam

AtmosphericSciencesatTexasA&MUniversity,CollegeStation

Intrinsic oceanic variability exhibits longer time scales and smaller spatial scales than atmosphericvariability.Thisisdue,inpart,tothelargerheatcapacityoftheoceanand,inthemiddlelatitudes,tothe smaller Rossby radius of deformation. Due to the slowly evolving nature of the ocean, theoceanicstateisusuallyassumedtobeinvariantforthepurposesofweatherprediction.However,onsubseasonaltimescales,theoceanicstatedoesevolve,especiallyonthefinescalesassociatedwithoceanic frontsandmesoscaleeddies. This talkwill survey recent studies thathave focusedon theinfluenceofthemidlatitudeonoceanicvariabilityonatmosphericstormtracks.Thisinteractionhaspotential implications for subseasonal forecast skill. Dry dynamics is unlikely to be responsible forthisinteraction,duetothemismatchintheRossbyradiusofdeformationbetweentheatmosphereandtheocean.Thissuggestsaroleformoistdynamicsinthisinteraction.


(A5-03)

Animprovedapproachtoland-surfaceinitializationintheMetOffice'sGlobalSeasonalForecastingSystem(GloSea)

Davis,Philip(1),Comer,Ruth(1),Fereday,David(1),Knight,Jeff(1),MacLachlan,Craig(1),Scaife,Adam(1)

MetOffice,HadleyCentre,UK(1)

The landsurface isa crucial component in theclimatesystem; theexchangeofheatandmoisturefluxbetweenthelandandatmospherehasanimportant impactonnear-surfacetemperaturesandprecipitation.

Here,wedescribe experiments to initialize the land surface (notably soilmoisture) in theUKMetOffice’s (MO) state-of-the-art Global Seasonal Forecasting System (GloSea). GloSea employs acoupledatmosphere-oceanmodel,usingMO’sUnifiedModelandNucleusforEuropeanModelingoftheOcean.LandinteractionsaremodeledusingtheJointUKLandEnvironmentSimulator(JULES).

Due to the challenge in obtaining consistent information for both the historical and real-timeperiods,wehavetoresorttousingaclimatologyforbothhindcastsandforecasts.Inconsistenciesinthe initialization (and therefore the forecast/ hindcast model climatology) can result in a biasedforecast.Thisworkhopestoimprovethecurrentinitializationscheme.

Owing to the availability of real-time data, we investigate land-surface initialization using theJapanese55-yearReanalysisProject(JRA-55),providedbytheJapaneseMeteorologicalAgency.Ourgoal is toreplacetheexistingclimatologyusedfortheforecastswithsoilmoisturecalculatedfromthedailydata.Hindcasts,asforourexperiments,willbeinitializedusingatime-seriesfromaJULESreanalysisforcedwiththeJRA-55data.

We discuss the impact of the new initialization scheme on standard skill scores, as well as casestudiesoftheEuropeanandRussianheat-wavesof2003and2010.


(A5-04)

Land-surfaceinitialisationaffectsIndianmonsoonsubseasonalpredictability

Tuinenburg,Obbe

UtrechtUniversity,TheNetherlands

Theinfluenceofthelandsurfacewetnessstateonprecipitationisknowntovaryacrosstheworld.Inareaswherethereexistssuchastrongcouplingbetweenlandandatmosphere,landsurfacewetnessstate information can be used to increase the predictability of precipitation on several timescale,varyingfromthedailytotheseasonaltimescale.

Whether there is a strong coupling between land surface and precipitation depends on how thesurfaceenergybalanceaffectsturbulentatmosphericboundarylayerprocessesandcloudformation.Iftheatmosphereistoodry,noprecipitationwilloccurregardlessoflandsurfaceconditions.Iftheatmosphere is too moist, precipitation will always occur regardless of land surface conditions.Moreover, there should be some conditional instability the to allow convective precipitation tooccur.

India is a location with a strong coupling between land and atmosphere.Moreover, atmosphericmodelshaveastrongdrybiasduringthemonsoonseason.

Here, we will study Indias summer monsoon predictability based on (sub-)seasonal atmosphericsimulations (S2S-project archive) up to the seasonal timescale. We specifically analyse theprecipitation difference between ensemble members and regress these differences against thedifferentlandsurfaceinitialisationsoftheensemblemembers.Thisregressionisusedtodeterminehowsensitivemonsoonprecipitationistolandsurfaceconditionsforeachmodel.

Finally,weanalyse this sensitivitybasedon local land-atmospherecouplingaswellasatmosphericmoisturetransport.

Initial results show that there is a strong relation between initialised surface wetness and Indiasummermonsoon precipitation. An analysis of land surface wetness-precipitation coupling in theS2S-database,aswellasitsatmosphericmoisturetransportpathways,willbepresented..

SESSION:(A6)Oceaninitializationandprocesses

(A6-01)

Impactofoceanobservationsystemsonoceananalysesandsubseasonalforecasts

Subramanian,Aneesh(1),Vitart,Frederic(2),Balmaseda,Magdalena(2),Zuo,Hao(2),Matsueda,Mio(3),Ralph,Marty(1)

SIO-UCSD,USA(1),ECMWF,UK(2),TsukubaUniversity,Japan(3)

We evaluate the relative merits of different ocean observation systems (moored buoys, Argo,satellite,XBTs,andothers)bytheirimpactonoceananalysesandsubseasonalforecastskill.Severalocean analyseswere performedwhere different oceanobservation platformswerewithheld fromtheassimilationinadditiontooneoceananalysiswhereallobservationswereassimilated.Wethenuse these ocean analyses products for initializing a set of subseasonal forecasts to evaluate theimpact of different ocean analyses states on the forecast skill. We use the European Centre forMedium-RangeWeather Forecasts (ECMWF) ensembleprediction system for the twenty-year sub-seasonal hindcast experiments. Results from these hindcast experiments will be presented tohighlight changes in the ocean analyses states and their impact on the forecast skill of theMJO,monsoon intraseasonal oscillations, atmospheric rivers as well as global temperature andprecipitation.

Coupled air-sea interaction processes relevant to intraseasonal variability (e.g. the MJO, MISO,atmospheric blocking) in the earth’s climate system are inadequately represented in regional andglobalcoupledmodels.Theseinaccuraciescouldberelatedtoeitherpoorparameterizationofmodelphysicsorinsufficientmodelresolutiontoresolvethecriticalprocesses.Neweffortsinobservations,process understanding, and translation into weather and climate models are necessary forimprovements in simulationandpredictionof the intraseasonal variability andassociatedweatherevents.Wewilldiscussthemeritsofdifferentoceanobservationplatformsinthiscontextandalsofutureobservationandmodelimprovementpathways.

SESSION:(A6)Oceaninitializationandprocesses

(A6-02)

SeaIceandFillingDataGapsforS2SPrediction

ChidongZhang

NOAAPMEL

SeaicehasbeenidentifiedasoneofthemajorsourcesofpredictabilityonS2Stimescales.Predictionof sea ice itselfhasbeenachallenge.HowS2Spredictiondependsonourabilityofunderstandingand predicting sea-ice variability has yet to be determined. Satellite observations of sea-iceenvironmentarerudimentalatbest. Insituobservationsfromopenwateradjacenttoseaicehavebeen unavailable up to recently. New observing technology has demonstrated that observationalmeasurementof theupperocean,atmosphereand their interfacenear theedgeof sea ice isnowpossible and promising. Further explorations are needed to optimally use such observations tobenefitS2Sprediction.ThisisbutoneexampleofissuesrelatedtodataneedforS2SpredictionthatshouldbeaddressedbytheS2SProject.

SESSION:(B1)MechanismsofS2Dpredictability

(B1-01)

OnthemechanismsthatgiverisetopredictabilityonSeasonal-to-decadaltime-scales

Robson,Jon

UniversityofReading

Providingthebestclimateinformationavailableforsmallerregions,andonshortertimescales, isacrucialgoalforClimateScienceinordertoaidplanningandadaptationdecisions.Therefore,overthelast10-15years,orso,therehasbeenanexplosionintheinterest insocalled“near-term”climatepredictions, which aim to predict both the regional impact of forced changes and the internalvariability. There isnow significantevidence showing that initialising coupled climatemodels fromobservations leads to improved retrospective predictions of a range of variables on inter-annual,multi-yearanddecadaltime-scales.However,toconfidentlypredictfuturevariability,itiscrucialthatthemechanismsthatgiverisetopredictiveskillonthesetime-scalesareunderstood.Aspredictionisalso theultimate testofourunderstandingandofourmodels,climatepredictionsonseasonal-to-decadaltime-scalesalsorepresentapowerfultooltoexploreourunderstandingofregionalclimatedynamics, and to improve our models. Therefore, in this talk I’ll give a brief overview of themechanisms that are important for delivering skill on different time-scales. I’ll also provide somecommentary on what mechanisms we think are more robust, and those which need moreunderstanding. Finally, I’ll discuss ways in which novel experiments using seasonal-to-decadalpredictionscanbeusedtoprobeourunderstandingfurther.


(B1-02)

InitializeddecadalpredictionfortransitiontopositivephaseoftheInterdecadalPacificOscillationandresumptionoflargerratesofglobalwarming

Meehl,GeraldA.(1),Hu,Aixue(1),Teng,Haiyan(1)

NCAR,USA(1)

Thenegativephaseof the InterdecadalPacificOscillation (IPO),adominantmodeofmulti-decadalvariabilityofseasurfacetemperatures(SSTs)inthePacific,contributedtothereducedrateofglobalsurface temperature warming in the early 2000s. A proposed mechanism for IPO multidecadalvariabilityindicatesthattherelativemagnitudeofdecadaltimescaleupperoceanheatcontentintheoff-equatorialwesterntropicalPacificcouldprovideconditionsforaninterannualElNiño/SouthernOscillation (ENSO) event to trigger a transition of tropical Pacific SSTs to the opposite IPO phase.EvidenceispresentedfromasetofinitializedhindcastswithCCSM4toshowthisroleforElNiñointhe1970stransitiontopositive IPO,andforLaNiña inthe late1990stransitiontonegative IPO.Adecadal prediction initialized in 2013 shows that the Niño3.4 SSTs qualitatively tracked theobservationsthrough2015.Theyear3-7averageprediction(2015-2019)fromthe2013initialstateshows a transition to the positive phase of the IPO from the previous negative phase, and aresumptionoflargerratesofglobalwarmingoverthe2013-2022periodconsistentwithapredictedpositiveIPOphase.


(B1-03)

TropicalAtlanticimpactsonsubdecadalvariabilityinthePacific

Mochizuki,Takashi(1),Watanabe,Masahiro(2),Kimoto,Masahide(2)

JAMSTEC,Japan(1),TheUniversityofTokyo,Japan(2)

We demonstrate the significant impact of the tropical Atlantic Ocean on the Pacific climate onsubdecadal timescales, by the so-called pacemaker experiments (partial data assimilationexperiments). In mid-2000s when the positive peak of the Atlantic Multidecadal Oscillation wasobserved, the high sea surface temperature over the tropical Atlantic Ocean strengthened theequatorial Pacific trade wind and worked to keep warm and cold tendencies in the western andeasternPacificOceans,respectively.Inaddition,ananti-cyclonicsurfacewindanomalyformedintheoff-equatorial area of the North Pacific works to raise the upper ocean heat content below andcontributes to the subsequent warming in the western Pacific Ocean. These changes of the off-equatorial surface winds are similar to common responses to seasonal warming of the tropicalAtlanticOcean,whilethedeepeningoftheoceanthermoclineisalsosimulatedasdynamicaloceanresponse on subdecadal timescales. Our CMIP5 decadal hindcasts with initialization insufficientlyreproducethissubdecadalmodulationinmid-2000sevenafewyearsinadvance.


(B1-04)

Multi-yearPredictabilityofTotalSoilWater,Drought,andWildfireovertheGlobe

June-YiLee(1),AxelTimmermann(2),YoshiitsuChikamoto(3)

PusanNationalUniversity(1),IBSCenterforClimatePhysics(2),UtahStateUniversity,Logan(3)

Severedroughtandincreasedchanceinwildfireoccurrencehavesignificantimpactstoawiderangeofsectorssuchasagriculture,energy,foodsecurity,forestry,drinkingwater,andtourism.Thisstudyaimstoassessmulti-yearpredictabilityandpredictionskillfortotalsoilwater,drought,andwildfireoccurrenceovertheGlobeusingamulti-yeardynamicalpredictionsystembasedontheCommunityEarth System Model version 1.0.3 and to better understand sources of their predictability. Wedemonstratethatthedynamicalpredictionsystemhasahighdegreeofskillinforecastingtotalsoilwater, drought, andwildfire probabilities up to 2~4-year lead time overmany parts of the GlobeincludingthesouthernpartofNorthandSouthAmerica,CentralAmerica,thenorthernpartofSouthAfrica, Maritime Continents, Europe and Asia. The important sources of multi-year predictabilityidentifiedinthestudyincludetheTrans-basinvariability(TBV)betweentheAtlanticandPacificseasurfacetemperature(SST),thelow-passfilteringcharacteristicsofsoils,andanthropogenicradiativeforcing.Inparticular,thepositivephaseofTBV,characterizedbytherelativelywarmerSSTovertheAtlanticthanthePacific,isfavorableforlessprecipitation,lesssoilwater,drought,andmorewildfireoccurrenceoverthesouthernpartofNorthandSouthAmerica,thenorthernpartofSouthAfricaandmanypartsofEuropeandAsia.However,theoppositeconditiontendstoprevailinCentralAmerica,the southern part of South Africa and the Maritime Continent. The multi-year predictability ofdroughtandwildfireoccurrencecanbeutilizedforHumanitarianCrisisManagement.


(B1-05)

ThePacificDecadalPrecession:Ourcurrentunderstandingofitsdynamics,regionalclimateeffects,andpredictability

Anderson,Bruce(1),Furtado,Jason(2),DiLorenzo,Emanuele(3),Capotondi,Antoinette(4)

BostonUniversity,USA(1),UniversityofOklahoma(2),GeorgiaInstituteofTechnology(3),UniversityofColorado(4)

Eventsof recentyearshighlight theprofound impactofdecadal-scaleclimate shiftsuponphysical,biological and socioeconomic systems. Previously, research tounderstand, anticipate, andpreparefor the regional effects that accompanydecadal-scale climate shifts invokedwell-knownmodesofdecadalclimatevariabilityandchange—e.g.,theAtlanticMultidecadalOscillation(AMO),thePacificDecadalOscillation (PDO), and theNorthPacificGyreOscillation (NPGO).Herewewill discuss thesourcesandphysicalprocessesgivingrisetoarecentlyrevealedmodeofdecadalclimatevariabilitytermed the Pacific Decadal Precession (PDP), a ~10-year counter-clockwise progression of anatmospheric pressure dipole around the North Pacific. During its progression, the PDP hasteleconnected links tomultiple climateextremes including: sustaineddroughts across thewesternandcentralUS;enhancedfireseverityinAlaskaandCalifornia;thepropensityformorefrequentcoldextremesovertheeasternUS;andtheformationandpersistenceofprolongedmarineheatwavesinthe Northeast Pacific. Further, it has signatures that extend from the tropical Pacific subsurfacethroughtotheArcticstratosphere.InthistalkwewillcharacterizethePDP’slocalandteleconnectedinteractionswith,andimpactson,multipleearthsystemcomponents,includingatmosphere,ocean,terrestrial,andcryospheric systems.Wewill alsoanalyzeanddiagnose theunderlyingphenomenaandprocessesthatsustainthePDPanditsregionalclimateeffects.Finally,wewillstarttoexaminewhich components of the PDP’s evolution are statistically and/or dynamically predictable; whichregional-scale effects are impacted by these predictable components; what oceanic, atmospheric,and/orterrestrialconditionssustainthesepredictablecomponents;andwhatunderlyingprocessesare fundamental for generating thesepredictable components.Overall, understanding the sourcesandphysicalprocesses giving rise to thepredictabilityof thePDP’sevolutionwill provide valuableinformationtomanycommunitiesandallowthemtoanticipateandprepareforthesocial,economic,environmentalimpactsofitsdecadal-andregional-scaleclimateeffects.


(B1-06)

DecadalvariabilityandpredictabilityintheSouthernOcean-implicationsforinterpretingrecentobservedtrends

Delworth,Thomas(1),Zhang,Liping(1,2),Cooke,William(1,3),Yang,Xiaosong(1,3)

GFDL/NOAA(1),PrincetonUniversity(2),UCAR,USA(3)

While decadal variability and predictability in the North Atlantic and North Pacific have receivedconsiderable attention, there has been less work on decadal variability and predictability in theSouthernOcean.Asshownpreviously,acoherentmodeofdecadaltocentennialvariabilityexistsinmultiple climate models. The mechanism involves a multidecadal accumulation of heat in thesubsurface of the Southern Ocean. This accumulation of heat in the subsurface tends to reduceocean stratification, eventually leading to the onset of intense oceanic convection and venting ofheat to the atmosphere. The discharge of heat from the interior ocean, combined with surfacefreshening, restratifies the water column and the cycle begins again with the accumulation ofsubsurface heat. During the phase when the accumulated subsurface heat is released throughoceanicconvection,therecanbeconsiderableregionalscaleclimaticimpacts,alongwithsubstantialimpactsonoceanheatuptake.Usinga largesuiteofperfectpredictabilityexperiments, inconcertwith long control simulations and experimental hindcasts,we show that this variability has a highdegree of predictability on decadal scales. We present further results that show this type ofvariability may play an important role for interpreting recently observed trends of sea ice andtemperature in the Southern Ocean. Specifically, observed trends over the last several decadesresemble a particular phase of this variability in which reduced oceanic convection leads tosubsurfacewarmingandsurfacecoolingandfreshening,withassociatedincreasesinseaiceextent.Thisphaseofnaturalvariabilitymaysubstantiallycontributetoobserveddecadaltrends,workinginconcertwithotherfactors.


(B1-07)

VariabilityandTeleconnectionsintheIndianOcean:Mechanisms,PredictabilityandClimaticInfluence

Chapman,Christopher(1),BernadetteSloyan(1)

CSIROOceansandAtmosphere(1)

TheIndianOceanbasinishosttovariabilityonavarietyofspatio-temporalscalesthatexertastronginfluence on the climate system as a whole, through influences on the Asian and AustralianMonsoons, the IndianOceanDipole (IOD) and theHadleyCell.However, howvarianceenters andpropagates throughout the IndianOcean,how itultimately imprintsonthegreaterclimatesystemand the potential predictability on time-scales exceeding 12 months are not currently wellunderstoodandimpedecurrentforecastingsystems.

Using a combination of ocean observation, a long, fully coupled control simulation, and a shorterdata-assimilatingensembleofsimulations,weinvestigatethephysicalmechanismsdrivingvariabilityon annual to decadal time-scales in the Indian Ocean. We demonstrate using band-selectiveperidograms that, in contrast to variability on sub-annual timescales, variance at time scalesexceeding 12months lies not in the tropics, but in the SouthernHemisphere sub-tropical regionsassociatedwithmode-water formation.Teleconnectionpathwaysbetweenthesesub-tropicalhigh-varianceregionsandthetropicswillbeelucidated,showingseveraldistinctpathwayslinkingmidandlow latitudes over longer time scales. By studying disturbances propagating along theseteleconnection pathways, we are able to determine characteristic time-scales of propagation andpersistence,aswellastheirdominantlength-scales.Thedisturbancesareshowntobelarge,multi-scaleandtopropagatesignificantlyslowerthanlinearRossbywaves,suggestiveofadominantrolebeingplayedbynon-linearityintheunderlyingdynamics.Thepotentialpredictabilityofdisturbancespropagating along these pathways is then briefly discussed using the results of the large, data-assimilatingensembleofsimulations.

Finally, the influenceofoceanic signals emanating from the subtropical Southern IndianOceanontheclimatesystemwillbediscussed,primarilythroughtheirinfluenceontheanomaloussea-surfaceheightsandsea-surfacetemperatures.


(B1-08)

DynamicalandthermodynamicalimpactsoftheAtlanticMultidecadalVariabilityontheEuropeanclimate

Cassou,Christophe(1),Qasmi,Saïd(1),Boé,Julien(1)

CECI,UniversitédeToulouse,CNRS,Cerfacs,Toulouse,France

The Atlantic Multidecadal Variability (AMV) is known for influencing the mid-latitudes climatevariability. The physical mechanisms between the AMV and the European winter climate areassessedwiththeCNRM-CM5modelwithlargeensemblesofsimulationsinwhichtheNorthAtlanticSeaSurfaceTemperature (SST) is restored toSSTanomaliescharacterizing theobservedAMV.Therestoringpattern isobtainedwith thesuperimpositionofobservedSSTanomalies (computed fromthe regression of the observed interannual SST on the standardized AMV index) on the model'smonthlyclimatology.TheinfluenceoftheAMVamplitudeontheteleconnectionhasbeenevaluatedwith three sets of 40-member of 10 years for each AMV phase. Each set corresponds to SSTanomalies related to one, two or three standard deviations of the observed AMV (respectivelynamed1*AMV,2*AMVand3*AMVexperiments).Eachmemberstarts froma randomdateof theso-called pre-industrial control run (a 850-yr long integrationwhere all external forcings are keptconstanttotheirestimated1850pre-industrialvalues).

We find that a change of the AMV phase in the 1*AMV experiment is not associated with asignificant anomaly of surface air temperature (SAT) over Europe. Conversely, in the 2*AMV and3*AMVexperiments,asignificantwarmingisfoundupto~0.5°CoverwesternEurope,andupto1°CoverScandinavia.Precipitationchangesinthe1*AMVexperimentarenotsignificantbutanincreaseof~6%isfoundovercentralEuropeinthe2*AMVand3*AMVexperiments.

WequantifytheimpactsoftheAMVontheEuropeanclimatewithananalogmethodseparatingtheinfluence of large-scale circulation changes from the thermodynamical effect due to the SSTvariability. The decomposition shows that SAT over Europe in the 1*AMV experiment is equallydriven by both the circulation anomalies and the residual (mainly containing the thermodynamiceffects).Inthe2*AMVand3*AMVexperiments,SATanomaliesaredrivenbytheresidualwhiletheamplitude of the dynamical response remains stable, at the 1*AMV level. For the precipitationanomalies, both dynamical and residual influences contribute to the total response in the threeexperiments.Weshowthat the residual component ismainlydrivenby theadvectionofheatandhumidity from the ocean by the mean westerly flow. The dynamical component related to thecirculationanomaliesoverEuropeisfoundtobelikelyexplainedbythetropicalAtlanticwhereaGill-like response influences the position and the speed of the jet, which enhances the northwardpropagationofRossbywavesduringapositivephaseoftheAMV.AninfluencefromtheNorthPacificisalsodetectedwiththepropagationoflongerRossbywavespotentiallyinteractingwiththeshorteronespropagatingfromthetropicalAtlantic.


