T. Meissner, F. Wentz, J. Scott, K. Hilburn Remote Sensing ... · Remote Sensing Workshop November...
Transcript of T. Meissner, F. Wentz, J. Scott, K. Hilburn Remote Sensing ... · Remote Sensing Workshop November...
T.Meissner,F.Wentz,J.Scott,[email protected]
ESAOceanSalinityScienceandSalinityRemoteSensingWorkshopNovember26- 28,2014MetOffice,Exeter,UK
Outline1. AquariusADPSVersion3.0
� MajorUpdatesfromV2.0
2. LocalBiases� RainEffectsversusGeophysicalModelFunction
3. RainImpact:Atmosphere
4. RainImpact:SurfaceSplashing
5. RainImpact:Freshening
6. FreshwaterLensing
7. Summary
AquariusADPSVersion3.0OfficialAquariusProduct.ReleasedJune2014.
ADPSV2.0 ADPSV3.0
� SurfaceRoughnessCorrection� BasedonNCEPwindspeeds
� ReflectedGalaxy� Physicalmodel:
Oceansurfacemodelledasensembleoftiltedfacets:GeometricOptics
� SurfaceRoughnessCorrection� BasedonAquariuswindspeeds� HH-polactiveandH-polpassive
� ReflectedGalaxy� GeometricOptics(about90%)� Empiricalcorrection:Symmetrization
betweenascendinganddescendingswaths
� Q/CFlagging� degradedconditions� undetectedRFI
� AntennaPatternCorrection� Reducedbiasesoverland+ice� Reducedbiasesin3rd Stokes
� ΔSSS(SST)biasadjustment� mitigateslargescalebiases
Ascending(PM)– Descending(AM)Biases
Month
Latitude
GO(V2.0) GO+empiricaladjustment(V3.0)
Performance:LocalBiasesAquarius- ARGO
� Freshbiasesintropicsandsubtropics� Saltybiasesatmid-highlatitudes
� BiggestinNPacific� Seasonalpattern
Month
Latitude
2-year:09/2011– 08/2013
CausesofRegionalBiasesandMitigation"Real"
Aquariusiscorrectshouldnotbecorrected
"Fake"shouldbecorrected/mitigated
� Raininducedfreshening� Mainlytropics(ITCZ)� Stratificationofupperoceanlayer.� Aquariusmeasureswithinfewcm
ofsurface.� Insitumeasurementsatvarious
depth:� ARGO:5m� HYCOM:upperlayer?� mooredbuoys:1m
� ErrorsinthegeophysicalmodelfunctionthatisusedintheSSSretrievalandthathasazonalsignature.
� Rain� splashingeffectonsurface� atmosphere
� UndetectedRFI� enteringthroughsidelobes� makesSSSartificiallylow� issueincertainareas(China,
Japan,Indonesia,…)
Itisessentialtoseparatetheeffects.
RainMeasurementandFiltering
� Approach:1. FilteroutrainatorclosetoAquariusobservation.2. ChooseobservationsthatarenotimpactedbyundetectedRFI.3. Dealwiththemodelfunctiondependentbiases.4. Estimateatmosphericeffectsandsurfacesplashing.5. Leftwiththerealfreshening.
� CONAEMicrowaveRadiometer(MWR):KandKa–band§ OnAquariusplatform§ WasoperatinguntilApril2014§ AllowsrainmeasurementatthelocationandtimeofAquariusobservation§ Rainratesarenotverygood(no18GHzchannel)§ Allowsrainflagging:Aquariusobservationrainfree§ AvailableatRSSftpserver
ftp://Aquarius_mwr:[email protected]/aquarius_mwr/rss_l2c_h5/§ WillbeavailableatPO.DAAC
� MicrowaveImagers§ SSMIS(F17),WindSat:1hourwithinAquarius(sameequatorialcrossingtime)§ TMI:crossingtimeshiftingrelativetoAquarius§ CandetectrainwithintimewindowofAquariusobservation
AquariusSSS:AllEvents– RainFiltered
About30%ofthefreshbiasesatlowlatitudesareduetorain.Thesaltybiasesathighlatitudesarenotduetorain.
CausesofSSTDependentBiases� GeophysicalModelFunctionGMFthatisusedintheretrievals
� 0.1KuncertaintyinGMFis0.2psuuncertaintyinSSS� Dielectricconstantofseawater� Oxygenabsorptionmodelusedforatmosphere(non-resonantcontinuum)� Surfaceroughnessmodel� ErrorinauxiliarySSTfieldthatisusedintheretrievals(coldwater)� UndetectedRFI� Acombinationofsuch� Wearecurrentlytrackingthatdown.ImprovementplannedforV4.0.
