Combined algorithm ATBD - NASA

GPM Combined Radar-Radiometer Precipitation Algorithm Theoretical Basis Document (Version 5)

William S. Olson

and the

GPM Combined Radar-Radiometer Algorithm Team

January 8, 2018

2

Table of Contents Page 1. Introduction ..................................................................................................... 3 2. Background ..................................................................................................... 3 GPM Instruments ........................................................................................... 3 Implications for Algorithm Design .................................................................. 4 3. Algorithm Architecture ................................................................................... 5 Overview ........................................................................................................ 5 Ku Radar Module ......................................................................................... 15 Forward Model Module................................................................................ 16 Radiance Enhancement Module.................................................................... 17 Filter Module ............................................................................................... 17 4. Ancillary Datasets ......................................................................................... 18 Geographic Data .......................................................................................... 18 Analysis Data ............................................................................................... 18 Databases Supporting Specification of Environmental Parameters ............... 19 Microwave Absorption and Single-Scattering Tables .................................... 20 5. Summary of Algorithm Input/Output ............................................................. 21 6. Algorithm Testing Plan ................................................................................. 22 Sensitivity Testing ......................................................................................... 22 Physics Testing ............................................................................................. 24 Pre-launch Validation .................................................................................. 26 Post-launch Validation ................................................................................. 28 Metrics ......................................................................................................... 29 7. References..................................................................................................... 30 Appendix A. Listing of Input/Output Parameters ............................................... 33 Input Parameters .......................................................................................... 33 Output Parameters ....................................................................................... 49 Appendix B. Output Product Volumes .............................................................. 63 Appendix C. Processing Requirements .............................................................. 63 Appendix D. Version Changes .......................................................................... 64

3

1. Introduction

The GPM Combined Radar-Radiometer Algorithm performs two basicfunctions: first, it provides, in principle, the most accurate, high resolutionestimatesofsurfacerainfallrateandprecipitationverticaldistributionsthatcanbe achieved from a spaceborne platform, and it is therefore valuable forapplications where information regarding instantaneous storm structure arevital. Second, a global, representative collection of combined algorithmestimates will yield a single common reference dataset that can be used to“cross-calibrate” rain rate estimates from all of the passive microwaveradiometers in the GPM constellation. The cross-calibration of radiometerestimates is crucial for developing a consistent, high time-resolutionprecipitation record for climate science and prediction model validationapplications. Because of the CombinedAlgorithm’s essential roles as accuratereferenceandcalibrator, theGPMProject is supportingaCombinedAlgorithmTeam to implement and test the algorithmprior to launch. In the pre-launchphase, GPM-funded science investigations led to significant improvements inalgorithm function, and thebasic algorithmarchitecturewas formulated. ThisalgorithmarchitectureislargelyconsistentwiththesuccessfulTRMMCombinedAlgorithmdesign,butithasbeenupdatedandmodularizedtotakeadvantageofimprovements in therepresentationofphysics,newclimatologicalbackgroundinformation,andmodel-basedanalysesthatmaybecomeavailableatanystageof the mission. Post-launch, algorithm physical parameterizations for effectssuchasthenon-uniformbeamfillingoftheradarfootprintbyrainandmultiplescatteringofradarpulsesbyice-phaseprecipitationhavebeenimproved.Also,theradiativeeffectsofnonsphericalice-phaseprecipitationandasemi-empricalmodel relating radar surface backscatter cross-section and multi-spectralmicrowave emissivities have been included. This document presents adescription of the GPM Combined Algorithm architecture, scientific basis,supportingancillarydatasets,inputs/outputs,andtestingplan. 2. Background GPM Instruments

The GPM coremission satellite observatory is shown in Fig. 1. From thisplatform, the Dual-frequency Precipitation Radar (DPR) scans cross-track inrelativelynarrowswathsatKuband (13.6GHz)andKaband (35.5GHz). Thedual-frequency radar reflectivity observations are nearly beam-matched overthe125kmKa-bandswath,withahorizontalresolutionofapproximately5km,and a vertical resolution of 250m in standard observingmode. TheKu bandradar scans over awider, 245 km swath. The GPMMicrowave Imager (GMI)scansconicallyoveran885kmwideswathat frequenciesof10.65,18.7,23.8,36.5, 89.0, 165.5, 183.31 ± 7, and 183.31 ± 3 GHz. Measured brightnesstemperaturesareintwopolarizations(verticalandhorizontal)atallbutthe23.8GHz and 183.3 GHz channels, which provide only vertical polarization

4

measurements. TheGMIobservationsarediffraction limited,with the lowest-resolution footprints (approx. 26 km) at 10.7 GHz and the highest-resolutionfootprints(approx.6km)atthe89.0GHzandhigherfrequencychannels.

Implications for Algorithm Design The current GPM Combined Radar-Radiometer Algorithm architecture is

descendedfromarichheritageofalgorithmsthatweredevelopedfortheTRMMmission, aswell as other algorithms developed and applied to airborne radar-radiometer data. In TRMM, only Ku-band radar observations were availablefrom the radar instrument (the Precipitation Radar, or PR), and only lower-frequency(10-85GHz)brightnesstemperaturemeasurementswereavailablefrom themicrowave radiometer (theTRMMMicrowave Imager, or TMI). TheTRMMFacilityCombinedAlgorithmusedradiometerinformationtoessentiallyreduce uncertainties in estimates of radar-derived total path-integratedattenuationtotheearth’ssurfacetoperformanimprovedattenuationcorrectionof the radar reflectivity vertical profile. The improved attenuation correctionwas effected by adjusting a single parameter of the precipitation particle-sizedistributionovertheentireprecipitationverticalprofile.Thissingleparameterrepresented a rain-normalized, mass-weighted mean particle diameter, whichwas assumed to be locally constant over the scale of TMI footprints. Addinghorizontal variations of this parameter would have introduced toomany freeparameterstotheinversionproblem.

The GPM Combined Algorithm takes advantage of the additional Ka band

Fig. 1. Configuration of the GPM core observatory, illustrating the scanning geometry of the DPR and GMI instruments.

5

radar channel to glean more specific information about the precipitation sizedistribution and associated attenuation in each gate. The estimation ofprecipitation size distribution parameters is further aided by precipitationattenuation information from the GMI channels, which have an extendedspectralrangerelativetotheTMI.However,iftheKabandreflectivitiesdonotprovideadditionalinformationduetoverylightrain(Rayleighlimit),ortheyareseverely attenuated in heavy precipitation, then the combined algorithmmustmake a natural transition to a single-frequency, Ku band solution in which amore approximate estimation of precipitation size distribution parameters isperformed.

Regardless of whether or not the Ka band data are applicable, however,

informationfromtheGMIbrightnesstemperaturescanbeusedtomakefurtheradjustmentsofpathattenuationduetonon-precipitatingcloudliquidwaterandwatervapor,whicharenotdirectly sensedby theDPR. Inaddition, thereareprecipitationmicrophysicalparameters,suchastheinterceptoftheparticlesizedistribution and the density of ice-phase precipitation that may be adjustedusingradiometerinformation.

Ultimately, the degree to which any precipitation or environmental

parameterscanbeadjustedislimitedbytheinformationcontentoftheDPRandGMIobservationsandanyadditionalinformationprovidedbyaprioridata,suchas the natural ranges of particle size distribution parameters, cloud watercontents,etc.,andhowtheseparameterscovaryspatially.Therefore,asoutlinedin section 3, the combined algorithm is designed to be able to accept bothdifferent physicalmodeling assumptions anda priori constraintson estimatedparameters. 3. Algorithm Architecture Overview

The current algorithm design is based upon an Ensemble Filtering (EnF)approach for inverting theDPRreflectivitiesandGMIbrightness temperaturestoestimateprecipitationprofiles;seeAnderson(2003)forageneraldescriptionofEnFapproaches.ThegeneralarchitectureoftheGPMCombinedAlgorithmisillustrated in Fig. 2a,b. There are four primary modules in the CombinedAlgorithm: the Ku Radar Module, which produces ensembles of Ku radar-consistent precipitation profile solutions at each DPR footprint location, theForward Model Module, that simulates the remaining DPR and GMImeasurements, the Radiance Enhancement Module, which estimates what theGMI brightness temperatures would be at the resolution of the DPR, and theFilterModule,thatmodifiestheKuradar-derivedprecipitationensemblestobemoreconsistentwithremainingobservations.Theoutputsofthealgorithmarethemean(bestestimate)andstandarddeviation(uncertaintyofestimate)oftheDPR-GMI filtered ensemble of estimated precipitation profiles at each DPR

6

footprintlocation.Followingisadescriptionofalgorithmflowfromtheingestofsatellitesensordatatotheoutputofprecipitationestimates.

The Combined Radar-Radiometer Algorithm first ingests Radar Algorithm

Level2calibratedreflectivitiesatKuandKabands(ifavailable)aswellasLevel1C intercalibrated brightness temperatures from the GMI. To stay withincomputermemory limitations, amaximumof 300 scan lines of DPR data andcorrespondingGMIdataareprocessedbythealgorithmatatime.Therefore,theflowdiagramofFig.2arepresentstheprocessingofoneswathsegment,whichisrepeateduntiltheentireorbitisprocessed. TheDPRfootprintlocationsdefinean

approximate5kmx5kmgridon theearth'ssurface, and these footprintsareused to represent the solution grid for algorithm-estimated precipitationprofiles.

Fig. 2a. Processing schematic for the Combined Radar-Radiometer Algorithm. Ku Radar Module is in blue (see also Fig. 2b), Forward Model Module is in green, Radiance Enhancement Module is in yellow, and Filter Module is in orange.

7

The first major process in the algorithm is the estimation of precipitation profiles

in the Ku Radar Module. The first step in this process is to determine whether precipitation is detected in the column. The Ku Radar has a minimum detectable signal of approximately 13 dBZ, and range gates with reflectivities higher than this threshold at altitudes above the surface clutter are interpreted as precipitating. Precipitation detection and surface clutter information is provided by the Radar Algorithm Level 2 output.

If precipitation is detected, then the near-vertical column of Ku reflectivities is

processed further to make initial estimates of precipitation. The precipitation estimation requires additional inferences of the pressure/temperature profile of the atmospheric column, the gaseous and cloud water absorption properties of the column, and whether or not the precipitation is convective or non-convective. The pressure/temperature is drawn from meteorological analysis data that have been interpolated to the locations of the Ku range bins, and these temperatures are provided by the Radar Algorithm Level 2 output. If a bright band of high Ku reflectivity is detected in the column, then this establishes a reference point in the phase-transition from ice to liquid in the column, with the transition starting 750 m above this point and ending 500 m below the point. Alternatively, if no bright band is detected, then the freezing level in the analysis temperature profile provides the reference point.

Gaseous absorption depends on the pressure, temperature, and humidity of the

atmosphere. Cloud absorption depends on the temperature and cloud water content. Since the water vapor and cloud water distributions are not well known a priori, and since first guess estimates of water vapor and cloud cannot be reliably determined from global analyses, prospective water vapor and cloud profiles are generated from EOF representations derived from cloud-system-resolving model simulations. Random weightings of the EOF components are used to generate ensembles of possible water vapor and cloud vertical profiles that could occur, a priori. An ensemble of 40 profiles is created in this way.

Along with the a priori ensembles of water vapor and cloud profiles, a priori

assumptions regarding the precipitation particle size distribution at each range bin are also made. The precipitation particle size distribution (PSD) in each bin isdescribedbyanormalizedgammadistribution(Testudetal.2001),

, (1)

where

. (2)

Here, Nw is the intercept of the normalized distribution, Dm is the volume-

n D( ) = Nw f µ( ) DDm

!

"#

$

%&

µ

exp −4 + µ( )Dm

D!

"#

$

%&

f µ( ) =6 4 + µ( )µ + 4

44 Γ 4 + µ( )

8

weightedmeandiameter,µisthedistributionshapefactor,Distheliquid-waterequivalentdiameterof theparticle, andn(D) is thespectralnumberdensityofparticleswithdiameterD.

Sincetherearethreeparameters thatdescribetheprecipitationPSDandatmost two independent radar-derived reflectivity observations associated witheach DPR range bin, assumptions regarding the PSD parameters are made toreduce the degrees of freedom in the estimation problem. In the currentalgorithm, µ is assumed constant, and it is currently set at a value of 2.Ensembles of prospective Nw profiles are generated using an autoregressivefunction,startingwitharandomvaluedrawnfromanassumedpdfatthetopofthe profile. The assumed pdf depends on the convective vs. non-convectiveclassification of the given profile; this classification is output from the RadarAlgorithmLevel2,anditisbaseduponthehorizontalandverticalstructuresofKureflectivityintheDPRobservations.EachensemblememberNwprofileandconstant µ profile is matched to a water vapor and cloud profile from theensemble ofwater vapor and cloud profiles. This creates a combineda prioriensemble of environmental and precipitationparameter profiles, and only thethird PSDparameter,Dm, is not specified in each ensemble profile. The thirdparameter, Dm, in each range bin is inverted from the profile of Ku bandreflectivitiesforeachensemblememberprofile,asdescribedforthwith.

ToestimateDm foreachensemblememberprofile,thereflectivityprofileat

Kubandisfirstcorrectedforgaseousandcloudwaterattenuationineachofthe40 ensemble member environmental profiles. The specific absorption byatmospheric gases and cloud water as a function of pressure, temperature,humidityandcloudwatercontentisdrawnfromtabulatedvaluesattheKubandfrequency. The remaining attenuation in each ensemble member reflectivityprofileisduetoprecipitation.

Using the gas/cloud absorption-corrected Ku reflectivity profile, ZKu,

associated with a given ensemble member, a generalized Hitschfeld-BordanmethodisappliedtosolvefortheprofileofDmforthatmember;seeGrecuetal.(2011). The generalized Hitschfeld-Bordan method is a single-wavelength,forwardrecursivemethodthatiterativelycomputestheeffectivereflectivityandextinction by precipitation in a given range bin; then attenuation-corrects thereflectivity in thenext range binof theprofile; seeFig.2b. In thegeneralizedapproach, the iteration of reflectivity and extinction calculations is madecomputationally efficient through an analytical manipulation of the radarequation;seeGrecuetal.(2011).Theiterationequationis

9

, (3)

Z r( ) =ZKu r( )

1− q ZKuβ s( )

k Z s( )( )Z β s( )

ds0

r∫

#

$%%

&

'((

1β

Fig.2b.SchematicforthegeneralizedHitschfeld-Bordanmethod,containedintheKuRadarModule;seeFig.2a.

10

where Z(r) is the attenuation-corrected reflectivity at range r, ZKu(r) is themeasuredreflectivityatthatrange,k(Z)isthespecificextinctioncorrespondingtoreflectivityZ,q=0.2ln(10),andk=a�Zbisanapproximatek-Zrelationthatisintroduced to reduce the number of iteration steps. The actual relationshipsbetween precipitation extinction and reflectivity are represented by static"scattering"tablesofthesequantitiesforarangeofNwandDmvalues,givenµ.Inhigh-attenuationregimes,numericalinstabilitiesareavoidedbyrescalingtheNwprofile;seeFig.2b. ThegeneralizedHitschfeld-BordanmethodisappliedtoallofthememberprofilesintheensembleateachKu-bandfootprintlocation.

At light rain rates where independent estimates of the column path-integrated attenuation from the Level 2 Radar algorithm (see Meneghini et al. 2000) are unreliable, the initial guess Nw is updated from the initial guess Dm using an empirical formula. The updated Nw will have the observed anti-correlated relationship with Dm; see Thompson et al. (2015). The updated Nw is used as a new first guess Nw, and the algorithm proceeds normally.

IfvalidKa-banddataand/orGMIbrightnesstemperaturedataareavailable,the algorithm passes to the Forward Model Module; see Fig. 2a. For eachensemble member profile passed from the Ku Radar Module, the singlescatteringpropertiesatKa-bandandtheGMIchannelfrequenciesateachradarbinlocationarecalculatedusingthegaseous/cloudabsorptionandprecipitationscatteringtablespreviouslydescribed.Thesingle-scatteringpropertiesareusedto simulate the Ka reflectivities,ZKa, and PIA’s at Ku andKa bands,PIAKu andPIAKa,respectively,foreachofthe40ensemblememberprofiles.