(B1-09)

AntarcticstratosphericozoneandseasonalpredictabilityoversouthernAfrica

Engelbrecht,Francois(1,2),Beraki,Asmerom(1)

CSIR,SouthAfrica(1),UniversityoftheWitwatersrand,SouthAfrica(2)

SeasonalpredictiveskilloversouthernAfricaishighestduringthemid-summerperiodofDecembertoFebruary,andparticularlysoduringseasonsassociatedwithElNiñoorLaNiñaevents.Forecastskill is significantly less for seasons associated with a neutral El Niño Southern Oscillation (ENSO)state.Most ElNiño events are associatedwith drought over southernAfrica, but the signal is notlinear,andthereareinstanceswheretheteleconnectionhasfailed.AkeyexampleofthelatteristhesuperElNiñoof1997/98,whichwasassociatedwithnormalrainfalloversouthernAfrica.Seasonalforecasts of mid-summer rainfall have also been demonstrated to be overconfident in predictingdroughtoversouthernAfricaduringseasonsassociatedwithElNiñoevents.InthisstudyweaddresstheroleofalargelyunexploredpotentialsourceofseasonalpredictabilityofsouthernAfricaclimate,namely the radiative forcing originating from anomalous Antarctic stratospheric ozoneconcentrations. The focus is specifically on the prediction of mid-summer rainfall and circulationpatterns over southern Africa, given that the critical period for radiative forcing from Antarcticstratospheric ozone is the austral springwhen sunlight returns to the South Pole. The signal thatoriginates in the Antarctic stratosphere in spring from anomalous ozone concentrations has longbeenknowntoaffect theSouthernHemisphere troposphere in lower latitudeswitha lagofa fewmonths. However, impacts onmid-summer climate variability, and associated predictability, haveremainedlargelyunexplored.OurfocusisondeterminingwhetherthisformofradiativeforcingcanprovideusefulskillinthepredictionofrainfalloversouthernAfricaduringENSOneutralseasons,andalso whether anomalous Antarctic stratospheric ozone forcing has an impact on the ENSOteleconnectiontosouthernAfrica(andthusonpredictability).

The investigation relies on a set of Atmospheric Model Intercomparison Project (AMIP) - stylesensitivity experiments performed using the conformal-cubic atmospheric model (CCAM) of theCommonwealthScientificandIndustrialResearchOrganisation(CSIRO).Themodelwasforcedwithsea-surfacetemperatures(SSTs)andsea-iceconcentrationsfortheperiod1871-2015obtainedfromtheCoupledModelIntercomparisonProjectPhaseSix(CMIP6)database.OzoneandcarbondioxideconcentrationsaswellasaerosolemissionsareprescribedinthesimulationsusingCMIP5forcings.Inthe simulations CCAM ran fully coupled to the dynamic land-surface model CABLE (CSIROAtmosphereBiosphereLandExchange)andutilisedaprognosticaerosolscheme.TheCCAM-CABLEsystem was applied at a resolution of about 200 km in the horizontal and with 27 levels in thevertical.Atwelvememberensembleofsimulationswasgenerated,witheachmemberusingslightlydifferentinitialconditions.Asecondensembleofsimulationswasgeneratedbyfollowingexactlythesame experimental design except that the time-varying ozone forcing files were replaced withseasonallyvaryingclimatologies (thisconstitutes thecontrolexperiment).Thesimulationsare thusdesigned to determine whether inter-annually varying prescribed ozone concentrations andassociatedradiativeforcingprovideenhancedskill insimulationsofmid-summerclimateanomaliesinthepresenceofaperfectdescriptionofhistoricalENSOevents(comparedtothecasewhereonlyclimatologicalozoneforcingisused).Thesimulationsweresubsequentlyanalysedtodeterminetheirabilitytorepresentinter-annualvariationsinspringandmid-summercirculationspatternsovertheSouthernHemisphereandthecorrespondingvariations in rainfalloversouthernAfrica.Theresultsstatistically differentiate inmodel skill for seasons of El Niño, LaNiña and neutral states, and aresupplementedbycasestudiesofthesimulationsobtainedforselectedindividualseasons,includingthe1997/1998ElNiñoevent.


(B1-10)

ImpactsoftheAtlanticMultidecadalVariabilityonthetropicalclimateandtropicalcycloneactivity

YohanRuprich-Robert(1,2,3),HiroyukiMurakami(2,3),TomDelworth(3),RymMsadek(4),FredericCastruccio(5),StephenYeager(5),GokhanDanabasoglu(5)

BarcelonSupercomputingCenter,Spain(1),PrincetonUniversity,USA(2),GFDL,USA(3),Cerfacs,France(4),NCAR,USA(5)

TheAtlanticMultidecadalVariability(AMV)isassociatedwithmarkedmodulationsofclimateanomaliesovermanyareasoftheglobe.ThisincludesdroughtsinAfricaandNorthAmerica,declineinseaice,changesoftropicalcycloneactivityintheAtlantic,andchangesintheatmosphericlarge-scalecirculation.However,theshortnessofthehistoricalobservationscomparedtotheAMVperiod(∼60-80yr)makesitdifficulttoshowthattheAMVisadirectdriverofthesevariations.ToisolatetheAMVclimateresponse,weuseasuiteofglobalcoupledmodelsfromGFDLandNCAR,inwhichtheNorthAtlanticseasurfacetemperaturesarerestoredtotheobservedAMVpattern,whiletheotheroceanicbasinsareleftfullycoupled.ToexploreandrobustlyisolatetheAMVimpactsonweatherextremes(e.g.,heatwaves,tropicalcyclones),weperformlargeensemblesimulations(between30and100members)thatareintegratedfor10years.

Duringborealsummer,inallmodelstheAMVwarmingalterstheWalkerCirculationandmodifiesthesurfacewindsoverthetropicalPacificOcean.Duringborealwinter,theAMVwarmingisassociatedwithlargeanomaliesoverthePacificthatprojectontoanegativephaseofthePacificDecadalOscillation(PDO).ThisPDOresponsecomesfromalaggedadjustmentofthetropicalPacifictothesummerAMVforcing,highlightingthenecessityof

usingaglobalcoupledframeworktofullycapturetheAMVclimateimpacts.Finally,itisshownthattheAMVwarmingleadstochangesinthetropicalcycloneactivityovertheentiretropicalbelt,withmoretropicalcyclonesovertheAtlanticandlesstropicalcyclonesoverthePacific.Thesechangesarevery similar to the fluctuations of the tropical cyclone activity observed over the last 40 years. Ithencesuggeststhatthe1995/96phaseshiftoftheAMVhaspartlydriventhesevariations.


(B1-11)

OceanicandAtmosphericSourcesofSeasonalTropicalCyclonePredictability

Patricola,ChristinaM.(1);Chang,Ping(2);Saravanan,R.(2);Hsu,Wei-Ching(2)

LawrenceBerkeleyNationalLaboratory,Berkeley,CA(1);TexasA&MUniversity,CollegeStation,TX(2)

Tropical cyclones (TCs) are among the costliest and deadliest natural hazards. The goal of thisresearchistoimproveseasonalpredictionandfutureprojectionofTCactivitybyunderstandingthephysicalrelationshipsbetweenTCs,modesofclimatevariability,andatmosphericinternalvariabilityusingobservationsandensemblesofhigh-resolutionclimatemodelexperiments.OnemajorsourceofseasonalTCpredictability issea-surfacetemperature(SST)patterns,whichcanbepredictable inadvanceofthehurricaneseasonandcaninfluenceenvironmentalfavorabilityforTCs.Wediscoveredthat:1)AtlanticandPacificSSTpatternsdrivecompensatingandconstructiveinfluencesonAtlantichurricaneseasons,withconcurrentLaNiñaandpositiveAtlanticMeridionalMode(AMM)conditionssupporting the most active hurricane seasons, and strong concurrent El Niño and positive AMMdriving near-average seasons; 2) The location of El Niño’s SSTwarming plays a critical role in thedegree of Atlantic TC suppression, with Central Pacific/Warm Pool El Niño substantially moreeffectiveatreducingAtlanticTCsthanEastPacific/ColdTongueElNiño,forequalwarmingintensity.ThisisphysicallyexplainedbythestrongzonalgradientinbackgroundtropicalPacificSST,togetherwiththe(climate-variant)SSTthresholdfordeepconvection.LesswarmingisneededintheCentralPacific to reach the SST threshold for deep convection,which is the key link to the teleconnectedAtlantic vertical wind shear response; 3) Variability in the typical Atlantic TC precursor, AfricanEasterly Waves (AEWs), provides little seasonal TC predictability, as suggested by mechanisticregionalclimatemodelsimulations inwhichAEWsareprescribedor removedthroughthemodel’slateral boundary conditions; and 4) The tropical SST biases common to generations of coupledatmosphere-oceanclimatemodelscausesubstantialerrors insimulatedTCactivity,withanunder-simulationof50%intheAtlanticandover-simulationof80%intheEastPacific.ThisworkhighlightskeyregionswherereducingSSTbiascouldimprovepredictionsandprojectionsofTCactivity.


(B1-12)

SourcesofskillindecadalpredictionsofSahelprecipitation

Maroon,Elizabeth(1),Yeager,Stephen(2),Danabasoglu,Gokhan(2)

CIRES,UniversityofColorado,USA(1),NCAR,USA(2)

The new, 40-member Large Ensemble of decadal predictions with the Community Earth SystemModel(CESM-DP-LE)showsconsiderableskill inpredictingprecipitationanomaliesovertheAfricanSahelregionseveralyearsinadvance.SuchskillfulpredictionsoftheSahelprecipitation

wouldprovidedecisionmakerswiththeabilitytomitigatetheimpactsofadrought.Here,weassessthe sourcesof this skill in theCESM-DP-LE through comparisonswith its predecessorbasedon anearlier version of the CESM; its uninitialized Large Ensemble counterpart; persistence; and withobservations.TherelationshipsofSahelprecipitationwithsurfaceanduppertroposphericconditionsare examined globally, but with a particular focus on regions and processes that have beenpreviously identified as important for Sahel rainfall predictability. We first examine how skill inAtlanticSSTassociatedwithAtlanticMultidecadalVariabilityaffectspredictionskillforSahelrainfall.Wealsoinvestigatetheinfluenceofocean-drivenhemisphericSSTasymmetryontheSahelthroughapplication of an energetic framework that connects high latitude surface heat fluxes to tropicalatmosphericoverturningandprecipitation.Wethenassess the roleofatmospheric stabilityacrossthe tropics and study whether improved representation of tropical tropospheric structure in theCESM-DP-LE is important.Thisconnection to theupper tropospheremayhelp toexplainhow longleadtimeskillinthetropicalIndianandPacificOceanscontributestoimprovedskillinpredictionsofSahelprecipitation.


(B1-13)

ProjectedChangesinS2DHydroclimatePredictabilityinNorthAmericainCESM-LE

Dewes,Candida(1,2),Newman,Matthew(1,2),Kumar,Sanjiv(3)

CooperativeInstituteforResearchinEnvironmentalSciences,UniversityofColoradoBoulder(1),PhysicalSciencesDivision,NOAA/ESRL(2),SchoolofForestryandWildlifeSciences,Auburn

University(3)

PotentialS2DpredictabilityofNorthAmericanhydroclimateanomalies,anditssensitivitytoexternalforcing,isevaluatedwithinthe40climaterealizationsfromtheCommunityEarthSystemModelingLargeEnsemble(CESM-LE).Eachensemblememberexperiencesthesameexternalforcingscenario(observed + RCP8.5) over the years 1950 to 2070. The hydroclimate variables examined are soilmoisture (0-1 m depth), Palmer Drought Severity Index (PDSI), and precipitation. As remotepredictors,weconsidertheleadingmodeofPacificseasurfacetemperaturevariabilitywithinthreedifferentdomains:theTropicalPacific(30°S-30°N;e.g.,ENSO),theIndo-Pacific(20°S-70°N;e.g.,theIPO), and the North Pacific (20-70°N; e.g. the PDO). We also include land surface memory byspecifying soil moisture or PDSI state 12months prior as a local source predictor. By performingsimple and multiple linear regression analyses across successive 20-year periods, where theexternally-forcedsignal representedby theensemblemean is first removed,wecanthenevaluatethe changing relative importance of Pacific forcing and land surface memory upon predictability,measuredbyasignal-to-noise(S2N)ratiobetweenexplainedandunexplainedvariances.Comparisonisalsomadewithasimilaranalysisappliedtotheobservationalrecordovertheyears1950-2015.

The relative importance of the remote Pacific and local memory predictors was found to havesignificantregionalvariationintheCESM-LE.ThePacificcontributionwaslargeacrossthesoutherntier of the United States, dominating predictability in the North American Southwest. In otherregions,thelandsurfacememoryprocesswasmoreimportant;inparticular,soilmoisturememorydominated central Canada predictability, yielding a S2N ratio of comparable magnitude (~0.5) asENSOovertheSouthwestUS.Forsomeregions,suchasthecentralUS,thetwosourceswereaboutequally important.Wealsoassessedtherelative importanceofthethreePacificpredictors, findingthatENSOyieldsthehighestS2Nratios,withtheIPOpredictoryieldingsimilarS2Npatterns,thoughslightly weaker, and the PDO slightly weaker still.We also found that precipitation is very poorlycorrelated to thepreviousyear’sprecipitationanomalies,while forannual soilmoistureanomaliesthe memory has a more significant footprint. PDSI predictors show similar patterns as for soilmoisture,albeitwithslightlylowerS2Nratios.

WefoundthatformostNorthAmericanregions,theCESMprojectshydroclimatepredictabilitytoincrease in thewarmerclimate,even though there isnosignificantchange inoverallhydroclimatevariability. This is due primarily to a strengthening of the Pacific-related predictable componentwhich coincides with a pronounced increase in the variance of tropical Pacific sea surfacetemperatures. In contrast, predictability due to land surfacememory remains the sameor slightlydecreasesovertime.ThesefindingssuggestthatinterannualhydroclimatepredictabilityovercertainregionsofNorthAmericacouldbeimprovedbycombiningremoteandlocalsourcesofhydroclimatepredictability.


(B1-14)

ENSO:towardsbreachingthespringtimepredictabilitybarrier

Webster,Peter(1),Toma,Violeta(2),Johnstone,James(3),Hirata,Fernando(4),Curry,Judith(5)

ClimateForecastApplicationsNetwork(1)(2)(3)(4)(5),GeorgiaInstituteofTechnology(1),(2)

The springtimepredictabilitybarrier inENSOpredictions (refWebster Song)arises fromstochasticprocesses occurring in the tropical Pacific that are tied to the annual cycle. As a result, forecastinitializedpriortoMay(andinsomeyears,aslateasJuly)haveshownlittleskillinENSOpredictionfromlatesummertotheendoftheyear.Recentadvancesinglobalseasonalforecastmodelsappeartobebreaching thepredictabilitybarrier tosomeextent.Weassess theENSOhindcastskillof thelatestversionof theECMWFSeasonalForecastingSystem(SEAS5), relative to thepreviousversion(SEAS4). Extended predictability of Niño 3.4 from SEAS5 shows correlation coefficients 0.7 for allinitializationmonths (including spring) for forecasts 5-6months in advance.Wehave conducted aclimatedynamicsanalysisseekingtoidentifythesourcesofthisextendedrangeENSOpredictability.WehaveidentifiedDJFprecursorsignals inuppertroposphericandstratosphericanomaliesathighlatitudesofbothhemispheres, consistentwith researchshowing importantextratropical forcingofsurface wind anomalies and SST responses in the equatorial and off-equatorial Pacific. We haveupdated the analysis of ENSO – precipitation relationships in the U.S. by analyzing shifts in thestatistical distribution of rainfall duringwarm (El Niño) and cold (LaNiña) phases of the SouthernOscillation for the period 1981-2016. An evaluation is presented of the skill of ECMWF SEAS5 inpredictingU.SprecipitationanomaliesassociatedwithENSO.�


(B1-15)

Multi‑scaleenhancementofclimatepredictionoverlandbyincreasingthemodelsensitivitytovegetationvariability

Alessandri,Andrea(1),Catalano,Franco(2),DeFelice,Matteo(2),VanDenHurk,Bart(1),Doblas-Reyes,Francisco(3),Boussetta,Souhail(4),Balsamo,Gianpaolo(4),Miller,Paul(5)

KNMI,Netherlands(1),ENEA,Italy(2),BSC,Spain(3),ECMWF,UK(4),LundUniversity,Sweden(5)

Theeffectivesub-gridvegetationfractionalcoveragevariesseasonallyandatinterannualtime-scalesin response to leaf-canopy growth, phenology and senescence. Therefore it affects biophysicalparameters such as the surface resistance to evapotranspiration, albedo, roughness length, andwaterfromsoilfieldcapacityexploitablebyvegetation.ToadequatelyrepresentthiseffectintheEC-EarthESM,weincludedanexponentialdependenceofthevegetationcoverontheLeafAreaIndexaccordingtotheLambertBeerlawofextinctionoflightundervegetationcanopy.

By comparing two sets of simulations performed with and without the new variable fractional-coverageparameterization, spanning fromcentennial (20thCentury) simulationsand retrospectivepredictions to the decadal (5-years), seasonal (2-4months) and weather (4 days) time-scales, weshowforthefirsttimeasignificantenhancementinclimatesimulationandpredictionoverlandthatisconsistentlyobtainedatthemultipletime-scalesconsidered.Particularly largeeffectsatmultipletime scales are shownover borealwintermiddle-to-high latitudes over Canada,WestUS, EasternEurope,RussiaandeasternSiberiaduetothe implementedtime-varyingshadowingeffectbytree-vegetation on snow surfaces. Over Northern Hemisphere boreal forest regions the improvedrepresentation of vegetation-cover consistently correct the winter warm biases, significantlyimproves the climate change sensitivity, the decadal potential predictability aswell as the skill offorecasts at seasonal and weather time-scales. Significant improvements of the prediction of 2mtemperature and precipitation are also shown over transitional land surface hot spots. Both thepotential predictability at decadal time-scale and seasonal-forecasts skill are enhancedover Sahel,North American Great Plains, Nordeste Brazil and South East Asia, mainly related to improvedperformanceinthesurfaceevapotranspiration.

Thisworkdemonstrates,forthefirsttime,thattheimplementationofarealisticrepresentationofvegetation in Earth System Models (ESMs) can significantly improve climate prediction acrossmultiple time-scales, and the details of the results are discussed in a peer-review paper justpublishedonClimateDynamics(Alessandrietal.,2017;doi:10.1007/s00382-017-3766-y).


(B1-16)

InvestigatingtheimpactofsoilmoistureonEuropeansummerclimatepredictions

Ardilouze,Constantin(1),Batté,Lauriane(1),Déqué,Michel(1),VanMeijgaard,Erik(2),VandenHurk,Bart(2)

CNRMUMR3589,Météo-France/CNRS,Toulouse,France(1),RoyalNetherlandsMeteorologicalInstitute(KNMI),DeBilt,Netherlands(2)

Duetothelimitedskillofstate-of-the-artpredictionsystems,highexpectationsonsummerseasonalforecastsoverEuropeareonlymarginallyfulfilled.Anumberofstudieshaveshowntheprominentimpact of soil moisture anomalies on summer mid-latitude climate variability and predictability.However,becauseofmodel systematicerrors,even thebestpossible initializationof soilmoistureconditions falls short in estimating the theoretical upper limit of predictive skill induced by soilmoisture boundary conditions. The present study aims at addressing this question by comparingidealizedensemblere-forecast-likesimulationsinwhichsoilmoistureconditionsareprescribedfromtheERA-InterimLANDreanalysiswithinitializeddynamicalre-forecastsinwhichsoilmoistureevolvesfreely. Two regional climate models with domains centered over Europe contribute to theseexperiments and generate very similar results. Simulations with constrained soil moisture displaysignificantlyincreasedcorrelationbetweenobservedandsimulatedseasonalanomaliesofmaximumtemperatureprecipitationandsurfacesolarradiation,ascomparedtothereferencere-forecast.Thiswidespread increase is not restricted to regions already known as hot-spots of land-atmospherecoupling. In spite of a limited change of the ensemble spread, the idealized simulations betterperformincapturinganomaliesexceedingadefinedthreshold.Afocusontwocasestudiesrevealscontrastedresultsbetweenthe2003and2010heatwaves.Theseresultssuggestthatsoilmoisturemaybealargersourceofsummerpredictabilitythanexpectedfrompreviousworks.


(B1-17)

ThenorthernhemispherecircumglobalteleconnectioninaseasonalforecastmodelanditsrelationshiptoEuropeansummerforecastskill

Beverley,Jonathan(1),Woolnough,Steven(2),Baker,Laura(2),Johnson,Stephanie(3),Weisheimer,Antje(3,4)

DepartmentofMeteorology,UniversityofReading,UK(1),NCAS,UniversityofReading,UK(2),ECMWF,UK(3),NCAS,UniversityofOxford,UK(4)

RecentresearchhasledtoimprovementsinEuropeanwinterseasonalforecasts,howevertherehasbeen less of a focus on the summer season and summer forecast skill remains relatively low. TheEuropeanclimate is affectedbya large rangeof influences,which include influences from tropicalregions, and better understanding of the mechanisms behind these tropical-extratropicalteleconnections can inform our evaluation of seasonal forecast systems and priorities for modeldevelopment.

Onepotential sourceof predictability for Europe is the Indian summermonsoon (ISM),which canaffectEuropeanweatherviaaglobalwavetrainknownasthe“CircumglobalTeleconnection”(CGT,DingandWang2005).HereweassesstheabilityoftheECMWFcoupledseasonalforecastmodeltorepresentthisteleconnectionmechanism.WeuseseasonalhindcastsforJJAwhichareinitialisedonthe 1st May, with 25 ensemble members, for the period 1981-2014. We show that therepresentationoftheCGTwavepatterninthemodelisweakerthanobserved,particularlyinAugustwhen the observed CGT wavetrain is the strongest, and the model has errors in forecastinggeopotentialheight in several key regions (“centresof action”) for the teleconnectionmechanism.Severalpossiblecausesoftheseerrorswillbeshown.First,modelvarianceingeopotentialheightinwest-centralAsia (an important region for themaintenanceof theCGT) is lower thanobserved inJuly andAugust, associatedwith apoor representationof the linkbetween this regionand Indianmonsoonprecipitation.Second,analysisoftheRossbywavesourceshowsthatthesourceassociatedwithmonsoonheatingisbothtoostronganddisplacedtothenortheastinthemodel.Thisisrelatedto errors inmonsoonprecipitationover theBayof Bengal andArabian Sea,where themodel hasmore precipitation than is observed. Third, the model jet is systematically shifted northwards byseveral degrees latitude over large parts of the northern hemisphere, which may affect thepropagationcharacteristicsofRossbywavesinthemodel.Inordertofurtherunderstandhowtheseerrors are related to errors in summer predictability over Europe, we will present results fromseveral relaxation experiments, including relaxing regions over west-central Asia and northwestEurope. These experiments have been designed to identify possible causes of errors in theteleconnectionpathway, and to explore how improving the representationof theCGT impacts onEuropeanSummerforecastskill.

