Transition of the Combined Radar-Radiometer Algorithm from V04 to V05

Bill Olson, Mircea Grecu, Joe Munchak, Lin Tian,Sarah Ringerud, Kwo-Sen Kuo, Ziad Haddad,

Ben Johnson, Bart Kelley, Dave Bolvin, Bob Morris

with support from

the Radar and Radiometer Algorithm Teams, the GV Team,WG’s, and Precipitation Processing System Personnel

Kuradar swath

Combined Radar-Radiometer Algorithm Input• Dual-Frequency Precipitation Radar (DPR); Ku & Ka bands• GPM Microwave Imager (GMI); 10 – 183 GHz.

GMI swath

Ku+Ka+GMIEstimate

Karadar swath

Algorithm “Concept” --- Ensemble FilterInput Ensemble Solution

Z

alt.observed Ku reflectivityprofile

assumed randomwater vapor,cloud water, µ, Nw profiles

precip.watercontentprofilesinvert DM

profiles

assumed randomsurface emissivity

simulatedKa reflectivityprofiles,Ku/Ka PIA’s

simulator

simulatedTB’s

EnsembleFilter

filtered precip.watercontentprofiles

usescovariancesofwatercontentsandsimulatedobservations

observed Kareflectivity profile, Ku/Ka PIA’s, and TB’s from GMI

Z

alt.

solutionprofile

(mean)

s

Algorithm “Concept” --- Nonuniform BeamfillingInput Ensemble Solution

Z

alt.observed Ku reflectivityprofile

assumed randomwater vapor,cloud water, µ, Nw profiles

precip.watercontentprofilesinvert DM

profiles

ZKu i

alt. ZKu i = fi ZKu

downscale

assumed randomwater vapor,cloud water, µ, Nw profiles invert DM

profiles

upscale

What we added with V04:

• GMI radiances were resolution-enhanced using regression-based filters; using all GMI channels.

• Hogan & Battaglia model for multiple-scattering in radar simulations was utilized where needed.

(note: V04 described in Grecu et al. 2016 JAOT article)

Comparison of GPM Mean Precip. vs. GPCPand MRMS Sep. – Aug. 2014/2015

Combined V4 Ku+Ka+GMI

V4 - GPCP

Dave BolvinCombined V4 and GPCP

Combined V4 vs MRMS

2.72 mm d-1

2.53 mm d-1

2.85 mm d-1

Ku+GMIKu+Ka+GMIGPCP

r = 0.84

Combined Algorithm V5 Updates and TRMM V8

• Changed initial assumptions on PSD’s and ensemble generation.

Revised modeling of path-integrated attenuation in response to non-uniform beamfilling effects.

In process:

• Nonspherical ice particle scattering tables.

• s0 – emissivity surface parameterization.

• Have begun interfacing Combined with TRMM input.

Radar-Based Simulations of Beamfilling-Affected Path Integrated Attenuation Mircea Grecu

PIA at Ku Band PIA at Ka Band

SRT

PIA

“Retrieved” PIA (uniform)SR

T PI

A“Retrieved” PIA (uniform)

• Use high-resolution ground radar to simulate SRT PIA andattenuated reflectivities over DPR footprint.

• Retrieve PIA’s, assuming uniform beamfilling.