Simulations of ZKa and PIA are subject to the effects of non-uniform

beamfilling of precipitation within the radar footprint, as well as multiplescatteringofradarpulseenergy.Non-uniformbeamfillingisaccommodatedbygenerating a set of “downscaled” reflectivity profiles from each observed Ku-bandprofile.Eachdownscaledprofileisasimple,scaledversionoftheobservedprofile,where the scaling is a random, lognormally-distributedparameter, andthemeanofthedownscaledprofilesisconstrainedtobeequaltotheobservedprofile.Thescalingleadstoasetofprofilesthathaveverticalcoherence,asseeninconvectiveprecipitationstructures,but theprofile-to-profilevariation inthesetrepresentstheeffectsofhorizontalvariabilitywithintheradarfootprint.Foreachensemblemember,thegeneralizedHitschfeld-Bordanmethodisappliedtoall the downscaled reflectivity profiles in the set, assuming the a priorienvironmentalandPSDparametersassociatedwiththeensemblemember.TheKa reflectivities and PIA’s are then simulated for each solution profile of thedownscaledset,andthenthemeanprecipitationparametersolutionprofileandthe mean simulated Ka profile/PIA’s of the set are computed. These mean,“upscaled” quantities are considered the precipitation parameter profile andsimulatedZKa,PIAKu,andPIAKaassociatedwiththegivenensemblemember.Thestandard deviation of the lognormally-distributed scaling parameter can beadjustedtoreflect the intrinsichorizontalvariabilityofprecipitationaccording

11

to the environmental state. Currently, only separate values of the standarddeviation for over-ocean and over-land applications of the algorithm areprescribed.

Although the lognormally-distributed setof reflectivityprofiles, justdescribed,fairly well represents the effects on footprint non-uniform beamfilling, it waslaterdeterminedthat thepath-integratedattenuationatKaband, inparticular,was being overestimated. Using off-line, high-resolution simulations ofattenuationbaseduponground-basedradarfields,itwasfoundthattheKa-bandpath-integrated attenuation in vertical columns over DPR-sized footprints,derived using aHitschfeld-Bordanmethod as it is done in the CMB algorithm,significantly overestimates Ka band path integrate attenuation in convectiveregionswherethefootprintsarepartiallyfilledwithprecipitation.However,thedegree of partial filling can be roughly estimated using a 3x3 array of DPRfootprints centeredon the footprintof interest. A scalingparameterbasedonthe 3x3 array is introduced to modify the Hitschfeld-Bordan derived path-integrated attenuation at Ka band to properly account for partial filling of theradar footprint by precipitation. At Ku band, the effects of partial footprintfilling on path-integrated attenuation are much smaller and are neglected atpresent.

Multiple scattering of radar pulses is important in precipitation columns

where bulk scattering by ice-phase precipitation is significant, as in the coreregions of strong convection (Battaglia et al. 2015). The effects of multiplescattering of radar pulses at Ka band are evaluated using the 1-D time-dependent radiative transfer method of Hogan and Battaglia (2008). Theensemble-meanaprioriconditionsareusedtoestimatetheprecipitationprofilefrom the original Ku-band observations. Then the Ka-band reflectivities aresimulatedusingbothasingle-scatteringassumptionandthemultiple-scatteringmethod of Hogan and Battaglia (2008). If the single and multiple-scatteringsolutions differ by more than a small tolerance, then the Ka reflectivities aresimulated for each ensemble member using the multiple-scattering method,utilizingthemeanscatteringpropertiesofthedownscaledsetofKu-bandprofilesolutionsasinput.Thisapproachlimitsthecomputationally-intensivemultiple-scatteringcalculationstosituationswheresuchcalculationsareneeded.

To simulate the GMI brightness temperatures (at DPR resolution), the

surface temperature and microwave emissivities must be specified for eachensemble member. The surface skin temperature is drawn from the modelanalysisdataset, and it is assumed tobe the same foreachensemblemember.Overwatersurfaces,thesurfaceemissivityiscalculatedusingtheMeissnerandWentz (2012)model, based upon the surface skin temperature and randomlygenerated10-meterwindspeedsforeachensemblemember. AconsistentsetofsurfacenormalizedradarcrosssectionsarealsogeneratedfromtheensembleofwindspeedsusingempiricalrelationshipsderivedbyMunchaketal.(2016).Over land surfaces, ensembles of surface normalized radar cross sections and

12

multi-spectralmicrowaveemissivitiesaregeneratedusinganEOFbasisforthecrosssections/emissivitiesderivedbyDr.S.Munchak(personalcomm.;2017).The EOF’s of the cross sections/emissivities are randomly weighted andcombined to yield ensembles of cross sections/emissivities that exhibit thenaturalcovariabilityoftheseparameters.

Microwave brightness temperatures are calculated for each ensemble

member using Eddington’s second approximation with delta scaling (seeKummerow 1993; Joseph et al. 1976), which include the effects of multiplescatteringofradiances.Asinthemultiple-scatteringKareflectivitysimulations,the mean scattering properties of the downscaled set of Ku-band profilesolutions are used as input to the microwave brightness temperaturesimulations.

ThebrightnesstemperaturessimulatedintheForwardModelModuleareat

DPRresolution(5km),yettheobservedbrightnesstemperaturesfromGMIareat lower spatial resolution (6 - 26 km). To accommodate the differences inresolutions, the observed GMI brightness temperatures are processed in theRadianceEnhancementModuletoestimatebrightnesstemperaturesatthesamefrequencies and polarizations, but at a resolution close to the DPR resolution.Offline,a largeswathofDPR-resolutionbrightnesstemperaturesaresimulatedat all of theGMI channel frequencies andpolarizationsusingoutputof theKuRadarModuleandForwardModelModule. Theseserve,effectively,asasetof“true”brightnesstemperaturesatDPRresolution,andtheyarethenconvolvedusing the GMI antenna patterns to GMI resolution. The DPR resolutionbrightness temperatures are regressed against a small neighborhood of GMI-resolution brightness temperatures to create filters for the resolutionenhancementoftheGMIradiances.TheregressionprocedurefindsaclosefitoftheDPR-resolutionbrightness temperatureswhile limitingnoiseamplification,asinRobinsonetal.(1992);seeGrecuetal.(2016).

The enhancement filters are not applied directly to the GMI observations,

however. Instead, they are applied to the error between the observed GMIbrightness temperatures and convolved, DPR-resolution brightnesstemperatures from the Forward Model Module. The deconvolved errors arethenused to correct thesimulated,DPR-resolutionbrightness temperatures toestimatethemostlikely“observed”brightnesstemperaturesatDPRresolution.These DPR-resolution brightness temperatures are further constrained to bewithin the bounds of the ensemble of simulated DPR-resolution brightnesstemperatures, and so the ensemble provides a physical constraint on theresolutionenhancementoftheGMIobservations.

In the Filter Module (Fig. 2a), the Ku-reflectivity-consistent ensembles ofprofilesof environmental andprecipitationPSDparametersareupdated tobeconsistent with the additional information provided by the Ka-bandreflectivities,PIA’s,andresolution-enhancedGMIbrightnesstemperatures.Note

13

thatthePIAKaestimateisnotuseddirectly,butrathertherain-affecteddelta-PIA(PIAKa – PIAKu), which is estimated from the difference of delta-PIA at thefootprint of interest and the background of Ka-Ku surface radar cross-sectiondifferences innon-raining footprints, isutilized. The rain-affecteddelta-PIA isprovidedintheoutputoftheRadarAlgorithmLevel2,anditislesscorruptedbynoise in the background surface cross-sections than traditional PIA estimatesderivedfortheindividualradarchannels.

In the filter update procedure, any physical variable associated with the

forwardmodelcouldbeincludedinthevectorofunknowns,orstatevariables.However,forthepurposeofcomputationalefficiency,onlyselectedvariablesareupdated. Included in the state vector are the profiles ofwater vapor,qv, andcloud liquid,qcld,andthe10-mwindspeedoverwatersurfaces,U10,aswellasprofileof the logarithmof the interceptof thePSD, log10(Nw). Othervariablesincludedaretheprofilesofprecipitationrate,R,andprecipitationwatercontent,LWC. In addition, for the purpose of testing the consistency of the brightnesstemperaturesassociatedwiththeupdatedprofilesrelativetotheGMIbrightnesstemperatures, the DPR-resolution brightness temperatures, TBsim, are alsoupdated. Theprofileofprecipitationmedianvolumediameters,Dm, isnotpartof the state vector because they are derived analytically from the Kureflectivitiesandaprioriinformation.

Anensemblefilteroperatorisconstructedbyfirstcomputingthecovariances

betweenthestatevariablesandthesimulatedZKa,PIA=[PIAKu,delta-PIA]T,andGMIbrightnesstemperatures,TB.Thecovariancesbetweentheunknownsandsimulated observations, combined with the actual observations, are used toupdatetheensembleofKu-bandsolutions.LetXi=[qv-iqcld-i,U10-i,log10(Nw-i),Ri,LWCi,TBsim-i ]Tbeavectorof theunknownparameters in the ithensemblememberat agivenDPR footprint location. Also, letY = [ZKa PIA,TB]Tbeavector of observed Ka-band reflectivities, path-integrated attenuations, andresolution-enhancedGMIobservations,allatagivenDPRfootprintlocation. Inaddition, let H(Xi) the simulation of the observations Y from the unknownparametersof the ithensemblemember. Then,at eachDPR footprint locationandforeachensemblemember,theEnFupdate, Xi’, is given by

, (4)

where

, (5)

, (6)

and

!Xi = Xi + PHT HPHT +R( )−1Y −H Xi( )#$ %&

PHT =1

N −1Xi −X( )

i =1

N

∑ H Xi( ) −H X( )#$

%&T

HPHT =1

N −1H Xi( ) −H X( )"#

$%

i =1

N

∑ H Xi( ) −H X( )"#

$%T

14

. (7)

Here, PHT is the ensemble covariance of the unknown parameters and thesimulated observations, HPHT is the ensemble covariance of the simulatedobservations,Risamatrixoftheuncertaintiesintheobservations,andNisthenumber of ensemble members. Estimated PIA from the surface referencetechniqueareprovidedbytheRadarAlgorithmLevel2output.

Equations(4)-(7)aresimilartoEqs.(8)-(11)inGrecuandOlson(2008),butappliedtothecombinedradar-radiometerestimationproblem.Thepurposeoftheupdateistomodifytheenvironmentalandprecipitationprofileparametersfrom the Ku-band solution to the extent that the observed ZKa, PIA, and TBcontain additional information. If the observed ZKa are not available or aretotallyattenuatedbyheavyrain,oriftherainfallislight(Rayleighregime,whereZKu=ZKa),thentheKu-bandsolutionwillnotbemodifiedbytheZKa,althoughthePIAandTBdatacanstillalterthesolution.

Equation (4) is applied to all 40 ensemble member profiles at each DPRfootprint location, creating an output ensemble of µ, Nw, and Dm profiles,environmental parameters, and estimated brightness temperatures that areconsistentwith theobserved reflectivitiesandpath-integratedattenuations, aswellastheobservedbrightnesstemperatures. TheoutputsoftheFilterModuleare ensembles of environmental/precipitation parameters and brightnesstemperatures consistent with both the DPR and GMI observations and theirerrors. The best estimate of the environmental/precipitation parameters andbrightnesstemperaturesatanyDPRlocationisgivenbythemeanofthefilteredensemble,

, (8)

and the uncertainty of the best estimate is given by the ensemble standard deviation,

. (9)

The uncertainty, sX, represents the error in the best estimate resulting fromerrorsintheobservationsaswellasambiguitiesduetothelimitedinformationcontentoftheobservations.

R =

WZKa0 0

0 WPIA 00 0 WTB

!

"

###

$

%

&&&

X =1N

!Xii =1

N

∑

sX =1

N −1"Xi −X( ) "Xi −X( )

T

i=1

N

∑$

%&

'

()

12

15

Finally,forthepurposeoftestingtheconsistencyofthesolutionbrightnesstemperatures and the GMI observed brightness temperatures, the solutionbrightness temperatures are convolved by theGMI antenna patterns to createsolutiontemperaturesatGMIresolution.

Oncetheoutputof theFilterModuleandtheconvolvedsolutionbrightnesstemperaturesissavedtodisk,thenext300scanlinesofDPRandcoincidentGMIdataareprocessed,andsoon,untilthewholeorbitisprocessed.

Inthefollowingsubsections,thebasicfunctionsofthefourprimarymodulesof the Combined Radar-Radiometer Algorithm, including input and outputparameters, are described. Supporting modules, datasets, and tables aredescribedinsection4.

Ku Radar Module

TheKuRadarModuleacceptsinputprecipitationdetectionandbrightbanddetection parameters, atmospheric environmental parameters, aswell as DPRKu reflectivity information. Its primary function is to estimate ensembles ofenvironmentalandprecipitationparametersconsistentwiththeseinputdataateachDPRfootprintlocation,usingthegeneralizedHitschfeld-Bordanapproach.Specifically, Radar Algorithm Level 2 input to the Ku Radar Module areprecipitation detection information and calibrated Ku-band reflectivities fromthePreparationModule(PRE),brightbanddetectionandconvective/stratiformclassification data from the Classification Module (CSF), and environmentalatmospheric pressure and temperature profiles from the EnvironmentModule(ENV).

TheKuRadarModulealsodrawsupontabulatedgaseous/cloudabsorption

coefficients and single-scattering parameters that have been pre-computed forthepurposeofalgorithmefficiency.TablesofgaseousabsorptioncoefficientsatKu and Ka bands, as well as the GMI channel frequencies, are currentlycalculated as functions of pressure, temperature, and vapor density. Tables ofcloud water/ice absorption coefficients at the same frequencies are currentlycalculatedasfunctionsoftemperatureandequivalentliquidwatercontent;seesection4fordetails.

Because precipitation of all phases produces scattering as well asabsorption/emission of microwaves, and since the particle size distribution,phase, and temperature of precipitation determine its bulk scattering andabsorption/emissioncharacteristics,separatedatabasesareusedtotabulatethesingle-scattering properties of precipitation. In these tables, values ofreflectivity, extinction coefficient, scattering coefficient, and asymmetryparameter are tabulated as functions of µ, Dm, and log10(Nw), which define anormalizedgammadistributionofprecipitationparticlesizes.Ice,liquidwater,andmixed-phaseparticlesarerepresentedinthetables.Therefore,theKu-bandextinctionandreflectivityofprecipitationcanbeaccessedfromthetablesinthe

16

KuRadarModule.Ensembles of Ku-band consistent profiles of µ, Dm, and log10(Nw) are

estimatedanalyticallyintheKuRadarModuleusingthegeneralizedHitschfeld-Bordanmethod. Theeffectsofbeamfillingarerepresentedbydownscalingtheoriginal observed Ku reflectivity profile to a set of lognormally-distributedreflectivityprofiles, andapplying thegeneralizedHitschfeld-Bordanmethod toeachprofileoftheset.

The output of the Radar Module are ensembles of pressure, temperature,vapor density, cloud water content, µ, Nw, and Dm, precipitation rate andprecipitationwatercontentprofiles,consistentwitheachDPRprofileofKu-bandreflectivitiesandtheiruncertainties.

Forward Model Module

The Forward Model Module accepts input ensembles of profiles ofenvironmental/precipitationparameters fromtheKuRadarModule,aswellassurfaceskintemperatureandemissivityinformation. ItsprimaryfunctionistoperformforwardradiativecalculationsofKa-bandreflectivities,path-integratedattenuations at Ku and Ka bands, and brightness temperatures at the GMIchannelfrequencies/polarizations,foreachensemblememberprofile.

In addition to the inputs from the Ku Radar Module, the Forward Model

Module utilizes the static gaseous/cloud absorption and scattering tablesdescribed in the previous subsection. Based upon the input ensembles ofpressure, temperature, vapor density, cloud water, µ, Nw, and Dm fields, theForward Model Module derives the corresponding fields of single-scatteringparametersforeachensemblememberattheDPRandGMIchannelfrequenciesusing these tables. In addition, the surface skin temperature from the RadarAlgorithm Level 2 Environment Module (ENV) and 10-meter wind speedassociatedwiththeensemblesareusedtocalculatethewatersurfaceemissivity,based upon the model of Meissner andWentz (2012). Over other surfaces,emissivitiesarecurrentlyspecifiedusingthegeographicdatabasedevelopedbyDr.S.Munchak(personalcomm.;2017);seesection4.