SESSION:(B2)ModellingissuesinS2Dprediction

(B2-01)

DemandsontheMPIEarthSystemModeltoperformseasonal-to-decadalclimatepredictions

Müller,WolfgangA.(1,2),Brune,Sebastian(2),Baehr,Johanna(3),Pohlmann,Holger(2),Modali,Kameshvar(2),Kröger,Jürgen(2),Fröhlich,Kristina(1),Pankatz,Klaus(1),Dobrynin,Mikhail(3),

Kadow,Christopher(4),Marotzke,Jochem(2)

DWD,Germany(1),MPI-M,Germany(2),UniversitätHamburg,Germany(3),FreieUniversitätBerlin,Germany(4)

The Max-Planck-Institute Earth SystemModel (MPI-ESM) is the baseline for the German ClimateForecast System (GCFS) and contributes to the upcoming CMIP6 experiments. An overview ispresented on the latest developments of MPI-ESM as a seasonal-to-decadal climate predictionsystem.Resultsareshownwithafocusontwotopics:(i)theincreasedspatialresolution,and(ii)theinitialization,eachwithrespecttothe(expected)subsequentpredictionskill.

Compared to its precursors provided forCMIP5, a higher-resolved versionofMPI-ESM (MPI-ESM-HR) has been developed,which represents in its atmospheric component a doubling in horizontalresolution (T127, ~100km). The model-tuning is straight-forward to targets of global meantemperatures,Arctic sea-ice, awell-balanced radiationbudget anda stableocean circulation. Twoexamplesarepresentedbywhichasimple increase inresolutionandastraight-forwardtuninghasand has not led to the desired improvement in prediction skill. First, the winter North AtlanticOscillation(NAO)isconsidered,forwhichhighpredictionskillforthefirstwinterhasbeenfoundinalower-resolved model version (T63, ~ 200km). Surprisingly, in MPI-ESM-HR the prediction skill isreduced. An examination of the predictants of the winter NAO reveal, that in particular thetroposphere-stratospherelinkageisonlyweaklyrepresentedinMPI-ESM-HR.Amechanismthatwasnot givenmuch attention during the tuning process. Second, the storm track bias is substantiallyreduced over the North Atlantic and Europe in the MPI-ESM-HR compared to its lower-resolvedversion. Initializedexperimentsrevealsubstantiallyhigherpredictionskill forstorm-trackdensity inMPI-ESM-HR,andpositiveskillisfoundforyears1-5.

Inaddition,progresshasbeenmadeintheinitializationofdecadalclimatepredictions.Approachesbased on a nudging procedure of atmosphere and ocean reanalyses in to MPI-ESM inducedimbalances in North Atlantic heat transport and budget. The induced transport modificationscontinue tobepresent in the freely runningpredictionsandare seen to reduce forecast skill. Theimbalances varywithdifferent reanalysis products andwith applicationof full fieldsor anomalies.Theassimilationof anomaly fields largely keeps theheat transportandbudget inbalanceand theaccordingpredictionskillissignificantlyhighercomparedtoinitializationsfromfullfields.Thisresultreflects the need for a “model-consistent” assimilation. In this respect an Ensemble Kalman Filter(EnKF)hasbeenimplementedintheoceaniccomponentofthelower-resolvedversionofMPI-ESM.Initializedexperimentswith theEnKFassimilatingEN4andHadISSToutperformexperimentsbasedonnudgingtechniquesintermsofpredictionskill inparticularinthePacificandtheNorthAtlantic.TheimplementationofEnKFinMPI-ESM-HRiscurrentlyinvestigated.


(B2-02)

Process-OrientedModelDiagnosistoImproveModelingSystems

Barrie,Daniel(1),Maloney,Eric(2),Gettelman,Andrew(3),Ming,Yi(4),Neelin,David(5),Mariotti,Annarita(6)

NOAACPO(1),ColoradoStateUniversity(2),NCAR(3),NOAAGFDL(4),UCLA(5),NOAACPO(6)

The NOAAModel Diagnostics Task Force has led an activity since 2016 to produce a community-designed, flexible softwarepackage that integratesprocess-orientedmetrics formodel evaluation.Process-oriented metrics in this case are model diagnostics that go beyond evaluation of rawperformanceofamodelandprovidephysicalinsightintothesourcesofmajorbiasesinmodels.Thedesired end result is pathways to improve model performance via process-focused diagnosis ofmodel error. The effort has been led by NCAR, GFDL, and academic community scientists, and abroad array of diagnostics have been contributed to the software package from a diverse pool ofinvestigators and institutions. This paperwill discuss process-orienteddiagnostics as a pathway tomodel improvementandprovideanupdateon theTaskForce'sprogress.Metricpackagessuchasthis one can help advance model and prediction system development, where improvements areoftenmadeinaheavilyperformance-orientedframework.


(B2-03)

DiagnosingthesourcesofsystematicSSTbiasesinCESMusingensembleseasonalhindcasts

Siongco,AngelaCheska(1),Ma,Hsi-Yen(1),Xie,Shaocheng(1),Klein,Steve(1),Karspeck,Alicia(2),Raeder,Kevin(2),Anderson,Jeffrey(2)

LawrenceLivermoreNationalLaboratory,USA(1),NCAR,USA(2)

In this study, we investigate the emergence and growth of systematic SST biases in ensembleseasonal hindcasts from the Community Earth System Model (CESM) version 1. Six-month long,twenty-fourmemberensemblehindcastsimulationscoveringtheperiod2001-2005areperformed,with initial conditions derived fromERA-Interim for the atmosphere and fromNCAR-DART for theocean.TheequatorialPacificandnorthernsubtropicalPacificandAtlanticoceansdevelopacoldbiasafter two to three months, reaching comparable magnitudes to climatological biases within sixmonthsof lead-time.Furtheranalysisof theequatorialPacificcoldbiasrevealsthathindcastswithstart dates during the upwelling period (boreal summer to fall) exhibit a strong drift from thereanalysesandobservationsaswellasa largeensemblespread. Incontrast, thosewithstartdatesoutside of the upwelling period show minimal drift and spread. This implies that the cold biasdevelopsquicklyduring theupwellingperiod,but takes longer than sixmonths toemergeoutsidethe upwelling period. An upper ocean heat budget analysis confirms that the anomalous coolingcomesfromtoostrongverticaladvectionintheocean.Theverticaladvectionbiasisassociatedwitheasterly wind stress anomalies, which emerge as early as the first two months of lead-time,preceding the onset of the cold SST bias. The too strong easterlies are accompanied by excessiveprecipitation north of the equator. The sensitivity of the wind and precipitation biases to therepresentationoflow-levelcirculationandmoistconvectioninthemodelarefurtherexplored.

[This study is funded by the Regional and Global Climate Modeling and Atmospheric SystemResearch Programs of the U.S. Department of Energy as part of the Cloud-AssociatedParameterizationsTestbed.ThisworkwasperformedundertheauspicesoftheU.S.DepartmentofEnergybyLawrenceLivermoreNationalLaboratoryunderContractDE-AC52-07NA27344.LLNL-ABS-747964]


(B2-04)

Estimatingerrorsinmodelvariability:acomparisonbetweenseasonalre-forecastsandcontinuousmulti-decadalsimulationswiththeECMWFcoupledmodel

Molteni,Franco(1),Johnson,Stephanie(2),Roberts,Chris(3),Senan,Retish(4),Stockdale,Tim(5)

ECMWF,Reading,U.K.(1,2,3,4,5)

In seasonal and decadal predictions initialised from real-world conditions, the drift of themodel'sclimate(includingbothmeanstateandvariability)fromtheobservedstatetoitslong-termattractormakes the estimation of systematic model errors particularly demanding, since errors during thedrifting phase are dependent on both the phase of the seasonal cycle and the forecast time.Typically,inordertocalibrateseasonalforecastsandassesstheirerrors,alargesetofensemblere-forecastsisrun,withinitialdatesspanningallmonthsforatleasttwoorthreedecades.Becauseofthe substantial overlap between integrations started from dates in consecutiveweeks ormonths,seasonal re-forecast require a total integration time possibly reaching thousands of years: forexample, the operational ECMWF system SEAS5 is calibrated using 25-member ensemble re-forecastsspanning36-years,foratotalof6,300yearsofsimulation.

Ininter-comparisonprojectsaimedatestimationofclimatechange,the“current”modelclimatologyis usually estimated by running continuous integrations spanning the historical period, initialisedfromamodelstateobtainedfromalongspin-uprun.Sinceall(oratleastmostof)theclimatedriftissupposedtooccurduringthespin-upphase,historicalmulti-decadalsimulationsprovideestimatesof the model mean-state and variability which are representative of the model's asymptoticattractor. A challenging question is to what extent diagnostics of modelled variability in multi-decadalhistoricalrunsmatchcorrespondingresultsobtainedfrominitialisedre-forecasts inastateofclimatedrift.

ECMWFhas runhistoricalmulti-decadal simulations for the1950-2014periodas a contribution totheEU-fundedH2020PRIMAVERAproject,usingaversionofthecoupledmodelalmostidenticaltothatusedintheseasonalforecastsystemSEAS5.Thistalkwillcomparediagnosticsoflow-frequencyvariability and teleconnections derived from the SEAS5 re-forecasts and the PRIMAVERAhistoricalruns, with a focus on tropical-extratropical interactions during the northern winter andteleconnections associated with monsoon systems. We show that a number of differences (orsimilarities)betweenobservedandmodelstatisticsonatmosphericvariabilityshowasimilarpatternin the initialised andmulti-decadal runs.On theotherhand,multi-decadal runs are able to revealerrors in the slow components (deep ocean, sea-ice) of the climate systemwhich are difficult todetectsignificantlyinseasonalexperimentsbecauseofasmallersignal-to-noiseratio.Therefore,weargue that continuous, multi-decadal historical simulations represent a valuable and conceptuallysimpleadditiontothetraditionalexperimentalset-upfortheoptimizationofcoupledmodelsusedininitialisedpredictions.


(B2-05)

Approachestoreducemodelbiasestoimproveinclimateprediction

N.Keenlyside(1,2,4),F.Counillon(1,2,4),M.Devilliers(2,4),S.Koseki(1,4),I.Bethke(3,4),T.Demissie(3,4),G.Duane(1),A.Gupta(3,4),M.-L.Shen(1,4),L.Svendsen(1,4),T.Toniazzo(3,4),Y.

Wang(2,4)

GeophysicalInstitute,UniversityofBergen(1);NansenEnvironmentalandRemoteSensingCenter(2);UniResearch(3),BjerknesCentreforClimateResearch(4)

Systematicmodelerroristhecauseoflargebiasesclimatemodels.Theaimofthispresentationistofirstdiscusstheimpactofthesebiasesonpredictionskillandsecond,tointroducethesupermodelapproachtoreducemodelsystematicerrors.

To illustrate the first point, we present seasonal prediction results for the equatorial Atlanticperformed with a standard and an anomaly-coupled configuration of the Norwegian ClimatePredictionModel (NorCPM). The biases in this region are particularly large and persistent amongmodels.AsignificantreductioninthesebiasesisachievedinNorCPMbyanomalycoupling(i.e.,byastatic correction of the momentum and SST fields exchanged between oceanic and atmosphericmodel components).Reductionof thebias is shown to improve the simulatedvariability, enhancethe quality of the ocean reanalysis, and significantly increase the skill in seasonal prediction ofequatorial Atlantic climate. These improvements are related to a better representation of ocean-atmosphereinteraction.

A superior representation is provided by a supermodel that can be constructed by interactivelycombininganumberofdifferentmodelsinruntimesothattheirindividualmodelerrorsaremadetocompensate.We show that the approach is able tomitigate the double intertropical convergencezone bias found in most climate models. Importantly, non-linear ocean-atmosphere interactionenables the super model to out perform the standard averaging of the coupled models runseparately. To demonstrate the broad applicability of the approach we have now used dataassimilationbetweenmodels to createa supermodel from three state-of-the-art climatemodels–NorESM,MPIESM,andECEARTH.Initialresultsfromthisnewsupermodelwillbepresented.


(B2-06)

SEAS5:ThenewECMWFseasonalforecastsystem

StephanieJohnson,TimStockdale,LauraFerranti,MagdalenaBalmaseda,FrancoMolteni,LinusMagnusson,SteffenTietsche,DamienDecremer,AntjeWeisheimer,AncaBrookshaw

ECMWF

In this presentation we will describe SEAS5, ECMWF's fifth generation seasonal forecast system,which became operational on Nov. 5, 2017. Compared to its predecessor, System 4, SEAS5 is asubstantiallynewsystemwhichincludesupgradedversionsoftheIFSatmosphereandNEMOoceanmodelsathigherresolutions,andintroducestheLIM2interactiveseaicemodel.

WewilldiscussSEAS5’sperformance,andhow it compares toSystem4.Manyaspectsof forecastskillhaveimproved,butthereareafewexceptionswhereforecastskilldecreases.SEAS5tropicalSSTbiases have significantly improved over System 4, including a two degree improvement in theequatorialPacific.Two-metretemperaturepredictionskillinthetropicshasimprovedandtherearealsoimprovementsinsomeaspectsofENSOforecastskill.TheincreasedoceanresolutioninSEAS5changes SST biases in the northern extratropics, especially in regions associated with westernboundarycurrents.IntheNorthwestAtlantic,SEAS5poorlycapturesobserveddecadalvariabilityofthesubpolargyre,whichresults ina localdegradationofDJF2-metretemperaturepredictionskill.Introducing the prognostic sea ice model gives SEAS5 the ability to forecast sea-ice cover in thecomingseasons.Insummary,SEAS5continuestobeastate-of-the-artseasonalforecastsystem,withaparticularstrengthinENSOprediction.

SEAS5 data and graphical forecast products are available to the public, under free access licence,through ECMWF's Copernicus Climate Change Service (C3S). We will present the multi-systemseasonal forecast service of C3S, to which ECMWF is a core contributor and SEAS5 is one of thecomponents.Abriefdescriptionoftheotherparticipatingforecastsystemswillbepresented,aswillthesetofgraphicalanddataproductscurrentlyavailableandplannedforthenearfuture.


(B2-07)

TheGermanClimateForecastSystemGCFS2.0

Fröhlich,Kristina(1),Dobrynin,Mikhail(2),Isensee,Katharina(1),Paxian,Andreas(1),Müller,Wolfgang(1),Gessner,Claudia(3),Baehr,Johanna(2),Früh,Barbara(1)

DWD,Germany(1),UniversitätHamburg,Germany(2),Goethe-UniversitätFrankfurt(3)

TheGermanseasonalforecastsystem(GCFS)hasbeenjointlydevelopedbyMaxPlanckInstituteforMeteorology (MPI-M), Universität Hamburg (UHH) and Deutscher Wetterdienst (GermanMeteorologicalService,DWD).ForecastsarepublishedstartinginOctober2016.Sincesummer2017DWD is one of 13 Global Producing Centres of Long-Range Forecasts to the WMO Multi-ModelEnsemble. Starting summer 2018 DWD provides its forecast data also to the Copernicus ClimateChangeServiceC3S.

ThefirstsystemstartedfromaCMIP5versionoftheclimatemodelMPI-ESM(MaxPlanckInstituteEarthSystemModel)andhasbeenadaptedforthedemandsofclimateforecasts,thereforebearingthenameGCFS1.0.ThesecondversionGCFS2.0isnowbasedonahigherresolvedMPI-ESMbothinatmosphereandoceanandgetsitshistoricalforcingdatafromCMIP6.Alongsidewithanimprovedversionoftheclimatemodeltheensembleconfigurationhasbeenalsoincreased.GCSF2.0nowuses30 members for the hindcasts and 50 members for the forecasts. Initial conditions for theatmosphere are taken from ERA-Interim or IFS-analyses while the ocean model is initialised byORAS5,thenewoceanreanalysisofECMWF.Wewillshowhowandwherethemodelclimateandforecasts have been improved, e.g. in terms of skill scores, ENSO forecasts and special weatherregimes,andwherethechallengesstillare.


(B2-08)

DevelopmentandcurrentS2DpredictionskilloftheNorwegianClimatePredictionModel

Wang,Yiguo(1,4),CounillonFrançois(1,2,4),KeenlysideNoel(1,2,4),KimmritzMadlen(1,4),BethkeIngo(3,4),LangehaugHelene(1,4)

NansenCenter,Norway(1),UniversityofBergen,Norway(2),UniResearchClimate,Norway(3),BjerknesCentreforClimateResearch,Norway(4)

The Norwegian Climate Prediction Model (NorCPM), developed within the Bjerknes Center,combines the Norwegian Earth System Model (NorESM) with the ensemble Kalman filter dataassimilationmethod. NorCPM can currently assimilate observations of ocean and sea ice into theocean and sea ice components, while the other model components are left unchanged by dataassimilation.We present the development and S2D prediction capacity of the version of NorCPMthat will contribute to CMIP6 DCPP.With the assimilation of SST data only, NorCPM can achievecompetitive skills at 6- and 12-month lead time compared to the North American MultimodelEnsemble(NMME). ItcanpredictthevariabilityofSST intheNordicSeasandthesea iceextent intheBarentsSeauptooneyearinadvance.Complementingthesystemwiththeassimilationofoceansubsurface data shows moderate improvements for seasonal prediction but shows greatimprovementsforinterannualtodecadalpredictionintheNorthAtlanticSubpolarGyreregionandintotheArctic.Thepredictionskilloftheprototypethatalsoassimilatesseaiceconcentrationwillbebrieflypresented.


(B2-09)

TheimportanceofstratosphericinitialconditionsonwintertimeseasonalpredictabilityintheEuro-Atlanticsectorandimplicationsforthesignal-to-noiseparadox

O'Reilly,Christopher(1),Woollings,Tim(1),Weisheimer,Antje(1,2)

UniversityofOxford,UK(1),ECMWF,UK(2)

InthisstudyweinvestigatetheinfluenceofatmosphericinitialconditionsonthepredictabilityoftheNAO in seasonal hindcast experiments. Three ensemble hindcast experiments are presented: anexperiment initalised from a reanalysis that assimilates a comprehensive set of observations, oneinitialised from a reanalysis that assimilates only surface observations and one initialised from arandom atmospheric initial condition. The skill of the NAO is analysed in the different hindcastexperiments. The stratospheric initial conditions, and in particular the QBO teleconnection to theNAO,emergeasan important sourceof skill in thehindcastexperiments.However, theQBO-NAOteleconnectionappearstobesomewhatweakerthanintheobservations,whichresults inasignal-to-noiseissueinthemostskillfulhindcastexperiment.Theseresultshavepotential implicationsforthesignal-to-noiseparadoxinoperationalseasonalforecastingsystems.


(B2-10)

SubtropicalNorthAtlanticpreconditioningkeytoskillfulsubpolargyreprediction

Bethke,Ingo(1,2),Wang,Yiguo(3,2),Counillon,Francois(3,2),Kimmritz,Madlen(3,2),Keenlyside,Noel(4,2)

UniResearchClimate,Norway(1),BjerknesCentreforClimateResearch,Norway(2),NansenEnvironmentalandRemoteSensingCenter,Norway(3),UniversityofBergen,Norway(4)

We compare decadal hindcast results for the Subpolar North Atlantic (SPNA) region from twoconfigurationsof theNorwegianClimatePredictionModel.Thefirstconfigurationobtains its initialconditionsbyanomaly-assimilatingSST-onlyobservationsintotheoceancomponentofthecoupledEarthsystemmodel.Thesecondconfigurationisidenticaltothefirstoneexceptthatitadditionallyanomaly-assimilates temperature and salinity profiles. Prior to 1995, both configurationsprecondition the SPNA in a cold state with a strong subpolar gyre (SPG) circulation. Differencesemerge,however,duringthehindcastperiods:whilethefirsthindcastsetexhibitsarapidlywarmingSPNAonce theassimilation is released, the second setmaintains theanomalous cold stateover alonger period of time and compares favourably with other prediction systems that havedemonstratedSPGhindcastcapability.TherapidSPNAwarming in the first setprimarilyoriginatesfromatoowarmSubtropicalNorthAtlantic(STNA)initialstate,leadingtoexcessivenorthwardheattransport that causes theSPG toprematurely rebound. Salinity anddynamical effects identified inpreviousresearchlikelycontributeaswellbuttoalesserextent.OurresultsillustratethatarealisticinitialstateintheSPNAdoesnotaloneguaranteedecadalforecastskillinthatregion,regardlessofadditionaldegradationduetomodelbias.Remoteeffects,inparticularthethermodynamicstateoftheSTNA,havetobeconsideredaswelltomaximiseSPNApredictioncapability.


(B2-11)

Candecadalclimatepredictionsbeimprovedbyoceanensembledispersionfiltering?Anyimpactonseasonalpredictions?

Kadow,Christopher(1),Illing,Sebastian(1),Kröner,Igor(1),Ulbrich,Uwe(1),andCubasch,Ulrich(1)

FreieUniversitätBerlin,Germany(1)

Decadal predictions by Earth system models aim to capture the state and phase of the climateseveral years in advance. Atmosphere-ocean interaction plays an important role for such climateforecasts. While short-term weather forecasts represent an initial value problem and long-termclimateprojectionsrepresentaboundaryconditionproblem,thedecadalclimatepredictionfallsin-betweenthesetwotimescales.Theoceanmemoryduetoitsheatcapacityholdsbigpotentialskillonthedecadalscale.Inrecentyears,morepreciseinitializationtechniquesofcoupledEarthsystemmodels (incl. atmosphere and ocean) have improved decadal predictions. Ensembles are anotherimportant aspect. Applying slightly perturbed predictions results in an ensemble. Instead of usingand evaluating one prediction, but the whole ensemble or its ensemble average, improves aprediction system. However, climate models in general start losing the initialized signal and itspredictiveskillfromoneforecastyeartothenext.HereweshowthattheclimatepredictionskillofanEarthsystemmodelcanbeimprovedbyashiftoftheoceanstatetowardtheensemblemeanofits individual members at seasonal intervals. We found that this procedure, called ensembledispersionfilter,resultsinmoreaccurateresultsthanthestandarddecadalprediction.Globalmeanandregionaltemperature,precipitation,andwintercyclonepredictionsshowanincreasedskillupto5yearsahead.Furthermore,thenoveltechniqueoutperformspredictionswithlargerensemblesandhigher resolution. Our results demonstrate how decadal climate predictions benefit from oceanensembledispersionfilteringtowardtheensemblemean.

