Operational Applications of Polarimetric Radar
Steven A. RutledgeDepartment of Atmospheric ScienceColorado State University
Acknowledgements—All current and former staff and students of the Radar Meteorology Group, CSU-CHILL staff, Profs. Bringi and Chandra from CSU/ECE, funding agencies especially NSF, NASA and NOAA, andmany colleagues in the community.
Benefits of Polarization Diversity….(based on several decades of research)
• Identification of anomalous propagation and clutter (non-meteorological echo)---data quality control
• Improved rain rate estimation, especially in presence of ice
• Remotely sense cloud and precipitation processes, especially when combined with Doppler measurements
• Detection of severe weather including hail• Attenuation correction, especially critical at sub S-band
wavelengths
Radar[ ]
TMatrixn Propagatio
S
⎥⎦
⎤⎢⎣
⎡Matrix
Scattering
H
V
The observations are a combination of backscatter and propagation characteristics of precipitation. Use various polarization schemes to remotely sense the precipitation medium.
Polarimetric Radar Variables
Horizontal (Vertical) Reflectivity (ZH,V) Size, concentration
Differential Reflectivity (ZDR) ZDR = 10 log (ZH/ZV)
Basic measure of mean shape; median volume diameter (D0) can be retrieved. N0 fixed by Z.ZDR ~ 3 dB rainZDR ~ 0 dB hail/graupel
Differential Phase (ΨDP) ϕdp = ϕh - ϕv
Specific Differential Phase (KDP)Filtered, range derivative of ϕdpLWC, oblateness; isotropic vs. anisotropic scatterers--- KDP very different between rain/hail
Oblate Raindrop
Small RaindropHail/Graupel
Polarimetric Radar Variables
Linear Depolarization Ratio (LDR)Orientation, canting, melting—not possible withNEXRAD polarimetric configuration
Correlation Coefficient (ρHV); measure of correlation between estimates of ZH and ZV.Mixed phase, melting—strong function of howdiverse the particle shapes are in pulse volume. Clutter or AP have very low correlations—useful!
Z ∝ Σ [N(D) • D6 ]
R ∝ Σ [N(D) • D3.67 ]
Kdp ∝ Σ [N(D) • D4.24 ]
Advantages of using Kdp for rainfall estimationLess sensitive to variations in DSD than Z (4.24th moment of DSD is closer
to 3.67th than 6th !)Independent of power calibration-phase measurementLess sensitive to beam blockingImmune to attenuation—provided enough signal!
Issues regarding Kdp
Trade-off between accuracy and spatial resolution (rain estimation). Filtering required from a noisy field ϕdp
Backscatter differential phase (Mie targets), range effects, gradient regions, choice of drop shape model
Issues regarding ZDR
Reflectivity gradients—power received through sidelobes that are mismatched can produce spurious values of ZDR in low reflectivity regions next to strong cores
Antenna performance is critical, main beam H/V matching
Three body scattering
Differential attenuation
Presence of hail---GOOD and BAD
NCAR S-pol
CSU-CHILL
Intense storm, rain and hailas viewed by the CSU-CHILLradar
Polarimetric conversion of NEXRAD
• Single transmitter, simultaneous transmit, two receivers
Seliga and Bringi, 1976 (JAM)---first discussion of single transmitter, dual-receiver configuration---now known as STSR
REMOVING NON-METEOROLOGICAL ECHOUSING POLARIMETRIC THRESHOLDS
Original ZV ρHV
σ(ΦDP) Filtered ZH
Basic thresholding of data can eliminate most non-meteorological echo
σ(ΦDP) – Clutter/AP, Noise
ρHV – Clutter/AP, Noise
ZH & ZDR – Biological Scatterers
LDR – Second Trip
(e.g., Ryzhkov & Zrnic 1998)
Ryzhkov et al. (2005)
REMOVING NON-MET ECHO:FUZZY LOGIC CLASSIFICATION (FHC)
Clutter/AP
Rain
Insects/Birds
Example from JPOLE of rain embedded in clutter/AP and biological scatterers
Remove using flags on FHC category Also simple thresholding techniques used
Mean clear-air power overseveral hours by azimuth0.8° elev.; clutter flagged white(from S-Pol, NAME 2004)
Major Blocks
MountainClutter
Ocean
MinorBlock
Invoke self-consistency of ZH and KDP in rain (Scarchilli et al. 1996)
ZH scatter plot for KDP between 1 and 2° km-1 by azimuth (~1 week’s data)
Depressions in ZHscatter denote blocks;Median dBZ reductionis +dBZ correction needed
OceanMtns
USING POLARIMETRIC RADAR TO CORRECT BLOCKAGE
Application for NEXRADin areas of rough terrain
Examples of rainfall estimation methods for polarimetric radar
1. R(KDP,ZDR) = 90.8 * (KDP)0.93 *10(0.169*ZDR) mm hr-1
2. R(ZH,ZDR) = 6.70 x 10-3 * (ZH)0.927 * 10(0.1*-3.433*ZDR)
mm hr-1
3. R(KDP) = 40.5 * (KDP)0.85 mmhr-1
4. R(ZH) = (ZH/300)0.7143 mm hr-1
• (1) and (2) assume Beard and Chuang equilibrium model (includes oscillations)
• Coefficient in (3) assumes equilibrium model of Pruppacher and Beard (no oscillations)
• (4) is the default NEXRAD Z-R
Equations Used in CSU Blended Algorithm
Intense storm over DIA
Fort Collins Flood Rain Gauge Total Map (inches)28 July 1997
Comparison of Radar-Derived Rainfall Totals
Cheyenne NEXRAD Z-R Rain Total CSU-CHILL Polarimetric Rain Total
• Storm totals using standard NEXRAD Z-R only 50-65% of gauge max
• Totals using polarimetric variables (Z, ZDR, KDP) are 95% of gauge max
• Polarimetric radar max location within ~0.5 km of gauge max
For rain events in Colorado……..
