Multimodel EnsemblePredictionSystem(MEPS):EnsembleModelingwithDataAssimilationModelsforSpaceWeatherScience,SpecificationsandForecasts
NASA-NSFSpaceWeatherModelingCollaborationNASAAmesResearchCenter,CA
May,20171
OverviewPresentedbyR.W.Schunk
MEPS Team
UtahStateUniversityR.W.Schunk,L.Scherliess,V.Eccles,L.C.Gardner,J.J.Sojka andL.Zhu
JetPropulsionLaboratoryX.Pi,A.J.Mannucci,andA.Komjathy
UniversityofSouthernCaliforniaC.WangandG.Rosen
MEPSModel
TheMultimodel EnsemblePredictionSystem(MEPS) coverstheIonosphere-Thermosphere-Electrodynamics(I-T-E)systemandincorporatesexisting,first-principles-based,dataassimilationmodelswithdifferentphysics,numericaltechniques,andinitialconditions.
MEPSallowsensemblemodelingwithdifferentdataassimilationmodels.
NationalHurricaneCentermulti-modelensembleforecastforhurricaneRita.
WhyEnsembleModeling
ModelsDisplayQualitative&QuantitativeDifferences:
• DifferentBackgroundPhysics-BasedModels
• DifferentAssimilationTechniques• DifferentSpatialandTemporal
Resolutions• DifferentDeducedElectrodynamics
Drifts• DifferentDeducedNeutralWinds
andO/N2 Ratios
Objectives
• Elucidatethefundamentalphysical,chemical,andcouplingprocessesthatoperateintheI-T-Esystemforarangeofactual,global-scale,spaceweatherevents,includingstorms&substorms.
• ConstructaMultimodel EnsemblePredictionSystem(MEPS)fortheI-T-Eenvironmentthatwillincorporateourexistingdataassimilationmodels.
• DeliverMEPS totheCommunityCoordinatedModelingCenter(CCMC)
MEPSDataAssimilationModels
GAIM-BLè Mid&LowLatitudesGAIM-GMè Mid&LowLatitudesGAIM-4DVARè Mid&LowLatitudes,withDriversGAIM-FPè Mid&LowLatitudes,withDriversMid-LowElectro-DAè IonospherewithDriversIDED-DAè HighLatitudes,withDriversGTM-DAè GlobalThermosphere
• Global,Regional&NestedGRIDCapabilities• GAIM-GM&GAIM-BLareOperationalModels• Science,Specifications&Forecasts
MEPSDataSources
MEPSAccomplishmentsandStatus
1. EnsembleModelAveraging2. ImprovedStatisticalTechniquesforMEPS3. ModelDeliveriestoCCMC4. Real-timeTWAM5. GlobalDataAssimilationStudyofSubstorms6. AdditionalTasksforIDED-DA7. PulsatingGeomagneticStorm8. FromDataAssimilationtoPrediction: Assimilative
ModelingofIonospheric Disturbances
1.EnsembleModelAveraging
ModelsUsed:GAIM-BLè Mid&LowLatitudesGAIM-GMè Mid&LowLatitudesGAIM-4DVARè Mid&LowLatitudes,withDriversGAIM-FPè Mid&LowLatitudes,withDriversMid-LowElectro-DAè IonospherewithDriversIFMPhysics-BasedModelè NoDA
DataAssimilated:SlantTECfromGroundReceiversNe ProfilesbelowF-RegionPeakfromDigisondesCOSMICOccultationDataInSituNe(SSIES)
ValidationData:NeutralwindsfromFabry-PerotInterferometers(FPI)COSMICOccultationDataJasonverticalTECovertheoceansDorisslantTECfromradiobeacons
GlobalDistribution
EnsembleModelAveragingApproach• DifferentNumberofDataAssimilationModels• DifferentDataTypesandAmounts• DifferentSeasonal,SolarCycle,StormandSubstorm Cases• DifferentDataAveragingTechniques
OriginalTest:March,201312-19(71-78)– Haveallmodelsfor21:00UTonMarch17.
Averagetestandpulsatingstormtest:May,20161-12(122-133)StormonMay8(129)
EnsembleModelAveragingSimulations
EnsembleModelAveragingExample
• 5DataAssimilation&1PhysicsModel• MidandLowLatitudes• GPSandOccultationData• SolarMedium,Equinox,Storm• SimpleAverage
March12-19,2013
AGUEosResearchSpotlight:Schunk,R.W.,etal.,SpaceweatherForecastingwithaMultimodel EnsemblePredictionSystem(MEPS),RadioSci.,51,doi:10.1002/2015RS005888,2016.
LeftpanelsareorGPS onlyrunandrightpanelsarefortheGPSandoccultationdatarun.Thetoppanelsshowthemean andthebottompanelsshowthestandarddeviation.Snapshotforthestormdayat2100UT.
