Beruflich Dokumente
Kultur Dokumente
Huffman
Algorithm3B42:TRMMMergedHQ/InfraredPrecipitation
AlgorithmOverview
ThepurposeofAlgorithm3B42istoproduceTropicalRainfallMeasuringMission(TRMM)mergedhighquality
(HQ)/infrared(IR)precipitationandrootmeansquare(RMS)precipitationerrorestimates.Thesegriddedestimates
areona3hourtemporalresolutionanda0.25degreeby0.25degreespatialresolutioninaglobalbeltextending
from50degreesSouthto50degreesNorthlatitude.
The3B42estimatesareproducedinfourstages;(1)themicrowaveprecipitationestimatesarecalibratedand
combined,(2)infraredprecipitationestimatesarecreatedusingthecalibratedmicrowaveprecipitation,(3)the
microwaveandIRestimatesarecombined,and(4)rescalingtomonthlydataisapplied.Eachprecipitationfieldis
bestinterpretedastheprecipitationrateeffectiveatthenominalobservationtime(Huffmanetal.2007).
a.HighQuality(HQ)microwaveestimates
Alloftheavailablepassivemicrowavedataareconvertedtoprecipitationestimatespriortouse,theneachdatasetis
averagedtothe0.25spatialgridoverthetimerange90minutesfromthenominalobservationtime.Allofthese
estimatesareadjustedtoa"best"estimateusingprobabilitymatchingofprecipitationratehistogramsassembled
fromcoincidentdata.ThealgorithmtakestheTRMMCombinedInstrument(TCI;2B31)asthecalibratingdata
source.However,thecoincidenceofTCIwithanyofthesensorsotherthantheTRMMMicrowaveImager(TMI;
2A12)ishighlysparse,soweestablishaTCITMIcalibration,thenapplythattoTMIcalibrationsoftheother
sensorstoestimatetheTCIcalibratedvalues.TheTCITMIrelationshipiscomputedona1x1gridforeach
monthusingthatmonthscoincidentdatatoaccommodatethesomewhatdifferentclimatologiesofthetwo
estimates.PreliminaryworkshowedthattheTMIcalibrationsoftheothersensorsestimatesareadequately
representedbyclimatologicallybasedcoefficientsrepresentinglargeareas.Offline,theGPROFSSMI,GPROF
SSMIS,GPROFAMSR,AMSUB,andMHSprecipitationestimateshavebeenclimatologicallyprobability
matchedto2A12.ThecalibrationsofAMSUB,MHS,SSMI,andAMSREtoTMIhaveonesetofcoefficientsfor
eachsensortypeforlandand14setsforocean,whileSSMISiscalibratedseparatelyforeachsatellite,againhaving
onesetforlandand14forocean.Theoceanlatitudebandsare15overlappinglatitudebandscenteredonthe5
bands3530S,3025S,2520S,,2025N,2530N,and3035N.Theoutermostbandsareusedintheir
respectivehemispheresforallhigherlatitudecalibrationsduetothelackofTMIdatabeyondabout38.The
SSMISarecalibratedindividuallyduetoparticularcalibrationissuesforeachsatellite.AMSREusesa2monthset
ofmatchupstoensuresufficientsampling,whilealloftheothersworkwithsinglemonthaccumulations.The
coefficientsarecomputedseparatelyforeachseason.Finally,avolumeadjustmentfactoriscomputedforeachset
toensurethattotalTMIprecipitationispreservedinthesetransformations.
Oncetheestimatesarecalibratedforeachsatelliteandauditedfor>40%"ambiguouspixels",thegridispopulated
bythe"best"datafromallavailableoverpasses,althoughthemostlikelynumberofoverpassesinthe3hrwindow
foragivengridboxiseitheroneorzero.Whentherearemultipleoverpasses,datafromTCI,TCIadjustedTMI,
TCIadjustedAMSRE,andTCIadjustedSSM/Iareaveragedtogether,andTCIadjustedAMSUB/MHSestimates
areusedifnoneoftheothersareavailableforthegridbox.Testsshowthatthehistogramofprecipitationrateis
somewhatsensitivetothenumberofoverpassesaveragedtogetherwhenthatnumberissmall.Accordingly,inthe
futureweexpecttotestaschemetakingthesingle"best"overpassinthe3hrperiod.
b.VariableRainRate(VAR)Infrared(IR)estimates
3B42usestwodifferentIRdatasetsforcreatingthecompleterecordof3hourly0.25griddedTbs.Intheperiod1
January1998to6February2000,10km(subsampled),3hourlygridsofIRdatafromtheNationalClimaticData
Center(NCDC)GridSatB1wereusedinthe3B42processing.Fortheperiod1January199803UTC16June
1998geoIRdatawerenotavailableintheIndianOceansector,buthighzenithangledatafromadjacentgeo
satellitesisgenerallysufficientforfillin.Fortheperiodfrom7February2000onwards,theClimatePrediction
Center(CPC)MergedIRisused.ForeachhourtheIRdataareaveragedto0.25resolutionandcombinedinto
hourlyfilesas30minutesfromthenominaltime.TheamountofimagerydeliveredtoNCDCandCPCvariesby
satelliteoperator,butinternationalagreementsmandatethatfullcoverageisprovidedforthe3hourlysynoptic
times(00Z,03Z,,21Z).HistogramsoftimespacematchedHQprecipitationratesandIRTbs,eachrepresented
onthesame3hourly0.25grid,areaccumulatedforamonth,andthenusedtocreatespatiallyvaryingcalibration
coefficientsthatconvertIRTbstoprecipitationrates.AsintheHQ,thecalibrationintervalfortheIRisacalendar