2.7 EVALUA TION OF POLA RIMETRIC CAPABILITY ON THE RESEARCH W SR-88D

Valery M. Melnikov*, Dusan S. Zrnic**, John K. Carter**, Alexander V. Ryzhkov*, Richard J. Doviak**

* - Cooperative Institute for Mesoscale Meteorological Studies/University of Oklahoma

** - NOAA/ National Severe Storms Laboratory

1. Introduction

One of the enhancements to the WSR-88D weather radar is a polarimetric capabilityto improve rainfall estimation and identifyprecipitation type. Implementation of theproof of concept scheme (Fig. 1) has beenmade on the NSSL’s Research &DevelopmentWSR-88D, and tests are being conducted todetermine the quality of this upgrade. Theradar transmits and receives horizontally andvertically polarized waves simultaneously. Ahigh voltage power splitter is used to formtwo channels in the WSR-88D’s transmitter,one for the horizontal H, the other for thevertical V mode. To process verticallypolarized waves, a second receiver, identicalto the existing one, has been added (Fig. 1). Acommercial ( Sigmet RVP-7) processor ispassively connected (in parallel with a powerPC based processor) to allow sooner test ofthe engineering quality of the system. Thisprocessor requires a sum of offset IF signals,one for the H the other for the V channel.Therefore we have retained the initial 57.54MHz IF for the H channel and have designedcircuits to generate a 63.30 MHz IF for the Vchannel. The following variables are availableon the RVP-7 processor: reflectivity Zh ,Doppler velocity V, and spectral width Fv, allthree at horizontal polarization (as is the caseon the WSR-88D network), differentialreflectivity ZDR , differential phase shift ndp , and correlation coefficient Dhv, betweenvoltages in horizontal and vertical channels____________________________Corresponding author address: ValeryMelnikov, 1313 Haley Circle, Norman, OK,73069. E-mail: Valery.Melnikov@noaa.gov

(see Doviak and Zrnic 1993, for the definitionof these variables).

Two waveguide switches in thetransmit chain are used to bypass the powersplitter to that only H waves can betransmitted. Nonetheless, in this mode bothH and V waves are received and processed toobtain the linear depolarization ratio Ldr andthe co to cross correlation coefficient Dxv inaddition to the three spectral moments and thedifferential phase. In this configuration, thetransmitter and the co-polar channel areessentially identical to the legacy system; thesecond channel receives the depolarizedwaves.

There are two simple polarizationparameters indicating the quality of the dual-polarization radar design: minimal value ofLDR and maximal value of Dhv measured inlight precipitation. Melnikov et al. (2001)reported minimal reliably measured LDR

values better than –30 dB in light rain whichis an indication of a good isolation betweentwo orthogonal channels. Typical maximalvalues of the cross-correlation coefficient Dhv

measured in light rain with strong SNR areabout 0.995. This ensures high accuracy ofthe measurements of the two basicpolarimetric variables that are used forrainfall estimation and hydrometeorclassification: KDP and ZDR (Bringi andChandrasekar, 2001).

The accuracy of the radar reflectivitymeasurements on the WSR-88Ds is 0.5 dB(Crum et al., 1993). In polarimetric mode, theaccuracy of the difference of reflectivities inthe horizontal and vertical channels shouldbe better than 0.2 dB. This accuracy isrequired for discrimination of frozen

precipitation and light rain. Thus the technicalissue to solve has to do with relativecalibration of the two receiver channels overthe full dynamic range of input signals. In thispaper, we describe various calibrationprocedures aimed to solve this problem andreport on stability of the channels.

To calibrate the WSR-88D radarreceivers we have used several schemes

including test signals of the built-in RF signalgenerators, external RF and IF generators,solar flux, and measurements in the near fieldof the antenna. The transmitted powers havebeen measured with the bolometers (ports 9 inFig. 1), and with horn antenna in the radarnear field.

