Dual-PolarizationQuantitative Precipitation Estimation

Group Discussion Report

Edward A. BrandesNational Center for Atmospheric Research

Dual-Polarization Radar

Impacts:● Data quality● Hydrometeor classification● Microphysics retrieval/understanding

of precipitation processes● Microphysics parameterization in numerical

models● Precipitation quantification

Group Discussion Objectives

Identify science issues related to quantitative precipitation estimation with polarimetric radar

Facilitate the transfer from research to operations

Discussion Topics● Measurement accuracy/Calibration● Drop axis ratios/DSD model● Role of KDP

● Hail● Estimator variable mix/functional form● Algorithm testing and verification criteria● Default algorithms● New techniques/hardware

Polarimetric Variables

• Radar Reflectivity:

• Specific Differential Propagation Phase:

• Differential Reflectivity:

_______________

• Correlation Coefficient:

4max

, 0 ,4 2 0

4 exp( ) ( )D

H V H Vw

Z N D D f D dDK

1010 log HDR

V

ZZ

Z

max

0

180 Re (0, ) (0, ) ( )D

DP H VK f D f D N D dD

1/ 222

( ) ( )*

( ) ( )

H VHV

H V

f D f D

f D f D

Measurement AccuracyParameter Error Error in R

Z 0.5 dB 10%ZDR 0.1−0.2 dB 5−25%KDP 0.2o km−1 5−100%

Model ErrorAxis ratio ZDR error of 0.2−0.3 dBDSD formAlgorithm Simulation or observationsCanting angle

KDP

Advantages● Immune to calibration problems● Unaffected by attenuation● Unaffected by beam blockage● Insensitive to dry tumbling hail

(a)

3060

90

120150

40

6080

(b)

30

60

90

120

15040

60

8060

12 June 1997Rainfall Accumulation 0600-1030 UTC

R(KDP)R(Z H )

KDP

Potential problems● Reflectivity gradients● Differential backscatter phase shift● Sensitivity to mismatched sidelobes● DSD sensitivity● High noise level in ΦDP measurements● Reduced spatial resolution in rainfall estimates

ZH (dBZ) KDP (o km-1)

30 May 19970058 UTC, 0.5o elevation

Estimator variable mix/functional form

Enhanced Z−R RelationsPower-law/Polynomial RelationsComposite AlgorithmsDrop-Size Distribution Retrieval

Florida-Tuned Power-Law Estimators(ZH and KDP in linear units, ZDR in dB)

Empirical axis ratios: r = 0.9951 + 0.02510D – 0.03644D2 + 0.005303D3 – 0.0002492D4

 Radar Reflectivity:

 Specific Differential Phase:

 Specific Diff. Phase/Diff. Reflectivity:

 Reflectivity/Diff. Reflectivity:

H

2 0.687R = 2.62x10 Z

0.806

DP DPR = sign( )54.3 KK

0.968DP DP DR

2.86R = sign( )136 KK Z

H DR

3 0.945 4.76R = 7.46x10 Z Z

Summary Results:Tuned Power-Law EstimatorsEmpirical Axis Ratios

(388 radar−gauge comparisons)

R(ZH) R(KDP) R(KDP,ZDR) R(ZH,ZDR)

Mean biasfactor

0.94 0.97 0.92 0.97

Bias factorrange

2.53 2.57 2.38 1.79

Correlationcoefficient

0.87 0.86 0.89 0.92

RMSE (mm) 7.9 8.0 7.4 6.4

RMSE (bias removed)

7.7 7.8 7.1 6.3

NSSL Composite Algorithm(linear units)

R(Z) < 6 mm h−1

6 < R(Z) < 50 mm h−1

R(Z) > 50 mm h−1

Where:

1.17( )/(0.4 5.05 1 )DRR R Z Z

1.72( )/(0.4 3.48 1 )DP DRR R K Z

( )DPR R K

2 0.714( ) 1.7 10R Z Z

0.786( ) 45.3 ( )DP DP DPR K K sign K

Point Estimates

Polarimetric Rainfall EstimationAreal Estimates

Drop-Size Distribution Retrieval with Polarimetric Radar Measurements

 Gamma drop size distribution: N(D)=N0Dμexp(-ΛD) N0 drop concentration parameter distribution shape term slope factor 

Solution:

21.935 0.735 0.0365

10log HDR

V

ZZZ

max4, ,04 2

0

4 exp( ) ( )| |H V H V

w

DZ N D D f D dD

K

Spatial Distribution of Retrieved DSD Parameters17 September 1998

Convective rain: 192646 UTC Stratiform rain: 222154 UTC

Comparison: Physical ProcessesConstrained-Gamma and Marshall−Palmer DSD Models

Recommendations

● Endorse the NSSL composite algorithmas an initial product

● Update and enhance methods for radar calibration

● Conduct a community-wide inter-comparison study to improve algorithms

Questions! Comments!