MoreinformationsaboutthisstudyintheAGUJournalofAdvancesinModelingEarthSystems:DOI:

10.1002/2016MS000787https://doi.org/10.1002/2016MS000787

This study is part of MiKlip (fona-miklip.de) - a major project on decadal climate prediction inGermany.

We focus on theMax-Planck-Institute Earth SystemModel using the low-resolution version (MPI-ESM-LR)andMiKlip’sbasic initialization strategyas in thepublisheddecadal climate forecast from2018:

http://www.fona-miklip.de/decadal-forecast/decadal-forecast-for-2018-2027

Inadditiontotheworkshopfocus,weaddsomenew(!)resultsoftheensembledispersionfilteranditsimpactonseasonalnear-surfaceairtemperaturepredictionofthefirstwinter(DJF).Thereferencepredictionsystemshows lowcorrelationvaluesoverEurope.Theensembledispersion filter showspositiveandsignificantvaluesoverEurope.Asthereferencesystemalreadylosttheinitializedsignalontheseasonalscaleinthefirstmonthsoftheprediction,theensembledispersionfilterkeepstheforecastontrackandshowsasignificantimprovementoverEuropeinthewinterprediction(DJF).


(B2-12)

Climate-modeinitializationfordecadalpredictions

Polkova,Iuliia(1),Köhl,Armin(1)andStammer,Detlef(1)

InstituteofOceanographyUniversitätHamburg,Germany(1)

Based on the Earth SystemModel from theMax Planck Institute forMeteorology (MPI-ESM) wedesignedaclimate-modeinitializationmethodforthedecadalpredictionsystemMiKlip.Theideaofthe initialization method is in improving the ESM’s prediction skill through the initialization ofbalanced components of the initial conditions and filtering out components inconsistent with thedynamicsofclimatemodel.AsthecurrentMiKlippredictionsystemisusinganomalynudgingtowardORAS4 reanalysis in the ocean, the proposed method is expected to eliminate inconsistenciesbetweenthepredictionsystemandthenon-nativeoceanreanalysisbysynchronizingtheiroscillationpatterns.

To thisend, the temperatureandsalinityanomalies fromtheORAS4reanalysisareprojectedontomodes of variability derived from an ensemble of historical simulations (15 ensemble members)performedwith theMPI-ESM.Theclimatemodesare calculatedas statisticalmodesbasedon thebivariate empirical orthogonal function (EOF) analysis. The explained standard deviation in thefiltered reanalysis amounts to 66%. As this value is somewhat lower thanwhat we expected, weassumethatmodesofvariabilityofthereanalysisarenotexactlycompatiblewiththemodesfromthe prediction system or that they are not yet sufficiently sampled by the available data used toconstruct theEOFs.Theanalysisof filteredandoriginal reanalysisanomalies shows that thesignaloverthewholePacificbasiniswellcapturedandrepresented,whileintheSouthernOceanandtheNorthAtlantic largefractionofthesignal is filteredout.Thefilteredreanalysis'anomaliesarethenadded to themodel’s climatology and are used as initial states for a set of retrospective decadalpredictions. The climate-mode initialization method is compared against the commonly usedanomalyinitializationmethod.

A comparison of the climate-modes initialization with the reference initialization indicates animproved surface temperature skill over the tropical Pacific Ocean at seasonal-to-interannualtimescalesintermsofaccuracyandcorrelationwiththeobservations.ThisresultshowsapotentialforimprovingseasonalforecastsoftheEl-NinoSouthernOscillation.Forthe2-5leadyearsaverages,climate-modeinitializationmethodoutperformstheskilloftheanomalyinitializationforthesurfacetemperatureaswellas theupperoceanheatcontentoverthecentralandnorthernPacificOcean.For the North Atlantic subpolar gyre region, the climate-mode initialization experiments ratherresemble a trend of the historical simulations than that of ORAS4 or the observations. Also theyshowsmalleramplitudesofvariabilityascomparedtothenon-filteredinitialization.Thissuggeststheneedtofurtherimprovethedesignoftheclimate-modeinitializationmethodattemptingtocapturebetterthevariabilitymodesintheNorthAtlanticinanlargerEOF-basis.


(B2-13)

ApplicationofnormalmodefunctionsfortheimprovedbalanceintheCAFEdataassimilationsystemandcharacterisationofmodesofvariability

Kitsios,Vassili(1),O'Kane,Terence(1)

CSIROOceansandAtmosphere,CastrayEsplanade,BatteryPoint,Tasmania7004,Australia(1)

Inthecontextofmulti-yearsimulationsoftheatmosphereandocean,thegoalofdataassimilation(DA) is to both shift the simulations toward the physical observations and attempt to achieve abalanced state. Ingeneral, theDAprocessof incrementing the simulated flow fields such that theresultinganalysisfieldsapproachtheobservations,canintroducefurtherimbalanceintothesystem.WithapplicationtotheCSIROClimateAnalysisForecastEnsemble(CAFE)decadalpredictionsystem,wedemonstratetheuseofnormalmodefunctions(NMFs)tofilterouttheimbalancedcomponentof the atmosphere. In the CAFE system an ensemble transform Kalman filter of 96 ensemblemembersisusedtoassimilateobservationsoftheatmosphereandoceaninafullycoupledmanner.That is all observations, regardless of their realm, impact both the ocean and atmosphere via adynamically evolving cross-variancematrix. TheNMF filtering is appliedonline to the atmosphericfieldsoftheindividualensemblemembersimmediatelyfollowingeachDAcycle.

The advantage of using NMFs is that they act as a physically based filter. They result from theeigensolutionsofaprimitiveequationmodelonasphere.ForaspecifiedstaticstabilityprofiletheNMFssimultaneously represent the fluctuations in themassandvelocity fields in theatmosphere.Note, in comparison, spherical harmonics are the eigenvectors of the global barotropic vorticityequation and hence represent a simplified physical view of the atmosphere. Each NMF ischaracterised by a zonalwavenumber,meridionalwavenumber, and verticalmode number. Fromthe properties of the eigenvectors, the NMFs are additionally classified as being either balancedmodes (Rossby wave like), imbalanced westerly propagating inertial gravity waves or imbalancedeasterlypropagatinginertialgravitywaves.Thebalancedmodesareslowerevolvingandpotentiallypredictable.Theinertialgravitymodeshavemuchshortertimescales(foragivenspatialscale)anddominatethepredictabilityerrorgrowth.WeperformtheNMFcalculationsusingtheMODEScodedevelopedattheNationalCenterforAtmosphericResearchasdescribedinŽagaret.al.(2015).

Inaddition,weusetheNMFsasanofflinediagnostictooltobothdeterminetheextentofbalanceinthe finaldataset,andalso tocharacterisevariousmodesofvariability. In fact,wehavedevisedanindex representing the Madden-Julian Oscillation (MJO) using the NMFs, which agrees with thewidely adopted index ofWheeler&Hendon (2004). However, as ourNMF based index is derivedfromtheequationsofmotionitadditionallyhastheabilitytointeractwithothermodesofvariabilityandpotentiallycapturetheonsetofMJOevents.

References:

M.C.Wheeler&H.H.Hendon,2004,Anall-seasonreal-timemultivariateMJOindex:developmentofanindexformonitoringandprediction,MonthlyWeatherReview,Vol.132,p1917-1932.

N. Žagar, A. Kasahara, K. Terasaki, J. Tribbia & H. Tanaka, 2015, Normal-mode functionrepresentation of global 3-D data sets: open-access software for the atmospheric researchcommunity,Geosci.ModelDev.,Vol.8,p1169-1195.

SESSION:(B3)S2Densemblepredictionsandforecastinformation

(B3-01)

Near-termhydroclimateoutlooksbasedonthe

CESMDecadalPredictionLargeEnsemble

Yeager,Steve

NCAR

The Community Earth SystemModel (CESM) Decadal Prediction Large Ensemble (DPLE) is a newpublic data resource for studying Earth System prediction that offers unprecedented statisticalpowerforassessinghindcastaccuracyandskill,quantifyingthebenefitsassociatedwithinitialization,andexploringtheprobabilisticattributesofdecadalpredictionsofclimateandthecarboncycle.The

CESM-DPLEconsistsof40-memberensemblesinitializedeachyearbetween1954and2015,anditrepresents the initialized counterpart to the 40+ member CESM Large Ensemble ofhistorical/projectionsimulationsthatwerespunupfrompreindustrialconditions(CESM-LE;Kayetal.2015).Inthistalk,Iwillpresentaglobalsurveyofdecadalprecipitationforecaststhrough2025basedonCESM-DPLE,focusingonlandregionscharacterizedbysignificanthindcastskillenhancedthroughinitializationandforwhichthereexistatleasttentativemechanisticlinkstoocean-drivenseasurfacetemperature signals. The survey will encompass both seasonal mean hydroclimate and, whereappropriate, outlooks for precipitation extremes. If resources permit and operational issuesassociatedwithnearreal timeocean initializationcanbesurmounted,morerecentstartdateswillbeaddedtoextendtheconsideredforecastrangethrough2027.


(B3-02)

HowearlycouldthecurrentLaNiñahavebeenpredicted?

Deser,Clara(1),DiNezio,Pedro(2),Okumura,Yuko(3)

NCAR,USA(1),UniversityofTexasInstituteforGeophysics(2,3)

HistoricalobservationsshowthatoneintwoLaNiñaeventshavelastedfortwoconsecutiveyears.Despite their outsized impacts on drought and flooding, little is known about the predictability ofthese2-yearLaNiñaevents.WeaddressedthisquestionusingahierarchyofsimulationsperformedwiththeCommunityEarthSystemModelVersion1(CESM1)–amodelthatsimulatesrealistic2-yearLaNiña.Firstweexploredmechanismsandpotentialpredictabilityunder“perfectmodel”conditions.Then we applied those results to the prediction of the current 2-year La Niña event. We usedinitializedpredictions from theCESMDecadalPrediction LargeEnsemble (CESM-DPLE)– a suiteofensemble forecasts initialized on everyNovember since 1954. These CESMhindcasts show similarmechanismsandpredictionskillforobserved2-yearLaNiñaasinour“perfectmodel”forecasts.ThisallowedustopredicttheevolutionofthecurrentLaNiñausingtheensembleinitializedinNovember2015, at the peak of the most recent El Niño. The CESM-DPLE predicted subsequent La Niñaconditionslastingfor2years.TheprobabilityofLaNiñaforthesecondyear,i.e.thecurrentwinter,were60%accordingtotheinitializedforecastsand80%accordingtoanempiricalmodelcombiningobserved and simulated predictors. Together these results demonstrate that, under specific initialconditions,LaNiñaeventscanbepredictedtwoyearsinadvance.


(B3-03)

Multi-yearENSOprediction

Jing-JiaLuo,HanhNguyen,HarryHendon,OscarAlves,GuoqiangLiu,NickDunstone,CraigMacLachlan

AustralianBureauofMeteorology(1),UKMetOffice(2)

El Niño and La Niña strongly affect the year-to-year variability of Australian climate,which exertsintensepressureonwaterresourcesandenvironmentalmanagement.Well-knownexamplesincludetheseverefloodsin2010-11and1998-2000inassociationwithmulti-yearpersistentLaNiñaevents,andstrongdroughtin2002inassociationwithElNiño.GrowingevidencesuggeststhatsomeENSOeventsmaybepredictableatleadtimesbeyondonetotwoyears,whichislongerthantheforecastscurrentlyissuedbymostoperationalforecastcentresworldwide.However,itisstillunclearhowwellENSOcanbepredictedatmulti-yeartimescalesandwhatessentialdynamicsunderpinssuchmulti-year predictability. In collaboration with the UK Met Office, three sets of ensemble multi-yearhindcastswereproducedusingthehigh-resolutionocean-atmospherecoupledmodel,ACCESS-S.Thefirst set of experiments contain 30-member ensemble predictions of 16 target months starting 1Novemberforeveryyearduringtheperiod1980-2014.Thesecondsetofexperimentsconsistof10-memberensemblepredictionsof66targetmonthsstarting1Novemberevery2-3yearsduringtheperiod1960-2014.Thethirdsetofexperimentsinvolvesusing23ensemblesof36-monthpredictionstoexaminethemulti-yearpredictabilityoftheback-to-backLaNiñaduring1998-2001andthe2002-03central-PacificElNiñoevent.Resultssuggestthat,despiteconsiderablemodelbias insimulatingtheannualmeanstatesandseasonalcycleoftheIndo-Pacificclimate,ENSOoverthepast35yearscanbeskilfullypredictedoutto16monthsahead.SomeENSOeventscanbewellpredictedwitha3-yearlead.Inaddition,theresultsalsoshowgoodskillinpredictingglobalwarmingrelatedsignalsatmulti-yeartimescales.Withoutanydownscaling,thehigh-resolutionmodelalsodisplaysencouragingskill inpredicting local climatologyandanomalies inAustraliaat seasonal tomulti-year timescales.


(B3-04)

Howskilfularedecadalclimatepredictions?

Smith,Doug(1),Eade,Rosie(1),Scaife,Adam(1,2),Caron,Louis-Philippe(3),Doblas-Reyes,Francisco(3),Dunstone,Nick(1),Hermanson,Leon(1),Müller,Wolfgang(4,5)

Pohlmann,Holger(4)andYeager,Stephen(6)

MetOffice,UK(1),ExeterUniversity,UK(2),BSC,Spain(3),MPI,Germany(4),DWD,Germany(5),NCAR,USA(6)

We assess the skill of decadal climate predictions using new state of the art model hindcastscombinedwithamulti-modelensemblefromCMIP5startingeachyearsince1960.Theresulting71memberensemblemeanshowssignificant skill at forecastingyear2 to9averagesofnear surfacetemperature, rainfall andmean sea level pressureovermany regions.Wepropose anddescribe anewapproach forassessing the impactof initialisationon forecastskill that ismorepowerful thanthe simple difference in correlation skill between initialised and uninitialized hindcasts usedpreviously.Weshowthatinitialisationsignificantlyimprovestheskillinmanyregions,includingoverland. However, the underlying patterns of skill are captured by the uninitialized simulations,suggesting that initialisation ismainly improving the simulated response to external factors ratherthanpredictinginternalvariability.Wealsofindthatthesignaltonoiseratioisanomalouslysmallinthemodelensemble,suggestingthatthesignaltonoiseparadoxrecentlydiscoveredinextratropicalseasonalforecastsalsoappliesondecadaltimescales.


(B3-05)

Evaluationofre-calibrateddecadalhindcastusingacommonverificationframework

Grieger,Jens(1),Pasternack,Alexander(1),Rust,HenningW.(1),Ulbrich,Uwe(1)

InstituteofMeteorology,FUBerlin,Berlin,Germany(1)

Decadal predictions deal with the time scale which is important for decision makers andinfrastructuralplannersforthenear-termfuture.Forusabilityofthosepredictionsitisimportanttoknow about their forecast skill. Recent work suggest verification frameworks to answer the keyquestionswhetherinitializationoftheforecastmodel leadstohigherskill incomparisontotheun-initializedsimulationsandwhetherthespreadoftheensemblerepresentstheforecastuncertainty?

Initializedmodelsimulationstypicallyhavetodealwithbiaseswhicharedependentonforecastleadtime,alsoknownasmodeldrift.Additionally,thisbehaviorcandependoninitializationtime,i.e.biasanddrift canbedifferent for simulationswhichare initialized in the1960s in comparison tomostrecenthindcasts.

This study uses a "Decadal Forecast Recalibration Strategy" (DeFoReSt) which adjusts mean andconditionalbiasaswellasensemblespread,takingleadtimeandinitializationtimeintoaccount.AcommonverificationframeworkisusedtoanalyzetheskillofdecadalhindcastssimulatedwiththegeneralcirculationmodelMPI-ESMundertheumbrellaofMiKlip,whichistheGermaninitiativefordecadal prediction. For near surface temperature and precipitation it is shown how the initializedsimulationsperformusingtheleadtimedependentanomalyadjustmentrecommendedbytheDCPP.Furthermore the improvement through the use of the sophisticated post processing procedure(DeFoReSt)isdiscussed.


(B3-06)

SkillassessmentoftheCSIROmulti-yearClimateAnalysisForecastEnsemblesystem

DougieSquire(1),JamesRisbey(2),CarlyTozer(3),DidierMonselesan(4),VassiliKitsios(5),TerryO'Kane(6)

CSIRO(all)

Meaningful decadal-scale forecasts are crucial for informing adaptation strategies in response toclimate variations. Yet, the predictability of the global state at multi-year time scales is not wellunderstood. CSIRO has recently begun development of a probabilistic, decadal-scale, ensembleclimate forecast system to investigate the predictability of oceanographic, atmospheric,biogeochemicaland iceprocesses. In this talkwewill assess theperformanceandpredictabilityofthenewforecastsystemusingavarietyofnewandexistingverificationmetrics.


(B3-07)

Predictionofshort-termclimateextremesusingtheNorthAmericanMulti-ModelEnsemble

Becker,Emily;vandenDool,Huug

NOAAClimatePredictionCenterandINNOVIM

Thisstudyexaminestheforecastskillofextrememonthlyandseasonalmean2mtemperature(t2m)andprecipitationusingtheNorthAmericanMulti-ModelEnsemble(NMME),anensembleofstate-of-the-art coupled global climate models. The NMME currently provides real-time guidance forNOAA’soperationalshort-termclimateforecasts,andincludesadatabaseofretrospectiveforecasts(1982-2010), used forbias correction, calibration, and skill studies. Sevenmodels from theNMMEcontribute to this study: NCEP-CFSv2, Environment Canada’s CanCM3 and CanCM4,GFDL’s CM2.1and FLOR, NASA-GEOS5, and NCAR-CCSM4. Both deterministic and probabilistic forecasts forextremesareconsidered,andforecastskillassessmentsfindthatskillscoresareequaltoorgreaterfor extremes than for non-extreme cases. Results are assessed for the real-time and hindcastperiods,andforvaryingcombinationsofmodelsinthemulti-modelensemble.


(B3-08)

CalibrationandCombinationofNMMEprecipitationforecastoverSouthAmericausingEnsembleRegression

Osman,Marisol(1),Vera,Carolina(1)

CentrodeInvestigacionesdelMarylaAtmosfera(CIMA-UBA/CONICET)

In this studywe calibrate the climatepredictions available at theNorthAmericanMultiModel Ensemble todevelopseasonalprecipitationforecasttoolsoverSouthAmerica.Forecastsvalidateachoverlappingseasonmadewith initialconditionsofthepriormonth(Lead1month)overtheperiod1982-2010areconsidered.AMulti-Model Ensemble (MME) using 8models with around 10 ensemblemembers was constructed and itsperformance was evaluated against CMAP and CPC-UNI database. The domain of application spans [15°N-60°S;275°E-330°E] and it is also divided in two subdomains: tropical South America and extratropical SouthAmerica.

TheEnsembleRegressiontechnique(EREG)appliesaregressionequationdevelopedfortheensemblemeantoeach ensemble member to obtain a probability density function (PDF) which represents the ensembleprediction.EREGisfirstappliedtoeachmodelanditsensemblememberstocalibratethem.Then,therelativeimportanceofeachmodel isdeterminedbyweightingthemaccordingtotheirhistoricalperformance.Todothis,themagnitudeofeachcalibratedPDFisevaluatedattheobservationpointandtheweightisproportionalto the number of times thismagnitudewasmaximum for eachmodel. Finally, two approaches are used toobtaintheconsolidatedPDF.ThefirstoneconsistsinusingtheweightedMMEinanewensembleregression,resultinginaweightedsuper-ensembleregression(WSEREG)togettheconsolidatedPDF.Theothertechniqueconsists inobtaining theconsolidatedPDFcomputing thenormalizedsummingof theweightedmodels’PDF(weighted PDFs, WPDFS). The consolidated PDFs obtained are used to forecast the three equally probablecategoriesbelow,nearandabovenormal.TheseforecastareconfrontedagainstthoseobtainedcountingtheproportionofensemblemembersoftheMMEfallingineachcategory(countingestimatetechnique,CE).

Results show thatbothWPDFSandWSEREGoutperformCE in termsof theRankedProbabilistic Skill Score(RPSS)andBrierSkillScore(BSS)inbothseasons.However,onlyinnorthernSouthAmericatheperformanceofboth consolidation techniques is slightly better than the climatological values of the predictand (threecategoriesequallyprobable).InextratropicalSouthAmericabothRPSSandBSSvalueschangefromlessthan-0.5forCEtonear0forbothWPDFSandWSEREG.Ontheotherhand,reliabilitydiagramscomputedovertheentiredomainshowsthatWPDFSandWSEREGsubstantiallyimprovetheforecastintermsofreliabilityrespecttothanobtainedwithCE.

Furtherstudiesarebeingconductedtotestthesensitivityoftheskilltochangesinthemodel’soriginalspreadaswellaschangesinthewaythatweightsbetweenmodelsarebeingdetermined.Theresultsofthesestudieswillbepresentedduringtheconference.


(B3-09)

EvaluatinganewcalibrationmethodforSeasonalProbabilisticPredictionforIndianSummerMonsoon

AcharyaNachiketa(1),RobertsonAndrew(1),VigaudNicolas(1),TippettMichael(1,2),GoddardLisa(1)andPaiD.S.(3)

IRI,ColumbiaUniversity,USA(1),DepartmentofMeteorology,KingAbdulazizUniversity,Jiddah,SaudiArabia(2),IMD,India(3)

Though a plethora of study exists to make deterministic model for predicting seasonal Indiansummer monsoon rainfall (ISMR), a few studies have described probabilistic prediction whichconveystheinherentuncertaintyoftheforecast.Probabilisticseasonalpredictioncanbedonebasedon the general circulation model (GCM)’s outputs, however the output from these ensembleprediction systems cannot be used directly and requires further calibration in order to producereliable forecasts. In this study, Extended Logistic Regression (ELR) based calibration methodimplemented in models from The North American Multi-Model Ensemble (NMME) project withgridded data from IndianMeteorological Department. Though, ELRmethod has been successfullyapplied in the past to ensemble weather and sub-seasonal forecast, this is the first time to ourknowledge that it hasbeenused toproduce seasonal probabilistic forecast of ISMR. ELRmethodsalso allowed generating forecast in more flexible format in addition to commonly used tercileprobability forecast.Theflexible formatenablesuserstoextract informationforthosepartsof theforecast distribution of greatest interest to them, especially the probability of extremely dry/wetconditions.TheskillofELR-basedforecastsisevaluatedover1982-2010followingaleave-one-year-outcross-validationandarefoundtobemoreskillfulthanthemorefamiliarapproachofestimatingtheforecastprobabilitiesbycountinghowmanymembersexceedacertainthreshold.


(B3-10)

Potentialofcombinedstatistical-dynamicalsub-samplingapproach.