2. Calibration with the test signal In these measurements, a low voltage

power splitter (box 10 in Fig. 1) supplies testsignals to the two channels simultaneously.The two test signals at the input to thedirectional couplers are equal. The differenceof attenuation of two couplers is less than 0.2dB (the manufacturer’s specification). Usingthe WSR-88D’s built-in CW generator wehave obtained the receiver response curves.Both curves are linear for sufficiently strongsignals; the bias of the ZDR is about -0.3 dB.The receiver noise levels are not equal andfurther at low SNR there is a question ofreceiver’s linearity which is being addressed.Throughout the warm season (May toSeptember) the receiver response curves havebeen very stable and we plan to monitor thesethrough the cold months ahead. Calibrationwith the test signal leaves out the circulators,rotary joints, and antenna. Thus the deducedZDR bias of -0.3 dB needs to be added to thepart caused by these unaccounted components.Nonetheless, frequent monitoring by the buildin generator of the partial bias isrecommended so that any change (in theseactive components) can be detected andcorrected.

3. Sun scansThe solar flux measurements can be

used to verify the coincidence of the radarbeam axes for horizontally and verticallypolarized waves and check the stability of thereceiver channels. To check alignment of theradar beams (for the H and V polarizations) inthe transverse direction, the antenna scans inazimuth through the solar disk while signalsin the two channels are recorded. Thepositions of signal maxima coincided hencewe concluded that the azimuthal alignment issatisfactory. Similar measurements of solarflux in the elevation direction confirmed that

the two beams are aligned. Here we report only on the imbalance

between the two receiver channels deducedfrom solar flux measurements. The solarsignal is 12 to14 dB above radar noise levelso this natural source provides onemeasurement point at the low end of thereceiver dynamic range. The noise levels inthe two channels are different and this isaccounted for in the ZDR measurements asfollows

,

where Ph and Pv are measured powers in Hand V channels respectively (they aredifferent due to different gains of thechannels), Nh and Nv are noise levels in thechannels, and SNRh and SNRv are themeasured signal-to-noise ratios. The ZDRn isthe ZDR for the atmospheric and system noise.We measure the ZDRn at the antenna parkposition (Az=0, El=22.5 deg). The measuredZDRn for the radar varies between 0.9 and 1.3dB.

We used the solar flux measurementsto monitor the stability of the receivercontribution to the system ZDR . Fig. 2presents the results of the measurementsduring the warm period of 2002. The solarflux values at 2700 MHz have been takenfrom the NRC/DRAO observatory web site atwww.spacew.com. One can see from Fig. 2that the variations of measured ZDR lie in theinterval of –0.2 to 0.2 dB.

Fig. 2. Measured SNRh and )ZDR for the solar flux. The solar flux is in sfu units.

4. ZDR calibration using theground clutter

The ZDR calibration using the testsignals and solar flux do not close thecalibration loop. The transmitter chains areleft out of the loop. The measurements ofoutgoing powers at the ports 9 Fig.1 have theaccuracy of 0.5 dB which is considerablylarger than required for precise ZDR

measurements. A metal sphere lifted with aballoon could be used for total calibration butthis is complicated and permission from theFAA is required for each case of the lift.Therefore we opted to use natural object in theradar vicinity.

Can the ground clutter be used as suchan object? The answer is no if a point ZDR isconsidered because the ground clutter hasvery variable ZDR properties; the ZDR canexceed 10 dB and be positive or negative.But a field of ZDR from ground clutter offersattractive possibilities. We have beenrecording the clutter since May 2002. Onefull rotation of the antenna is used to collectdata at low elevation. Only echoes with theSNR larger than 30 dB were processed so thatthe effects of noise are insignificant. An

Fig. 3 Histogram of measured ZDR of ground clutter for one full antenna revolution

example of the output histogram of ZDR isshown in Fig. 3.

The RVP-7 processor presents the ZDR

values in the interval of -7.94 to 7.94 dB.Because ground clutter has some ZDR valuesoutside of this interval the histogram has thelong “horns” due to clipping in the processor.The median value of the histogram, M, wascalculated for the whole data set with clippedvalues. The M2 is for data in the interval of -7to 7 dB, t.e without clipped values. The Meanvalue is also for the data in the interval -7 to

7 dB. In Fig. 4, these three values areshown over the time from May 2002 toSeptember 2002. Clearly the mean valuesover this time period are close to 0.1 dB. Themean values have the lowest standarddeviations of 0.15 dB. These measurementssuggest that the ground clutter could be agood candidate for ZDR calibration. We arecontinuing the measurements to include thecold season so that seasonal statistics andlong term variation of the ZDR could bedetermined.