Comparison of GPM Mean Precip. vs. GPCPand MRMS Sep. – Aug. 2014/2015

Combined ITE Ku+Ka+GMI

ITE - GPCP

Dave BolvinCombined ITE and GPCP

Combined ITE vs MRMS

Ku+GMIKu+Ka+GMIGPCP

2.67 mm d-1

2.57 mm d-1

2.85 mm d-1

r = 0.86

Comparison of GPM Mean Precip. vs. GPCPand MRMS Sep. – Aug. 2014/2015

Combined V4 Ku+Ka+GMI

V4 - GPCP

Dave BolvinCombined V4 and GPCP

Combined V4 vs MRMS

2.72 mm d-1

2.53 mm d-1

2.85 mm d-1

Ku+GMIKu+Ka+GMIGPCP

r = 0.84

Comparison of ITE vs. MRMS Sep. - Aug. 2014/2015

Combined ITE Errors vs. MRMSCombined ITE – MRMS

Combined ITE vs. MRMS

r = 0.86

Comparison of ITE Ku+Ka+GMI and MRMS Rain Ratesat Footprint Resolution

MRMS RAIN RATE [mm h-1]Ku+

Ka+

GM

IRA

IN R

ATE

[mm

h-1

]MRMS RAIN RATE [mm h-1]K

u+K

a+G

MIR

AIN

RAT

E [m

m h

-1] Stratiform Convective

Comparison of GPM V4 vs. ITE Corrected Ku Reflectivity on May 11, 2015Using “Validation Network” Matched Data

KFWS

V4 reflectivity

ITEreflectivity

reflectivity

V4 - KFWSreflectivity

ITE - KFWSreflectivity

Bob Morris

Final Remarks:

• Combined Algorithm V5 should be ready fordelivery 1 month after final Radar Algorithm is delivered.

• In short term, examine relationships between PSD/non-uniform beamfilling assumptions and attenuation correction.

• Non-uniform beamfilling, multiple scattering, ice/mixed-phase particle properties are currently parameterized, but comprehensive descriptions will require longer-term efforts.

• Will work with radiometer team on high latitude estimates. Generate databases for algorithm.

extras

Combined Radar-Radiometer Algorithm Input• Dual-Frequency Precipitation Radar (DPR); Ku & Ka bands• GPM Microwave Imager (GMI); 10 – 183 GHz.

GMI swath

Ku+GMIEstimate

(likeTRMM)

Kuradar swath

Comparison of ITE Ku+GMI and MRMS Rain Rate at Footprint Resolution

MRMS RAIN RATE [mm h-1]K

u+G

MIR

AIN

RAT

E [m

m h

-1]

MRMS RAIN RATE [mm h-1]

Ku+

GM

IRA

IN R

ATE

[mm

h-1

]

Stratiform Convective

Comparison of GPM V4 vs. ITE Sep. - Aug. 2014/2015

Combined ITE - V4Combined V4 - GPCP

Radiometer Database GenerationSimulated vs. Observed TB’s

• brightness temperatures aresimulated using 1 year of retrieved profiles from V4 combined algorithm code.

• ice column is adjusted to getbetter agreement with high-frequency GMI channels (uses DDA ice).

• over ocean comparisons atright.

Sarah Ringerud

Nw Study

Rsfc

Rsfc

Rsfc

2BCMBITE104NS

2BCMBITE106NS

2AKuITE104

Observations

Zcorr 2.5 km

Zcorr 2.5 km

Zcorr 2.5 km

Zm 2.5 km

Zcorr 5.0 km

Zcorr 5.0 km

Zcorr 5.0 km

Zm 5.0 km

est PIA

est PIA

est PIA

SRT PIA

Nw Study

Rsfc

Rsfc

Rsfc

2BCMBITE104NS

2BCMBITE106NS

2AKuITE104

low lev Nw Dm 2.5 km Dm 5.0 km

low lev Nw Dm 2.5 km Dm 5.0 km

Nw 1.25 km

low lev bin #

Dm 2.5 km Dm 5.0 km

Comparison of GPM V4 vs. ITE Corrected Ku Reflectivity on Sep. 2, 2014Using VN Matched Data

KEAX

V4 reflectivity

ITEreflectivity

reflectivity

V4 - KEAXreflectivity

ITE - KEAXreflectivity

Comparison of GPM and MRMS Radar (Q3) Sep. - Feb. 2014/2015

Combined ITE – MRMS Combined ITE vs. MRMS

Resolution-Enhancement of GMI Radiancesradar-simulatedmicrowave

reconstructingfilter

TB

microwave

antennapattern

TB

Mircea Grecu

Non-Uniform Precipitation Beamfilling

Uniform Filling

rangegate

5-km footprint

surface

Non-Uniform Filling

5-km footprint

“Downscaling” to Represent Non-Uniform Beamfilling

ZKu

alt.observed Ku reflectivityprofile estimate simulate

Uniform Beamfilling

ZKa

alt.simulated Ka reflectivityprofile

aggregate aggregate

LWC

alt.estimatedprecipitationprofile

Non-Uniform Filling“downscale”

estimate simulate

ZKu i

alt. ZKu i = fi ZKu

LWCi

alt.