� "Biasadjusted"SSSproduct� providedinV3.0inadditiontostandardproduct
SSSAQ- HYCOM
ImpactonPerformanceLocalBiases
V3.0norainfilter
biasadjustedrainfilter
ImpactonPerformanceTripleColocationStatistics
Global.RainFiltered
EstimatedIndividualErrorsσ[psu]AQ– HYCOM– ADPRCARGOσ[psu]
TripleCollocation:Assumethat errorsinthe3productsareindependent
3o monthlyaverages1.5o monthlyaverages
“Errors”inHYCOM/ARGOduetosamplingAQUARIUS HYCOM ARGO
V3.0 0.24 0.27 0.16 0.19SSTbiasadjusted 0.18 0.22 0.16 0.19
Std.Dev.ofMonthlyAveragesvs.ARGO
ADPSwithSST-biasadjustment
CAP
ADPSwithoutSST-biasadjustment
RainInducedEffects(1)1. Realfreshening
� Aquarius(fewcmofsurface)versusARGOorHYCOM(5mdepth)
2. Atmosphericabsorption� AquariusL2SSSretrieval
usescloudwaterfromNCEPforatmosphericcorrection.
3. Surfacesplashing?
Wewillnowshowthatabout80-90%oftheobservedeffectismostlyduetorainfreshening.
LowwindsSurfacelayerstratified
HighwindsSurfacelayermixed
AtmosphericCorrection
� AquariususesNCEPliquidcloudwaterforatmosphericraincorrection.
� NCEPmissesmostoftherain.Itunderestimatesthecloudwaterbycloseto100%andthusitisbasicallyuselessinrain.
� FutureversionwilluseCMORPHaccumulatedrainproduct.
ImpactonSalinityRetrieval� AssumethatSSM/Igivescorrect
cloudwater.
� AtmosphericabsorptionincreasesTBandthusgivesaSSSthatistoolow.
� Theatmosphericraincorrectionalgorithmissupposedtocorrectthat.
� NCEPunderestimatesthecloudwatercontentandthereforeundercorrectstheatmosphericeffect.
� OurcurrentAquariusV3.0SSSistoolow/freshinrain.� 0.2psuinhighrain.
� FuturereleaseswilluseCMORPH
NCEP
CMORPH3hr
RainSplashing:WindSpeed� WhencomparedtoNCEPthescatterometer(HH- pol)windspeedshowsanincreaseatlowwindspeedsandhighrainrates.
� Likelyduetosurfacesplashing.
� NCEPwindspeedbestgroundtruthinrain?
� Howdoesthateffectthesalinityretrievalsinrain?
Rainsplashing:SurfaceemissivityAnalyzeX-pol:X=β·H - VCombinationofV-polandH-polthatisinsensitivetoSSSβ= 1.18 (innerhorn),1.33 (middlehorn),1.54 (outerhorn)Itishowever,verysensitivetoroughness
V-polH-polX-pol
SSS
∆TB
Wind
∆TB
RainEffectonX-polTBmeasured- model
� ThereisverylittleraininducedexcessemissivityonX-pol.
� Partofthelittleeffectweareseeingcouldbeactuallyduetoundercorrectingtheatmosphericeffect.
� Scatterometerwindisusedinroughnesscorrection.Scatterometerwindisproxyforroughness.
� TheresultsindicatethatitisNOTwarrantedtoperforma"rainsplashcorrection"intheADPSalgorithmbyforcingtheAquariusSSStoHYCOM.
horn1 horn2 horn3
RainInducedEffects(2)
About80-90%oftheobservedeffectisduetorainfreshening.Therestisduetoatmosphere+surfacesplashing.
Lowwinds:fresheningisabout- 0.13to- 0.15psu/(mm/h)
LowwindsSurfacelayerstratified
HighwindsSurfacelayermixed
RainFresheninginTropicsAquariusversusHYCOM,ARGO,mooredBuoys
FreshwaterLensingRainRatesfromMicrowaveimager
Rainwithin1hourofAquarius
Fresheningatsurfacecomparedwith1m/5m
NorainatAquariusRain0- 1hourbeforeAQFreshwaterlensingvisible
Norainwithin1hourbeforeAquarius
Rain1- 12hoursbeforeAQupperlayerhasmixed
Aquarius– PMEL(1m)
Aquarius– HYCOM(5m?)
Aquarius– ARGO(5m)
SummaryandConclusions� Whencomparingsatellitesalinitywithgroundtruthweneedtocarefully
separatebiasesthatareduetoinaccuraciesintheGMFfrombiasesthatareduetostratificationwithintheupperlayer.