EnsembleAveraging(6 models)
GPS&OccGPS
StandardDeviation
MeanTEC
EnsembleAveraging(6 models)
EnsembleMeanVerticalTECFromGPS&OccultationRun
VerticalTECData
EnsembleMeanVerticalTECVersusTECData
EnsembleAveraging(6 models)
EnsembleMeanVerticalTECFromGPS&OccultationRun
VerticalTECData
EnsembleMeanVerticalTECVersusTECData
ThemeanoftheEnsembleineachcasePerformedBetterthantheIndividualDataAssimilationModels
2.ImprovedStatisticalTechniquesforMEPS
• 5 DataAssimilationModels&1PhysicsmodelusedinAverage
• GPSandOccultationData• SimpleAverage– Summodels,divideby
numberofmodels• WeightedAverage– Summodelsweighted
byfittoGPSdata,dividebynumberofmodels
MEPSEnsembleAverage- Simple
ModeledmaximumTEC(red)lowerthanmeasuredTEC(pink)
EnsembleMeanVerticalTECVersusTECData
MEPSEnsembleAverage- Weighted
TheWeightedmeanoftheEnsembleofDAModelsIsbetterthantheSimplemean.
EnsembleMeanVerticalTECVersusTECData
3.ModelDeliveriestoCCMC
GAIM-GM(Latestupgradedversion)è delivered Spring2016IDED-DA(HighLatitudeGAIM)è deliveredSpring2017GAIM-FPè Fall2017Mid-LowElectro-DA è Winter2017
AllDeliveries Include:– Backgroundionospheremodels– Connectionstorelevantdatasources– USUinstallationonCCMCcomputers– User’sManual
• VariableNumberofGroundGPS/TECSites
• VariableNumberofDISSStations
• DMSPinsituNe (SSIESF13,F14,F15,F16,…)
• UVRadiances(SSUSI,SSULI)
• COSMICOccultationData
• QualityControlAlgorithms
• DataLatency(upto3hours)
• HotStartCapability
• 24-HourForecastAlgorithm
• User’sManualandTraining
GAIM-GMandGAIM-FPmodeldeliveriestoCCMCinclude
CCMCIDED-DAModelDelivery
• DeliveredSpring2017– Physics-basedmodels– DatareductionutilitiesforSuperMAG andSuperDARN data– Kalman-filterdataassimilation
• IssuesduringWinter2016– ComputersizerequiredsmallerIDED-DAVersion
• WorkedtominimizeIDEDfootprint• OnlyonemonthofSuperMag andSuperDARN dataprovided(March2013)fortesting
• 2018Upgrades– IncludeAMPEREsatellitemeasurements– HelpCCMCwithWebsiteinterface
§ TWAMisbasedonafirst-principlesmodelforthethermosphericwind.
§ DataareassimilatedusinganimplicitKalmanfiltertechnique.
§ PreviouslyTWAMhasbeenusedtodeterminewindclimatology.
§ InitialresultsusingTWAMonaday-to-daybasishavebeenobtained.
4.Thermospheric WindAssimilationModel(TWAM)
5.GlobalDataAssimilationStudyofSubstorms
• Focusisonsmall-scalesubstorm structures• Lookingfornewundiscoveredfeatures(like
terminatorfield-alignedcurrent)• Selected23substorm cases
• Differentsolar,seasonalandsubstorm conditions
• Strong,moderateandweakintensities
• Singleandmultiplesubstorms
• CompletedallIDED-DArunsforsinglesubstorms
SelectedSubstorm Cases
• Averyquiethigh-latitudeionosphere(usedasabaseline)(1)• Singlesubstorms withstrong,medium,weakintensities(3)• Substorms withmultipleonsets/brightenings (1)• Multiplesubstorms withvariouscharacteristics(4)• Substorms withclassicalfeatures(3)• Substorms withmultipleprecursors(1)• Directlydrivenelectrojet enhancementevent(1)• Substorm withnogrowthphaseandaverydisturbedrecoveryphase(1)• Longdisturbingperiodswithirregularmultiplesubstorms (2)• Longstableperiodswithenhancedwestwardelectrojet,butnosubstorms (1)• Longlastingsubstorms withhugedisturbancesingrowthphaseandrecoveryphase(1)• Multiplesubstorms withveryshortgrowthphase(1)• Longlastingenhancedwestwardandeastwardelectrojets withnosubstorms (1)• Substorms withmultiplesurgesinrecoveryphase(1)• Longlastingdisturbingperiodwithmultipleauroral brightening,butnosubstorms (1)
IsolatedSubstorm MultipleSubstorms
Substorm withMultipleOnsets
Substorm Types
IsolatedSubstormsExpansionPhaseUpperLeft:Weak,WinterUpperRight:Moderate,EquinoxLowerLeft:StrongSingleSubstorm
DuringStorm
6.AdditionalIDED-DATasks
• AdddensityparameterstoIDED-DAdata-assimilation scheme(GPS-TEC)– Tendstosmootharcs,patches,blobs,etc.
7.PulsatingGeomagneticStorm• IdentifiedRecentPulsatingStormswithDataAvailable• RunwithMEPSDAModels• TrackTADDynamics• ApplyEnsembleAveraging
Summary
• MEPSè ensemblemodelingwithdifferentdataassimilationmodels
• Dataassimilationonmultiplespatial&temporalscales• Widerangeofgroundandspacedata• Animportanttoolforstudyingbasicphysics• Cancombinedifferentdatasetsintoacoherentpicture• Fillsinregionswheretherearenodata• Newapproachtospecificationsandforecasts
AdditionalMEPSSlides
OriginalTest:March,201312-19(71-78)– Haveallmodelsfor21:00UTonMarch17.
Averagetestandpulsatingstormtest:May,20161-12(122-133)StormonMay8(129)
EnsembleModelAveragingSimulations