ZKa i

alt.

Ku and Ka Band Radar

single-scatter return range gate

multiple-scatterreturn

Impact of Multiple Scattering

Observed and Modeled DPR Reflectivities

single-scatteringcontribution

Ka observedKu observed

fromBattaglia,Tanelli,Mroz, Tridon

HEI

GH

T [k

m]

REFLECTIVITY [dBZ]

Precip. Estimates from Hurricane Edouard

Ku+GMI Rain Rate Ku+Ka+GMI Rain Rate

TB Simulations from Hurricane Edouard10 GHz 19 GHz 37 GHz 89 GHz

Obs.

Ku+GMIest.

Ku+Ka+GMIest.

Algorithm Theoretical BasisGeneralized Hitschfeld-Bordan Method

(applied to Ku-band data only)• original Hitschfeld-Bordan fast, but reqs. k = a Z b .

• iterative techniques typically slow.

• alternative interative procedure, assuming No(r) and approximate approximate b from k-Z relation:

€

Z(r) =ZKu r( )

1 − q α s( )ZKuβ s( )ds

0

r

∫&

' (

)

* +

1β, q ≡ 0.2 β ln 10( )

€

Z(r) =ZKu r( )

1 − q ZKuβ s( )

k Z s( )( )Z β s( )

ds0

r

∫%

& ' '

(

) * *

1β

Algorithm Theoretical Basis

Correct DPR ZKu for attenuation due to cloud and water vapor.

Set true Z = DPR ZKu .

qS(rs)<ζmax?Scale N0 by and S(r) by

€

ζmax /(qS(rs))( )11−β

€

ζmax /(qS(rs))( )

Z(r) = ZKu(r)/(1-qS(r))(1/β)

Conver-gence ?

Retrieve D0 from Z and N0 .

Yes

No

Yes

No

Calculate .

€

S(r) = ZKuβ

0

r

∫ s( )k Z s( )( )Z β s( )

ds

Generalized Hitschfeld-BordanMethod• procedure is fastbecause iterativeequation is a closeapprox. to H-B solution.

• note procedureavoids instability byrescaling No(r), if needed.

• yields Do(r), given No(r),µ, and ZKu .

0 oC

Evaluating Snow Physics Using HIWRAP and CoSMIR in MC3E

• Retrieve precip profile(PSD’s) using HIWRAP. HIWRAP

insitu

• Compute consistentmicrowave scatteringproperties in profile.

• Simulate upwelling brightness temperaturesat 89, 165.5 GHz.

• Compare to CoSMIR obs.

Note: brightness tempsaren’t sensitive to variationsof surface emission andliquid precip if light rain is present => scattering signatures discriminate snow particle models.

CoSMIRTB’s

• Assign scattering model. W. Olson,K.-S. Kuo,L. Tian,M. Grecu,B. Johnson,A. Heymsfield,G. Heymsfield,J. Munchak

(Ku/Ka)

Radar Retrieval and Simulation of TB’s Using Spherical/Aggregate Ice

Kuo aggregatesr = 0.1 g cm-3 spheres

(Aggregates)

r = 0.1 g cm-3 spheres Kuo aggregates

Comparison of GPM Mean Precip. vs. GPCPand MRMS Sep. – Aug. 2014/2015

Combined ITE Ku+Ka+GMI

ITE - GPCP

Dave BolvinCombined ITE and GPCP

Combined ITE vs MRMS

Ku+GMIKu+Ka+GMIGPCP

2.60 mm d-1

2.52 mm d-1

2.85 mm d-1

r = 0.85

Comparison of GPM V4 vs. ITE Sep. - Aug. 2014/2015

Combined ITE - V4Combined V4 - GPCP

Issues with V04:

• estimates over land, particularly in climatologicallyconvective regions, were overestimated.

• overestimation was made even greater by the DPR calibration change.