Theatmospherictemperatureandsingle-scatteringproperties,aswellasthe

surface skin temperature and emissivities, at each DPR footprint location areused to simulate theKa-band reflectivitiesandpath-integratedattenuationsatKu and Ka bands, and they are also input to a radiative transfer model tocalculatetheupwellingmicrowavebrightnesstemperaturesatthatlocation.Toaccount for beamfilling effects within the radar footprint, the sets ofprecipitationprofilesderivedfromthedownscaledKu-reflectivityprofileintheKuRadarModuleareusedtocreatecorrespondingsetsofKareflectivityprofilesthat are averaged (or “upscaled”) to theDPR footprint resolution. Eddington'sSecondApproximation,whichaccountsformultiple-scatteringeffectsbutwhich

17

isalsocomputationally-efficient,isutilizedtocalculatetheupwellingbrightnesstemperatures; see Kummerow (1993). However, the delta scaling of bulkscatteringparameters thataccounts forstrong forwardscatteringby ice-phaseprecipitationisalsoapplied;seeJosephetal.(1976).

The Forward Model Module passes along the ensembles of the profiles of

pressure, temperature, vapor density, cloud water content, µ, Nw, and Dm,generated by the Ku Radar Module. To these profiles, it adds ensembles ofsurface skin temperature, 10-meter wind speed (over water surfaces) andassociatedensemblesofsimulatedKa-bandreflectivityprofiles,path-integratedattenuations at Ku and Ka bands, and surface emissivities and upwellingbrightnesstemperaturesattheGMIchannelfrequencies/polarizations.

RadianceEnhancementModule

The Radiance Enhancement Module receives convolved, simulatedbrightness temperature ensembles from then Forward Model Module andobservedGMIbrightnesstemperaturesfromtheRadiometerAlgorithmLevel1Casinputs.Itestimatesnear-DPR-resolutionGMIbrightnesstemperaturesusinga regression-based filter, and outputs these enhanced-resolution brightnesstemperatures. It also passes along the ensembles of profiles of pressure,temperature, vapor density, cloud water content, µ,Nw, andDm, precipitationratesandprecipitationwatercontents,surfaceskintemperaturesand10-meterwindspeeds(overwatersurfaces),andsimulationsofsurfaceemissivitiesandupwellingbrightnesstemperatures,generatedbypreviousModules.FilterModule

TheFilterModuleingeststheensemblesofenvironmentalandprecipitationparameters as well as the simulated Ka-band reflectivities, path-integratedattenuations at Ku and Ka bands, and surface emissivities and upwellingbrightness temperatures at theGMI channel frequencies/polarizations. It alsorequiresobservedKa-bandreflectivities,path-integratedattenuationsatKuandKa bands, as well as the resolution-enhanced GMI observations from theRadiance Enhancement Module, as input. Its primary function is to filter theprofileensemblestocreateupdatedprofileensemblesthatareconsistentwithall valid DPR and GMI observations, and their uncertainties. Based on theupdated ensembles, it also computes “best estimates” of the precipitationparametersandtheiruncertainties.

Specifically, the Filter Module uses the covariances between the input

ensemble profiles and the simulated observations from the Forward ModelModule, andcombines thesewithactualobservationsofKa-band reflectivities,path-integratedattenuationsatKuandKabands,andresolution-enhancedGMIbrightness temperatures to perform an ensemble filter update of the inputensembles of precipitation parameter profiles. The brightness temperatureensembles are also filtered for the purpose of evaluating the fitting of the

18

simulated brightness temperatures to the observed brightness temperatures.Ka-bandreflectivitiesaredrawnfromtheRadarAlgorithmLevel2Preparation(PRE)Module, and Ku/Ka path-integrated attenuation observations are takenfrom theRadarAlgorithmLevel2SurfaceReferenceTechnique (SRT)Module.GMI brightness temperatures are drawn from the Radiance EnhancementModule output of observed brightness temperature data that have beenresolution-enhanced to near-DPR resolution. Details of the ensemble filteringmethodmaybefoundinGrecuandOlson(2008),Grecuetal.(2011),andGrecuetal.(2016).

Output of the Filter Module includes best estimates of the environmental

parameters,suchasprofilesofvapordensityandcloudliquidwater,aswellasestimates of 10-meter wind speed. With respect to precipitation, profiles ofprecipitation size distribution parameters (µ, Nw, Dm), precipitation rate andwater content, aswell assurfaceprecipitation rateareestimated. Inaddition,the fractions of liquid in the profiles of precipitation water content andprecipitation rate and surface precipitation rate are output. Estimates of theuncertainties of the precipitation rates and water contents, as well asuncertaintiesof surfaceprecipitation rates, areproduced. Alsooutputareby-products of the estimation method, including profiles of corrected radarreflectivity factor at Ku and Ka bands, and the estimated path-integratedattenuation at Ku and Ka bands and surface emissivities/brightnesstemperaturesattheGMIchannelfrequencies/polarizations. 4. Ancillary Datasets

In the current algorithm formulation, only the Analysis Data, describedbelow, are ingested from an external source during Combined Algorithmprocessing.Theotherdatabasesandtablesarestaticandarereadintomemoryupontheexecutionofthealgorithmsoftware.GeographicData

Ageographicdatabasecontainingwatercoverageandelevationinformationat5kmresolutionisrequiredbytheCombinedAlgorithm;seesection3.Waterfractions in the database are derived from the Moderate Resolution ImagingSpectroradiometer (MODIS)250mresolution land-watermask, andelevationsare derived from the Shuttle Radar Topography Mission 30" (SRTM30) data,bothre-projectedtotheNASALandInformationSystem(LIS)1kmgrid.Finally,the 1 km resolution data are averaged to 5 km tomatch the resolution of theDPR.

Analysis Data

Analysis data are required to produce initial estimates of environmentalparameterssuchaspressure,temperature,andsurfaceskintemperature. Thecurrent algorithmdesign requires space-time interpolation of these data fromthe Japanese Meteorological Agency’s (JMA) global analyses (GANAL) during

19

standard algorithm processing. The data are interpolated to the DPRfootprint/range bin locations and overpass times in the Environment Module(ENV) of the Radar Algorithm Level 2 and then output. For near real-timeprocessing, the JMA analysis is supplemented with JMA forecast fields, but ifthesefieldsarenotreceivedintimeforanyreason,Japanese25-yearRe-analysis(JRA-25) data are substituted for the JMA analysis/forecast data in the ENVprocessing.

DataSupportingtheSpecification of Environmental Parameters

Atmosphericvapordensityandcloudwater contentprofilesutilized in thecreationoftheaprioriprofileensemblesarederivedfromrandomcombinationsof EOF’s that are based upon cloud-system-resolving model simulations. TheEOF’s are drawn from Weather Research and Forecasting model (WRF,Michalakesetal.2001)simulationsrepresentingdiversemeteorologicalsystems(e.g.mid-latitude cyclones, tropical convection, etc.). Cloud ice is currently notrepresentedintheCombinedAlgorithm.

Water surface emissivities are modeled as a function of surface skin

temperature and 10-meter wind speed over water surfaces (Meissner andWentz 2012. These emissivities are empirically related to surface normalizedradar cross sections byMunchak et al. (2016). Land surface emissivities and

Fig.3.Meancolumnwatervapor(upperleft)andsurfaceemissivitiesat10GHz(upperright),37GHz(lowerleft),and166GHz(lowerright)inthehorizontalpolarizationforthemonthofDecember,derivedbyDr.S.Munchak(personalcomm.,2017).Missingdatainhighcloudinessareasareduetoalackofanadequateuncontaminatedradiancesampleinthoseareas.

20

normalized radar cross sections are derived from the geographic empiricaldatabase developed by Dr. S. Munchak (personal comm.; 2017). Examples ofmonthly-meanlandsurfaceemissivitiesinthehorizontalpolarizationareshowninFig.3.

Microwave Absorption and Single-Scattering Tables

Two types of tables are produced by the GPM Radar Algorithm andCombinedAlgorithmTeams.Thefirsttabletypecontainsmicrowaveabsorptioncoefficients for atmospheric gaseous constituents indexed by pressure,temperature, and humidity, and also cloud water/ice absorption coefficientsindexed by temperature and liquid-equivalent cloud water content. Tables ofgaseousabsorptioncoefficientsatKuandKabands,aswellastheGMIchannelfrequencies,arecurrentlycalculatedasfunctionsofpressure,temperature,andvapordensityusingthemodeldescribedinRosenkranz(1998).Tablesofcloudwater/ice absorption coefficients at the same frequencies are currentlycalculatedasfunctionsoftemperatureandequivalentliquidwatercontentusingMietheory.

Because precipitation of all phases induces scattering as well asabsorption/emission of microwaves, and since the particle size distribution,

Fig. 4. Graphical illustrations of Ku-band (top) and Ka-band (bottom) scattering table entries for radar reflectivities within the melting layer. Color plots show the variation of reflectivity (given Nw) with both Do and distance below the 0oC level. Line plots show the same information for specific depths below the 0oC level.

21

phase, and temperature of precipitation determine its bulk (size distributionintegrated)scatteringandabsorption/emissioncharacteristics,separatetablesareusedtostorethesingle-scatteringpropertiesofprecipitation. Thissecondtable type contains bulk reflectivities, extinction coefficients, scatteringcoefficients,andasymmetryparametersofprecipitationatKuandKabands,aswellastheGMIchannelfrequencies.Thebulkscatteringparametersarederivedby integrating the single-scattering properties of precipitation particles overassumed gamma distributions of particles, indexed by µ,Nw, and Dm, whichdefine the normalized gamma distribution. Separate tables are generated forprecipitation at different temperatures. In addition, for mixed-phaseprecipitation instratiformregions, tablesarealsocreated fordifferentverticaldisplacementsrelativetothefreezinglevel.

To estimate the single-scattering properties of raindrops, all drops are

assumedtobesphericalliquidparticles,andMietheoryisappliedtoobtainthescattering properties of the drops. Mixed-phase particles are assumed to beconcentricshellsofice,air,andliquidwater,forwhichtheice-airdensityis0.1gcm-3. Theproportionof liquidwater ineachshell isassumedto increasewithradius within the particle according to an analytical formula (Liao andMeneghini, 2005), summing to a prescribed total melt fraction. In stratiformregions the total melt fraction is determined using a thermodynamic meltingsimulation; see Yokoyama andTanaka (1984),while in convective regions themelt fraction is assumed tovary linearly from0at the topof themixed-phaseregionto1attheitsbase.Multi-shellelectromagnetictheoryisusedtocomputethesingle-scatteringpropertiesofthemixed-phaseparticles;seeWuandWang(1991).Agraphicalillustrationofentriesforthemixed-phaseprecipitationtypeis shown in Fig. 4 for Ku- and Ka-band reflectivities of melting snow particledistributionsatvariousdepthsbelowthefreezinglevel.Ice-phaseprecipitationparticles are assumed to be nonspherical, computationally simulated asaggregates of pristine crystals, as described in Kuo et al. (2016). The single-scattering properties of the nonspherical ice particles are derived using thediscretedipoleapproximation;seeDraineandFlatau(1994).

5. Summary of Algorithm Input/Output

Input to the CombinedAlgorithm isderived fromRadarAlgorithmLevel 2(2AKu, 2AKuENV, 2ADPR) products and from theRadiometerAlgorithmLevel1CGMI(1CGMI)product,aswellastheancillarydatasetsdescribedinsection4.

TheoutputoftheCombinedAlgorithmisthe2BCMBproduct,whichcontains

twoswathsofdata.PrecipitationestimatesintheNSoutputswatharederivedfrom coincident DPR Ku band reflectivities/PIA's and GMI brightnesstemperatures,andtheseestimateswillextendacrosstheentireKubandswath.PrecipitationestimatesintheMSoutputswatharederivedfromcoincidentDPRKu-andKa-bandreflectivities/PIA'sandGMIbrightnesstemperatures.TheMS

22

precipitation estimates are therefore limited to theDPRKu-Ka overlap swath.SincetheMSswathestimatesdrawuponthemaximumamountof informationfrom the GPM Core sensors, these estimates will be the main tool for cross-calibrating the GPM constellation radiometer precipitation estimates throughthe creation of a priori databases. The same Combined Algorithm softwarearchitectureisusedtocreateboththeNSorMSoutputswaths.

A complete listing of Combined Algorithm input/output parameters is

included in Appendix A. Output volumes and algorithm processingrequirementsareincludedinAppendixBandC,respectively.

6. Algorithm Testing Plan

Prior to the GPM Core Observatory launch, testing of the CombinedAlgorithm or Algorithm components fell into three categories: SensitivityTesting, Physics Testing, and Pre-launch Validation. After GPM Core launch,Sensitivity Testing and Physics Testing continued, and Pre-launch ValidationactivitiesevolvedintoPost-launchValidation.

Sensitivity Testing

Sensitivity tests basically quantify the impact of different algorithmmodificationsonoutputproducts. So, it ispossible that even if theCombinedAlgorithminvestigatorssuspectthataparticularalgorithmmodificationshouldhaveasignificantimpactbaseduponpreviousworkorintuitivereasoning,thatmodificationmayactuallyhavelittleimpactoranimpactthatwasnotforeseen.Sensitivity testing can therefore be used to prioritize or re-focus areas ofalgorithmdevelopmentbaseduponthespecificimpactalgorithmmodificationshaveonoutput.

23

Input data for sensitivity testing can vary depending on the algorithmmodification.Typically,weuseKubandradardatafromeitherairbornesensorsor TRMM observations to synthesize Ka band radar reflectivity and GMIbrightness temperature observations. In this way we can ensure physicalconsistencybetweentheradarandradiometerchannelsusingthesameforwardmodeling assumptions. The following are some ongoing and planned areasofsensitivitytesting:•impactofchangesinalgorithmmethodology•impactofdifferentinputdata(e.g.,radarvs.radar-radiometer)•impactofPSDdescription•impactofassumedenvironmental/precipitationparametercorrelations•impactofparticlescatteringassumptions•impactofprecipitationphasetransitionassumptions•impactoflandsurfacecharacterizationandphysicalparameterizations•impactofsourceofancillary(analysis)data•impactofradarnon-uniformbeamfillingassumptions•impactofradarmultiple-scatteringeffects

Thislistisbynomeansexhaustive.

Anexampleofsensitivity testing is thestudyof the impactofDPRandGMIinputdata on CombinedAlgorithm estimates relative to estimates based uponthe DPR input data alone. Shown in Fig. 5 are 2D histograms (similar toscatterplots) of estimated surface rainfall rates from algorithm applications to

Fig. 5. Shown in the top panels are 2D histograms of synthetic retrievals of surface rain rate vs. reference rain rates, using only Ku band input data (left panel) and using both Ku band and TMI input data (right panel). The bottom panels are the same, except that the mean of the initial guess Nw values is assumed to be 4x the mean of the actual reference values.

24

synthetic data versus the known reference rainfall rates. Only the Ku bandreflectivitydataareutilizedinthealgorithmapplicationsillustratedontheleft-handsideofthefigure,whilebothKubandandTMIbrightnesstemperaturedataare utilized in the algorithm applications illustrated on the right-handside. Ifthe initial guess ensemble ofNw values has the same mean as the referencevalues, then the addition of the TMI observations reduces the scatter ofestimated rainfall rates relative to the reference values. If the initial guessensembleofNwvaluesisbiasedbyafactorof4relativetothereferencevalues,thereisagreaterbiascorrectionifboththeKubandandTMIobservationsareutilized.

Sensitivity testing isexpectedtobea long-termactivity thatwillhelpusto

improveourunderstandingoftheCombinedAlgorithm'sresponsetoavarietyofpotentialmodifications.

Physics Testing

The objective of Physics Testing is to verify assumptions in the forwardmodels that relate environmental/precipitation parameters to sensorobservations. Since the currentTRMMandGPMradarsand radiometershaverelatively low resolution, field campaign observations from airborne andground-based instrumentation generally provide superior data for physicstesting. These data may include remote sensing radar and radiometerobservationsaswellasinsitumeasurements.