MikhailDobrynin(1),KristinaFröhlich(2),WolfgangA.Müller(2),TimStockdale(3),AdamScaife(4),JohannaBaehr(1)

InstituteofOceanography,CenterforEarthSystemResearchandSustainability(CEN),UniversitätHamburg,Germany(1),DeutscherWetterdienst(DWD),OffenbachamMain,Germany(2),ECMWF,

UK(3),MetOfficeHadleyCentre,UK(4)

Thepotentialofacombinedstatistical-dynamicalsub-samplingapproachisanalysedintheseasonalforecasts currently contributing to Copernicus Climate Change Service (C3S). For four seasonalforecastsystems,weanalysetheenhancementofwinterNorthAtlanticOscillation(NAO)predictionskill. First, Teleconnections between the autumn state of sea surface temperature, sea ice, snowdepth, and stratospheric temperature and the subsequent winter NAO are established as NAOpredictorsintheERA-Interimreanalysisfrom1979-2000.Second,theseNAOpredictorsareusedtoderivea statistical “firstguess” foreachwinterNAO from2001 to2017.Third,every ``firstguess''NAO is used as a reference for the sub-sampling of each dynamical ensemble for every seasonalpredictionsystem.Forallusedsystem,NAOpredictionskill isconsiderablyimproved,whilesurfacepropertiessuchassealevelpressureorsurfacetemperaturemostlyshowaregional improvement,alsodependentontheforecastssystemsabilitytosimulateNAOteleconnections.


(B3-11)

RoomforImprovementinSeasonal-to-DecadalClimatePrediction

Sang-IkShin,MatthewNewman

CIRES,UniversityofColoradoandPSD,NOAA/ESRL

Thehindcastskill, for leadsfromaseasontoadecade,ofthecurrentgenerationofcoupledGCMs(CGCMs)isassessedandbenchmarkedwithskillofanempiricalmodel,aLinearInverseModel(LIM).Both systems produce sea surface temperature (SST) and sea surface height (SSH) forecasts. Forseasonal forecast leads, the CGCM ensemble mean hindcasts come from the North AmericanMultimodel Ensemble (NMME), while for interannual-to-decadal leads the CMIP5 hindcasts areanalyzed. The LIM, constructed fromnear global (60oS-65oN) observedmonthly anomalies duringtheperiod1961-2010,producesforecastsfrom1monthto9yearslead.

TheLIMskill iscomparabletoorbetterthantheCGCMensemblemean,aswellaslocalunivariateAR(1)process,atall timescalesandall locations.Notably, theLIM issignificantlymoreskillful thanthe CGCM ensemblemean over the extratropics, especially at longer forecast lead. For example,significantskillinthePacificDecadalOscillation(PDO),inbothPDOphaseandamplitude,upto6-9yrleadinLIMarewellcomparedtorelativelylowinsignificantskillsinCMIP5models.

Analysisof the LIM suggests thatpossibleerror in representing the SST-SSH coupling, rather thanuncertaintyinforecastinitialization,isamajorcauseofreducedSSTandSSHskillintheCGCMs.Thisalso suggests that reducing thismodel error should improvemodel prediction skill of seasonal-to-decadalSSTandSSHanomalies.

SESSION:(B4)S2Dforecastsfordecisionmaking

(B4-01)

ClimatePredictionsforFisheriesApplications

Tommasi,Desiree(1),CharlesStock(2)

NOAASWFSC,USA(1),NOAAGFDL,USA(2)

Theproductivityanddistributionoffishpopulationsisstronglyinfluencedbyclimatevariability.Theinability of fisheriesmanagers to anticipate such environment-driven fluctuations in fish dynamicscan lead to overfishing and stock collapses. Herewe demonstrate how recent advances in globaldynamical climate prediction systems have allowed for skillful sea surface temperature anomalypredictions at a scale useful to understanding and managing fisheries. Such predictions presentopportunities for improved fisheriesmanagementand industryoperations.Pioneeringcasestudiesdemonstrating the utility of seasonal climate predictions to inform fisheries decisions will behighlighted.Weconcludebyoffering remarksonprioritydevelopments required for theexpandeduseofclimatepredictionsforfisheriesmanagement.


(B4-02)

Applicationofoperationalseasonalpredictionsystemsforseasonalpredictionoffiredanger:acasestudyoftheextremewildfireeventsinCalifornia,SpainandPortugalof

2017

EtienneTourigny(1),JoaquinBedia(2),RaulMarcos,OmarBellprat,FranciscoJ.Doblas-Reyes(1,3)

BarcelonaSupercomputingCenter(1),SantanderMeteorologyGroup,PredictiaIntelligentDataSolutions(2),ICREA(3)

TheextremewildfireeventsinCalifornia,PortugalandSpainduringthefireseasonsof2017causedconsiderableeconomic, environmental andhuman life lossesandgatheredmuchmediaattention.Theseeventshighlightedtheneedforbothshortandmedium-termforecastsofwildfiredanger,thelatterusefulforraisingawarenessandpreparingforwildfirepreventionandsuppressionstrategies.TheEuropeancountriesmostaffectedbywildfiresareintheMediterraneanbasin,withsummerfiresoccurringduringperiodsofdrought.WhilecountriessuchastheUnitedStates,CanadaandAustraliahavedevelopedextensiveandreliableshort-termandseasonalwildfireforecastingsystems,similarsystemsinEuropearecomparativelylesswellestablished.111

Forexample,theEuropeanForestFireInformationSystem(EFFIS)producesshort-term(1-10days)firedangerforecasts,butdoesnotprovideseasonalforecasts.Theseshort-termforecastsrelyonCanadianFireWeatherIndex(FWI)valuescomputedfromtheoperationalweatherpredictionsfromECMWF.TheMcArthurForestFireDangerIndex(FFDI)isasimilarsystemoriginatingfromAustralia.FWIandFDDIarecomputedfromdailyvaluesoftemperature,precipitation,relativehumidityandwinddata,accountingforfactorsthatareimportantforfireseverityandspread.Thesedailymeteorologicalvariablescanbeobtainedfromobservations,climatepredictionsorclimateprojections.

Ourapproach to seasonalpredictionof fire risk is touseoperational forecasts suchas those fromECMWF’sSystem5forecastsystemto issuepredictionsofthefiredanger indicesFWIandFFDI.Asoperational forecasts produce ensemble predictions of daily values of the relevantmeteorologicalvariables,weareabletocomputeensembledailypredictionsoffiredangerindicesataglobalscale,whichcanbeused to formulateprobabilisticpredictions.Globalobservationsofburnedarea fromtheMCD64globalburnedareaproductandtheGlobalFireEmissionsDatabaseversion4(GFED4)areusedtoevaluatetheskillofthepredictions.WewillshowtheskillofthepredictionsfortheIberianPeninsulaandCalifornia,withafocusontheextremewildfireseasonssuchasthatof2017.1

1


(B4-03)

UsingSubseasonaltoSeasonalForecastGuidancetoSupportFamineEarlyWarningSystemsNetworkInternationalFoodSecurityAssessments

Hoell,Andrew(1),Verdin,James(2)

NOAA/EarthSystemResearchLaboratory(1),FamineEarlyWarningSystemsNetwork(2)

Earlywarningofacutefoodinsecurityrequiresrobustagro-climatologicalassessmentsatmanytimehorizons,fromthemonitoringofcurrentconditionstoforecastsrangingfromweeklytointerannualtime scales. This poses a unique challenge to FamineEarlyWarning SystemsNetwork (FEWSNET)physicalsciencepartnerscomprisedofindividualsfromNOAA,USGS,NASAandUSDA.Oneaspectofthis unique challenge lies in the integration of all relevant forecasts in the context of emergingpredictability information gathered through research and the clear communicationof key forecastinformationtothesocialscientists,whoalsoconsidermarkets,tradenutritionandhealthwhentheyproduce food security outlooks. The food security outlooks are ultimately used to inform theprogrammingoflimitedaidresourcesforthetimesandplaceswheretheywillbemostneeded.

We summarize how FEWS NET physical scientists provide agro-climatological guidance over thesubsequent 9 months for countries in sub-Saharan Africa, Central Asia, Central America andCaribbean.FEWSNET’sphysicalsciencepartners:(1)monitorcurrentagro-climatologicalconditions,suchassoilmoisture,precipitation,temperatureandcrophealth, (2)gatherandinterpretweatherandclimateprojections tobetterunderstand the likely future conditionsonweekly to interannualtime scales, and (3) research the drivers of regional agro-climatic variability on sub-seasonal tointerannualtodecadaltimescalesinthecontextofachangingclimate.

KeypositivefeaturesofFEWSNETclimateservicesinclude:

(1)Directcommunicationofagro-climatologicalquestionsbyfoodsecurityanalyststothephysicalscientists,resulting inreducedneedtospeculateaboutwhatagro-climatological informationmightberelevantand(2)frequentdirectcommunicationbetweenFEWSNETphysicalscientistsandfoodsecurity analysts generates highly relevant research questions, and the resulting research findingsconsequentlyhaveafasttracktoapplicationandgenerationofsocietalbenefits.


(B4-04)

8-MonthSnowpackPredictionPotential

SarahB.Kapnick1,XiaosongYang1,2,GabrielA.Vecchi3,ThomasL.Delworth1,RichGudgel1,SergeyMalyshev4,P.C.D.Milly5,ElenaShevliakova1,4,SethUnderwood1,StevenA.Margulis6

(1)GFDL/NOAA(2)UCAR(3)PrincetonUniversity(4)USGS(5)UCLA(6)

WesternU.S. snowpack—snow that accumulates on the ground in themountains—plays a criticalrole in regional hydroclimate and water supply with 80% of snowmelt runoff being used foragriculture. Utilizing observations, climate indices, and a suite of global climate models, wedemonstrate the feasibility of seasonal snowpack predictions and quantify the limits of predictiveskill 8 months in advance. This physically-based dynamic system outperforms observation-basedstatistical predictionsmadeon 1 July forMarch snowpack everywhere except the Southern SierraNevada, a region where prediction skill is nonexistent for every predictor presently tested.Additionally, in the absence of externally forced negative trends in snowpack, narrow maritimemountain rangeswithhighhydroclimatevariabilityposea challenge for seasonalprediction inourpresentsystem;naturalsnowpackvariabilitymayinherentlybeunpredictableatthistimescale.Thiswork highlights present prediction system successes and provides a roadmap for testing futureseasonal prediction systems to improveprediction skill.We find that a prediction systemmustbespecificallydesignedforuserneeds(e.g.leadtimes,choicevariables,eventsofinterest),highlightingtheneedforengagementwithstakeholdersforfuturepredictionsystemdevelopment.

桔 �潬眠 �楴整 � ʹ業� 潴獮漠 �批慥獮愠楬湯湯 �景


(B4-05)

HarnessingNMMEpredictionstosupportseasonalhydrologicprediction

Lehner,Flavio(1),Wood,Andrew(1),Llewellyn,Dagmar(2),Blatchford,Douglas(2),Goodbody,Angus(3),Pappenberger,Florian(4)

NCAR,USA(1),BureauofReclamation,USA(2),NRCSNWCC,USA(3),ECMWF,UK(4)

RecentstudiesdocumentanincreasinginfluenceoftemperatureonstreamflowacrosstheAmericanWest,includingsnow-meltdrivenriverssuchastheColoradoorRioGrande.Atthesametime,somebasinsarereportingdecreasingskillinseasonalstreamflowforecasts,termedwatersupplyforecasts(WSFs), over the recent decade.While the skill in seasonal precipitation forecasts fromdynamicalmodels remains low, the little skill there is in seasonal temperature forecasts could potentially beharvested for WSFs in these temperature-sensitive basins. Here, we show that WSF skill can beimprovedbyincorporatingseasonaltemperatureforecastsfromtheNationalMulti-ModelEnsemble(NMME) into traditional statistical streamflow forecastmodels.We find improved skill relative totraditionalWSF approaches in amajority of headwater locations in the Colorado and Rio Grandebasins at lead times of 1-5 months. Incorporation of temperature into WSFs can increase theresilience of streamflow forecasting and water management systems in the face of continuingwarming aswell as decadal-scale temperature variability and thus help tomitigate the impacts ofclimatenon-stationarityonstreamflowpredictability.Wediscuss thepotential for incorporationofthese results into dynamic hydrologic forecasting models and explore NMME model weightingschemestopotentiallyfurtherboosthydrologicpredictionskillacrosstheAmericanWest.


(B4-06)

UDECIDE:UnderstandingDecision-ClimateInteractionsonDecadalScales

JamesDone(1),JeffreyCzajkowski(2),TapashDas(3),MingGe(1),JoshuaHewitt(4),JenniferHoeting(4),HeatherLazrus(1),RebeccaMorss(1),ArminMunévar(3),ErinTowler(1),MariTye(1)

andAlexVanZant(5)

NationalCenterforAtmosphericResearch,USA(1),Wharton,UniversityofPennsylvania,USA(2),CH2M,USA(3),DepartmentofStatistics,ColoradoStateUniversity,USA(4),RutgersBusinessSchool,

USA(5).

Waterresourceandfloodmanagers increasinglyrequirepredictiveclimate information indecision-relevant terms to enable appropriate planning and adaptation to future conditions. A number ofdecisions stand to benefit from predictive information on decadal scales. The UDECIDE(Understanding Decision Climate Interactions on Decadal Scales) project brings togetherpractitioners,engineers,statisticians,socialscientistsandclimatescientiststounderstandtheroleofdecadalclimateinformationforwaterresourceandfloodmanagementdecisions.

UDECIDE proceeds in two overlapping parts. Part 1 explores current information needs and use,through in-depthunderstandingdevelopedacross floodandwater resourcemanagers inColoradoand California. Part 2 explores what can be skillfully predicted on decadal scales through thedevelopmentofnewstatistical-dynamicalmodelingtechniques.Prototypepresentationsofdecadalpredictions are developed at the intersections of what is needed and what is skillful. Thesepresentationsaretestedwithstakeholdersand iteratedupontobuildunderstandingoftheroleofdecadalclimatepredictionfordecisions.

Thispresentationwilloutlinetheemergingpointsofintersection.Potentialrolesfordecadalclimateprediction based on interviewswithwater resource and floodmanagerswill be presented. Thesepotential roleshave informed thedevelopmentofanewgeostatisticalmodelofprecipitation thatsimultaneously accounts for the effects of local and remote covariates, and global dynamicalmodelingofatmosphericrivercharacteristicsandpredictabilityunderdecadalmodesofvariability.Implications forwaterandfloodriskmanagementpracticeover thenextdecadewillbediscussed.


(B4-07)

Incorporatingdecadalpredictionsintowatermanagement

Towler,Erin(1),Done,James(1),Yates,David(1)

NCAR,USA(1)

Tounderstandtheirclimatesensitivities,watermanagersusehydrologicmodelstotranslateclimateinformationintoparametersthatarerelevanttotheiroperationsanddecision-making.Manywatermanagersarefamiliarwithseasonalclimateforecastsandlong-termclimatechangeprojections,andhave started to consider this information in their planning and management. Decadal climatepredictionsarelesswell-knownamongwatermanagers,butmeetaplanninghorizonneed.Decadalpredictionusage canbenefit fromsomeof the lessons learned fromusageofpredictions atothertimescales,butarealsounique,warrantingnewstudy.Thispresentationwilldrawuponarecentlydevelopedframeworkthatexploreshowdecadalpredictionscanbeused.Specifically,wewilluseacasestudyapproachinthewatersector,wherebydecadalpredictionsarestatisticallydownscaledtobe used as inputs to a hydrological model. By comparing current practices for using seasonalforecastsandclimatechangeprojectionsinwatermanagement,wewilldiscussthedifferencesandsimilarities,needsandbenefits,aswellasremainingchallenges,ofincorporatingdecadalpredictionsinwatermanagementdecisions.ThisworkispartofanongoingNationalScienceFoundation(NSF)-fundedproject,UnderstandingDecision-ClimateInteractionsonDecadalScales(UDECIDE),thataimstounderstandtheroleofdecadalclimateinformationforwatermanagementdecisions.


(B4-08)

SeasonalanddecadalpredictionservicesoftheCopernicusClimateChangeService(C3S)-currentstatusandplansforthefuture

Vamborg,FrejaS.E.(1),Brookshaw,Anca(1),Buontempo,Carlo(1),DeeDick(1),PenabadRamos,Eduardo(1),Thépaut,Jean-Noël(1)

EuropeanCentreforMedium-RangeWeatherForecasts(1)

CopernicusistheEuropeanCommission’sflagshipEarthobservationprogrammethatdeliversfreelyaccessibledatatoanoperationalscheduleandinformationservices.ECMWFhasbeenentrustedtooperatetwokeypartsof theCopernicusprogramme,whichwillbringaconsistentstandardtothemeasurement,forecastingandpredictionofatmosphericconditionsandclimatechange.

TheCopernicusClimateChangeService(C3S)currentlyprovidesaportfolioofproductsandservicesthat includeseasonalaswellascentennial timescales,aswellas theuseofsuchproducts forkeyeconomic sectors in need of tailored climate information. Here we present the current role ofseasonalpredictionintheservice,aswellasfutureplansforclimatepredictionintheservice.

An important element of C3S, at present, is a seasonal forecast service, based on amulti-systemframework.Datafromstate-of-the-artseasonalforecastsystemswithoperationalstatusatanumberof European institution is collected by C3S, where forecast products are generated and madeavailable to the public in graphical and digital format. Some of the elements of the participatingforecastsystemshavebeenharmonised, tooptimise thebenefitsavailable tousers (e.g.minimumhindcastperiodandensemblesize).Theoutputs-whichincludereal-timeforecastsandhindcasts,aswell as graphical products from the individual contributing systems and as a combination - areavailableatastandardresolution,instandardisedformats.ThesystemscurrentlyparticipatingintheC3Sseasonalserviceare fromECMWF,UKMetOffice,MeteoFrance,CMCCandDWD; inthenextyearnon-EuropeanCentresarealsolikelytojointheserviceasin-kindcontributors.

A largefractionofthesectoralusersarekeentoobtain informationontheshorttime-scales(e.g.,seasonalandsub-seasonal). Seasonalpredictions fromC3Smodelshavebeenused forenergyandwaterapplications. Inparticular,oneC3Scontract implementedanoperationalhydrologicalserviceoperatingatapan-Europeanlevel.Theanalysisoftheskillofthissystemhasrevealedareaswherethe skill is in river-flow is significantly larger than the skill thatexists in theatmosphericdriversofriver-flowvariability.

The seasonal component of the C3S will continue to expand: in the near-term by increasing thenumber of graphical and digital products, and in the longer term by increasing the number ofcontributorstothemulti-modelsystem,thusprovidingenhancedinformationinsupportofclimateservices(e.g.WMORegionalClimateCentres,RegionalClimateOutlookForums).

Even thoughmanyusershavealso shownan interest inobtaining informationon the longer timescale,decadalpredictionproductswerenot included in the initialC3Sportfolio,due to the lackofmaturityofthesciencetomakeitanoperationalservicefromtheoutset.C3Sisnowtakingstockoftheprogressthathasbeenmadesincethen,toreconsiderwhetherdecadalpredictionscouldbepartof the new C3S portfolio. An important element of this component would be the provision of ascientifically sound framework for the evaluation of decadal hindcasts and the statistical post-processingofdecadalpredictions.


(B4-09)

Co-productionofseasonalforecasts:experiencesfromNorway

Kolstad,ErikW.(1)

UniResearchClimate,BjerknesCentreforClimateResearch,Bergen,Norway(1)

TheSeasonalForecastingEngine(SFE)researchprojectaimstodevelopastate-of-the-artoperationalseasonalclimatepredictionsystemforNorthernEuropeandtheArctic.Ourmotivationisthatmanycompanies and public stakeholders face climate-related risks that must be managed to staycompetitive and to protect life, property and the environment. To our users, the SFE will beaccessiblethroughaflexibleinterfacewhichcanbequeriedtoobtainpredictionsofrelevantclimateindicesandvariables.Underthehood,our‘engine’consistsofstatisticalalgorithmsthatmergevastamountsofdataintounifiedforecasts.InthistalkIwillsharesomeofourexperiencesfromworkingwithourusersfromthepublicsectorandlargebusinesscorporations.Fromthepublicsector,IwillfocusontheNorwegiancoastguard,whoareprimarilyinterestedinseaiceforecastsintheBarentsSea.Howdowereconciletheirdemandsforhigh-resolution(bothspatiallyandtemporally)forecastswiththeforecastskillontheseasonaltimescale?Thesamebalancehastobefoundindialoguewithourbusinessusers,whichcome fromhydropower,windpower,weather services,and insurance. Ibelievethattheseexperiencesareofgeneralinteresttoprovidersofseasonalforecastsandclimateservicesmoregenerally.

SESSION:(B5)Hindcastandforecastqualityassessment

(B5-01)

RecentDevelopmentsinForecastQualityAssessment

TimothyDelSoleandMichaelK.Tippett

GeorgeMasonUniversity(1)andColumbiaUniversity(2)

Thistalkgivesageneraloverviewofsomerecentdevelopmentsinforecastqualityassessment.First,I discuss the problem of deciding whether one forecast is superior to another, which is a basicquestioninmulti-modelforecastingandinmodeldevelopment.Iwillshowthatstandardstatisticaltestsbasedondifferencesincorrelation(ormeansquareerror)donotgivecorrectresultswhentheskill measures are computed from data over a common period or with a common set ofobservations. I thendiscuss several other tests that give correct results, including a relativelynewtest called the randomwalk test. In probabilistic forecast verification, I discuss the importance ofproperscoresandhownon-localscorescanleadtocounter-intuitiveresults.Ialsoshowhowproperscores can be subjected to rigorous hypothesis tests. Finally, I discussmultivariate techniques forassessingforecastfields.


(B5-02)

Robustevaluationofseasonalforecastqualityusingteleconnections

Volpi,Danila(1),Ardilouze,Constantin(1),Batté,Lauriane(1),Guérémy,Jean-François(1),Déqué,Michel(1)

CentreNationaldeRecherchesMétéorologiques(CNRM),Météo-France,CNRS,Toulouse,France

In response to the high demand formore reliable climate information at the seasonal timescale,innovative climate prediction systems are developedwith improved physics and increased spatialresolution.Alongsidethemodeldevelopmentprocess,seasonalpredictionsneedtobeevaluatedonpastyearstoproviderobustinformationontheforecastperformance.ThisworkpresentsthequalityassessmentofanintermediateversionoftheMétéo-FrancecoupledclimatepredictionsystembasedonCNRM-CM.Inordertohavearobustevaluation,theexperimentisperformedwith90ensemblemembersovera37-yearre-forecastperiodfrom1979to2015.Wefocusontheborealwinterseason(December to February) initialised in November. Beyond typical skill measures we evaluate themodelcapabilityinreproducingENSOandNAOteleconnectionswithprecipitationandnearsurfacetemperaturerespectively.Suchanassessmentiscarriedoutfirstthroughacompositeanalysis,andshowsthatthemodelsucceedsinreproducingthemainpatternsfornearsurfacetemperatureandprecipitation.Acovariancemethodleadstoconsistentresults.Finallywefindthatshorteningto23years the verification period and reducing to 30members the ensemble size does not impact therepresentation of teleconnections in themodel re-forecasts. In a second stage of the study, twodifferent versions of the model used in the operational seasonal forecasting systems 5 and 6 atMétéo-Franceareshownandcompared,usingexperimentsthatfollowtheoperationalstandards(30members,24yearverificationperiod).


