Upload
duongnhan
View
214
Download
0
Embed Size (px)
Citation preview
GPMandAirborneRadarInvestigations:PIAEstimationPI:StephenL.Durden
SimoneTanelliJetPropulsionLaboratory,CaliforniaInstituteofTechnology
Email:[email protected]
1
6
2 3
• TheAPR-2airborneprecipitationradarhasbeeninoperationsince2001,collectingdatawithsamefrequenciesandgeometryasDPR
• In2015APR-2wasaugmentedwithW-band(referredtoasAPR-3)OriginallydueforcompletionJuneof2016AcceleratedscheduletoaccommodateOLYMPEX
• APR3acquireddataonallDC8flights,Nov12– Dec19• Amongaccomplishmentsincludetwounderflights ofGPMfor
observationsofprecipitation• TheAPR3Ku/Ka-bandchannelsallowdirectcomparisonwithDPR• TheW-bandchannelprovideshighresolutionobservationsoflight
precipitationandclouds,allowingdual-frequency(Ka/W)retrievals
5
Introduction
4
The authors gratefully acknowledge support by the Precipitation Measurement Missions Program. The research described here was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. Copyright 2016 California Institute of Technology. Government sponsorship acknowledged.
• NUBFeasiertodetectwithDPRbecauseNUBFchangesexpectedratiooftheKu/Ka PIAs
• Calculationbelow(left)ofexpectedvalueofPIAasafunctionofrainrateforlog-normalrain
• Atright,GPMdatashowsamebehavior
NUBFisimportanttounderstandandcorrectsinceitcausesbiasesinradarestimates:
• WhenthereisNUBF,wewouldliketogettheaverageoftheprecipitation• Instead,wemeasuretheaverageofradarreflectivityandsurfacebackscatter• Nonlinearrelationsbetweenprecipitationquantitiesandradarquantitiescausebiases
whenusingaverageradarquantitiestoestimateaverageprecipitation
7 8
9
ExamplefromNovember18:Ku,Ka, W,DFRKu/Ka,DFRKa/W
Shallowpost-frontalconvectionwithanoverrunningaltostratusice-cloudlayer
• APR-3Ka/WDFRiniceshouldbeindicativeoficeparticlesize;plotsbelowcompareMASCbrightnesstemperatureswithDFR(horizontalaxisistime)– Dec3duringGPMunderflight
• Generally,largerDFRisassociatedwithlowerTb;datasetshouldallowmulti-frequencyand/ormulti-instrumentretrievalsforcomparisonwithGPM
(Co-alignmentofdatacourtesyJ.Turk)
Left:dataforGPMunderflight Dec3Below:Dec19underflight
Underflights allowdirectcomparisonofAPR3andDPRforassessmentofnon-uniformbeam-filling(NUBF)effects
StudiesusingairborneradardatashowthatPIAestimatedwithSurfaceReferenceTechnique(SRT)areespeciallyimpactedbyNUBF
Right:plotofPIAErrorversusstandarddeviationofPIAinsidesimulatedGPMfootprint(Durden&Tanelli 2008)
ExamplecasesuggestsPIAKA~5xPIAKUwhenNUBFissmall(blue)
Datainprecip notshown
• Hence,theory,simulations,andDPRdataconfirmimpactofNUBFonSRT-estimatedPIA
• AnalternativemethodofPIAestimationusesonlytheradarprofile,notthesurfacemeasurement(suggestedbyZ.Haddad)
• Advantages:• Shapeofverticalprofileshouldbeless
impactedbyNUBF• Couldprovidealternativeoverland,
whereSRTislessreliable
Why?Consideracaseinwhichhorizontalvariationisthesameatallaltitudes;simplestexampleishalf-filled,half-empty
Near-surfacerangebinswillbeattenuatedbythePIAduetotherain;surfacebinattenuatedbyonly3dB
10
Todevelopprofile-onlyPIAestimate,weneededadatabasewithknownprecipitation,PIA,etc;syntheticdatabase:–Usedmorethan1yearofTRMMPRreflectivityprofiles–Foreachprofileoverocean,allowdropsizedistributiontovary–SimultaneouslyestimateR,Zka,attenuations,PIAs–Result:manyrainprofilesthatareconsistentwiththemeasuredTRMMKu-bandZprofile
• Examinedvariousprofileshapemetrics,includingprincipalcomponents,slope,maxZ,maxdual-frequencyratio(DFR)• BestpredictorofPIAisDFRnearsurface• SmallererrorforGPMPIAestimatewhenusingpredictortrainedfromAPRKu/Ka databasethansynthetic• ScatterplotisKa-bandSRTPIA(vert axis)versuspredictedKa PIAforGPM cases;NUBFcasesareblue• RMSerrorusingGPMKa-bandis2.6dBforcaseswithsmallNUBF• Increasesto4.2dBforlargerNUBF(>5dBstd dev)• SRTbiascanbe15dBinsimilarcases(Durden&Tanelli2008andplotinpanel5,above)
• Summary• APR3collecteddatasetinOLYMPEXthatwillbeusefulformulti-frequency/instrumentretrievals,includingdirectcomparisonwithGPM• Initialanalysisofnon-SRTPIAestimateindicatesmaybepossibletopredictPIAtowithinafewdBusingonlytheobservedreflectivityprofile• OurhypothesisisthatPIAestimationfromprofilemayhavelesserrorthanSRTincaseswithlargeNUBF• TestsofPIAestimationusingDFRnearsurfaceprovides2.6dBRMSerroratKa-band(smalleratKu-band)• ErrorforNUBFcasesislargerbutlikelybetterthanSRTinsomecases
MASCradiometer
APR3W-band
Secondapproachtodatabase:usedAPRKu/Ka-bandmeasurementsoveroceanconvection
NUBF
Windshearwillviolatethisassumption;however,impactmaybelessthanwithSRT
SyntheticdatabaseKa PIA(verticalaxis)versusmeasuredDFRnearsurface(horizontalaxis)
Profile-onlyPIA • Plans:directlyassessNUBFviaOLYMPEXcases,furtherinvestigateutilityofPIAestimationusingprofile-onlyapproach