Ongoingorplannedareasofphysicstestinginclude•assessmentofappropriatephysicalmodelsforice-phaseprecipitation•assessmentofproperparameterizationsfortheice-to-liquidphasetransition•assessmentofappropriatephysicalmodelsformixed-phaseprecipitation•assessmentofphysicalmodelsforlandsurfaceemissivities•assessmentofparameterizationsfordescribinginhomogeneityofprecipitationwithintheradarfootprint•assessmentofphysicalparameterizationsformultiplescatteringofradarpulses

25

Fig. 6 illustrates a precursor of physics testing to determine appropriatephysical parameterizations of ice-phase precipitation in the CombinedAlgorithm.Inthistest,differentphysicalmodelsareassumedfordescribingthesingle-scattering properties of ice-phase precipitation in a precipitationestimation algorithm. The estimation algorithm is applied to Dual-frequencyAirborne Precipitation Radar (APR-2) Ku and Ka band observations from the

Genesis and Rapid Intensification Processes (GRIP) field campaign. Thereflectivityobservations (not shown) indicatea stratiformprecipitation regionwithembeddedconvectiveelements.Itisevidentfromthefigurethatthechoiceof ice-phaseprecipitation scatteringmodelhasan impacton theestimationofthe precipitationmass-weighted particle diameter, water content, and rainfallrate.

During theMidlatitude Continental Convective Clouds Experiment (MC3E),the GPM Cold-season Precipitation Experiment (GCPEX), the IntegratedPrecipitation and Hydrology Experiment (IPHEx), and the Olympic MountainsExperiment (OLYMPEX), airborne dual-frequency radar observations arecoupledwithcoincidentmicrowavebrightnesstemperaturemeasurementsandin situ microphysics probe observations of precipitation. The precipitationalgorithm will be applied to these combined radar-radiometer data to see if

Fig. 6. Top row are estimated mass-weighted mean diameter, DM, liquid water content, LWC, and rain rate, R, based upon airborne APR-2 Ku and Ka band radar data. Ice-phase precipitation is assumed to be spherical snow particles with mixed ice-air dielectric properties. Shown in the second row are the differences between Ku/Ka band estimates assuming spherical, mixed-dielectric graupel particles and mixed-dielectric snow particles. Shown in the third, fourth, and fifth rows are the differences between estimates assuming structured snow particles composed of multiple dendritic flakes and spherical mixed-dielectric snow particles.

26

reasonablefitstothedatacanbeachieved,andtoseeifagreementwithinsituobservations is possible, given different assumed ice scatteringparameterizations.

Physics testing is expected to be a long-term activity that will help us to

improveourunderstandingoftheCombinedAlgorithm'sforwardmodelanditsuncertainties. Pre-launch Validation

Prior to the GPM Core Observatory launch, the Combined Algorithm wasexamined to determine (a) in TRMM applications, how Combined Algorithmprecipitation estimates compared to TRMMV7 Algorithm estimates andwell-calibrated ground-based radar estimates, and (b) whether or not CombinedAlgorithm estimates would be expected to meet the GPM Level-1 ScienceRequirements. It is stated in the Science Requirements that the algorithmshould be capable of estimating instantaneous surface rainfall rates at 50 kmresolutionwithabiasandrandomerrorwithin50%at1mmh-1rainrateandwithin25%at10mmh-1rainrate.

With respect to TRMM applications, the Combined Algorithm performance

should be similar towhatmight be expected outside theGPMKa bandswath,whereonlyKubandandGMIbrightnesstemperatureobservationsareavailable.ShowninFig.7arepreliminarycomparisonsofsurfacerainrateestimatesfromthe TRMM V7 Radar Algorithm (2A25), the TRMM V7 Combined Algorithm(2B31), and theGPM prototype Combined Algorithm (EnF). Note the generalsimilarity between the rain rate estimates, which is attributed to the similarphysicalbasisofallthreealgorithms.

Althoughcoincidentground-basedradarobservationswerenotavailablefortheprecipitationsystemsshowninFig.7,wellcalibratedradarwithraingagessited within the radar observing domain were available from the Melbourne,FloridaandKwajaleinAtoll,RepublicofMarshallIslandsgroundvalidationsites;seeWolffetal.(2005).BothsitesfeatureS-band,nearlynon-attenuatingradars;theKPOLradaratKwajaleinispolarimetricandtheMelbourneWSR-88DradarwasupgradedtopolarimetricinJanuary,2012.Polarimetriccapabilityhelpstoimprove quality control of the data and provide more definitive rain rateestimates. ThesehighqualitygroundvalidationestimateswillbecomparedtoCombined Algorithm estimates to provide evidence for whether or not theScienceRequirementswillbemetbythealgorithm.

27

Inaddition,theGPMValidationNetwork(VN)andMulti-RadarMulti-Sensor(MRMS)quantitativeprecipitationestimationproductformerlycalledNMQ,willprovide large-scale coverage of three dimension reflectivity distributions andsurface rainfall rates, respectively, for identifying locally large discrepanciesbetween satellite and ground-based measurements. Both products are basedupontheUSNEXRADradarnetwork.TheVNnetworkcoversthecontinentalUSand several sites outside theUS, and these reflectivity data are geometrically"matched" to reflectivity observations from the TRMM PR and GPM DPR; seeSchwallerandMorris(2011).OfparticularinterestishowwelltheattenuationcorrectedreflectivitiesfromtheCombinedAlgorithmagreewiththeVNS-bandreflectivities, that are essentially unaffected by attenuation. To do theseintercomparisons properly, a smallMie correction is applied to the Combinedestimates of attenuation-corrected Ku band reflectivities; see Liao andMeneghini (2009). The NOAAMRMS product is a 1 km, 5 minute resolutioninstantaneous rain rate product derived from the National Weather ServiceNEXRAD radar network and Environment Canada radars in lower Canada, aswellascoincidentraingages;seeKirstetteretal.(2012).LiketheVN,theMRMSdatahavetheadvantageofbroadspatialcoverage,withobservationsextendingoverthecontinentalUSstartingin2006.

Fig. 7. Top row, estimated surface rainfall rates in Tropical Cyclone Floyd at 09 UTC 13 September 1999 from the TRMM V7 2A25 (radar-only), TRMM V7 2B31 (radar-radiometer) and TRMM EnF (prototype GPM radar-radiometer) algorithms. Bottom row, same as top row but TRMM estimates are for a wintertime cold frontal band over the Eastern Pacific Ocean at 00 UTC 19 February 2001.

28

Post-launch Validation NowthattheGPMCoreObservatoryhasbeenputintoorbitandcalibration

adjustments are nearly complete, ground validation of Combined Algorithmestimates of precipitation continues using data from well calibrated andmonitoredground radardata fromsites suchasKwajalein,Dallas/FortWorth,and Houston. These sites will be augmented with observations from NASA'sNPOLradarandsupportingraingagenetworkatNASAWallopsFlightFacility,aswellashigher-latitudesites inFinland,Canada,andS.Korea. TheUSNEXRADNetwork has been completely upgraded to dual-polarization, leading toimprovedaccuracyofquantitativeprecipitationestimates from theMRMSandVN. Validationstrategiesaremuchthesameas thoseusedtoevaluateTRMM-basedalgorithmestimatesinthePre-launchera;seetheprecedingsubsection.

ShowninFig.8arescatterplotsof instantaneous,0.5oresolutionCombined

AlgorithmV5 estimates of surface precipitation rate vs.MRMS gage-calibratedradar precipitation rates over the continentalUS from the period Sep. 2014 –Aug. 2015. The MRMS precipitation rates are matched to the CombinedAlgorithm resolution footprints (5 km resolution) and binned with theCombinedAlgorithmestimates in0.5o x0.5oboxes. Note that there is a slightincrease in the correlationofCombinedAlgorithmMSestimatesrelative toNSestimates.

Global distributions of Combined Algorithm MS mode (Ku+Ka+GMI)

estimates of surface precipitation rate and Global Precipitation ClimatologyProject(GPCP)estimatesandtheirdifferences,alongwithcorrespondingzonalmeanrainratedistributions,fortheperiodSep.2014–Aug.2015,areshowninFig. 9. Note that the CombinedAlgorithm andGPCPdistributions are similar,

Fig.8.Scatterplotsofinstantaneous,0.5oresolutionNSmode(Ku+GMI;atleft)andMSmode(Ku+Ka+GMI;atright)estimatesofsurfaceprecipitationratesvs.MRMSraingage-calibratedradarestimatesfortheperiodSep.2014–Aug.2015.ThecorrelationofNSmodeprecipitationratesis0.85,whilethecorrelationofMSmodeestimatesis0.86.

29

with themost obvious differences at the high latitudes. At high latitudes, theweaksignaloflow-density,ice-phaseprecipitationmaynotbedetectablebytheDPR, and alternative strategies for estimating precipitation with the radar-radiometer combination are now being pursued for those cold regimes.Fluctuations of differences at lower latitudes are partly due to samplingdifferences of the narrow-swath MS product and the GeoIR-based GPCP, butsomebiasesofMSoverlandareevident.Thesebiasesarethesubjectofcurrentalgorithminvestigations.

Metrics

AsshowninOlsonetal. (2006), theuseof instantaneous,50kmresolutionrain rate estimates represents a reasonable compromise in resolution forevaluating algorithm errors. Although validation at resolutions down to5 km(the nominal DPR resolution) may be attempted, the influence of satellite vs.groundradarcollocationerrorsdegradesanyderivedstatistics.Accumulationofthe50-kmestimatesoverseasonscanrevealspatialpatternsofbias(relativetoreferenceestimates)thatarehelpfulfordiagnosingsystematicalgorithmerrors.

Standard bivariate statistics of the "errors" between Combined Algorithm

rainestimatesandground-basedestimates,suchasthemeanerror(bias),errorstandard deviation, and correlation coefficient will be computed, generallystratifiedbytherainintensity. These statistics can be used to assess the

Fig.9.GlobalmeanprecipitationdistributionsfromtheCombinedAlgorithmMSmode(Ku+Ka+GMI;upperleft)andGPCP(upperright),theMS–GPCPdifference(lowerleft),aswellaszonalmeans(lowerright),fortheperiodSep.2014–Aug.2015.

30

algorithm's performance relative to the Level 1 Science Requirements, forexample.However,informationmorerelevanttoalgorithmimprovementcanbegainedifthestatisticsarestratifiedbyothervariablesthatindicatespecificstatedependenciesoftheerror.Forexample,aretheerrorsfunctionsofsurfaceskintemperature,totalprecipitablewater,orothervariablesthatindicatetheclimateregime,or,aretheerrorsfunctionsoftheprecipitationsystemortype,suchasthe convective/stratiform class? Stratification of statistics to reveal statedependencies of algorithm error will be important for diagnosing algorithmdeficiencies,especiallyintheearlyphasesoftheGPMCoremission.

InadditiontomeetingtheL1ScienceRequirements,ameasureofthesuccess

ofoureffortwillbethequantificationofanyimprovementsincombinedradar-radiometerestimatesrelativetoradar-onlyestimateswithintheGPMCombinedAlgorithmframework.

7. References Aires, F., C. Prigent, F. Bernardo, C. Jimenez, R. Saunders, and P. Brunel, 2011: A

Tool to Estimate Land Surface Emissivities in the Microwave (TELSEM) for use in numerical weather prediction. Quart. J. Royal Meteorol. Soc., 137, 690-699.

Anderson, J. L., 2003: A local least squares framework for ensemble filtering. Mon.

Wea. Rev., 131, 634-642. Battaglia, A., S. Tanelli, K. Mroz, and F. Tridon, 2015: Multiple scattering in

observations of the GPM dual-frequency precipitation radar: Evidence and impact on retrievals. J. Geophys. Res. Atmos., 120, 4090-4101.

Draine, B. T., and P. J. Flatau, 1994: Discrete-dipole approximation for scattering

calculations. J. Opt. Soc. Am. A, 11, 1491-1499. Grecu, M., and W. S. Olson, 2008: Precipitating snow retrievals from combined

airborne cloud radar and millimeter-wave radiometer observations. J. Appl. Meteor. and Climatol., 47, 1634-1650.

Grecu, M., L. Tian, W. S. Olson, and S. Tanelli, 2011: A robust dual-frequency radar

profiling algorithm. J. Appl. Meteor. and Climatol., 50, 1543-1557. Grecu, M., W. S. Olson, S. J. Munchak, S. Ringerud, L. Liao, Z. S. Haddad, B. L.

Kelley, and S. F. McLaughlin, 2016: The GPM Combined Algorithm. J. Atmos. and Oceanic Tech., 33, 2225-2245.

Hogan, R. J., and A. Battaglia, 2008: Fast lidar and radar multiple-scattering models.

Part II: Wide-angle scattering using the time-dependent two-stream approximation. J. Atmos. Sci., 65, 3636-3651.

31

Joseph,J.H.,W.J.Wiscombe,andJ.A.Weinman,1976:Thedelta-Eddington

approximationforradiativefluxtransfer.J.Atmos.Sci.,33,2452-2459.Kirstetter,P.-E.,Y.Hong,J.J.Gourley,S.Chen,Z.Flamig,J.Zhang,M.Schwaller,W.

Petersen,andE.Amitai,2012:TowardaframeworkforsystematicerrormodelingofspaceborneprecipitationradarwithNOAA/NSSLgroundradar-basednationalmosaicQPE.J.Hydrometeorol.,13,1285-1300.

Kummerow, C., 1993: On the accuracy of the Eddington Approximation for radiative

transfer in the microwave frequencies. J. Geophys. Res. – Atmos., 98, 2757-2765. Kuo, K.-S., W. S. Olson, B. T. Johnson, M. Grecu, L. Tian, T. L. Clune, B. H. van

Aartsen, A. J. Heymsfield, L. Liao, and R. Meneghini, 2016: The microwave radiative properties of falling snow derived from nonspherical ice particle models. Part I: An extensive database of simulated pristine crystals and aggregate particles, and their scattering properties. J. Appl. Meteor. and Climatol., 55, 691-708.

Liao, L., and R. Meneghini, 2005: On modeling air/spaceborne radar returns in the

melting layer. IEEE Trans. Geosci. Remote Sens., 43, 2799 - 2809. Liao, L., and R. Meneghini, 2009: Changes in the TRMM Version-5 and Version-6

Precipitation Radar products due to orbit boost. J. Met. Soc. Japan, 87A, 93-107. Meissner, T., and F. J. Wentz, 2012: The emissivity of the ocean surface between 6

and 90 GHz over a range of wind speeds and earth incidence angles. IEEE Trans. Geosci. Remote Sens., 50, 3004-3026.

Meneghini, R., T. Iguchi, T. Kozu, L. Liao, Ken’Ichi Okamoto, J. A. Jones, and J.

Kwiatkowski, 2000: Use of the Surface Reference Technique for path attenuation estimates from the TRMM Precipitation Radar. J. Appl. Meteor., 39, 2053-2070.

Michalakes, J., S. Chen, J. Dudhia, L. Hart, J. Klemp, J. Middlecoff, and W.

Skamarock, 2001: Development of a next generation regional weather research and forecast model. Developments in Teracomputing: Proceedings of the Ninth ECMWF Workshop on the Use of High Performance Computing in Meteorology, W. Zwieflhofer and N Kreitz, Eds., World Scientific, 269-276.

Munchak, S. J., R. Meneghini, M. Grecu, and W. S. Olson, 2016: A consistent

treatment of microwave emissivity and radar backscatter for retrieval of precipitation over water surfaces. J. Atmos. Oceanic Tech., 33, 215-229.

Olson, W. S., C. D. Kummerow, S. Yang, G. W. Petty, W.-K. Tao, T. L. Bell, S.

A. Braun, Y. Wang, S. E. Lang, D. E. Johnson and C. Chiu. 2006: Precipitation and latent heating distributions from satellite passive microwave radiometry. Part

32

I: Improved method and uncertainties. J. Appl. Meteor. and Climatol., 45, 702–720.

Prigent, C., F. Aires, and W. B. Rossow, 2006: Land surface microwave emissivities

over the globe for a decade. Bull. Amer. Met. Soc., 87, 1573-1584. Robinson, W. D., C. Kummerow, and W. S. Olson, 1992: A technique for enhancing

and matching the resolution of microwave measurements from the SSM/I instrument. IEEE Trans. Geosci. Remote Sens., 30, 419-429.