(B5-03)

Signalandnoiseinregimesystems:understandingNAOpredictability

Strommen,Kristian(1),Palmer,Tim(2)

OxfordUniversity

Overthelastdecade,NWPmodelshavebeguntoexhibitsignificantskillatpredictingthewintertimeNorthAtlanticOscillation(NAO)index.Inparticular,recentstudiesconductedbytheUKMetOfficereportedsignificantskilloverahindcastperiodrangingfrom1980to2015.Atthesametime,averylow signal-to-noise ratio was observed. This was captured using the `ratio of predictablecomponents' (RPC) metric, designed to measure the extent to which reality is more or lesspredictable than themodel itself.AnRPC less than1 indicates that themodel ismorepredictablethanreality,andgreaterthan1thatitislesspredictable.A`signal-to-noiseparadox'wasdescribedinSiegertet.al(2016)astheobservationthat,whiletheensemblemeanoftheirhindcastscorrelatedstronglywithobservations,theRPCwassignificantlygreaterthan1,suggestingthemodelisinsomesensehighlyunderconfidentdespiteitsnotablelevelofskill.

ManystudieshaveemphasizedtheroleplayedbyregimesinmodulatingNorthAtlanticweather.Wewillexaminehowactualskill,potentialpredictabilityandRPCareexpectedtobehaveinanidealizedtwo-stateregimesystem.Herethetwostatescanbelooselybethoughtofas`positiveNAO'(zonaljetstructure)and`negativeNAO'(morewavyjetstructure),andregimetransitionsareassumedtotakeplaceondailytimescales.Predictabilityisassumedtobeinducedasaconsequenceofseasonaldeviationsofthepersistenceprobabilitiesofthetwostatesfromtheirclimatologicalmeans.Modelskill isthenassumedtobeaconsequenceofthemodelsabilitytocorrectlycapturetheseseasonalvariations.We show that in this framework, the combination of high levels of skill and large RPCvalues can be expected for any model which captures the signal, but systematically reduces theoverall level of regime persistence. In particular, a high RPC value on seasonal timescalesmay beexpectedevenwhenthemodelslevelofinternalnoiseondailytimescalesisperfectlyrealistic.ThislackofpersistenceisshowntobeacommonfeatureofseveralcurrentNWPmodels.

Finally, we will compare the RPC metric with another commonly used method of evaluatingover/under-confidenceofmodels,namelytheratiooftheRMSerrortoensemblespread.Itisshownthat for models that are not statistically perfect, these metrics may not give the same answer,depending on if the underlying system is assumed to be highly linear, or e.g. a more non-linearregime-likesystem.Thisemphasizestheimportance,whenevaluatinghindcastquality,ofthechoiceofstatisticalmodelusedtodescribethesystem.


(B5-04)

CanonicalskillanalysisoftropicalPacificvariabilityintheCCCmadecadalhindcasts

ReinelSospedra-Alfonso

ECCC/CCCma

This study identifies the predictable components and the canonical skill components of wintertropicalPacific sea surface temperatureanomalies (SSTAs) indecadalhindcasts fromtheCanadianCentreforClimateModellingandAnalysis(CCCma)CanCM4climatemodel.Predictablecomponentsareobtainedbyapplyingaprincipal componentanalysis to thehindcast ensemble thatmaximizesthe signal-to-noise ratio, a procedure known as predictable component analysis (PrCA). Canonicalskillcomponentsarederivedbyfindingthespatialweightsofthesignalportionofthehindcasts(inthesenseofPrCA)andthoseoftheverifyingobservationsthatmaximizethetemporalmeansquareskill score of modeled and observed averaged SSTAs, a procedure that is analogous to canonicalcorrelationanalysis.Theconditionalbiasoftheseaveragedhindcastsvanishesresultingincanonicalmean square skill scores that equal the square of the canonical correlations. It is shown that theactualcorrelationofthefilteredhindcastsandtherawobservationscanbeexpandedexactlybythesum of the correlation of the canonical patterns multiplied by the canonical correlations, plus aresidual representing a portion of the observed variance that can bemade arbitrarily small. Thisdecompositionenablesonetodeterminethecomponentsofthehindcaststhatcontributetoactualskill.Thehindcastsfilteredbypredictablecomponentanalysisandcanonicalskillanalysisareshowntobegenerallymoreskilfulthantherawhindcastsonseasonaltomulti-annualtimescales.


(B5-05)

Anadvancedscoreforevaluatingseasonalforecastskill

Düsterhus,André(1),Wahl,Sabrina(2,3)

Inst.ofOceanography,CEN,Univ.ofHamburg,Germany(1),MeteorologicalInst.,Univ.ofBonn,Germany(2),HansErtelCenterforWeatherResearch,ClimateMonitoringBranch(3)

Acommontool toevaluate theperformanceof seasonalanddecadalpredictionsare theAnomalyCorrelation Coefficient (ACC) and the root-mean-square error (RMSE). Both methods rely onassumptions, among others normality assumptions of the data or uncertainty. Failing theseassumptions can lead tounder-oroverestimationof the skill.Also,RMSEandACC showdifferentcharacteristicsoftheforecastskill,butdoitinanon-comparableway.

Wehavedevelopedanewscoreforpredictions,whichisindependentofthedatadistribution.Thescorecategorisesthemodelandobservationaldataandcreatesamulti-categoricalcontingencytable(MCT).Onthisametricfromimageprocessing,thetwo-dimensionalEarthMover'sDistance(EMD),isappliedanda scoreestimated.Applyingacategorisationallows toevaluateclearlynon-gaussiandatasetslikeprecipitation,butalsoforvariablesliketemperature,whichareassumedtobegaussian.

We demonstrate this at the hand of an analysis of seasonal prediction. We show strategies togeneratespatialpatternswiththisnewscore,whicharesimilartotheACCandRMSEandsobringbothprocedurestothesamebasisandmakingthemcomparable.WealsoshowdifferencesfortheACC,whenthenormalityassumptionisdropped.


(B5-06)

Howskillfularethemulti-annualforecastsofAtlantichurricaneactivity?

Caron,Louis-Philippe(1),Hermanson,Leon(2),Dobbin,Alison(3),Imbers,Jara(3),Lledo,Llorenç(1),Vecchi,GabrielA.(4)

BarcelonaSupercomputingCenter,Spain(1),MetOfficeHadleyCenter,UnitedKingdom(2),RiskManagemetnSolutions,UnitedKingdom(3),PrincetonUniversity,USA(4)

The recent emergence of near-term climate prediction has prompted the development of threedifferentmulti-annualforecastingtechniquesofNorthAtlantichurricanefrequency.Descriptionsofthese three different approaches, aswell as their respective skill, are available in the peer-reviewliterature, but because these various studies are sufficiently different in their details (e.g. periodcovered,metricusedtocomputetheskill,measureofhurricaneactivity), it isnearly impossible tocompare them. Using the latest decadal re-forecasts currently available, we present a directcomparison of these three multi-annual forecasting techniques with a combination of simplestatisticalmodels,withthehopeofofferingaperspectiveonthecurrentstate-of-the-artresearchinthis field and the skill level currently reached by these forecasts. Using both deterministic andprobabilisticapproaches,weshowthattheseforecastsystemshaveasignificantlevelofskillandcanimproveonsimplealternativessuchasclimatologicalandpersistenceforecasts,andhaveskillsimilarto,orbetterthan,statisticalforecastscurrentlyusedbythecatmodelingindustry.


(B5-07)

Makingsenseofseasonalsea-iceforecasts

Tietsche,Steffen(1),Balan-Sarojini,Beena(1),Mayer,Michael(1,2),AlonsoBalmaseda,Magdalena(1)

(1)ECMWF,UK(2)UniVienna,Austria

Withtherecentintroductionofprognosticsea-icemodelsintoseasonalforecastingsystems,routineseasonalsea-iceforecaststhatareusefultosocietyarewithinreach.Thesearedirectlyrelevantforplanningmarineoperationsinhighlatitudes,andscientificevidenceisaccumulatingthattheyhavefar-reaching implications for predicting large-scale anomalies of circulation and temperatures.However,currentsystemssufferfromstrongbiasesinthesea-icestate,whicharedifficulttocorrecta-posterioribecauseoftheirnon-Gaussian,non-linearnature.Deficienciesincurrentsea-iceforecastare often rooted in sea-ice thickness.Here,wemake the case thatmodel biases and initializationmethodsforsea-icethicknessneedtobeimprovedinordertomakeprogresswithseasonalsea-iceforecasts, and that these improvements need to be guided by innovative satellite observations ofsea-ice thickness. We discuss the skill of the recently implemented ECMWF seasonal forecastingsystem SEAS5 in predicting sea ice, using innovative metrics of integrated ice-edge error andintegratedicethicknesserror.Wefindthatforecasterrorsiniceedgeandiceconcentrationcanbequantifiedwith high certainty, but forecast errors of ice thickness aremuch less certain, becauseobservational products and reanalyses have large discrepancies. We illustrate this by discussingexperimentswheresea-icethicknesshasbeeninitializedfromobservationaldatasetsandconcludethatprogresswithsea-icepredictionsrequiresclosecollaborationbetweendifferentcommunities,inordertoimprovethefidelityofbothearthobservationsandforecastingmodels.

SESSION:(B6)Frontiersinearthsystemprediction

(B6-01)

Decadalpredictabilityoftheoceancarbonuptakevariation

Li,Hongmei(1),Ilyina,Tatiana(1),Müller,Wolfgang(1),Landschützer,Peter(1)

MaxPlanckInstituteforMeteorology,Germany(1)

The global oceans, as an important sink of CO2 emitted by humans, play an essential role inmodulating the global carbon cycle and hence the global climate change. To achieve the ParisAgreement goals of limiting surface warming below 2.0°C and even pursuing to limit it to 1.5°Crelative to the preindustrial level, one major requirement will be discerning the anthropogeniccarbon emission pathway in the Earth System in order to verify the effectiveness of fossil fuelemissions reduction measures. Recent observation-based studies have revealed unprecedentedstrengthindecadalvariationoftheoceancarbonuptakeinthelastdecades,challengingourabilityto predict the variability of the ocean carbon sink and the future global warming on decadaltimescales.

GrandensemblesimulationsstartingfromdifferentinitialstatesrevealedpossibilityforEarthsystemmodels (ESM) to produce the decadal variation of ocean carbon sink that is comparable toobservations, and suggested an essential effects of initial conditions on the evolution of oceancarbon sink [Li and Ilyina 2018]. By applying decadal prediction framework to the Max PlanckInstituteESM(MPI-ESM),wefoundapotentialpredictiveskillofNorthAtlanticCO2uptakeupto4-7yearsthroughimprovementoftheinitialstates intheoceanphysics[Lietal.,2016].Séférianetal.[2018] confirmed the potential predictive skill of the ocean carbon sink up to 6 years based on a‘perfect model’ approach. Such a skill was confined to the model world due to large gaps inobservations.

We extend our decadal predictions by assessing the skill of the global ocean carbon sink againstobservation-based estimates. Based on a high resolution configuration of MPI-ESM, assimilatingobservedtemperatureandsalinity,weareabletoreproducetheobservedcarbonuptakevariability.Wedemonstrateapredictiveskillof theglobaloceancarbonsink in2years throughensemblesofhindcast simulations starting from the assimilation state, which is close to observations. Such apredictiveskillismoreprominentinwintermonths,mainlybecauseoftheimprovedrepresentationoftheoceandynamicsduetoinitialization.

Decadalpredictionstudyoftheoceancarbonuptakeandcorrespondingprocessesisstillatitsearlystageandfacinganumberofchallengesarisingfromlimitedoceanbiogeochemicalobservationsandlackofproperinitializationstrategies.

Li, H., and T. Ilyina (2018), Current and future decadal trends in the oceanic carbon uptake aredominatedbyinternalvariability,Geophys.Res.Lett.,45,916–925.

Li,H.M.,T.Ilyina,W.A.Muller,andF.Sienz(2016),DecadalpredictionsoftheNorthAtlanticCO2uptake,NatureCommunications,7,11076.

Séférian, R., S. Berthet, andM. Chevallier (2018), Assessing thedecadal predictability of land andoceancarbonuptake,Geophys.Res.Lett.,doi:10.1002/2017GL076092.


(B6-02)

IntegrationofcarboncyclecomponentsintoESM-basedpredictionsystems

TatianaIlyina(1),HongmeiLi(1),RaffaeleBernardello(2),LaurentBopp(3),JohnDunne(4),PierreFriedlingstein(5),ChrisJones(6),AndrewLenton(7),NicoleLovenduski(8),Jong-yeonPark(4),

RolandSeferian(9),StephenYeager(10)

MaxPlanckInstituteforMeteorology(1),BarcelonaSupercomputingCenter(2),CNRS/EcoleNormaleSuperieure/IPSL(3),NOAAGFDL(4),UniversityofExeter(5),MetOffice(6),CSIRO(7),

UniversityofColorado(8),Meteo-France/CNRS(9)NCAR(10)

AdvancementofESM-basedpredictionsystemsbyintegratingthecarboncyclecomponentsenablespredictions of variations of the ocean and land carbon sinks. This new knowledge is pivotal forpredicting the Advancement of ESM-based prediction systems by integrating the carbon cyclecomponents enables predictions of variations of the ocean and land carbon sinks. This newknowledge is pivotal for predicting the fate of anthropogenic CO2 emissions and for facilitatingverificationofnear-termemissiontrendsinsupportoftheUNFCCCglobalstocktakes.

Such prediction systems are typically forced with prescribed historical atmospheric CO2concentrationsandassimilateonlythephysicalclimatedata,albeitusingdifferentassimilationandinitializationdesigns.Thelandandoceancarboncyclemodelsrunaspassivecomponentsadjustingtothestateandthephaseofthephysicalclimatesystem.

Duetothelackofcoherentbiogeochemicaldatasets,recentstudiesmostlyfocusedonthepotentialpredictionskillofthecarbonsinks.Theseperfectmodelstudiesprovidemechanisticunderstandingofthesourcesofpredictabilityofthecarboncyclevariationswithassessmentofuncertainties.Theyfurthermoresuggesta largecontributionofnaturalvariability thattakeplace inthebackgroundofthe forced signal driven by rising CO2 emissions.Despite substantial uncertainties associatedwithquantitativeestimatesofyearlytodecadalvariationsinthelandandoceancarbonsinks,thereisarobust potential prediction skill established in a number of prediction systems. For instance,potentialpredictionskillforlandandoceancarbonsinkisabout4years(Seferianetal.2018)andregionallyintheNorthAtlanticitisupto7years(e.g.Lietal.,2015).

Available observational synthesis products based on neural networks or satellite observations oflandandoceanbiogeochemistryenableassessmentoftheeffectivepredictionskill.Moreover,newlyemergingoceanandlandbiogeochemicalmeasurementsfacilitatethedevelopmentofnovelcarboncycleinitializationtechniques.

Itisbecomingstandardacrossseveralmajormodelingcenterstoincludethelandandoceancarboncyclecomponentsintotheirdecadalandseasonalpredictionsystems.Yet,thesedifferentpredictionsystemsuse various approaches regarding initialization, data assimilation, and spin up techniques.What are the implications of these different methodologies for the carbon cycle predictability?Furthermore,todatemulti-modelassessmentsarenotyetavailable,butfirst isightsfromdifferentprediction systems suggest consistent features of the carbon cycle predictability. What are thesourcesandtimescalesofpredictabilityofthecarboncycle,aswellasoftheotherbiogeochemicalvariablesindifferentpredictionsystems?ThesequestionswillbeaddressedinthecontextofcarboncyclepredictionsavailablefromdifferentESM-basedpredictionsystems..


(B6-03)

Seasonaltomulti-annualmarinebiogeochemicalpredictionusingGFDL’sEarthSystemModel

Park,Jong-Yeon(1,2),Stock,CharlesA.(2),Dunne,JohnP.(2),Yang,Xiaosong(2),Rosati,Anthony(2),John,Jasmin(2),Zhang,Shaoqing(3)

PrincetonUniversity(1),NOAA/GFDL(2),PhysicalOceanographyLaboratory/CIMST(3)

Whilephysicaloceanpredictionsystemsroutinelyassimilateobservationsandproduceseasonaltodecadalforecasts,oceanbiogeochemical(BGC)predictionsystemsarelessmatureduetoadditionalchallenges.A first impediment is thehighBGC sensitivity to transientmomentum imbalances thatariseduringphysicaldataassimilation.Inthisstudy,wedevelopastrategytorobustlyintegratetheGFDL’soceanBGCmodelwiththeensemblecoupled-climatedataassimilation(ECDA)systemusedfor GFDL’s seasonal to decadal global climate predictions. The ocean and atmosphere dataconstraints in the assimilation system are optimally modified to reduce BGC biases caused bymomentum imbalances while retaining the information of observed physical states. We thenperformedretrospectivepredictionrunsbyinitializingthemodelwiththeoutputfromourECDAruncoupledwithBGCmodelandinvestigatedseasonaltomulti-annualpredictionskillsofBGCvariablesover 1991 to 2016.We found that our earth system prediction system can provide skillful globalmarine biogeochemistry predictions about one year in advance in many ocean basins althoughforecastskillvariedbyregionandinitializationmonth.Wefurtherinvestigatedpotentialutilityofourearthsystempredictionsystemformarineresourcemanagementandfoundthatreportedtemporalvariability of annual fish catch in the Large Marine Ecosystem around the United States is wellexplainedbypredictedfishcatchestimatedfrompredictedBGCvariables.


(B6-04)

Achangeintheforecast:Oceanbiogeochemistryoverthenextdecade

Lovenduski,Nicole(1),Yeager,Stephen(2),Lindsay,Keith(2),Long,Matthew(2)

UniversityofColoradoBoulder,USA(1),NCAR,USA(2)

Observationscollectedoverthepastfewdecadesandprojectionsfromclimatemodelssuggestthatanthropogenic climate change has influenced and will continue to alter ocean biogeochemistry.Superimposed on the background of long-term changes in ocean biogeochemistry is naturalvariability in the physical and biogeochemical system that manifests on timescales of years todecades.Whiletherehasbeenconsiderableresearchinthefieldofdecadalclimatepredictionofthephysical oceanographic system, decadal predictability of ocean biogeochemistry remains relativelyunexplored.

Here,weexplore thepredictabilityandpredictive skillofoceanbiogeochemistrygeneratedby theCommunity Earth SystemModel Decadal Prediction System (CESM-DP),with a particular focus onoceancarbonuptakeandmarineplanktonproductivity.Ourpromisingfirstresultsindicateregionalpredictabilityinthesequantitiesoninterannualtodecadaltimescales.StandardanomalycorrelationanalysisofCESM-DPretrospectiveforecastsrevealssignificantpredictabilityinoceancarbonuptakeintheCaliforniaCurrent,easternsubtropicalsouthPacific,andNorthAtlanticbasinsonforecastleadtimes of 1-6 years.Mean square skill score statistics suggests that model initialization engenderspredictabilityofcarbonuptakeintheseregions.Formarineproductivity,theAtlanticbasin,westernboundary current regions, and eastern boundary upwelling systems have particularly long-lastingpredictiveskillthatderivesfrominitialization.Ourfindingssuggestpotentiallyimportantimplicationsforthedevelopmentofcarbondioxideemissionandfisheriesmanagementstrategies.


(B6-05)

Predictingoceanoxygen:capabilitiesandpotential

Long,Matthew;Yeager,Stephen

NCAR

Advanced prediction of ocean biogeochemistry fields has the potential to provide actionableinformationformarineconservationandfisheriesmanagers.Wepresentresultsfromrecentdecadalprediction experiments conducted with the Community Earth System Model (CESM)—the firstinstanceofsuchexperimentswiththismodelthatincludeoceanbiogeochemistry.Wedemonstratethatadvancedpredictionatmulti-annualforecastleadtimesispossibleforkeyquantitiessuchasnetprimaryproductivityandinteriordissolvedoxygenconcentrations.Inthispresentation,wefocusondissolved oxygen predictability and discuss the mechanisms enabling advanced prediction ofanomalies.Theresultssuggestthatdissolvedoxygenconcentrationsinthermoclinedepthranges,forinstance,havesubstantialpredictabilitywith forecast leadtimesofseveralyearsovermuchof theglobal ocean. Comparisons with uninitialized forecasts demonstrate that predictability can beattributed to initialization in many regions. Prediction is enabled in some cases by initializinganomalies within mode and intermediate water formations; forecast skill arises as the modeltransports anomalies in the interior.Weconsiderpredictability inbothaperfectmodel senseandwithdirect comparison toavailableobservations. Inaddition todemonstratingexistingcapabilitieswithCESM,wehighlightkeyissueswiththeoxygensimulationthatarecommontocoarseresolutionmodelsandspeculateondefiningrequirementsforsuccessfuloxygenpredictionmoregenerally.


(B6-06)

ApplicationofEarthsystemmodelingtoolstoexplorepredictabilityofmarineecosystemstressors

Rodgers,Keith(1),Fröhlicher,T(2),andTimmermann,Axel(1)

IBSCenterforClimatePhysics,SouthKorea(1),andUniversityBern,Switzerland

Marine ecosystems are increasingly stressed by human-induced changes. The ocean is gettingwarmer,moreacidic,anditsoxygencontentisdeclining.Thesestressorsmaycauselarge-scaleshiftsin marine biodiversity, with consequences for potential fisheries catches, livelihood, and foodsecurity.Inadditiontolong-termtrendssustainedthroughanthropogenicperturbationstotheEarthsystem, potential marine ecosystem stressors are known to exhibit variations over wide range ofspatial and temporal scales associatedwith natural variability in the climate system. For resourcemanagementofmarineecosystems,itwouldbeofgreatvalueiffluctuationsofstressorsweretobepredictablewithmodelingresources.