� AquariusADPSV3.0mitigatestheGMFbiasesbutkeepsthestratification.� Estimatedaccuracy:
0.22psu(1.5degmonthly).0.18psu(3.0degmonthly).
� RainfresheningfromAquarius� Observeddifferencebetweensurfaceand5mlayer:0.17psu/(mm/h)atlowwindspeeds.
� Smallerrorduetoinadequateatmosphericcorrectionandsurfacesplasheffects.
� Thescatterometerwindspeedisagoodproxyforsurfaceroughnessincludingrainsplashing.Nooronlyverysmallcorrectionforrainsplashingeffectsiswarranted,ifscatterometerwindspeedsareusedinthesurfaceroughnesscorrection.
� Instantaneousfresheningestimatedtoabout0.13to0.15psu/(mm/h)atlowwindspeeds.
� ConsistentwithSMOSresults.
� SurfacefreshwaterlensingisvisibleinAquariusdata1– 3hoursafterrainevent.
BackupSlides
AquariusSSS:AllEvents– RainFilteredGlobalAverageasTimeSeries
� Globaldifferencerain– no-rainissmall� 0.04psu/0.02K.� Timedependence(month)veryweak.
� Globallyitdoesnotrainthatmuch!
� Compensationtoaccountofthefactthatwedonotrainfilterincalibrationcanbedoneeeasily.� Addone-timesmallTAnumberforeachofthe6channelsincalibration.
RainImpactonSensorCalibration� V3.0senorcalibration:
� Matchesglobal7-dayaverageofAquariuswithHYCOM� Doesnotdorainfiltering.
� Consequence:TheV3.0AquariusSSSaretoosalty� Globalbias.� Stratificationintropics:
Aquariusseesrainfreshening,whichisreal,butHYCOMdoesnot.Itthereforeblamesthemismatchtothesensorcalibration.
� ThecorrectwaywouldbetofilterAquariusobservationsforrainbeforematchingthemtoHYCOM(orARGO).
� Howmuchdoesthatamountto?
� Study:Computedmonthlyclimatologyofnon-rainfilteredminusrainfilteredAquariusobservations.� TimeframewhereMWRisavailableforrainflagging.� InadditionuseMWimagers(SSMIS,WindSat,TMI)tofilteroutrainwithin1hour
beforeAquariusobservation.
MWRRainRates� AvailableatRSSFTPsite
� ftp://Aquarius_mwr:[email protected]/aquarius_mwr/rss_l2c_h5/
� SpaceandtimecollocatedtoAquariusL2swath.
� Format:� HDF5� samestructureasAquariusADPSL2files� samenameconventionforgranules
� Variables� time,latitude,longitude(matchesADPSL2)� cloudwater,rainrate,liquidwaterattenuationatL-band� windspeed,watervapor
� Documentationonprocessingalgorithmandfilecontent
� PO.DAAC?
� UseinAquarius� Foranalysisonly.NotpartofL2processing.� RainFlagging� RainRates:notverygood(37GHzchannelsaturates)
AquariusADPSversusCAPPerformanceAnalysis
� Exactlythesameensembleofsamplesusedforallcomparisons� SameQ/C
� Two-typesofStandardDeviations� 1.44Secsingleretrievals� Retrievalsfirstaveragedoveramonth
� CAPisVersion2.5.1� ADPS
� withoutandwithSST-biasadjustment
MeanandStd.Dev.(1.44sec)vs.SSTErrordominatedbyroughness
StdDev
Bias
ADPSwithSST-biasadjustment
CAP
ADPSwithoutSST-biasadjustment
1. CAPisnoisieratADPSon1.44seccycle.2. CAPhasappliedanSSTflatteningbymakinganadjustmentofthe
dielectricconstant.Thisoneofmanywaystodothat.
MeanandStd.Dev.(1.44sec)vs.WindErrordominatedbyroughness
StdDev
Bias
ADPSwithSST-biasadjustment
CAPADPSwithoutSST-biasadjustment
• CAPisnoisieratADPSon1.44seccycle.• Noiseisincreasingwithincreasingwindspeed.
Std.Dev.ofMonthlyAveragesvs.HYCOMErrordominatedbylocalSST-dependentbiases
ADPSwithSST-biasadjustment
CAP
ADPSwithoutSST-biasadjustment
Std.Dev.ofMonthlyAveragesvs.ARGO
ADPSwithSST-biasadjustment
CAP
ADPSwithoutSST-biasadjustment