Rosenkranz, P. W., 1998: Water vapor microwave continuum absorption: A

comparison of measurements and models. Radio Sci., 33, 919-928. Schwaller, M. R., and K. R. Morris, 2011: A ground validation network for the

Global Precipitation Measurement Mission. J. Atmos. Oceanic Tech., 28, 301-319.

Testud, J., S. Oury, P. Amayenc, and R. A. Black, 2001: The concept of normalized

distributions to describe raindrop spectra: A tool for cloud physics and cloud remote sensing. J. Appl. Meteor., 40, 1118–1140.

Thompson, E. J., S. A. Rutledge, B. Dolan, and M. Thurai, 2015: Drop size

distributionsandradarobservationsofconvectiveandstratiformrainovertheequatorialIndianandWestPacificOceans.J.Atmos.Sci.,72,4091-4125.

Wolff, D. B., D. A. Marks, E. Amitai, D. S. Silberstein, B. L. Fisher, A. Tokay, J.

Wang, and J. L. Pippitt, 2005: Ground validation for the Tropical Rainfall Measuring Mission (TRMM). J. Atmos. Oceanic Tech., 22, 365-380.

Wu, Z. S., and Y. P. Wang, 1991: Electromagnetic scattering for multilayered

sphere: Recursive algorithms. Radio Sci., 26, 1393-1401. Yokoyama, T., and H. Tanaka, 1984: Microphysical process of melting snowflakes

detected by two-wavelength radar. J. Met. Soc. Japan, 62, 650-666.

33

Appendix A. Listing of Input/Output Parameters Input Parameters ThislistingincludesonlythoseparametersthatwillbeingestedfromGPMAlgorithmsthatareexternaltotheCombinedRadar-RadiometerAlgorithmcode.Thegivenarraysizeargumentscorrespondto:nscan=numberofDPRscanspergranule,approximately7900nray=49raysofNormalSwath(NS)KubanddatanrayMS=25raysofMatchedSwath(MS)KabanddatanrayHS=24raysofHighsensitivitySwath(HS)Kabanddatanbin=176rangebinsofNSorMSradardataperraynbinHS=88rangebinsofHSKabanddataperraymethod=6methodsforestimatingpath-integratedattenuationfromtheSRTnNode=5binnodesidentifiedintheradar-definedDSDstructurenwind=2windcomponents:u,vnwater=2watervaporandcloudliquidwaterprofileparametersnscan1=numberoflower-frequency(S1)GMIscansinthegranule,approx.2954nscan2=numberofhigher-frequency(S2)GMIscansinthegranule,approx. 2954npixel1=221pixelsperlower-frequency(S1)GMIscannpixel2=221pixelsperhigher-frequency(S2)GMIscannchannel1=9channelsoflower-frequency(S1)GMIdataperpixelnchannel2=4channelsofhigher-frequency(S2)GMIdataperpixelnchUIA1=1numberoflower-frequency(S1)uniqueincidenceanglesnchUIA2=1numberofhigher-frequency(S2)uniqueincidenceangles

from the 2AKu Radar Algorithm Year

yearoftheKuscan(2byteinteger,nscan);fromthe2AKuRadarAlgorithm.

Month monthoftheKuscan(1byteinteger,nscan);fromthe2AKuRadar

Algorithm.DayOfMonth

dayofmonthoftheKuscan(1byteinteger,nscan);fromthe2AKuRadarAlgorithm.

Hour houroftheKuscan(1byteinteger,nscan);fromthe2AKuRadar

Algorithm.

34

MinuteminuteoftheKuscan(1byteinteger,nscan);fromthe2AKuRadarAlgorithm.

Second secondoftheKuscan(1byteinteger,nscan);fromthe2AKuRadar

Algorithm.MilliSecond

millisecondoftheKuscan(2byteinteger,nscan);fromthe2AKuRadarAlgorithm.

DayOfYear

dayoftheyearoftheKuscan(2byteinteger,nscan);fromthe2AKuRadarAlgorithm.

SecondOfDay

secondofthedayoftheKuscan(8bytefloat,nscan);fromthe2AKuRadarAlgorithm.

Latitude

latitudeofKufootprint(4bytefloat,nrayxnscan);fromthe2AKuRadarAlgorithm.

Longitude

longitudeofKufootprint(4bytefloat,nrayxnscan);fromthe2AKuRadarAlgorithm.

elevation

altitudeabovetheEarthellipsoidofthesurfacegateinKuray(4bytefloat,nrayxnscan);fromthe2AKuRadarAlgorithm(PREModule).

landSurfaceType

water/land/coastandsurfacetypeatKufootprintlocation(4byteinteger,nrayxnscan);fromthe2AKuRadarAlgorithm(PREModule).Notethatthisinformationisusedtointerpretsurfacereferencetechniqueoutput.

localZenithAngle

localincidenceanglesofDPRrayrelativetolocalzenithontheEarthellipsoid(4bytefloat,nrayxnscan);fromthe2AKuRadarAlgorithm(PREModule).

flagPrecip

flagindicatingdetectionofprecipitationornoprecipitationinKuray(4byteinteger,nrayxnscan);fromthe2AKuRadarAlgorithm(PRE

35

Module).binRealSurface

surfacerangebininKuray(2byteinteger,nrayxnscan);fromthe2AKuRadarAlgorithm(PREModule).

binStormTop

rangebinofstormtopinKuray(2byteinteger,nrayxnscan);fromthe2AKuRadarAlgorithm(PREModule).

heightStormTop

altitudeofstormtopinKuray(4bytefloat,nrayxnscan);fromthe2AKuRadarAlgorithm(PREModule).

binClutterFreeBottom

rangebinofthelowestclutter-freebinofKuray(2byteinteger,nrayxnscan);fromthe2AKuRadarAlgorithm(PREModule).

sigmaZeroMeasured

measuredsurfacenormalizedradarbackscatteringcross-sectionatKu(4bytefloat,nrayxnscan);fromthe2AKuRadarAlgorithm(PREModule).

zFactorMeasured

measuredreflectivityatKu(2byteinteger,nbinxnrayxnscan);fromthe2AKuRadarAlgorithm(PREModule).

ellipsoidBinOffset

offsetalongKuraybetweenearthellipsoidandmidpointofsurfacerangebin(4bytefloat,nrayxnscan);fromthe2AKuRadarAlgorithm(PREModule).

binZeroDeg

rangebinofthezerodegreeisotherminKuray(2byteinteger,nrayxnscan);fromthe2AKuRadarAlgorithm(VERModule).

heightZeroDeg

altitudeofthezerodegreeisotherminKuray(4bytefloat,nrayxnscan);fromthe2AKuRadarAlgorithm(VERModule).

flagBB

flagindicatingthedetectionofabright-bandinKuray(4byteinteger,nrayxnscan);fromthe2AKuRadarAlgorithm(CSFModule).

binBBPeak

rangebinofthebright-bandmaximumreflectivity,ifdetected,inKuray(2byteinteger,nrayxnscan);fromthe2AKuRadarAlgorithm(CSF

36

Module).heightBB

altitudeofthebright-bandmaximumreflectivity,ifdetected,inKuray(4bytefloat,nrayxnscan);fromthe2AKuRadarAlgorithm(CSFModule).

qualityBB

qualityflagforbrightbanddetectioninKuray(4byteinteger,nrayxnscan);fromthe2AKuRadarAlgorithm(CSFModule).

type_Precip

classificationofprecipitationtypeinKuray(4byteinteger,nrayxnscan);fromthe2AKuRadarAlgorithm(CSFModule).

qualityTypePrecip

qualityofclassificationofprecipitationtypeinKuray(4byteinteger,nrayxnscan);fromthe2AKuRadarAlgorithm(CSFModule).

PIAalt

total2-waypath-integratedattenuationtothesurfacebaseduponsurfacereferencetechniquemethodsforKu(4bytefloat,methodxnrayxnscan);fromthe2AKuRadarAlgorithm(SRTModule).

RFactorAlt

reliabilityfactorsoftotal2-waypath-integratedattenuationestimatesbaseduponsurfacereferencetechniquemethodsforKu(4bytefloat,methodxnrayxnscan);fromthe2AKuRadarAlgorithm(SRTModule).

PIAweight

weightsofindividual2-waytotalpath-integratedattenuationestimatestoformeffectiveestimateforKu(4bytefloat,methodxnrayxnscan);fromthe2AKuRadarAlgorithm(SRTModule).

pathAtten

effectivetotal2-waypath-integratedattenuationtothesurfacebaseduponweightedaveragesofsurfacereferencetechniquemethodsforKu(4bytefloat,nrayxnscan);fromthe2AKuRadarAlgorithm(SRTModule).

reliabFactor

reliabilityfactorofeffectivetotal2-waypath-integratedattenuationestimatebaseduponsurfacereferencetechniquemethodsforKu(4bytefloat,nrayxnscan);fromthe2AKuRadarAlgorithm(SRTModule).

reliabFlag

reliabilityflagforcompositetotal2-waypath-integratedattenuation

37

estimatebaseduponsurfacereferencemethodsforKu(2byteinteger,nrayxnscan);fromthe2AKuRadarAlgorithm(SRTModule).

phase

particlephasebasedonKu(1byteinteger,nbinxnrayxnscan);fromthe2AKuRadarAlgorithm(DSDModule).

binNode

binnodeforpartitioningradarprofilebasedonKu(4byteinteger,nNodexnrayxnscan);fromthe2AKuRadarAlgorithm(DSDModule).

flagParticle

particleflagbasedonKu(1byteinteger,nbinxnrayxnscan);fromthe2AKuRadarAlgorithm(DSDModule).

from the 2AKuENV Radar Algorithm airTemperature

air temperature interpolated to Ku range bins (4 byte float, nbin x nray x nscan); from JMA data using the 2AKuENV Radar Algorithm (VER Module).

airPressure

air pressure interpolated to Ku range bins (4 byte float, nbin x nray x nscan); from JMA data using the 2AKuENV Radar Algorithm (VER Module).

waterVapor

water vapor density interpolated to Ku range bins (4 byte float, nwater x nray x nscan); from JMA data using the 2AKuENV Radar Algorithm (VER Module).

cloudLiquidWater

cloud liquid water content interpolated to Ku range bins (4 byte float, nwater x nray x nscan); from JMA data using the 2AKuENV Radar Algorithm (VER Module).

surfacePressure

surface air pressure interpolated to Ku footprint location (4 byte float, nray x nscan); from JMA data using the 2AKuENV Radar Algorithm (VER Module).

groundTemperature

surface skin temperature interpolated to Ku footprint location (4 byte float, nray x nscan); from JMA data using the 2AKuENV Radar Algorithm (VER Module).

surfaceWind

38

10-meter wind speed interpolated to Ku footprint location (4 byte float, nwind x nray x nscan); from JMA analysis using the 2AKuENV Radar Algorithm (VER Module).

from the 2ADPR Radar Algorithm (NS) Year

yearoftheKuscan(2byteinteger,nscan);fromthe2ADPRRadarAlgorithm(inNS).

Month

monthoftheKuscan(1byteinteger,nscan);fromthe2ADPRRadarAlgorithm(inNS).

DayOfMonth

dayofmonthoftheKuscan(1byteinteger,nscan);fromthe2ADPRRadarAlgorithm(inNS).

Hour

houroftheKuscan(1byteinteger,nscan);fromthe2ADPRRadarAlgorithm(inNS).

Minute

minuteoftheKuscan(1byteinteger,nscan);fromthe2ADPRRadarAlgorithm(inNS).

Second

secondoftheKuscan(1byteinteger,nscan);fromthe2ADPRRadarAlgorithm(inNS).

MilliSecond

millisecondoftheKuscan(2byteinteger,nscan);fromthe2ADPRRadarAlgorithm(inNS).

DayOfYear

dayoftheyearoftheKuscan(2byteinteger,nscan);fromthe2ADPRRadarAlgorithm(inNS).

SecondOfDay

secondofthedayoftheKuscan(8bytefloat,nscan);fromthe2ADPRRadarAlgorithm(inNS).

Latitude

latitudeofKufootprint(4bytefloat,nrayxnscan);fromthe2ADPRRadarAlgorithm(inNS).

Longitude

39

longitudeofKufootprint(4bytefloat,nrayxnscan);fromthe2ADPRRadarAlgorithm(inNS).

elevation

altitudeabovetheEarthellipsoidofthesurfacegateinKuray(4bytefloat,nrayxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinNS).

landSurfaceType

water/land/coastandsurfacetypeatKufootprintlocation(4byteinteger,nrayxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinNS).Notethatthisinformationisusedtointerpretsurfacereferencetechniqueoutput.

localZenithAngle

localincidenceanglesofDPRrayrelativetolocalzenithontheEarthellipsoid(4bytefloat,nrayxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinNS).

flagPrecip

flagindicatingdetectionofprecipitationornoprecipitationinKuray(4byteinteger,nrayxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinNS).

binRealSurface

surfacerangebininKuray(2byteinteger,nrayxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinNS).

binStormTop

rangebinofstormtopinKuray(2byteinteger,nrayxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinNS).

heightStormTop

altitudeofstormtopinKuray(4bytefloat,nrayxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinNS).

binClutterFreeBottom

rangebinofthelowestclutter-freebinofKuray(2byteinteger,nrayxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinNS).

sigmaZeroMeasured

measuredsurfacenormalizedradarbackscatteringcross-sectionatKu(4bytefloat,nrayxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinNS).

zFactorMeasured

40

measuredreflectivityatKu(2byteinteger,nbinxnrayxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinNS).

ellipsoidBinOffset

offsetalongKuraybetweenearthellipsoidandmidpointofsurfacerangebin(4bytefloat,nrayxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinNS).

binZeroDeg

rangebinofthezerodegreeisotherminKuray(2byteinteger,nrayxnscan);fromthe2ADPRRadarAlgorithm(VERModuleinNS).

heightZeroDeg

altitudeofthezerodegreeisotherminKuray(4bytefloat,nrayxnscan);fromthe2ADPRRadarAlgorithm(VERModuleinNS).

flagBB

flagindicatingthedetectionofabright-bandinKuray(4byteinteger,nrayxnscan);fromthe2ADPRRadarAlgorithm(CSFModuleinNS).

binBBPeak

rangebinofthebright-bandmaximumreflectivity,ifdetected,inKuray(2byteinteger,nrayxnscan);fromthe2ADPRRadarAlgorithm(CSFModuleinNS).

heightBB

altitudeofthebright-bandmaximumreflectivity,ifdetected,inKuray(4bytefloat,nrayxnscan);fromthe2ADPRRadarAlgorithm(CSFModuleinNS).

qualityBB

qualityflagforbrightbanddetectioninKuray(4byteinteger,nrayxnscan);fromthe2ADPRRadarAlgorithm(CSFModuleinNS).

type_Precip

classificationofprecipitationtypeinKuray(4byteinteger,nrayxnscan);fromthe2ADPRRadarAlgorithm(CSFModuleinNS).

qualityTypePrecip

qualityofclassificationofprecipitationtypeinKuray(4byteinteger,nrayxnscan);fromthe2ADPRRadarAlgorithm(CSFModuleinNS).

PIAalt

total2-waypath-integratedattenuationtothesurfacebaseduponsurfacereferencetechniquemethodsforKu(4bytefloat,methodxnrayxnscan);fromthe2ADPRRadarAlgorithm(SRTModuleinNS).

41

RFactorAlt

reliabilityfactorsoftotal2-waypath-integratedattenuationestimatesbaseduponsurfacereferencetechniquemethodsforKu(4bytefloat,methodxnrayxnscan);fromthe2ADPRRadarAlgorithm(SRTModuleinNS).