HereweexplorethepredictabilityofoceanecosystemstressorswithacomprehensiveglobalEarthsystemmodelthathasbeenextensivelyevaluatedagainstobservationalconstraints,namelyGFDL’sESM2M.Potentialpredictability isexploredthroughtheapplicationof infinitesimal initialconditionperturbations to a preindustrial control simulation with ESM2M in large ensemble (40 members)mode,using6differentrandomlychoseninitialconditionsfromthepre-industrialrun.Thusatotalof240 separate simulations have been performed, each of these for a duration of 10 years. Thevariablesconsideredareseasurfacetemperature,seasurfacepH,subsurfaceoxygenconcentrations,and net primary productivity. In addition to identifying the upper limits of predictability of thesevariableswithintheEarthsystemmodel,itisalsoourobjectivetoidentifytherelativepredictabilityof the four variables under consideration. Given the potential importance of acidification as astressor, the analysis will include the broader range of carbon dioxide-related variables for thesurfaceocean.


(B6-07)

RegionalArcticsea-iceprediction:Potentialversusoperationalseasonalforecastskill

MitchBushuk(1,2),RymMsadek(3),MichaelWinton(1),GabrielVecchi(4),XiaosongYang(1,2),AnthonyRosati(1,2),RichGudgel(1)

GFDL,USA(1),UCAR,USA(2),CNRS/CERFACS,France(3),PrincetonUniversity,USA(4)

SeasonalpredictionsofArcticseaiceonregionalspatialscalesareapressingneedforabroadgroupofstakeholders,however,mostassessmentsofpredictabilityandforecastskilltodatehavefocusedon pan-Arctic sea-ice extent (SIE). In this work, we present the first direct comparison of perfectmodel (PM)andoperational (OP) seasonalprediction skill for regionalArctic SIEwithina commondynamical prediction system. This assessment is based on two complementary suites of seasonalpredictionensembleexperimentsperformedwithaglobalcoupledclimatemodel.First,wepresentasuiteofPMpredictabilityexperimentswithstartdatesspanningthecalendaryear,whichareusedtoquantifythepotentialregionalSIEpredictionskillofthissystem.Second,weassessthesystem'sOPpredictionskillfordetrendedregionalSIEusingasuiteofretrospectiveinitializedseasonalforecastsspanning1981-2016.InnearlyallArcticregionsandforalltargetmonths,wefindasubstantialskillgapbetweenPMandOPpredictionsofregionalSIE.ThePMexperimentsrevealthatregionalwinterSIE ispotentiallypredictableat lead timesbeyond12months, substantially longer than theskilloftheirOP counterparts. Both theOPandPMpredictionsdisplay a springprediction skill barrier forregional summer SIE forecasts, indicating a fundamental predictability limit for summer regionalpredictions.Wefindthatasimilarbarrierexistsforpan-Arcticsea-icevolumepredictions,butisnotpresent forpredictionsofpan-Arctic SIE. The skill gap identified in thiswork indicatesapromisingpotentialforfutureimprovementsinregionalSIEpredictions.


(B6-08)

SkillfulseasonalforecastsofArcticseaiceretreatandadvancedatesinadynamicalforecastsystem

Sigmond,Michael(1),Reader,Cathy(1),Flato,Gregory(1),Merryfield,William(1),Tivy,Adrienne(2)

CCCma,ECCC,Canada(1),CIS,ECCC,Canada(2)

TheneedforskillfulseasonalforecastsofArcticseaiceisrapidlyincreasing.Technologytoperformsuchforecastswithcoupledatmosphere-ocean-seaicesystemshasonlyrecentlybecomeavailable,withpreviousskillevaluationsmainlylimitedtoarea-integratedquantities.Hereweshow,basedona large set of retrospective ensemblemodel forecasts, that a dynamical forecast systemproducesskillfulseasonalforecastsoflocaldatesatwhichiceforms(theadvancedate)andmelts(theretreatdate)-variablesthatareofgreatinteresttoawiderangeofend-users.Advancedatescangenerallybeskillfullypredictedat longer leadtimes(~5monthsonaverage)thanretreatdates (~3months).Theskillofretreatdateforecastsmainlystemsfrompersistenceofinitialseaiceanomalies,whereasadvance date forecasts benefit from longer timescale and more predictable variability in oceantemperatures. Initial steps taken and challenges encountered in translating these results intooperational products will be described. These results suggest that further investments in thedevelopment of dynamical seasonal forecast systems may result in significant socio-economicbenefits.


(B6-09)

ENSOpredictionusinganearthsystemmodelincorporatingahigh-resolutiontropicaloceannestingmodel

YukikoImada(1),GoroYamanaka(1),HiroyukiTsujino(1),ShogoUrakawa(1),YasushiTakatsuki(1)

(1)MRI,JMA,Japan

Mesoscale eddies in the tropical oceans have significant impacts on the oceanic mean states,atmospheric circulation, ENSO characteristics, and other natural variabilities. Here, we found asignificantimprovementofENSOpredictionskillbyincorporatingahigh-resolution(eddy-permitted)tropicaloceannestingmodelintoaseasonalpredictionsystembasedonanearthsystemmodelMRI-ESM1.Becauseoftherealisticrepresentationoftropicalinstabilitywaves(TIWs),thesimulatededdyheat flux improved not only tropical oceanic mean states but also spatial distributions of meansurface wind stress and precipitation in the nested version of MRI-ESM1. ENSO characteristics(amplitude,period,spatialpattern,teleconnection)werealsomodifiedthroughthechangesofmeanstate.NonlineareddyheattransportsduetoTIWsincreasedtheENSOskewnessandimprovedthecharacteristics of ENSO asymmetricity. These improvements resulted in more accurate ENSOpredictionseveralmonthsahead.

SESSION:(C1)Initializationinitializationshockandmodelerror(includesdataassimilation)

(C1-01)

Climate-modelinitializationfornear-termclimate-prediction:asurveyofrecentadvancesandanticipatedtrends

Karspeck,Alicia

JupiterIntelligence,SKAnalytics,NationalCenterforAtmosphericResearch

Over15groupsparticipatedinthedecadalpredictionexperimentalprotocoldesignedunderCMIP5.Forsomeofthesegroups,thismarkedthefirstforayintoclimate-modelinitializationforthepurposeofnear-termprediction.Forothergroups,theexperimentscalledforintheCMIP5DPprotocolwereachieved thought the extension of initialization procedures already in place to support eitherseasonalpredicitonactivitiesorexistingresearch-focusedlong-termprediction.Inthisoverviewtalk,wesurveythemethodsthathavebeenusedinthelastdecadetomovethescienceandskillofneartherm climate prediction forward. As production of CMIP6 commences, we also consider howinitialization procedures have evolved (andwhy) anddiscuss anticipated trends, including coupledassimilation.


(C1-02)

Non-linearandnon-stationaryforecasterrors:shouldwerevisitthecurrentforecaststrategies?

MagdalenaA.Balmaseda,LauraFerranti,StephanieJohnson,MichaelMayer,ChristopherRoberts,TimStockdale,SteffenTietsche,FredericVitart,HaoZuo.

ECMWF

Currentapproachesforsub-seasonalandseasonalforecastuseasetof initializedcoupledhindcasttobiascorrecttheforecastproducts.Thisapproachimplicitlyassumesstationarityofforecasterrorsandneglectsnon-linearinteractionbetweenmeanstateandvariability.Herewereviewanumberofcaseswheretheseassumptionsdonothold,resultingindegradationofforecastskill.Thelimitationsof the a-posteriori linear bias approach in ENSO have discussed in previous studies, and aresummarizedhere.WealsodiscussresultsfromthelatestECMWFseasonalforecastingsystem,whichexhibitsa cleardecadalmodulationof the forecasterrorsassociatedwithdifferent regimesof theAtlantic Meridional Circulation. These pathological cases raise the question of whether betterforecasts could be obtained by modifying the current forecast or initialization strategies. Aframeworkforanalternativeforecaststrategy,wherethemodelerrorsareexplicitlyaccountedforduring themodel integration via empiricalmodelling is presented. This frameworkwill exploit theinformationfromcoupleddataassimilation incrementstotraintheerrormodel.Thestrengthsandweaknessofsuchanapproachwillbeoutlined.


(C1-03)

Coupleddataassimilationandensembleinitializationwithapplicationtomulti-yearENSOprediction

O'Kane,Terence(1),Sandery,Paul(1),Monselesan,Didier(1),Chamberlain,Matt(1),Sakov,Pavel(2),Matear,Richard(1)

CSIROOceansandAtmosphere(1),AustralianBureauofMeteorology(2)

We develop and compare variants of coupled data assimilation (DA) systems based on ensembleoptimalinterpolation(EnOI)andensembletransformKalmanfilter(ETKF)methods.Theassimilationsystem is first tested on a simplemultiscale paradigmmodel of the coupled tropical-extratropicalclimatesystem,thenimplementedforacoupledgeneralcirculationmodel(GCM).StronglycoupledDAwasemployedtoperformanobservingsystemsimulationexperiment(OSSE)andtoassesstheimpact of assimilating ocean observations on the atmospheric state analysis update via the cross-domain error covariances from the coupled-model background ensemble. We examine therelationship between ensemble spread, analysis increments and forecast skill in multi-year ENSOprediction experiments with a particular focus on the atmospheric response to tropical oceanperturbations. Various approaches to generating flow dependent initial forecast perturbations,either in terms of ETKF and bred vectors (BV) whose norms are based on 3-D inband variancestructures(withinisosurfaces).

Our focus is on seasonal and longer timescales, and in particular ENSO. Therefore, our premiseunderpinningtheOSSEsisthatpredictabilityprimarilyresidesintheoceansandthefastatmosphereacts as a stochastic driver on the longer timescale ocean variability.While the ETKF analysis wasgenerally less biased and with lower errors, both systems performed comparably in the tropics.Outside the tropics theETKF systemproduceddramatically lower forecastbias and forecastmeanabsolute deviation error than the EnOI system. The reason for the low analysis error in the EnOIsystem in the tropicswas found to be a result of larger ensemble spread in the EnOI backgroundcovariances,relativetotheETKF,systematicallyweightingobservationsmore.

BVs representative of growing coupled tropical instabilities were found to modify tropicalconvection,inparticularlyintheregionofthemaritimecontinent,whichinturngenerateacoherentmodulation of the Hadley circulation. A direct renormalization of thermocline disturbances wasfoundtobemosteffectiveincommunicatinginformationfromthetropicaloceantothemidlatitudeatmosphereontimescalesofacoupleofweekstoamonth.Comparisonofensembleforecastsbasedon bred perturbations centred about the EnOI analysed state reveal a substantial reduction inuncertainties,wheredisturbancesnotdirectlyassociatedwiththermoclinevariabilityaremasked.Inparticular, excluding SSTdisturbances led toa significant reduction in forecast errors inmulti-yearENSOpredictions.EnsembleforecastsinitialisedwithETKFperturbationswereverymuchlessskilful,evenafterbiascorrectionwasapplied.

TheOSSEsandmethodsdiscussedformasystematicbasisforcoupledDArelevanttomulti-yearnearterm climate forecasts. The masked isosurface BV approach allows for the specific targeting ofregions of large scale variability pertinent to dynamical processes that determinepredictability onseasonaltointerannualspatio-temporalscales.Beyondaseason,stronglycoupleddataassimilation,where the slow oceanmodes are explicitly constrained including projection onto the backgroundatmospheric states (i.e. jets, cells etc)while leaving the fast atmospheric dynamics free, includingtargeted forecastperturbations, offers a systematic approach todetermining themechanismsandpredictabilityofthekeyclimatemodes.


(C1-04)

Sub-seasonaltoDecadalPredictabilityandPredictionwithanOceanEddyResolvingGlobalCoupledModel

BenKirtman

UniversityofMiami,USA

Increasingly, high resolution observations and coupled model experiments with eddy-resolvingoceansindicatethatwesternboundarycurrentsareregionsofstrongocean-atmosphereinteractionsthatarecriticalcomponentsoftheclimaticmeanstateandvariability.ThehighSSTsandstrongSSTgradientscouplewiththeatmospheretopumpmoistureintothemarineboundarylayer,acceleratewinds,sharpenSSTfronts,andintroducesignificantsub-seasonaltodecadalclimatevariabilitythataffectsthefrequencyandintensityofextremeevents(e.g.,heatwaves,coldspells,droughts,floods,extremewinds) at remote locations. These extreme events are embeddedwithin sub-seasonal todecadalvariabilitythatmaybepredictedbyglobalmodels.Thistalkdocumentsa30-yearsetofsub-seasonal to seasonal retrospective forecasts with an ocean eddy-resolving coupled model incomparison with a parallel set of retrospective forecasts with an ocean eddy-permitting coupledmodelusedfortheNorthAmericanMulti-ModelEnsemble(NMME)project.Theforecastfrombothsystems use identical initial conditions derived from CFSR and use the same modeling systems(CCSM4) except for resolution. On the decadal time-scale, the talk emphasizes identical twinpredictabilityexperimentswiththeoceaneddy-resolvingmodel.Resultsincludebothsetsofanalysisfocusonhowtheresolvedoceaneddiesaffectpredictionskillandpredictability,particularemphasisisplaceonextremeevents.

SESSION:(C2)Researchtooperation(includesseamlessprediction)

(C2-01)

Fromreliableinitialisedforecaststoskilfulclimateprojection:adynamicalsystemsapproach

Christensen,Hannah(1,2),Berner,Judith(1)

NCAR,USA(1),UniversityofOxford,UK(2)

Whilemodels for initialised forecasts canbe rigorously testedbyperformingandevaluatingmanyhindcasts, the limitedobservational record restricts thedegree towhich climatemodels and theirprojectionscanbeevaluated.Thereforeakeyquestionofinterestis:towhatdegreecanweevaluatethepotential skill of a climatemodel's projectionsby evaluating short-range initial value forecastsproducedbythesamemodel?Weaddressthisquestionusingadynamicalsystemsframework.Wederive themean climate response of a general dynamical system to a small external forcing, andrelatethis responsetothereliabilityof initialvalue forecasts.Wefindthat inordertocapturethemean climatic response, the forecast model must correctly represent the slowest modes in thesystem.Reliable forecastsonseasonaland longertimescalescouldthereforebe indicativethattheclimatemodelof interestwill respondcorrectly toanappliedanthropogenic forcing.This indicatesthe potential of using the 'seamless prediction' framework in evaluating climatemodels.We alsohighlight some important caveats: anunreliable seasonal forecastdoesnotnecessarily indicateanincorrectclimateprojection,ascorrectrepresentationoffastprocessesisalsonecessaryforreliableseasonalforecasts.


(C2-02)

Movingfromconcepttoclimateservice;howwecanmeettheneedsofaclimatesmartcommunity

AndrewB.Watkins,DavidJones,CatherineGanter,DavidWalland

BureauofMeteorology,MelbourneVictoria,Australia

While many of us know how to make a climate outlook, and what products our research couldultimatelyproduce,thequestionofwhyweproduceitissometimesnotfrontofmind.The"why"isultimately to help decision makers in weather and climate sensitive sectors - from farmers toinsurancepricerstogovernments–makebetterchoices.This iswhereresearchbecomesaclimateservice.Theultimategoalofaclimateserviceistotakesometimes-complexclimatescienceandrawmodel output, and convert that combination into true climate intelligence that can be easilyinterpreted by a non-specialist. But a climate service is not (just) a fancy web page or dazzlinggraphics. It involves supporting reliable, on-time, clearly communicated climate information thatmeets a clear and demonstrable user need. It can involve partnerships between users and metservicesorscienceagencies.Wewillaimtotakeyoualongthejourneyfromthegermofanideatoitsresearchphaseandthenthefinalclimateservice.Ultimatelyitshouldbecomeclearerhowpeoplecanmaketheirresearch'climateserviceready.


(C2-03)

USNavy’sEarthSystemPredictionCapabilityEffort

ReynoldsC,BartonN,BishopC,FlatauM,FrolovS,JanigaM,McLayJ,RidoutJ,RustonB,WhitcombT(1),EleuterioD(2),HoganP,JacobsG,MetzgerEJ,RogersE,RowleyC(3),RichmanJ(4)

NRLMarineMeteorology,USA(1),OfficeofNavalResearch,USA(2),NRLOceanography,USA(3),FloridaStateUniversity,USA(4)

TheNationalEarthSystemPredictionCapability(NationalESPC) isaU.S.multi-agencycollaborativeeffort to leverage resources to develop the next generation environmental forecasting system. Aspartof thiseffort,U.S.Navy isdevelopinga fullycoupledglobalsystem includingtheNavyGlobalEnvironmental Model (NAVGEM), the HYbrid Coordinate Ocean Model (HYCOM), the Los AlamosCommunityIceCodE(CICE),andtheWavewatchIIIoceansurfacewavemodel.Thissystemisbeingdeveloped to meet Navy needs for high-resolution global environmental forecasts on timescalesfrom days tomonths. The design and implementation of the coupled architecture uses the EarthSystem Modeling Framework (ESMF) with the National Unified Operational Prediction Capability(NUOPC) standard in order to maximize flexibility in adopting future models. Initial operationalcapability isplannedfor2019andwill includedailyhigh-resolutiondeterministicforecasts(with19kmatmospheric resolution, 1/25oocean and sea ice resolution, and 1/8owavemodel resolution)andweekly extended-range ensemble forecasts (with 37 km atmospheric resolution, 1/12o oceanandseaiceresolution,and1/4owavemodelresolution).Oneaspectthatmakesthesystemuniqueis the relatively high resolution of the ocean and ice models, reflecting the Navy’s strategic andtactical interests in theserealms.A17-yeararchiveof45-day forecasts four timesperweekat theensembleresolutionhasbeenproducedaspartoftheNavy’sparticipationintheNOAASubseasonaleXperiment (SubX) project. The performance of the system in simulating the Madden-JulianOscillationandothertropicalphenomenaiscomparabletootherstate-of-the-artsystems.TheNavysystemhasalsoshowngoodperformancerelativetoothersystemsinmulti-monthforecastsofarcticsea ice as part of the Sea Ice PredictionNetwork September Sea IceOutlook. Plans for ensembledesign,coupleddataassimilation,andotherfuturedevelopmentswillalsobepresented.


(C2-04)

TRANSFERRINGSCIENCETOPRACTICE:NEARLYTWODECADESOFSEASONALFORECASTINGFORWEATHER-SENSITIVEINDUSTRIES

Crawford,Todd

TheWeatherCompany,anIBMBusiness

In2000,aseasonalforecastingprogramwasstartedatwhatwasthenWSICorporation,insupportofanewbusinesslineaimedattherecently-deregulatedenergytradingindustry.A$500KinvestmentwasmadeinanAlphacomputingcluster,whichrana16-memberensembleoftheNCARCCM3.6.6climatemodel.Fromthere,variousstatisticalforecastingalgorithmsweredevelopedtosupplementtheclimatemodelforecasts.Overtime,theCCMwasabandonedinfavorofthenewergenerationofclimatemodels,includingtheECMWF,CFS,andmorerecentlythecalibratedNMMEsuite,insupportoftwice-monthlyseasonalforecastsforNorthAmerica,Europe,andAsia.Further,statisticalmodelshavebeenrefinedbasedonthelatestscienceandourexperience.

More recently, aswehavebeen acquiredby IBM,wehave addedpurely automated and griddedseasonalforecastsfortherestoftheworld,usingablendofNMMEandECMWFoutputs.Finally,weare currently working towards our first formal attempt at probabilistic seasonal output since ourinitialfailedattemptin2000.

Alongtheway,therehavebeennumeroussuccesses,failures,andlessonslearned,asthetrickytaskof translating seasonal weather information into a language and format easily digestible for not-always-rational energy traders has been a challenging experience. This talk will provide apractitioner’sviewofseasonalforecastingandsomesuggestionsforwhat’sahead.

SESSION:(C3)Timescaleinteraction(includesteleconnections)

(C3-01)

MULTI-SCALEINTERACTIONSINAHIGH-RESOLUTIONTROPICAL-BELTEXPERIMENTUSINGWRFMODEL

Tieh-YongKOH(1),RicardoFONSECA(2),Chee-KiatTEO(1)

SingaporeUniversityofSocialSciences,Singapore(1),LuleaUniversityofTechnology,Kiruna,Sweden(2)

TheWeatherResearchandForecasting (WRF)model isused todynamicallydownscale27yearsofthe Climate Forecast SystemReanalysis (CFSR) in a tropical belt configuration at 36 kmhorizontalgrid spacing.WRF is found to give a good rainfall climatology as observedby the Tropical RainfallMeasuringMission (TRMM)and to reproducewell the large-scalecirculationandsurface radiationfluxes. The impact of conventional andModoki-type El Niño-Southern Oscillation (ENSO) and theIndianOceanDipole(IOD)areconfirmedbylinearregressioninthemodel.Madden-JulianOscillation(MJO)andBorealSummerIntra-seasonalOscillation(BSISO)arealsowell-simulated.However,WRFdoes not capture well the diurnal cycle of precipitation over the Maritime Continent. For theinvestigationofmulti-scaleinteractionsthroughthelocaldiurnalcycle,TRMMdataisusedinstead.

The WRF simulation shows that in the boreal summer, conventional ENSO modifies the MJOamplitudewhileModoki-type ENSO and IOD impacts areMJO-phase dependent; in borealwinter,inter-annualvariationshave little impactonMJOamplitude.TheTRMMobservationsshowthat intheMaritimeContinent,moderateENSOmodifiestheMJO’sinfluenceonthediurnalcycleinspecificways;strongENSOleadstonon-linearimpactsonthediurnalcycle.


(C3-02)

NorthAmericawintercirculationregimechangeandimplicationsonS2S/S2D

SimonWang

UtahStateUniversity,USA

The2018NewYearextremecoldnessintheeasternU.S.createdhazards,whilewesternstateslikeCalifornia and CO underwent a wild swing from drought to flooding and to fires. The heat/coldcontrastor“dipole”maybeanewnormal.WeshowprogressinunderstandingthedipolevariationinthedecadaltimescaleandtheeffectoftropicalandArcticforcingsattheS2Sspectrum.


(C3-03)

Predictabilityofblockingandtropicalcycloneactivities?--Anassessmentwithalargeensemblesimulation--

Kimoto,Masahide(1),Chiba,Joutaro(2)

(1)AtmosphereandOceanResearchInstitute,TheUniversityofTokyo(2)JapanMeteorologicalAgency

Potentialseasonalpredictabilityisinvestigatedbyalargeensembleexperimentwithanatmosphericgeneral circulationmodel (AGCM)with prescribed sea surface temperature (SST) and sea ice (thedatabase forpolicydecisionmaking for futureclimatechange;d4PDF;Mizutaetal.2017).Weusethepresent-daycomponentofthedatasetwhichconsistsof100ensemblemembersforobserved1961-2010 boundary conditions. The large ensemble enables us to explore potential seasonalpredictabilityofnotonlyseasonalaverages,butalsomodulationofday-to-dayweatherdisturbances.Herewetakeupmidlatitudeblockingandtropicalcycloneactivitiestoseeifthereisusefulseasonalpredictabilityonthemeasuresoftheiractivities.SignificantforcedsignalsarefoundforNorthPacificblockingandnorthwestern tropicalcycloneactivities in thedataset,andtheydoshowcorrelationwithobservations.


