gyuwon lee, aldo bellon, and isztar zawadzki

Error structure in rain estimation by radar1. Radar Calibration.2. Attenuation.3. Variability of drop size distribution.

GyuWon Lee, Aldo Bellon, and Isztar Zawadzki

Radar Calibration Error in the current calibration by gages

Gage calibration

34%

Disdrometer calibration

7%

Perfect gage

Per

fect

rad

ar

Radar Calibration Error in the polarimetric calibration

KDP: 1 dB

KDP & ZDR: 0.3 dB

The (specific) differential phase shift, DP (KDP) is immune to the radar calibration error whereas the reflectivity (Z) is affected by the calibration error.

Radar Calibration Calibration by disdrometer and polarimetry

by a disdrometer by polarimetry

Bias=1.5

Lee and Zawadzki (2002) Submitted to J. Hydrology

Polarimetric radar calibration Sensitivity to the drop deformation

date KDP calibration (dB) Disdrometric calibration

(dB)Pruppacher and Beard (1970)

Illingworth and Johnson

(1999)

Goddard et al. (1995)

Andsager et al. (1999)

2001. 9. 13. -1.92 -3.16 -4.27 -4.10 -1.73 2001. 9. 25.

1.72 1.24 0 -0.21 1.51

2001.10. 25.

-1.58 -2.14 -3.34 -3.58 -1.15

KDP = (3.5 ~ 8.8)x10-5 D4.6~5.2

Zh D6

Zh=(3.9 ~ 6.5)x105 KDP(1.0~1.2)

Ex: Due to the drop deformation,

dZh~2 dB at KDP=1 deg/km

Quantification of Wet-radome Attenuation Method 1 : Monitoring ground echoes with time

King City C-band radar

Quantification of wet-radome Attenuation Method 2 : Natural variability over a large area

< the variability caused by the precipitation at the radar

site

Franktown C-band radar

Attenuation & Calibration error.

H-B Correction

Gage Simulation

+ Random error

Minimization of

Cost function.

ATTENUATION SIMULATIONAND CORRECTION

How many parameters are needed to describe the variability of DSDs?

N D N D( ) exp( ) 0 R

N D N D D( ) exp( ) 0

2/12)(1

R

RR

NT

R-Z

R-(Z,M2)

Scale dependence of the variability of drop size distribution

1. Climatological variability

A single Climatological Z-R relationship

SDfe ~ 37 %

Scale dependence of the variability of drop size distribution

2. Day-to-Day variability

A single Climatological Z-R relationshipSDfe= 34%

3. Variability within a day

Daily Z-R relationships

SDfe~ 31 %

Scale dependence of the variability of DSD

UHF profiler Reflectivity

Between different physical processes

Within a quasi-homogeneous

physical process

SDfe= 30%

SDfe= 10%Collocated disdrometer

Conclusion and Discussion

1. Three different methods of radar calibration

- accuracy

- consistency between methods

2. Quantification of attenuation & Correction method.

- QPE in C-band (?)

3. The variability of drop size distribution

- scale dependence of DSD variability

- categorization of different physical processes

: using morphological characteristics

& polarimetric information

Error structure in rain estimation by radar4. AP & Ground Echoes5. Range effects6. VPR & Optimal Surface Precipitation (OSP)

Aldo Bellon, GyuWon Lee, and Isztar Zadwadzki

Eliminate GE and AP

a) Radial velocity at one or more elevation angles

b) Vertical gradient of reflectivity c) Horizontal gradient of reflectivity

d) Preferred location of AP echoes.

AP

GE

AP & GE Elimination

From Polarimetric Information

a) SD of ZDR

b) SD of ΦDP

c) SD of Z

d) Radial Velocity

Radial velocity & Reflectivity

(1km x 1deg.)

Polarization (150 m)

VPR CORRECTION ===> SURFACE RAINFALL

1. Accumulations based on CAPPIs at a pre-defined height (1.5 to 2.5 km) (Correction applied to the 1-h accumulations as a function of range)

a) dBZ = dBZ [ CAPPI height ] - dBZ [ Ref. height ]

b) Beyond 110 km, apply Gaussian smoother (vertical) to last profile

dBZ = dBZG [ H(range) ] – dBZLast VPR [ Ref. height ]

c) Interpolate dBZ at every km and convert it into a rainfall rate factor

- multiply uncorrected 1-h accumulations and integrate for total rainfall

2. Accumulations based on OSP maps(Optimum Surface Precipitation)

(Correction performed at every radar cycle and for every pixel)

a) Lowest pixel (~1.0 km) above the 3-D average ground echo mask

- Radial velocity used to reach any precip. above stationary targets - If this height is close to echo top, perform horizontal interpolation ELSE

- dBZ is obtained as in (1) to modify reflectivity at selected height

b) Automatic VPR identification and correction

- Avoids rather than corrects for bright band Data is taken from height sufficiently below BB bottom or above BB top c) No dBZ adjustment for convective pixels ( > 30 dBZ 1.5 km above BB height)

- Separate Z-R for stratiform and convective pixels

d) Knowledge of 0º isotherm would prevent correction in cases of low bright band - identifies low level growth due to warm rain or snow

3. Simulations of high resolution volume scan (3-D) data

Rainfall Accumulation Correction

Original 1hr accum.

Optimal 1hr accum.

BB contamination

VPR correction

Range Effects

CAPPI SimulationH= 1.1 km H= 1.9 km

H= 2.7 km H= 3.7 km

120 km

240 km

Error Structure: Instant. F(range, height)

Before Correction

Aft

er C

orr

ecti

on

1x1 km2

9x9 km2

2/12)(1

R

RR

NSDfe E

Range-Dependent Errors

Bias

BIAS

Measurement in snow

Contamination by the BB

Random error after bias correction

A stratiform case with a weak bright band at 3.5 km

Animation of 1-hr Accum.

Error Structure: 1-hr Accum. F (range, height)Before Correction After Correction

1 km x 1 km

9 km x 9 km

CONVECTIVE

A Convective Case Error Structure: 1-hr Accum. F (range, height)

1 km x 1 km 9 km x 9 km

A Convective Case Error Structure: 1-hr Accum. F (range, height)

Before Correction After Correction

A Snow Case

A Snow Case Error Structure: 1-hr Accum. F (range, height)

Before Correction After Correction

Snowfall Correction

Before Correction After Correction

Thank you!

Low Bright Band (Worst Case Scenario)

Uncorrected Erroneously Corrected

Radar Composite