PIAweight

weightsofindividual2-waytotalpath-integratedattenuationestimatestoformeffectiveestimateforKu(4bytefloat,methodxnrayxnscan);fromthe2ADPRRadarAlgorithm(SRTModuleinNS).

pathAtten

effectivetotal2-waypath-integratedattenuationtothesurfacebaseduponweightedaveragesofsurfacereferencetechniqueestimatesforKu(4bytefloat,nrayxnscan);fromthe2ADPRRadarAlgorithm(SRTModuleinNS).

reliabFactor

reliabilityfactorofeffectivetotal2-waypath-integratedattenuationestimatebaseduponsurfacereferencetechniquemethodsforKu(4bytefloat,nrayxnscan);fromthe2ADPRRadarAlgorithm(SRTModuleinNS).

reliabFlag

reliabilityflagforeffectivetotal2-waypath-integratedattenuationestimatebaseduponsurfacereferencemethodsforKu(2byteinteger,nrayxnscan);fromthe2ADPRRadarAlgorithm(SRTModuleinNS).

phase

particlephasebasedonKu(1byteinteger,nbinxnrayxnscan);fromthe2ADPRRadarAlgorithm(DSDModuleinNS).

binNode

binnodeforpartitioningradarprofilebasedonKu(4byteinteger,nNodexnrayxnscan);fromthe2ADPRRadarAlgorithm(DSDModuleinNS).

flagParticle

particleflagbasedonKu(1byteinteger,nbinxnrayxnscan);fromthe2ADPRRadarAlgorithm(DSDModuleinNS).

from the 2ADPR Radar Algorithm (MS) Year

yearoftheKascan(2byteinteger,nscan);fromthe2ADPRRadarAlgorithm(inMS).

42

MonthmonthoftheKascan(1byteinteger,nscan);fromthe2ADPRRadarAlgorithm(inMS).

DayOfMonth

dayofmonthoftheKascan(1byteinteger,nscan);fromthe2ADPRRadarAlgorithm(inMS).

Hour

houroftheKascan(1byteinteger,nscan);fromthe2ADPRRadarAlgorithm(inMS).

Minute

minuteoftheKascan(1byteinteger,nscan);fromthe2ADPRRadarAlgorithm(inMS).

Second

secondoftheKascan(1byteinteger,nscan);fromthe2ADPRRadarAlgorithm(inMS).

MilliSecond

millisecondoftheKascan(2byteinteger,nscan);fromthe2ADPRRadarAlgorithm(inMS).

DayOfYear

dayoftheyearoftheKascan(2byteinteger,nscan);fromthe2ADPRRadarAlgorithm(inMS).

SecondOfDay

secondofthedayoftheKascan(8bytefloat,nscan);fromthe2ADPRRadarAlgorithm(inMS).

Latitude

latitudeofKafootprint(4bytefloat,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(inMS).

Longitude

longitudeofKafootprint(4bytefloat,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(inMS).

elevation

altitudeabovetheEarthellipsoidofthesurfacegateinKaray(4bytefloat,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinMS).

landSurfaceType

43

water/land/coastandsurfacetypeatKafootprintlocation(4byteinteger,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinMS).Notethatthisinformationisusedtointerpretsurfacereferencetechniqueoutput.

localZenithAngle

localincidenceanglesofDPRrayrelativetolocalzenithontheEarthellipsoid(4bytefloat,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinMS).

flagPrecip

flagindicatingdetectionofprecipitationornoprecipitationinKaray(4byteinteger,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinMS).

binRealSurface

surfacerangebininKaray(2byteinteger,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinMS).

binStormTop

rangebinofstormtopinKaray(2byteinteger,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinMS).

heightStormTop

altitudeofstormtopinKaray(4bytefloat,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinMS).

binClutterFreeBottom

rangebinofthelowestclutter-freebinofKaray(2byteinteger,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinMS).

sigmaZeroMeasured

measuredsurfacenormalizedradarbackscatteringcross-sectionatKa(4bytefloat,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinMS).

zFactorMeasured

measuredreflectivityatKa(2byteinteger,nbinxnrayMSxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinMS).

ellipsoidBinOffset

offsetalongKaraybetweenearthellipsoidandmidpointofsurfacerangebin(4bytefloat,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(PREModuleinMS).

binZeroDeg

44

rangebinofthezerodegreeisotherminKaray(2byteinteger,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(VERModuleinMS).

heightZeroDeg

altitudeofthezerodegreeisotherminKaray(4bytefloat,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(VERModuleinMS).

flagBB

flagindicatingthedetectionofabright-bandinKaray(4byteinteger,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(CSFModuleinMS).

binBBPeak

rangebinofthebright-bandmaximumreflectivity,ifdetected,inKaray(2byteinteger,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(CSFModuleinMS).

heightBB

altitudeofthebright-bandmaximumreflectivity,ifdetected,inKaray(4bytefloat,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(CSFModuleinMS).

qualityBB

qualityflagforbrightbanddetectioninKaray(4byteinteger,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(CSFModuleinMS).

type_Precip

classificationofprecipitationtypeinKaray(4byteinteger,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(CSFModuleinMS).

qualityTypePrecip

qualityofclassificationofprecipitationtypeinKaray(4byteinteger,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(CSFModuleinMS).

PIAalt

total2-waypath-integratedattenuationtothesurfacebaseduponsurfacereferencetechniquemethodsforKa(4bytefloat,methodxnrayMSxnscan);fromthe2ADPRRadarAlgorithm(SRTModuleinMS).

RFactorAlt

reliabilityfactorsoftotal2-waypath-integratedattenuationestimatesbaseduponsurfacereferencetechniquemethodsforKa(4bytefloat,methodxnrayMSxnscan);fromthe2ADPRRadarAlgorithm(SRTModuleinMS).

PIAweight

weightsofindividual2-waytotalpath-integratedattenuationestimates

45

toformeffectiveestimatesforKa(4bytefloat,methodxnrayMSxnscan);fromthe2ADPRRadarAlgorithm(SRTModuleinMS).

pathAtten

effectivetotal2-waypath-integratedattenuationtothesurfacebaseduponweightedaveragesofsurfacereferencetechniqueestimatesforKa(4bytefloat,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(SRTModuleinMS).

reliabFactor

reliabilityfactorofeffectivetotal2-waypath-integratedattenuationestimatebaseduponsurfacereferencetechniquemethodsforKa(4bytefloat,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(SRTModuleinMS).

reliabFlag

reliabilityflagforcompositetotal2-waypath-integratedattenuationestimatebaseduponsurfacereferencemethodsforKa(2byteinteger,nrayMSxnscan);fromthe2ADPRRadarAlgorithm(SRTModuleinMS).

phase

particlephasebasedonKa(1byteinteger,nbinxnrayMSxnscan);fromthe2ADPRRadarAlgorithm(DSDModuleinMS).

binNode

binnodeforpartitioningradarprofilebasedonKa(4byteinteger,nNodexnrayMSxnscan);fromthe2ADPRRadarAlgorithm(DSDModuleinMS).

flagParticle

particleflagbasedonKa(1byteinteger,nbinxnrayMSxnscan);fromthe2ADPRRadarAlgorithm(DSDModuleinMS).

from the 1CGMI Algorithm (S1) Year

yearoftheGMIscan(2byteinteger,nscan1)fromthe1CGMIAlgorithm(inS1).

Month

monthoftheGMIscan(1byteinteger,nscan1)fromthe1CGMIAlgorithm(inS1).

DayOfMonth

dayofmonthoftheGMIscan(1byteinteger,nscan1)fromthe1CGMI

46

Algorithm(inS1).Hour

houroftheGMIscan(1byteinteger,nscan1)fromthe1CGMIAlgorithm(inS1).

Minute

minuteoftheGMIscan(1byteinteger,nscan1)fromthe1CGMIAlgorithm(inS1).

Second

secondoftheGMIscan(1byteinteger,nscan1)fromthe1CGMIAlgorithm(inS1).

MilliSecond

millisecondoftheGMIscan(2byteinteger,nscan1)fromthe1CGMIAlgorithm(inS1).

DayOfYear

dayofyearoftheGMIscan(2byteinteger,nscan1);fromthe1CGMIAlgorithm(inS1).

SecondOfDay

secondofdayoftheGMIscan(8bytefloat,nscan1);fromthe1CGMIAlgorithm(inS1).

Latitude

latitudeoftheGMIfootprint(4bytefloat,npixel1xnscan1);fromthe1CGMIAlgorithm(inS1).

Longitude

longitudeoftheGMIfootprint(4bytefloat,npixel1xnscan1);fromthe1CGMIAlgorithm(inS1).

Quality

qualityofthelower-frequencycalibratedbrightnesstemperatures(1byteinteger,npixel1xnscan1);fromthe1CGMIAlgorithm(inS1).

incidenceAngle

earthincidenceangleoftheGMIlower-frequencydata(4bytefloat,nchUIA1xnpixel1xnscan1);fromthe1CGMIAlgorithm(inS1).

sunGlintAngle

sunglintanglesoftheGMIlower-frequencydata(1byteinteger,nchUIA1xnpixel1xnscan1);fromthe1CGMIAlgorithm(inS1).

47

incidenceAngleIndexindexoftheincidenceanglearrayforeachlower-frequencychannel(1byteinteger,nchannel1xnscan1);fromthe1CGMIAlgorithm(inS1).

Tc

commoncalibratedGMIbrightnesstemperaturesinthelower-frequencydataswath(4bytefloat,nchannel1xnpixel1xnscan1);fromthe1CGMIAlgorithm(inS1).

from the 1CGMI Algorithm (S2) Year

yearoftheGMIscan(2byteinteger,nscan2)fromthe1CGMIAlgorithm(inS2).

Month

monthoftheGMIscan(1byteinteger,nscan2)fromthe1CGMIAlgorithm(inS2).

DayOfMonth

dayofmonthoftheGMIscan(1byteinteger,nscan2)fromthe1CGMIAlgorithm(inS2).

Hour

houroftheGMIscan(1byteinteger,nscan2)fromthe1CGMIAlgorithm(inS2).

Minute

minuteoftheGMIscan(1byteinteger,nscan2)fromthe1CGMIAlgorithm(inS2).

Second

secondoftheGMIscan(1byteinteger,nscan2)fromthe1CGMIAlgorithm(inS2).

MilliSecond

millisecondoftheGMIscan(2byteinteger,nscan2)fromthe1CGMIAlgorithm(inS2).

DayOfYear

dayofyearoftheGMIscan(2byteinteger,nscan2);fromthe1CGMIAlgorithm(inS2).

SecondOfDay

secondofdayoftheGMIscan(8bytefloat,nscan2);fromthe1CGMIAlgorithm(inS2).

48

Latitude

latitudeoftheGMIfootprint(4bytefloat,npixel2xnscan2);fromthe1CGMIAlgorithm(inS2).

Longitude

longitudeoftheGMIfootprint(4bytefloat,npixel2xnscan2);fromthe1CGMIAlgorithm(inS2).

Quality

qualityofthehigher-frequencycalibratedbrightnesstemperatures(1byteinteger,npixel2xnscan2);fromthe1CGMIAlgorithm(inS2).

incidenceAngle

earthincidenceanglesoftheGMIhigher-frequencydata(4bytefloat,nchUIA2xnpixel2xnscan2);fromthe1CGMIAlgorithm(inS2).

sunGlintAngle

sunglintanglesoftheGMIhigher-frequencydata(1byteinteger,nchUIA2xnpixel2xnscan2);fromthe1CGMIAlgorithm(inS2).

incidenceAngleIndex

indexoftheincidenceanglearrayforeachhigher-frequencychannel(1byteinteger,nchannel2xnscan2);fromthe1CGMIAlgorithm(inS2).

Tc

commoncalibratedGMIbrightnesstemperaturesinthehigher-frequencydataswath(4bytefloat,nchannel2xnpixel2xnscan2);fromthe1CGMIAlgorithm(inS2).

49

Output Parameters Notestandardoutputproductsaresampledat250mverticalresolution.Thegivenarraysizeargumentscorrespondto:nscan=numberofDPRscanspergranule,approximately7900nrayNS=49raysineachKuband(NS)scannrayMS=25raysineachmatchedKu-Ka(MS)scannbinC=88verticalrangebinsat250mintervalsnbinEnv=10rangebinsforenvironmentalparametersamplingnbinLow=9rangebinsforlow-resolutionPSDparametersamplingnbinPhase=5rangebinsindicatingphasetransitionsnbinTrans=10rangebinsdescribingtheprecipitationliquidphasefractionthroughthemixedphaselayernPSDhigh=1parametersfordescribingtheprecipitationparticlesizedistributionat250mresolutionnPSDlow=2parametersfordescribingtheprecipitationparticlesizedistributionatlowverticalresolution.nAB=2powerlawparameterstodescribeparticledensitiesnKuKa=2indicesfortheKuandKachannelsnchan=15GMIchannels,includingseparateaccountingforthe doubleside-bandchannels.fromthe2BCMBCombinedRadar-RadiometerAlgorithm(NS)Year

yearoftheKubandscan(2byteinteger,nscan);fromthe2BCMBCombinedAlgorithm(inNS/ScanTime).

Month

monthoftheKubandscan(1byteinteger,nscan);fromthe2BCMBCombinedAlgorithm(inNS/ScanTime).

DayOfMonth

dayofmonthoftheKubandscan(1byteinteger,nscan);fromthe2BCMBCombinedAlgorithm(inNS/ScanTime).

Hour

houroftheKubandscan(1byteinteger,nscan);fromthe2BCMBCombinedAlgorithm(inNS/ScanTime).

Minute

minuteoftheKubandscan(1byteinteger,nscan);fromthe2BCMBCombinedAlgorithm(inNS/ScanTime).

Second

secondoftheKubandscan(1byteinteger,nscan);fromthe2BCMB

50

CombinedAlgorithm(inNS/ScanTime).MilliSecond

millisecondoftheKubandscan(2byteinteger,nscan);fromthe2BCMBCombinedAlgorithm(inNS/Scantime).

DayOfYear

dayofyearoftheKubandscan(2byteinteger,nscan);fromthe2BCMBCombinedAlgorithm(inNS/ScanTime).

SecondOfDay

secondofdayoftheKubandscan(8bytefloat,nscan);fromthe2BCMBCombinedAlgorithm(inNS/ScanTime).

Latitude

latitudeofKubandfootprint(4bytefloat,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

Longitude

longitudeofKubandfootprint(4bytefloat,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

ioQuality

6-digitflagdescribingthequalityofinputdataandprecipitationestimate(4byteinteger,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS/FLG).

1’sdigit: 0ifestimateisvalid 9ifnoestimate 10’sdigit: 0ifKudatavalidandraindetectedatKu 1ifKudatavalidbutnoraindetected 9ifbadKuinputdata 100’sdigit: 0ifKuSRTgivesviablePIAestimate 1ifsoKuiswithinnoiseofbackground 2ifsoKuiscompletelyattenuated 9ifbadKuinputdata 1000sdigit: 0iffreezinglevelderivedfromKubrightband

1iffreezinglevelderivedfrommeteorologicalanalysis

9ifbadKuinputdata 10000’sdigit: 0ifKuclassifiedasstratiformorconvective 1ifKuclassifiedasindeterminate 2ifprecipnotdetectedatKu(nofeature) 9ifbadKuinputdata

100000’sdigit: 0ifatleastsomemeasuredbrightnesstemperaturesarevalid

51

9ifnomeasuredbrightnesstemperaturesarevalidmultiScatCalc

flagindicatingwhetherornotmultiple-scatteringradarcalculationsareutilizedatKaband(1)ornot(0),orthatnorainwasdetected-9999.AlthoughoutputinNSmode,thisvariableisonlyapplicabletoMSmode,andsoaflagvalueisoutputinNSmodel(inNS/FLG).

surfaceElevation

altitudeabovetheEarthellipsoidofthesurfacegateinKubandray(4bytefloat,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS/Input).

surfaceType

water/land/coastandsurfacetypeatKubandfootprintlocation(4byteinteger,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS/Input).

localZenithAngle

localincidenceangleofKubandrayrelativetolocalzenithontheEarthellipsoid(4bytefloat,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS/Input).

precipitationFlag

flagindicatingdetectionofprecipitationornoprecipitationinKubandray(4byteinteger,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS/Input).

surfaceRangeBin

surfacerangebininKubandray(2byteinteger,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS/Input).

lowestClutterFreeBin

lowestclutterfreebininKubandray(2byteinteger,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS/Input).

ellipsoidBinOffset

offsetalongKuraybetweenearthellipsoidandmidpointofsurfacerangebin(4bytefloat,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS/Input).

stormTopBin

stormtoprangebininKubandray(2byteinteger,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS/Input).

stormTopAltitude

52

stormtopaltitudeinKubandray(4bytefloat,nrayNSxnscan);fromthe 2BCMBCombinedAlgorithm(inNS/Input).zeroDegBin

freezinglevelbininKubandray(2byteinteger,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS/Input).