Fig.4. Variations of the M, M2 , and Mean values of measured ZDR for ground clutter

5. ZDR calibration using weatherobjects

Measurements at vertical incidence inrain are often used to establish the overallsystem bias of ZDR (Bringi and Chandrasekar2001). Due to mechanical constraints, theWSR-88D radar can elevate its antenna onlyup to 60º. Still, there are means to check ZDR

calibration at this less than ideal elevation asfollows.

The differential reflectivity decreaseswith elevation angle for all types ofhydrometeors. For oblate spheroidal particleswith a mean vertical orientation, thisdependence is expressed by the followingformula that can be simply derived usingBringi, Chandrasekhar 2001:

Fig. 5. Azimuthal variations of ZDR above the melting layer for the highest elevation of the WSR-88D (the upper curves are displaced by 1 and 2 dB )

where Zdr(0) and Zdr(2) are differentialreflectivities at elevation angles of 0 and 2respectively. Here Zdr is expressed in linearscale. It can be easily shown from aboveequation that ZDR(2 = 60º) . 0.25 ZDR(2 = 0º),where ZDR is expressed in logarithmic units.Atmospheric scatterers with low variability ofintrinsic ZDR can serve as a natural target forZDR calibration. Dry aggregated snow isprobably the best choice because its ZDR

usually varies between 0 and 0.5 dB at grazingangles (Ryzhkov and Zrnic 1998 a,b) and itcan be easily identified within the cloud in theregions slightly above the melting level. Ourobservations show that the melting level

can be easily detected at higher elevation tilts(including 60º) by a sharp drop of the cross-correlation coefficient. Thus, in the 2 – 3 kmheight interval above the melting layer, theexpected value of ZDR at the 60º elevationangle should vary in the narrow rangebetween 0 and 0.15 dB. Fig. 5 shows an example of suchmeasurements. According to radar data themelting layer was at height below 3 km.Three curves of ZDR as functions of azimuthare presented in the figure wherein there is noazimuth dependency. The upper curves aredisplaced by 1 dB from each other to easyviewing. The brackets denote azimuthalaveraging. The system ZDR sys is estimated tobe near to 0 dB (after introducing -0.3 db ofthe ZDR bias described in section 2).

6. ConclusionsThe NOAA’s research WSR-88D with

polarization capabilities has high isolationbetween polarization channels. High values ofthe Dhv, (0.995) in light rain exhibit goodpolarization purity of the radar. To measurethe ZDR with the accuracy of 0.2 dB, precisecalibration of the two receive chains isneeded. To find the system ZDR, severalmethods were used. From test signals wedetermined the receive channel contributionto the system ZDR. These measurements alongwith calibrations by the solar flux showsatisfactory stability of the system at least fora warm season. To close the transmit - receive ZDR calibration loop, two methods have beenapplied: 1) histogram of the ZDR of groundclutter in the interval of -7 to 7 dB and 2)high elevation azimuthal scans in cloudsabove the melting layer. By measuring theZDR in clouds at the highest possible elevationangle ( 60 degrees for the WSR-88D) andabove the melting layer, the system ZDR canbe determined with the accuracy of 0.1 dB.

Polarimetric system parameters of theresearch WSR-88D and their stability showthat the radar can be successfully used forpolarimetric measurements.

References

Bringi, V. N. and V. Chandrasekhar, 2001:Polarimetric Doppler WeatherRadar. Principles and Applications.Cambridge University Press. 636pp.

Crum, T.D., R. L. Alberty, and D. W.Burgess, 1993: Recording,archiving, and using WSR-88Ddata. Bull. Amer. Meteorol. Soc., 74, 645 – 653.

Doviak, R. J. and D. S. Zrnic, 1993: Doppler radar and weatherobservations, 2nd ed., AcademicPress, 562 pp.

Melnikov, V., D.S. Zrnic, R. J. Doviak, J. K.Carter, 2002: Status of the dualpolarization upgrade on the NOAAresearch and development WSR-88D. 18-th Internat. Conf. IIPS.Boston, AMS, p. 124 - 126.

Ryzhkov, A., and D. Zrnic, 1998a:Discrimination between rain and snowwith a polarimetric radar. Journal ofApplied Meteorology, 37, 1228-1240.

Ryzhkov, A., D. Zrnic and B. Gordon, 1998b:Polarimetric method for ice watercontent determination. Journal ofApplied Meteorology, 37, 125-134.