(C3-04)

RelatingwinterNAOskilltojetvariabilityacrosstimescales

Woollings,Tim(1),Parker,Tess(1),Baker,Laura(2),Weisheimer,Antje(1),Shaffrey,Len(2),O'Reilly,Chris(1)

UniversityofOxford,UK(1),NCASandUniversityofReading,UK(2)

There has been encouraging progress in recent years in the skill of seasonal forecast models topredictthewinterNorthAtlanticOscillation(NAO).TheNAOreflectsacombinationofvariabilityinthestrengthandpositionoftheNorthAtlanticeddy-drivenjetstream.HereweinvestigatetheNAOskill in operational hindcasts and a long, atmosphere-only re-forecast of the ECMWF system. Thisshowsthattheskillinthesesystemslargelyarisesfrompredictingtheposition,ratherthanspeedofthejet,ontheinterannualtimescale.

Asthetimescale lengthenstodecadal,thevariability in jetspeedmakesupa largerfractionoftheobservedNAOvariability.Hence,thepredictableNAOsignalsonthistimescalewillhaveadifferentbalanceofmechanismsthantheseasonaltimescale,sothatseasonalskill inaforecastsystemmaynotnecessarilytranslateintodecadalskill.

Inaddition,weshowthattheslowdecadalvariationsinjetspeedacttomodulatetheamountofjetlatitudevariabilityonsubseasonaltoseasonaltimescales.Hence,thepotentiallypredictablesignalsontheS2Stimescalesmightbemodulatedbythedecadalvariability.


(C3-05)

SeasonalForecastsofthe20thCentury:Multi-DecadalVariabilityinPredictiveSkilloftheWinterNAO

AntjeWeisheimer(1,2),NathalieSchaller(2,3),ChrisO’Reilly(2),DavidMacLeod(2)andTimPalmer(2)

ECMWF,Reading,UK(1),UniversityofOxford,UK(2),Cicero,InternationalCentreforClimateResearch,Oslo,Norway(3)

Basedonskillestimates fromhindcastsmadeover the lastcoupleofdecades, recentstudieshavesuggested that considerableprogresshasbeenmade in forecastingwinter climateanomaliesoverthe Euro-Atlantic area using current-generation forecast models. However, previous-generationmodelshadalreadyshownthatforecastsofwinterclimateanomaliesinthe1960sand1970swerelesssuccessfulthanforecastsofthe1980sand1990s.Giventhatthemorerecentdecadeshavebeendominated by theNAO in its positive phase, it is important to knowwhether the performance ofcurrentmodelswouldbesimilarlyskilfulwhentestedoverperiodsofapredominantlynegativeNAO.

To this end, newensembles of retrospective seasonal forecasts covering the period 1900 to 2009have been created with uncoupled and coupled versions of the ECMWFmodel, providing uniquetools to exploremanyaspectsof seasonal climateprediction. In this studywe focuson themulti-decadalvariabilityinpredictingthewinterNAO.Theexistenceofrelativelylowskilllevelsduringtheperiod1950s-1970shasbeenconfirmedinthenewdataset.Thisskillappearstoincreaseagainforearlier and later periods. Whilst these interdecadal differences in skill are, by themselves, onlymarginallystatisticallysignificant,thevariationsinskillstronglyco-varywithstatisticsofthegeneralcirculationitselfsuggestingthatsuchdifferencesareindeedphysicallyreal.Themid-Centuryperiodof low forecast skill coincides with a negative NAO phase but the relationship between the NAOphase/amplitudeandforecastskillismorecomplexthanlinear.


(C3-06)

Theroleoftropical-extratropicalinteractionsontheoptimalgrowthofMadden-JulianOscillationevents

Henderson,Stephanie(1),Vimont,Dan(2)

CenterforClimaticResearch,UniversityofWisconsin-Madison(1,2)

The influence of tropical-extratropical interactions on the Madden-Julian Oscillation (MJO) isinvestigated using linear inverse modeling (LIM). Extratropical circulation and tropical heatingoptimal initial conditions forMJOeventsare identified forallMJOphases.Theextratropical initialconditionforsomeMJOphasesincludesadipolecirculationanomalyovertheAtlanticbasin,aswellascirculationanomaliesnearthePacificsubtropical jet likelyassociatedwithpreviousMJOphases.The impact of tropical-extratropical interactions on MJO growth is examined using a LIM thatremoves these interactions. Results suggest that removing the influence of the extratropicalcirculationreducestheamplitudeofMJOheatinganddecreasesMJOvariance.


(C3-07)

ImpactofintraseasonaloscillationsononsetanddemiseoftheIndiansummermonsoonrainfall.

Karmakar,Nirupam(1),Misra,Vasubandhu(1,2,3)

FloridaStateUniversity(1),FloridaClimateInstitute(2),CenterforOcean-AtmosphericPredictionStudies,Tallahassee(3)

TwoofthemostimportanthydroclimaticfeaturesoftheIndiansummermonsoonrainfall(ISMR)areitsonset/demiseandintraseasonalvariabilitymanifestedbytheactive-breakcycles.Inthisstudy,weaim to understand the quantitative association between these two phenomena. An objectivedefinitionoflocalonsetanddemiseofISMRbasedonmorethanacenturylongIndiaMeteorologicalDepartment(IMD)rain-gaugeobservationistakenintoconsiderationinthisstudy.Onsetanddemisearedefinedoverthenativegrids(0.25degx0.25deg)oftheIMDdata.Thisdefinitionofonsetanddemise avoids false onset and advantageous as it has a close correspondence with the seasonalrainfallanomaly.

Using a non-parametric spectral approach, we identify intraseasonal oscillatory (ISO) modes inrainfallover Indiaandextractedtwoclassesofvariability: low-andhigh-frequency ISO(LF-ISOandHF-ISO).Theyexhibitperiodicitiesof25–60-daysand10–20-days,respectively.RainfalloverIndiaishighlymodulatedbythesetwomodesandactivespellwithcopiousrainfallatalocationismarkedbyconcurrent peaks of these two oscillations. LF-ISO is characterized by northwardmarch of rainfallbands from the southern-most part of India to the foothills of the Himalayas. HF-ISO showscomparativelymorecomplexandsmall-scalestructures,withanorth-westwardpropagationalongcentralIndia.

UsingthedefinitionoflocalonsetanddemiseandLF-ISOandHF-ISOcharacterizedateachlocation,weaimtounderstandtheassociationbetweenthem.Inotherwords,howISOmodesmodulatetheonsetanddemiseateachlocation.WecalculatethephasesoftheISOmodesbasedonitsperiodicnatureanddeterminedthephasesofISOduringonsetanddemiseoccurredateachgridpoint.Itisobserved that theprobabilityofoccurrenceof localonset is remarkablyhighwhenLF-andHF-ISOexhibitfavorableconditionsoverthatlocation.Onsetdatesaremostlymarkedbytheascendinglimbof the positive part of the ISO cycle,with LF-ISO showing strongermodulation. Similar results areseen for local demise,which ismostly seenover the descending limb. The results presented hereestablish an important aspect of the predictability of monsoon onset and demise. Associationbetweenthetwophenomenaraisesthehopeforthepredictabilityoflocalonsetanddemise,astheycouldberelatedtoslowlyevolvinglarge-scalecirculation.Thiscouldbeanimportantavenueintheproblemofforecastingonset/demiseusingnumericalsimulations.

SESSION:Plenary

(Pl-05)

SubseasonaltoSeasonalScienceandPredictionsInitiativesoftheNOAAMAPPProgram

AnnaritaMariotti(1)andDanBarrie(1)

(1)NOAA,OAR/ClimateProgramOffice

Scientific communities have historically organized themselves around the weather and climateproblems, but the subseasonal to seasonal (S2S) time scale range is overall recognized as newterritory,forwhichaconcertedsharedeffortisneeded.Forinstance,theclimatecommunity,aspartof programs like CLIVAR, has historically tackled coupled phenomena and modeling, keys toharnessingpredictabilityon longer timescales. Incontrast, theweathercommunityhas focusedonsynopticdynamics,higher-resolutionmodeling,andenhancedinitialconditions,ofimportanceattheshorter timescales and especially for the prediction of extremes. The processes and phenomenaspecific to the intermediate S2S range require a unified approach to science, modeling, andpredictions. Internationally, the WWRP/WCRP S2S Prediction Project is a promising catalyzer forthese types of activities. Among the various contributing U.S. research programs, the NOAAModeling, Analysis, Predictions and Projections (MAPP) program, has put in place a set ofcoordinatedresearchactivitiestohelpaddressthesub-seasonal-seasonalpredictiongap.

This presentation will describe ongoing MAPP program S2S science and prediction activities,specificallytheMAPPS2SPredictionTaskForceandtheSubXpredictionexperiment.Anoverviewofresultsfromtheseresearchinitiativeswillbepresented.

SESSION:Plenary

(Pl-07)

WGSIP/DCPP–Projectachievementsandfutureplans

Merryfield,William,DougSmith

CCCma,Canada(1),INM/RAS,Russia(2),HydrometcentreofRussia(3),BarcelonaSupercomputingCenter,Spain(4),MRI/JMA,Japan(5)

SESSION:Plenary

(PL-08)

SubseasonaltoDecadalPrediction:Lookingbackon40yearsofprogress-andprojectingforwardanother40years

Palmer,Tim

UniversityofOxford

In this talk I want to reflect back on some of the key successes across the range of initialisedpredictionsonthesub-seasonaltodecadaltimescalesoverthe last40years.This includesthefirstensembleforecastsystems,thefirstdynamicalpredictionsofElNinoandthedevelopmentof linksbetweendecadalmodesofseatemperaturevariabilityandSahelDrought.However,overthese40years, predictive skill has been compromised by persistent model systematic error. Modelintercomparisonprogrammeshaveshownthatmodelbiasesareonlyslowlyreducingovertime.Forthe sake of society, wemust start accelerated programmes of research to eliminate them. Somesuggestionsonhowtodothiswillbeoutlined.

SESSION:Plenary

(PL-09)

ResearchNeedsforAdvancingOperationalS2DForecastingInfrastructure

KUMAR,Arun

ClimatePredictionCenter

Advances in understanding of predictability of climate variability on sub-seasonal to decadal timescale has led to the establishment of operational infrastructure in support of providing routinepredictions to interested stakeholders. This is highlighted by the designation of Global ProducingCenters (GPCs) for seasonal and annual to decadal predictions under the purview of the WorldMeteorologicalOrganization(WMO).TheinfrastructureofGPCsisfurtheraugmentedbyassociatedlead-centers that play the role of collecting, synthesizing and disseminating forecast informationfrom individualGPCs.Despite tremendousadvances in thedevelopmentofoperational forecastingonS2Dtime-scales,challengesremainthatcompromisetherealizationof inherentpredictabilityasprediction skill. Thesechallenges spanconsiderations suchas configurationofoperational forecastsystems, development and verification of forecast products, strategies for the communication offorecastinformation,etc.Inthistalk,researchneedstoaddresschallengesforadvancingoperationalS2Dforecastinginfrastructurewillbediscussed.

SESSIONS:(A7/A8)Stratosphere/Chemistry

(A7-01)

ImpactsofNASA’sEarthObservationsonSubseasonalandSeasonalForecasts

StevenPawson,AndreaMolodandEricHackert

GlobalModelingandAssimilationOffice,NASAGoddardSpaceFlightCenter

NASA’ssuiteofEarth-observingsatellitesprovidesauniqueviewofmanyprocessesonEarth,withrelevanceontimescalesrangingfromhourstoweeksandevenyears.NASA’sobservationsspanallpartsoftheEarthsystem:atmospheric,ocean, landandcryosphere,andincludephysical,chemicalandbiologicalcomponents.ThispresentationexploresuseofNASAobservationsinextended-rangeprediction, from many days to months, using the Goddard Earth Observing System (GEOS)assimilationandpredictivemodelingcapabilities.

Theskillofaweatherforecastislinkedtothefidelityoftheinitializationprocess(dataassimilation)andtherealisticrepresentationof“fast”physicalprocesses inthemodel. At longertimehorizons,the slower “feedback” processes in the Earth System begin to take amore prominent role in theaccuracy of the forecast. Simultaneously, forecast-skill attribution transitions away from feature-basedmetricsthatemphasizesmallerscales(e.g.,representationsoffrontsandvortices)tometricsthat emphasize the statistical distributions of large-scale features (e.g., ENSO diagnostics andteleconnections).

This presentationwill summarize studiesperformedusing theGEOS-S2S (subseasonal to seasonal)systemthatexploretheimpactsofNASAobservationsonthefidelityoftheforecasts.TheGEOS-S2Ssystem is configured for the atmosphere-ocean-land-ice model and is initialized using in-situ andspace-based observations, including atmospheric aerosols and ozone which are not typicallyanalyzed in such systems. The GEOS-S2S model routinely includes aerosol feedbacks, whichsystematically impact the realismof the forecasts, providinga first exampleofhow suitableNASAobservations impact the performance of the GEOS-S2S system. Studies in which a stratosphericchemistry module is activated in the GEOS-S2S system allow the impacts of ozone radiativefeedbacks to be isolated. Space-basedobservations of sea-surface temperature and altimetry areroutinely analyzed for the initialization of the GEOS-S2S system; recent advances allow the useofNASA’ssea-surfacesalinitydata,whichareshowntoimpactthelong-rangeskilloftheforecasts.


(A8-01)

Theroleofthestratosphereinsub-seasonaltoseasonalpredictability

Butler,A.(1),Charlton-Perez,A.(2),Domeisen,D.(3),Garfinkel,C.(4),Gerber,E.(5),Hitchcock,P.(6),Karpechko,A.(7),Maycock,A.(8),Sigmond,M.(9),Simpson,I.(10),Son,S.(11)

CIRES/NOAA(1)

Knowledge of the state of the stratosphere has the potential to enhance predictability of thetroposphere on sub-seasonal to seasonal timescales. Here, we provide a broad overview of ourcurrent understanding of how the stratosphere couples to the troposphere in the tropics andextratropics, via processes such as the Madden-Julian Oscillation and the El Niño-SouthernOscillation.Wediscuss progress to date in trying to harness stratosphere-troposphere coupling toenhance predictability on the sub-seasonal to seasonal (S2S) timescale, a key focus of theWCRP/SPARCStratosphericNetworkfortheAssessmentofPredictability(SNAP)project.Finally,weexamineopenquestionsandprovidesomesuggestionsonwhereandhowimprovedunderstandingandsimulationofstratosphere-tropospherecouplingismostlikelytoleadtoimprovedskill.


(A8-02)

EffectofSuddenStratosphericWarmingsonSubseasonalPredictionSkillintheNASAS2SForecastSystem

M.JoanAlexander(1),LawrenceCoy(2),LauraHolt(1),ZhaoLi(3),AndreaMolod(3),andStevenPawson(3)

(1)NWRA,(2)USRA/NASA,(3)NASA/GMAO

Sudden stratospheric warmings (SSW) represent a major mode of subseasonal variability in thewinteratmosphere,representingapotentialfor“ForecastsofOpportunity”inS2Ssystems.So-called“majorSSWs”appearasdramaticpolarcapwarmingsof20oormoreaccompaniedbyweakeningofthecircumpolar zonalwindand reversal fromthenormalwesterly toeasterly. SSWanomaliescandevelop rapidly over days, descend in time over 1-2weeks, and persist at the tropopause for 1-2months.Thesestratosphericwindchangesaffecttheverticalandlatitudinalpropagationofplanetarywaves throughout the atmosphere. Thus extreme stratospheric wind events, whether extremelyweak or strong, are statistically associated with anomalous southward or northward shifts in thestormtrackswithassociatedsurfacetemperatureandprecipitationpatternshifts.Previousworkhassuggested enhanced prediction skill at the 16-60 day forecast rangewhen forecasts are initializedduringSSWevents.However,predictionskilloftheSSWsthemselvesremainspoorbeyond10days.

NASA’sGlobalModelingandAssimilationOffice(GMAO)atGoddardSpaceFlightCenterhasrecentlyreleased a new subseasonal to seasonal forecast system, GEOS-S2S version 2.1. Compared toGMAO’s previous system, the new version runs at higher atmospheric resolution (approximately1/2o globally), contains a substantially improved model of the cryosphere, includes additionalinteractive aerosolmodel components, and theoceandata assimilation systemhasbeen replacedwithaLocalEnsembleTransformKalmanFilter.Asetofretrospective45-dayforecastswasinitializedbasedontheMERRA2reanalysisat5-dayintervalsthroughoutyears1999to2015,with4ensemblemembers per initialization date. Our analysis of these retrospective forecasts reveals surfaceanomalypatternsintheperiodfollowingSSWevents,andcompareskeymeasuresofforecastskill,contrasting forecasts initialized during an SSW to those initialized during normal conditions. Wecomparetheseresultstothosepreviouslypublishedusingotherforecastsystems,andweexaminethepotentialroleofstratosphericwindbiasesandorographicgravitywavedragontheforecastskill.


(A8-03)

PredictabilityofSuddenStratosphericWarmingsinsub-seasonalforecastmodels

Karpechko,Alexey

FMI

We analyze the skill of the Arctic stratospheric retrospective ensemble forecasts (hindcasts) ofsubseasonal forecastmodels fromthes2sdatabasewitha focusonthepredictabilityof themajorsudden stratospheric warmings (SSWs) during the period 1981-2013. Predictability is assessed inboth deterministic, based on ensemble-mean forecasts, and probabilistic sense. To estimateforecastedSSWprobabilityweusethespreadofensemblemembers.WeshowthatsomeSSWscanbepredictedwithhigh(0.9)probabilityatleadtimesof12-13daysifadifferenceof3daysbetweenactualandforecastedSSWisallowed.FocusingonSSWswithsignificantimpactsonthetroposphericcirculationwe find that,onaverage, the forecastedSSWprobability is smallat lead timesofmorethan twoweeksand then increases rapidly tonearly1atday7beforeSSWs.Theperiodbetweendays8-12iswhenmostoftheSSWsarepredictedbythemodelswithaprobability0.5-0.9whichisconsiderably larger than the observed SSW occurrence frequency. Therefore this period can bethoughtofasanestimateoftheSSWpredictabilitylimit.Wealsofindindicationsthatpredictabilitylimit for some SSWs may be longer than two weeks; however more studies are needed tounderstandwhenandwhysuchlongpredictabilityispossible.


(A8-04)

Theroleofstratosphere-tropospherecouplinginsub-seasonaltoseasonalpredictionusingtheS2Sdatabase

DanielaI.V.Domeisen*(1),AmyH.Butler(2),AndrewJ.Charlton-Perez(3),BlancaAyarzagüena(4),MarkP.Baldwin(5),EtienneDunn-Sigouin(6),JasonC.Furtado(7),ChaimI.Garfinkel(8),Peter

Hitchcock(9),AlexeyYu.Karpechko(10),HeraKim(11),JeffKnight(12),AndreaL.Lang(13),Eun-PaLim(14),AndrewMarshall(14),ChenSchwartz(8),IslaR.Simpson(15),Seok-WooSon(11),

MasakazuTaguchi(16)

(1)ETHZurich,Switzerland(2)CIRES/UniversityofColorado&NOAA,USA(3)UniversityofReading,UK(4)UniversidadComplutensedeMadrid,Spain(5)UniversityofExeter,UK(6)Universityof

Bergen/BjerknesCentreforClimateResearch,Norway(7)UniversityofOklahoma,USA(8)HebrewUniversity,Israel(9)LMD,EcolePolytechnique,France(10)FinnishMeteorologicalInstitute,Finland(11)SeoulNationalUniversity,SouthKorea(12)MetOffice,UK(13)UniversityatAlbany,SUNY,USA

(14)BureauofMeteorology,Australia(15)UCAR,USA(16)AichiUniversityofEducation,Japan

Overthepastdecades,thestratospherehasbeenfoundtostronglycouplewithsurfaceprocesses,especiallyinwinter.Inthelightofimprovingpredictionsonsub-seasonaltoseasonaltimescales,thestratospherehasbeenfoundtopotentiallyrepresentacrucialsourceofpredictability, inparticularafter stratospheric extreme events. It however remains to be quantified to what extent thispredictabilityisreproducedinsub-seasonalmodelpredictions.Inacommunityeffort,weexplorethepredictabilityarisingfromstratosphere–tropospherecouplingonsub-seasonaltimescalesinawiderangeofmodelsfromtheS2Sdatabase.Surfacepredictabilityarisingfromarangeofstratosphericeventssuchassuddenandfinalstratosphericwarmings,strongvortexevents,andnegativewave-1heatfluxeventsisquantified,aswellaspredictabilityofthestratosphereitself,arisingfromremoteconnectionsintheclimatesystem.AcomparableanalysisisperformedfortheSouthernHemisphere.Thiscontributionwillprovideanoverviewofthestate-of-the-artofthecurrentlyavailableforecastskillarisingfromthecouplingbetweenthetroposphereandthestratosphere.


(A8-05)

Asignalandnoiseanalysisofstratosphere-tropospherecouplingintheS2Smodels

Charlton-PerezA(1)

UniversityofReading,UK

On sub-seasonal timescales, coupling between the stratosphere and troposphere represents asignificant sourceof skill fornorthernmidandhigh-latitudes. Previous studieshaveexamined thisskill either on a case-by-case basis or from the perspective of the additional skill present duringsudden stratospheric warming (SSW) or strong polar vortex events. Here we complement theseapproaches by fitting a simple statistical model to the full hindcast set available from the S2Sdatabase. The statisticalmodel usedenableus to separate thepredictable signal andnoise in theannularmodepresentineachhindcastset.WhileallmodelsintheS2Sdatabaseexhibitsomedegreeofstratosphere-tropospherecouplingintheannularmode,therearesignificantdifferencesbetweenthem. In themiddle and lower stratosphere, models have high skill out to week four, with largesignaltonoiseratio.Inthetroposphere,annularmodeskillisweakerinweeksthreeandfour.Inthelowerstratosphere,manymodelshavelowspreadandareover-confident.Inthetroposphere,thereissimilaroverconfidence,particularlyinweek3.Modelswiththelargestoverconfidenceinthelowerstratosphere also exhibit the largest overconfidence in the troposphere. In the troposphere, theoverconfidenceisduetoanoverestimationofthesizeofthepredictablesignalinmostmodels.

Correlationbetween theextracted signal in the lower stratosphere and surface varies significantlybetweenmodels.Takentogetherwiththeover-estimationofthesizeofthepredictablesignalinthetroposphere in these models suggests that some models have excessively strong stratosphere-tropospherecouplingonsub-seasonaltimescales.

abstract book for oral presentations · identifying wave processes associated with predictability...

Documents