zeroDegAltitude

freezinglevelaltitudeinKubandray(4bytefloat,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS/Input).

precipitationType

classificationofprecipitationtypeinKubandray(4byteinteger,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS/Input).

precipTypeQualityFlag

qualityofclassificationofprecipitationtypeinKubandray(4byteinteger,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS/Input).

piaEffective

effectivetotal2-waypath-integratedattenuationtothesurfaceatKuband,baseduponweightedaveragesofsurfacereferencetechniqueestimates(4bytefloat,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS/Input).

piaEffectiveSigma

uncertaintyofeffectivetotal2-waypath-integratedattenuationatKubandbaseduponsurfacereferencetechniquemethods(4bytefloat,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS/Input).

piaEffectiveReliabFlag

reliabilityflagoftheeffectivetotal2-waypath-integratedattenuationatKuband,baseduponweightedaveragesofsurfacereferencetechniqueestimates(2byteinteger,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS/Input).

snowIceCoverflagindicatingthepresence(2)orabsence(1)ofsnowontheearth’ssurface(4-byteinteger,nrayNSxnscan).Oceanisindicatedby0.(inNS/Input)

surfaceAirPressure

surfaceairpressureattheKubandfootprintlocation(4bytefloat,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

53

surfaceAirTemperaturesurfaceairtemperatureattheKubandfootprintlocation(4bytefloat,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

surfaceVaporDensity

surfacevapordensityattheKubandfootprintlocation(4bytefloat,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

skinTemperature

surfaceskintemperatureattheKubandfootprintlocation(4bytefloat,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

envParamNode

binnumberswhereenvironmentalpressure,temperature,andvapordensityaresampled(2byteinteger,nbinEnvxnrayNSxnscan)from2BCMBCombinedAlgorithm(inNS).

airPressure

airpressurealongtheKubandrayattheenvParamNodelocations(4bytefloat,nbinEnvxnrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

airTemperature

airtemperaturealongtheKubandrayattheenvParamNodelocations(4bytefloat,nbinEnvxnrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

vaporDensity

vapordensityalongtheKubandrayattheenvParamNodelocations(4byteinteger,nbinEnvxnrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

cloudLiqWaterCont

cloudliquidwatercontentalongtheKubandrayat250msamplingresolution(4bytefloat,nbinCxnrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

cloudIceWaterCont

cloudiceliquid-equivalentwatercontentalongtheKubandrayat250msamplingresolution(4bytefloat,nbinCxnrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).Currentlythisparameterisnotassigned.

phaseBinNodesbinnumbersindicating(0)stormtop,(1)topofmixed-phaselayer,(2)maximumreflectivityinmixed-phaselayerifbrightbanddetected;

54

otherwise,thefreezinglevelfromanalysis,(3)bottomofmixed-phaselayer,and(4)bottomofrainlayer(2byteinteger,nbinPhasexnrayNSxnscan);from2BCMBCombinedAlgorithm(inNS).

liqMassFracTrans

fractionoftheprecipitationmassthatisliquidinthetransitionbetweeniceandliquid-phaseprecipitation,startingfromthetopofthemixed-phaselayer(phaseBinNode1)andproceedingdownwardalongtheKubandrayat250msamplingresolution(4bytefloat,nbinTransxnrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

liqRateFracTrans

fractionoftheprecipitationratethatisliquidinthetransitionbetweeniceandliquid-phaseprecipitation,startingfromthetopofthemixed-phaselayer(phaseBinNode1)andproceedingdownwardalongtheKubandrayat250msamplingresolution(4bytefloat,nbinTransxnrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

precipTotRateprecipitationratealongtheKubandrayat250msamplingresolution(4bytefloat,nbinCxnrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

precipTotRateSigma

uncertaintyofprecipitationratealongtheKubandrayat250msamplingresolution(4bytefloat,nbinCxnrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

precipTotWaterCont

precipitationwatercontentalongtheKubandrayat250msamplingresolution(4bytefloat,nbinCxnrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

precipTotWaterContSigma

uncertaintyofprecipitationwatercontentalongtheKubandrayat250msamplingresolution(4bytefloat,nbinCxnrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

precipTotPSDparamHigh

precipitationdrop-sizedistributionmeanvolumediameter(Dm)alongtheKubandrayat250msamplingresolution(4bytefloat,nPSDhighxnbinCxnrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

PSDparamLowNode

binnumberswherelow-resolutionprecipitationPSDparametersaresampled(2byteinteger,nbinLowxnrayNSxnscan)from2BCMB

55

CombinedAlgorithm(inNS).precipTotPSDparamLow

precipitationdrop-sizedistributionparameters(log(Nw),µ)alongtheKubandrayatreducedsamplingresolution(2byteinteger,nPSDlowxnbinLowxnrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

surfPrecipTotRate

surfaceprecipitationrateatKufootprintlocation(4bytefloat,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

surfPrecipTotRateSigma

surfaceprecipitationrateuncertaintyatKufootprintlocation(4bytefloat,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

surfLiqRateFrac

fractionofthesurfaceprecipitationratethatisliquidattheKufootprintlocation(4bytefloat,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

tenMeterWindSpeed

10-meterwindspeedatKufootprintlocation(4bytefloat,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

tenMeterWindSigma

estimateduncertaintyof10-meterwindspeedatKufootprintlocation(4bytefloat,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm.Estimatesareforwatersurfacesonly(inNS).

surfEmissivitymicrowavesurfaceemissivitiesattheGMIchannelfrequencies/polarizationsandviewingangleatKufootprintlocation(4bytefloat,nchanxnrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

surfEmissSigmaestimateduncertaintiesofmicrowavesurfaceemissivitiesattheGMIchannelfrequencies/polarizationsandviewingangleatKufootprintlocation(4bytefloat,nchanxnrayNSxnscan);fromthe2BCMBCombinedAlgorithm.Estimatesoverland,only(inNS).

simulatedBrightTemp

upwellingmicrowavesurfacebrightnesstemperaturesattheGMIchannelfrequencies/polarizationsandviewingangle(4bytefloat,nchanxnrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

56

pia

totalpath-integratedattenuationatKu(4bytefloat,nrayNSxnscan);fromthe2BCMBCombinedAlgorithm(inNS).

correctedReflectFactor

attenuation-correctedradarreflectivityfactoralongKurayat250msamplingresolution(4bytefloat,nbinCxnrayNSxnscan);fromthe2BCMBCombinedAlgorithm.

multiScatMaxContribthemaximumcontributionbymultiplescatteringtoareflectivitysimulation(4bytefloat,nrayNSxnscan)inthegivenradarprofile.AlthoughoutputinNS,multiple-scatteringisonlycalculatedforsimulatedKabandreflectivities,andsoaflagvalueisoutput(inNS).

nubfPIAfactorthefractionalreductionoftheHitschfeld-Bordanestimatedpath-integratedattenuationforsimulatingtheSurfaceReferenceTechniquepath-integratedattenuation(4bytefloat,nrayNSxnscan).Sincethenon-uniformbeamfillingfractionisonlycalculatedforKaband,aflagvalueisoutputforthisparameter(inNS).

from the 2BCMB Combined Radar-Radiometer Algorithm (MS) Year

yearoftheKabandscan(2byteinteger,nscan);fromthe2BCMBCombinedAlgorithm(inMS/ScanTime).

Month

monthoftheKabandscan(1byteinteger,nscan);fromthe2BCMBCombinedAlgorithm(inMS/ScanTime).

DayOfMonth

dayofmonthoftheKabandscan(1byteinteger,nscan);fromthe2BCMBCombinedAlgorithm(inMS/ScanTime).

Hour

houroftheKabandscan(1byteinteger,nscan);fromthe2BCMBCombinedAlgorithm(inMS/ScanTime).

Minute

minuteoftheKabandscan(1byteinteger,nscan);fromthe2BCMBCombinedAlgorithm(inMS/ScanTime).

Second

57

secondoftheKabandscan(1byteinteger,nscan);fromthe2BCMBCombinedAlgorithm(inMS/ScanTime).

MilliSecond

millisecondoftheKabandscan(2byteinteger,nscan);fromthe2BCMBCombinedAlgorithm(inMS/Scantime).

DayOfYear

dayofyearoftheKabandscan(2byteinteger,nscan);fromthe2BCMBCombinedAlgorithm(inMS/ScanTime).

SecondOfDay

secondofdayoftheKabandscan(8bytefloat,nscan);fromthe2BCMBCombinedAlgorithm(inMS/ScanTime).

Latitude

latitudeofKabandfootprint(4bytefloat,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

Longitude

longitudeofKabandfootprint(4bytefloat,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

ioQuality

6-digitflagdescribingthequalityofinputdataandprecipitationestimate(4byteinteger,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS/FLG).Currently,thisisjustareplicaoftheNSmodeioQualityflag.

1’sdigit: 0ifestimateisvalid 9ifnoestimate 10’sdigit: 0ifKudatavalidandraindetectedatKu 1ifKudatavalidbutnoraindetected 9ifbadKuinputdata 100’sdigit: 0ifKuSRTgivesviablePIAestimate 1ifsoKuiswithinnoiseofbackground 2ifsoKuiscompletelyattenuated 9ifbadKuinputdata 1000sdigit: 0iffreezinglevelderivedfromKubrightband

1iffreezinglevelderivedfrommeteorologicalanalysis

9ifbadKuinputdata 10000’sdigit: 0ifKuclassifiedasstratiformorconvective 1ifKuclassifiedasindeterminate 2ifprecipnotdetectedatKu(nofeature) 9ifbadKuinputdata

100000’sdigit: 0ifatleastsomemeasuredbrightnesstemperatures

58

arevalid 9ifnomeasuredbrightnesstemperaturesarevalidmultiScatCalc

flagindicatingwhetherornotmultiple-scatteringradarcalculationsareutilizedatKaband(1)ornot(0),orthatnorainwasdetected-9999(4byteinteger,nrayMSxnscan).Location(inMS/FLG)

surfaceElevation

altitudeabovetheEarthellipsoidofthesurfacegateinKabandray(4bytefloat,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS/Input).

surfaceType

water/land/coastandsurfacetypeatKabandfootprintlocation(4byteinteger,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS/Input).

localZenithAngle

localincidenceangleofKabandrayrelativetolocalzenithontheEarthellipsoid(4bytefloat,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS/Input).

precipitationFlag

flagindicatingdetectionofprecipitationornoprecipitationinKabandray(4byteinteger,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS/Input).

surfaceRangeBin

surfacerangebininKabandray(2byteinteger,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS/Input).

lowestClutterFreeBin

lowestclutterfreebininKabandray(2byteinteger,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS/Input).

ellipsoidBinOffset

offsetalongKaraybetweenearthellipsoidandmidpointofsurfacerangebin(4bytefloat,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS/Input).

stormTopBin

stormtoprangebininKabandray(2byteinteger,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS/Input).

stormTopAltitude

59

stormtopaltitudeinKabandray(4bytefloat,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS/Input).

zeroDegBinfreezinglevelbininKabandray(2byteinteger,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS/Input).

zeroDegAltitude

freezinglevelaltitudeinKabandray(4bytefloat,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS/Input).

precipitationType

classificationofprecipitationtypeinKabandray(4byteinteger,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS/Input).

precipTypeQualityFlag

qualityofclassificationofprecipitationtypeinKabandray(4byteinteger,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS/Input).

piaEffective

effectivetotal2-waypath-integratedattenuationtothesurfaceatKuandKabands,baseduponweightedaveragesofsurfacereferencetechniqueestimates(4bytefloat,nKuKaxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS/Input).

piaEffectiveSigma

uncertaintyofeffectivetotal2-waypath-integratedattenuationatKuandKabandsbaseduponsurfacereferencetechniquemethods(4bytefloat,nKuKaxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS/Input).

piaEffectiveReliabFlag

reliabilityflagoftheeffectivetotal2-waypath-integratedattenuationatKuandKabands,baseduponweightedaveragesofsurfacereferencetechniqueestimates(2byteinteger,nKuKaxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS/Input).

snowIceCoverflagindicatingthepresence(2)orabsence(1)ofsnowontheearth’ssurface(4-byteinteger,nrayMSxnscan).Oceanisindicatedby0.(inMS/Input)

surfaceAirPressure

surfaceairpressureattheKabandfootprintlocation(4bytefloat,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

60

surfaceAirTemperature

surfaceairtemperatureattheKabandfootprintlocation(4bytefloat,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

surfaceVaporDensitysurfacevapordensityattheKabandfootprintlocation(4bytefloat,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

skinTemperature

surfaceskintemperatureattheKabandfootprintlocation(4bytefloat,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

envParamNode

binnumberswhereenvironmentalpressure,temperature,andvapordensityaresampled(2byteinteger,nbinEnvxnrayMSxnscan)from2BCMBCombinedAlgorithm(inMS).

airPressure

airpressurealongtheKabandrayattheenvParamNodelocations(4bytefloat,nbinEnvxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

airTemperature

airtemperaturealongtheKabandrayattheenvParamNodelocations(4bytefloat,nbinEnvxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

vaporDensity

vapordensityalongtheKabandrayattheenvParamNodelocations(4byteinteger,nbinEnvxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

cloudLiqWaterCont

cloudliquidwatercontentalongtheKabandrayat250msamplingresolution(4bytefloat,nbinCxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

cloudIceWaterCont

cloudiceliquid-equivalentwatercontentalongtheKabandrayat250msamplingresolution(4bytefloat,nbinCxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).Currentlythisparameterisnotassigned.

phaseBinNodesbinnumbersindicating(0)stormtop,(1)topofmixed-phaselayer,(2)

61

maximumreflectivityinmixed-phaselayerifbrightbanddetected;otherwise,thefreezinglevelfromanalysis,(3)bottomofmixed-phaselayer,and(4)bottomofrainlayer(2byteinteger,nbinPhasexnrayMSxnscan);from2BCMBCombinedAlgorithm(inMS).

liqMassFracTrans

fractionoftheprecipitationmassthatisliquidinthetransitionbetweeniceandliquid-phaseprecipitation,startingfromthetopofthemixed-phaselayer(phaseBinNode1)andproceedingdownwardalongtheKabandrayat250msamplingresolution(4bytefloat,nbinTransxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

liqRateFracTrans

fractionoftheprecipitationratethatisliquidinthetransitionbetweeniceandliquid-phaseprecipitation,startingfromthetopofthemixed-phaselayer(phaseBinNode1)andproceedingdownwardalongtheKubandrayat250msamplingresolution(4bytefloat,nbinTransxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

precipTotRate

precipitationratealongtheKabandrayat250msamplingresolution(4bytefloat,nbinCxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

precipTotRateSigma

uncertaintyofprecipitationratealongtheKabandrayat250msamplingresolution(4bytefloat,nbinCxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

precipTotWaterCont

precipitationwatercontentalongtheKabandrayat250msamplingresolution(4bytefloat,nbinCxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

precipTotWaterContSigma

uncertaintyofprecipitationwatercontentalongtheKabandrayat250msamplingresolution(4bytefloat,nbinCxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

precipTotPSDparamHigh

precipitationdrop-sizedistributionmeanvolumediameter(Dm)alongtheKabandrayat250msamplingresolution(4bytefloat,nPSDhighxnbinCxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

PSDparamLowNode

binnumberswherelow-resolutionprecipitationPSDparametersare

62

sampled(2byteinteger,nbinLowxnrayMSxnscan)from2BCMBCombinedAlgorithm(inMS).

precipTotPSDparamLow

precipitationdrop-sizedistributionparameters(log(Nw),µ)alongtheKabandrayatreducedsamplingresolution(2byteinteger,nPSDlowxnbinLowxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

surfPrecipTotRate

surfaceprecipitationrateatKafootprintlocation(4bytefloat,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

surfPrecipTotRateSigma

surfaceprecipitationrateuncertaintyatKafootprintlocation(4bytefloat,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

surfLiqRateFrac

fractionofthesurfaceprecipitationratethatisliquidattheKafootprintlocation(4bytefloat,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

tenMeterWindSpeed

10-meterwindspeedatKafootprintlocation(4bytefloat,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

tenMeterWindSigmaestimateduncertaintyof10-meterwindspeedatKafootprintlocation(4bytefloat,nrayMSxnscan);fromthe2BCMBCombinedAlgorithm.Estimatesareforwatersurfacesonly(inMS).

surfEmissivitymicrowavesurfaceemissivitiesattheGMIchannelfrequencies/polarizationsandviewingangleatKafootprintlocation(4bytefloat,nchanxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

surfEmissSigmaestimateduncertaintiesofmicrowavesurfaceemissivitiesattheGMIchannelfrequencies/polarizationsandviewingangleatKafootprintlocation(4bytefloat,nchanxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm.Estimatesoverland,only(inMS).

simulatedBrightTemp

upwellingmicrowavesurfacebrightnesstemperaturesattheGMIchannelfrequencies/polarizationsandviewingangle(4bytefloat,nchan

63

xnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).pia

totalpath-integratedattenuationatKuandKabands(4bytefloat,nKuKaxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

correctedReflectFactor

attenuation-correctedradarreflectivityfactoratKuandKabandsat250msamplingresolution(4bytefloat,nKuKaxnbinCxnrayMSxnscan);fromthe2BCMBCombinedAlgorithm(inMS).

multiScatMaxContribthemaximumcontributionbymultiplescatteringtoareflectivitysimulation(4bytefloat,nrayNSxnscan)inthegivenradarprofile(inMS).

nubfPIAfactorthefractionalreductionoftheHitschfeld-Bordanestimatedpath-integratedattenuationforsimulatingtheSurfaceReferenceTechniquepath-integratedattenuation(4bytefloat,nrayMSxnscan).Location(inNS).