Need for good Z, ZDRcalibrations
NSSL Synthetic Rainfall Algorithm
R = R(ZH)/f1(ZDR), if R(ZH) < 6 mm hr-1
R = R(KDP)/f2(ZDR), if 6 < R(ZH) < 50 mm hr-1
R = R(KDP) , if R(ZH) > 50 mm hr-1
Where,R(ZH) = (ZH/300)0.714
R(KDP) = 44.0|KDP|0.822 sign(KDP)
f1(ZDR) = 0.4 + 5.0|Zdr-1|1.3 ; Zdr is in linear unitsf2(ZDR) = 0.4 +3.5|Zdr-1|1.7
• R(ZH), R(KDP) and ZDR are averaged over 1 km X 1º area prior to calculating point estimates
• Negative KDP is used to compensate spurious large positive KDP’s in regions of high reflectivity gradients
• f1 and f2 attempt to compensate for DSD variability in a mean sense• Exhibited best performance among all polarimetric (16) and non
polarimetric (1) relations used in JPOLE
Performance of NSSL Synthetic Algorithm
Ryzhkov et al. 2005
Ryzhkov et al. 2005
• Vast improvement over R(ZH)• Performance needs to be evaluated for other locations
To what fraction of total rainfall will polarimetric rainfall retrievals apply?
Steve Nesbitt, CSU
Winter Spring
Summer Fall
Fraction of rainfall exceeding 35 dBZ
Nw, Dm for stratiformrain
Derive these quantitiesfrom polarimetricobservations
Tune Z-R
Nw, Dm for convective
rain
Derive these quantities from polarimetric observations
Tune Z-R
⎥⎦
⎤⎢⎣
⎡= 4
467.3
oww D
WNπρ
Estimation of rainfall via polarimetric tuning technique—derive Gamma DSD parameters from polarimetric data
Bringi, Gorgucci and colleagues
Fuzzy-Logic Hydrometeor ID
Hydrometeor ClassesLarge Hail (D > 2 cm; LH)Small Hail (D < 2 cm; SH)Rain (R)High-Density Graupel (HG)Low-Density Graupel (LG)Drizzle (Drz)Wet snow (WS)Dry Snow (DS)Vertical Ice (VI)
1. Examine polarimetric parameters and temperature at each grid point
2. Score each hydrometeor category based on observations relative to known range of values for each hydrometeor class (determined from field obs, scatter modeling)
3. Highest score wins
Algorithm produces “dominant” hydrometeor type---thiscan be summed to provide storm volumes of hydrometeortype-crude information can be derived on mixing ratios.
June 29: STEPS Lim et al. (2005)C
B
Polarimetric dataused to diagnosehail and graupelcontents and relate toelectrical properties
K. Wiens and S. Tessendorf
Polarimetricradar as a cloud physicstool
Ryzhkov et al. 2005, J. Appl. Met.
PolarimetricTornado Detection
LoweredDiff ReflectivityandCorrelationCoefficient
Brandes and Ikeda, JAM, 2004
Mapping of meltinglevel
March 2003 blizzard, microphysical processesrevealed by polarimetric datacollected by the CSU-CHILLFacility—courtesy Pat Kennedy
CSU-CHILL, Z Rain-snow, Vr
ZDR LDR
Conclusions• Much research has been done to date using polarimetric radars
and much has been learned. More work is needed to investigate application of polarimetric rain estimators to broad spectrum ofrain regimes
• Use of polarimetric S-band radars for cool season precipitation is relatively unexplored compared to warm season precipitation—this work needs to be accelerated in advance of the NEXRAD polarimetric upgrade
• Also need to further explore role of surface based measurements such as disdrometers, profilers, mesonets, etc to enhance polarimetric NEXRAD network
• Need more study of short wavelength polarimetric radars combined with S-band data
S-Band and X-Band Polarimetric Radars Complement One Another
Example: Mesoscale Convective System
Trailing Stratiform
LeadingConvection
•X-band allows greater accuracy in light rain events (e.g., stratiform)•S-band provides excellent rainfall estimates in heavy rain (convective)
X-band S-band
Matrosov and Martner
© 1998 Prentice-Hall, Inc. -- From: Lutgens and Tarbuck, The Atmosphere, 7th Ed.
Collaborative Adaptive Sensing of the Atmosphere
• Network of small radars• Low to the ground• Adaptive• Low cost
$$$$$$$$$$$$$$
$ $ $0-3 km
Courtsey J. Brotzge, OU
The Future• Polarimetric NEXRAD’s on line---improved rain and snow
estimation for a wide range of hydrological applications, improved data quality, better detection of convective and mesoscale phenomena
• NASA GPM mission launches in 2010-2011 and provides 3 hourly, 4 km rain estimates as backdrop to the finer scale measurements
• NEXRAD network supplemented by smaller, shorter wavelength radars in key areas for detailed wind and precipitation information
• Widespread use of 3-D lightning mapping networks for aid in storm nowcasting and basic research; integration of other surface-based sensors
• Massive data assimilation effort to couple with forecast models—serious challenge