Appendix B. Output Product Volumes

The volume of the Combined Algorithm output product, based upon theoutput parameters listed above, is approximately 150 MB per orbit file ininternallycompressedHDF5format.

Appendix C. Processing Requirements

ThecurrentconfigurationoftheCombinedAlgorithmrequiresinputfromsixmodulesoftheGPMRadarAlgorithm:thePreparationModule,VerticalProfileModule, Classification Module, DSD Module, Surface Reference TechniqueModule,andtheEnvironmentModule.OutputofthesemodulesisexpectedfromtheLevel2RadarAlgorithmsoftware;however,thecomputationalrequirementsofthisroutinecouldaddsignificantlatencytoCombinedAlgorithmprocessing.

A primary input to the Vertical Profile Module is JMA global analyses

(GANAL) for standard processing and JMA global analyses/forecasts for near-real-time processing. Therefore, this input should be accommodated in PPSoperations.

64

Thecurrentversionofthealgorithmrequiresapproximately15minutesonasingle PPS processing node to process a typical orbit of data, using multi-processor capability. Parallel processingwill be achieved using POSIX threadlibraries. Although some economies in the coding can reduce this processingtime,theymayresultindegradedperformanceofthealgorithm.GranulesofDPRandGMIdatawillbesubdividedintoroughly30(300DPRscanline)segmentstokeeprequiredmemorywithinlimitsonagivenprocessingnode.

AppendixD.VersionChangesV03toV04Changes

Many updates have beenmade to the Combined Algorithm Level 2 in thetransitionfromV03toV04,andthesignificantupdatesaresummarizedhere.Itmay be noted at the outset, however, that the basic algorithmmechanics (i.e.,estimation methodology) and output file structure have not changed. Theestimation method filters ensembles of DPR Ku reflectivity-consistentprecipitation profiles using the DPR Ka reflectivities, path integratedattenuations at Ku and Ka bands, and GMI radiances. The filtered profileensembles are consistentwith all of the observations and their uncertainties,andthemeanofthefilteredensemblegivesthebestestimateoftheprecipitationprofile.

In the Combined Algorithm V03 and V04, input data are passed from the

Radar Algorithm Level 2 and Radiometer Algorithm Level 1C. However, toobtainbetterresponsivenessofprecipitationprofileestimatestotheGMIdatainV04, inputradiancesare firstresolution-enhancedtoapproximately thespatialresolutionoftheDPRresolution(~5km).Thisenhancementisaccomplished,ateachchannel frequencyandpolarization,usingastatisticallyderived filter thatpredicts the DPR-resolution radiance from a weighted average of native-resolution GMI radiances in a small neighborhood of the observation to beenhanced. Filter weights are derived from regressions on synthetic radiancedata,andthedegreeofenhancementistradedagainstnoiseamplification,withan optimal balance between enhancement and noise determined by cross-validation. Use of the resolution-enhanced data leads to a greaterresponsiveness of precipitation estimates to the GMI radiometer data, and abetterfittingofthosedata.Moreover,datafromallthirteenoftheGMIchannelsareutilized intheV04CMBalgorithm,whereasdata fromonlysevenchannelswereusedintheV03algorithm.

In V03, the impact of multiple scattering on simulated reflectivities wascrudely represented by typical reflectivity corrections (relative to single-scatteringcalculations)asfunctionsofbulkscatteringopticaldepth.Thissimplecorrectionofreflectivitiesisreplaced inV04bythe fullsimulationofmultiple-scatteringaffectedreflectivitiesusingthe1Dtime-dependentradiativetransfermodel of Hogan and Battaglia (2008). This model is fully invoked only insituations where single- and multiple-scattering reflectivity simulations based

65

upon the ensemble-mean, Ku-consistent precipitation profile are significantlydifferent,inwhichcasethemultiple-scatteringmodelisappliedtoallensemblemember profiles to simulate the Ka reflectivities. The impact of multiplescatteringonKureflectivities isgenerallymuchsmaller thanatKabandand isnotconsideredinV04.

The general parameterization of the effects of radar footprint non-uniformbeamfilling by precipitation is the same in Combined Algorithm V03 to V04;however,theimpactofnon-uniformbeamfillingonsimulationsofaveragepath-integratedattenuationattheearth’ssurfaceisnowproperlyrepresentedinthisparameterizationinV04.Thisallowsmoreconsistentcomparisonsofsimulatedandsurface reference technique (SRT)derivedpath-integratedattenuations inthealgorithm.

Further, the use of individual SRT-based estimates of path-integrated

attenuation at Ka band in V03 has been replaced by differential Ka-Ku path-integratedattenuationintheMS(Ku+Ka+GMI)modeoftheCombinedAlgorithmV04. The precipitation-free differential Ka-Ku path-integrated attenuationreferenceismuchmorestablethantheKa-bandreference,particularlyoverlandsurfaces, and this leads to less uncertainty in SRT-derived, differential Ka-Kupath-integrated attenuation estimates in precipitation regions. The SRTdifferentialpath-integratedattenuationisusedtodirectlyfiltertheprecipitationprofileensembles,ratherthaninferringtheindividualKuandKapath-integratedattenuationsfromthedifferentialpath-integratedattenuation,andthenfilteringwiththoseindividualpath-integratedattenuations.

The expected uncertainties of forward model simulations (relative to

observations)prescribedintheensemblefilterkernelarechangedfrom1.4dBto3dB forKa-bandreflectivitiesand from5 oK to6.1 oK forGMI radiancesatfrequencies above 37 GHz, going from V03 to V04. Expected uncertainties ofpath-integrated attenuations are maintained at 4 dB in the filter, anduncertaintiesofGMIradiancesatfrequenciesupto37GHzaremaintainedat5oK.V04toV05Changes

NumerousmodificationshavebeenmadetotheCMBLevel2algorithminthetransitionfromV04toV05,andthesignificantupdatesaresummarizedhere.Itmay be noted at the outset, however, that the basic algorithmmechanics (i.e.,estimation methodology) has not changed. The estimation method filtersensemblesofDPRKureflectivity-consistentprecipitationprofilesusingtheDPRKa reflectivities, path integrated attenuations and attenuated surface radarcross-sections at Ku and Ka bands, and GMI radiances. The filtered profileensembles are consistentwith all of the observations and their uncertainties,andthemeanofthefilteredensemblegivesthebestestimateoftheprecipitation

66

profile.TheoutputfilestructureisessentiallythesameasinCMBV04,butafewadditionalvariablesareincludedfordiagnosticpurposes.

In the CMBV03 andV04 algorithms, estimated precipitation profileswere

constrainedbyestimatesof totalpath-integratedattenuation fromthesatelliteto the earth’s surface, derived from the DPR algorithm’s surface referencetechnique(SRT)module;Grecuetal.(2016).However,analternativeapproachistodevelopamodelforthenormalizedradarcross-section(s0)ofthesurfaceattheKuandKachannelfrequenciesoftheDPRandrelatethattoamodelofthesurface emissivities (e) at the GMI frequencies. Such a s0/e model wasdevelopedbyMunchaketal.(2016).Themodelisusedtoeffectivelyconstrainthesimulatedsurfaces0/e in thealgorithm’ssimulationsofattenuatedsurfacecross-section and upwelling brightness temperatures, which are compared tothe observed attenuated cross-sections and brightness temperatures. In theCMB V05 algorithm, both the path-integrated attenuations and attenuatedsurface cross-sections are utilized to constrain solutions, even though there issomeredundancybetweenthesetwoobservables.Itshouldbenoted,however,that some redundancy in the information content of observations leads togreatersuppressionofuncorrelatednoiseinalgorithmestimates.

Another new feature of the V05 algorithm involves the algorithm’s

simulation of path-integrated attenuation at Ka band. Using off-line, high-resolution simulations of attenuation based upon ground-based radar fields, itwas determined that the Ka-band path-integrated attenuation in verticalcolumnsoverDPR-sized footprints,derived using aHitschfeld-Bordanmethodas it isdone in theCMBalgorithm, issignificantlyoverestimated in convectiveregionswherethefootprintsarepartiallyfilledwithprecipitation.Thedegreeofpartial filling, however, can be estimated using a 3x3 array of DPR footprintscentered on the footprint of interest. In the CMB V05 algorithm, a scalingparameter based on the 3x3 array is used to modify the Hitschfeld-Bordanderivedpath-integratedattenuationatKaband toproperlyaccount forpartialfillingoftheradarfootprintbyprecipitation. AtKuband,theeffectsofpartialfootprint filling on path-integrated attenuation are much smaller and areneglectedinCMBV05.

The CMB V04 algorithm estimates exhibited a lack of sensitivity to path-

integratedattenuation,suchthatthescalingofestimatedattenuationrelativetoreflectivitywas sometimes inappropriately high (i.e., the scalingwas adjustedlittle from the initial guess), leading to overestimation of rain rates. TwochangesareintroducedintotheCMBV05algorithmtoobtainmoreappropriatesensitivity. First, the prescribed uncertainties of SRT-derived path-integratedattenuationsarereduced,forcinggreaterfidelityofsolutionstoobservedpath-integratedattenuations. Second, aweakempirical constraintbetweenparticlesizedistributionmass-weightedmeandiameters(Dm)andnormalizedintercepts(Nw) is imposed, such that larger Dm values tend to correlate with lowerNw

67

values. Thisconstraintisimplementedbymakingthenormal first-guessofDmbut then using an empirical formula to calculate the correspondingNw. Theempirical Nw has the observed anti-correlated relationship with Dm (seeThompson et al. 2015), and it is used as an updated first guess forNw. Thealgorithmthenproceedsnormallybutwiththeupdatedfirstguess.ThisNw-Dmconstraint is important at low rain rates, where uncertainties in estimates ofpath-integrated attenuation estimated from the Level 2 radar algorithm (seeMeneghinietal.2000)make it impossible toadjustNw (andthecorrespondingattenuation-reflectivity relationship) using path-integrated attenuationinformation.ThetwochangesdescribedheregenerallyleadtolowerrainratesusingCMBV05.

AnotheraspectofthealgorithmthatisimprovedinV05isthedescriptionof

scatteringbyice-phaseprecipitationparticles.InallversionsthroughV04,ice-phase precipitation particles were represented as spherically shaped,homogeneousmixtures of ice and air. In CMB V05, ice-phase precipitation instratiform regions is represented using nonspherical particles with realisticgeometries, as described in Kuo et al. (2016) and Olson et al. (2016). Therigorously computedmicrowave single-scattering properties of these particlesareincludedinthealgorithm’sscatteringtables. Thenonsphericaliceparticlesare less strongly forward scattering than sphericalparticlesof the samemass,leadingtosubstantially lowersimulatedupwellingmicrowaveradiancesat thehigher-frequencyGMI channels. The impact is to reduce CMBV05 algorithm-estimatedsnowwatercontents,sincelesssnowisrequiredtoproducethesamesignal at the higher frequency channels. Mixed-phase particles are stilldescribedusingsphericalgeometrymodelsinV05.

The prescribed uncertainty of any observation in the CMB algorithm

represents both the noise in the observation as well as the error in thesimulationofthatobservationbythealgorithm’sforwardmodel,andthereforeitdeterminesthedegreetowhichtheobservation impactsestimatesproducedbytheensemblefilter.Aspreviouslymentioned,theprescribeduncertaintiesofKa-band reflectivities and Ku- and Ka-band path-integrated attenuations aremodified in CMB V05. In addition, attenuated s0 observations are alsointroduced, and these observations are assigned uncertainties based on thevariances of s0 for the given earth surface type, incidence angle, and windconditionsbaseduponaclimatologyofs0;seeMunchaketal.(2016).

TheprescribeduncertaintiesofKa-bandreflectivitiesarereducedfrom3dB

in V04 to 2 dB in CMB V05. The uncertainty of Ku-band path-integratedattenuationsisreducedfrom4dBto3dB.Ifpath-integratedattenuationatKa-bandisavailable,thedifferenceofthepath-integratedattenuations(Ka–Ku)isusedasanobservable,withaprescribeduncertaintyreducedfrom4dBto2dBinV05.ThisreductionofuncertaintyisinrecognitionofthefactthattheKa-Kupath-integratedattenuationdifferenceinnon-precipitationsituationsprovidesa

68

more stable reference relative to thatof eitheroneof the twochannels. Overopenwater surfaces, the uncertainties of thes0 at Ku andKa band are set totheir climatological variabilities, given the 10-m wind speed derived fromreanalysisdata.Forothersurfaces,theuncertaintyofKu s0isalsoderivedfromits climatological variability, but it is limited to values above 2 dB, while theuncertainty of the Ka s0 is limited to values above 4 dB. Uncertainties inbrightnesstemperaturesaremaintainedattheV04valuesof5K(atorbelow37GHz)and6.1K(above37GHz).Grecu, M., W. S. Olson, S. J. Munchak, S. Ringerud, L. Liao, Z. S. Haddad, B. L.

Kelley, and S. F. McLaughlin, 2016: The GPM Combined Algorithm. J. Atmos. and Oceanic Tech., 33, 2225-2245.

Hogan, R. J., and A. Battaglia, 2008: Fast lidar and radar multiple-scattering

models.PartII:Wide-anglescatteringusingthetime-dependenttwo-streamapproximation.J.Atmos.Sci.,65,3636–3651.

Kuo, K.-S., W. S. Olson, B. T. Johnson, M. Grecu, L. Tian, T. L. Clune, B. H. van

Aartsen, A. J. Heymsfield, L. Liao, and R. Meneghini, 2016: The microwave radiative properties of falling snow derived from nonspherical ice particle models. Part I: An extensive database of simulated pristine crystals and aggregate particles, and their scattering properties. J. Appl. Meteor. and Climatol., 55, 691-708.

Meneghini, R., T. Iguchi, T. Kozu, L. Liao, Ken’Ichi Okamoto, J. A. Jones, and J.

Kwiatkowski, 2000: Use of the Surface Reference Technique for path attenuation estimates from the TRMM Precipitation Radar. J. Appl. Meteor., 39, 2053-2070.

Munchak, S. J., R. Meneghini, M. Grecu, and W. S. Olson, 2016: A coupled

emissivity and surface backscatter cross-section model for radar-radiometer retrieval of precipitation over water surfaces. J. Atmos. Oceanic Technol., 33, 215-229.

Olson, W. S., L. Tian, M. Grecu, K.-S. Kuo, B. T. Johnson, A. J. Heymsfield, A.

Bansemer, G. M. Heymsfield, J. R. Wang, and R. Meneghini, 2016: The microwave radiative properties of falling snow derived from nonspherical ice particle models. Part II: Initial testing using radar, radiometer and in situ observations. J. Appl. Meteor. Climatol., 55, 709 – 722.

Thompson, E. J., S. A. Rutledge, B. Dolan, and M. Thurai, 2015: Drop size

distributionsandradarobservationsofconvectiveandstratiformrainovertheequatorialIndianandWestPacificOceans.J.Atmos.Sci.,72,4091-4125.