МЕТОДЫ ПОВЫШЕНИЯ КАЧЕСТВА ДАННЫХ ПОЛЯРИМЕТРИЧЕСКИХ МЕТЕОРОЛОГИЧЕСКИХ РАДАРОВ Текст научной статьи по специальности «Физика»

УДК 621.396

МЕТОДЫ ПОВЫШЕНИЯ КАЧЕСТВА ДАННЫХ ПОЛЯРИМЕТРИЧЕСКИХ

МЕТЕОРОЛОГИЧЕСКИХ РАДАРОВ

Рыжков Александр Васильевич

кандидат физико-математических наук, адьюнкт-профессор университета Оклахомы1, старший научный сотрудник Национальной лаборатории по исследованию сильных штормов2.

E-mail: [email protected].

Мельников Валерий Михайлович

кандидат физико-математических наук, адьюнкт-профессор университета Оклахомы1, старший научный сотрудник Национальной лаборатории по исследованию сильных штормов2.

E-mail: [email protected].

Душан Зрнич

академик Национальной инженерной Академии США, старший научный сотрудник Национальной лаборатории по исследованию сильных штормов2.

E-mail: [email protected].

1,2Адрес: 120 David L Boren Blvd., Norman, Oklahoma, USA, 73072.

Abstract: An overview of the methods for improving data quality of polarimetric weather radars is presented herein. The issues with data quality addressed in the paper include absolute calibration of radar reflectivity factor Z, absolute calibration of differential reflectivity ZDR, the need for correction for attenuation/differential attenuation in precipitation, and mitigation of partial beam blockage of the radar. Various methodologies are suggested for utilization on weather radars operating at S, C, and X bands. A data-based method for absolute calibration of Z capitalizes on the consistency between Z, ZDR, and specific differential phase KDP in rain. Different techniques for absolute calibration of ZDR are discussed: (1) system internal hardware calibration, (2) "birdbath" calibration with vertically pointing radar, (3) Z - ZDR consistency in light rain, (4) using dry aggregated snow as a natural calibrator for ZDR, and (5) using Bragg scatter as another natural target for calibration. Attenuation and radar beam blockage correction of Z and ZDR is performed using KDP and specific attenuation A which are immune to these factors.

Keywords: Polarimetric weather radars, Absolute calibration, Attenuation correction, Beam blockage mitigation, Data quality.

Introduction

Dual-polarization Doppler radars become a standard for operational networks of weather radars. Weather applications of dual-polarization radars are summarized by Ryzhkov et al. [1]. First network of polarimetric weather radars operating at S band has been completed in the US in 2013. Since then, similar operational weather radar systems have been either implemented or remain under development in Europe, Asia, and Australia. The Russian Federation follows a trend and, starting from 2011, а full-scale modernization of existing weather radar network by replacing old radars with C-band polarimetric Doppler radars (ДМРЛ-С) is underway (Ефремов и др., [2]; Дядюченко и др., [3]; Жуков и Щукин, [4].

Providing high quality of weather radar data is essential for producing reliable and robust hydro-logical and meteorological information useful to the scientific and operational communities. The accuracy of quantitative precipitation estimation (QPE) and hydrometeor classification directly depends on the quality of different radar variable estimates. Modern operational Doppler polarimetric radars directly measure radar reflectivity Z, differential reflectivity ZDR, differential phase Odp, cross-correlation coefficient phv, Doppler velocity v, Doppler spectrum width av, and linear depolarization ratio LDR (in the LDR mode of operation). Specific differential phase KDP is not directly measured but derived from 0DP. The

meaning of listed radar variables is explained in Bringi and Chandrasekar [5], Ryzhkov et al. [1].

The estimates of all these radar variables are obtained in the radar data processor from the time series of successive radar samples within the dwell time interval and are subject to random fluctuations caused by the statistical nature of the radar signal. The uncertainty of such estimates is characterized by bias (or accuracy) and standard deviation (or precision). The latter one is the measure of the "noisiness" of the estimate or the intensity of its temporal and spatial fluctuations. Several factors may cause bias in the estimates of different radar variables. These include (1) radar miscalibration, (2) impact of wet antenna radome, (3) attenuation in atmospheric gases and precipitation, (4) partial beam blockage (PBB), (5) ground clutter contamination, (6) low signal-to-noise ratio (SNR), (7) nonuniform beam filling (NBF), (8) depolarization from propagation in oriented ice crystals, and (9) multipath propagation (three-body scattering). These factors affect differently the biases of various radar variables. In this paper, a brief summary of the measurement errors and the methods to reduce such errors is presented.

2. Absolute calibration of Z

For most important practical applications of polarimetric weather radar, the radar reflectivity

factor Z should be calibrated with the accuracy of

1 dB, and differential reflectivity ZDR with the accuracy of 0.2 dB. These generally enable estimating rainfall within 15% accuracy (Ryzhkov et al.,

[6]). Better accuracy of the ZDR calibration (0.1

dB) might be needed for measurements of light rain or snow.

Polarimetric diversity provides a new method for absolute calibration of Z which was a longstanding problem for single-polarization radars. This methodology rests on the idea that Z, ZDR, and KDP are interdependent in rain and Z can be estimated from KDP and ZDR which are independent of absolute radar calibration. The difference between computed and measured values of Z is considered to be the Z bias. The consistency of Z, ZDR, and KDP in rain can be formulated as a dependence of the ratio KDP/Z on Zdr :

K

DP

z

= f (Z dr).

(1)

In (1), Z and KDP are in linear scale (i.e., mm6m-3 and deg km-1 respectively). The scatterplots of the ratio KDP/Z versus ZDR simulated from large DSD dataset in Oklahoma for three radar wavelengths and two temperatures, 0°C and 30°C, are illustrated in Fig. 1. It is evident that the dependence in (1) on temperature is negligibly small at S band where the effects of resonance scattering are insignificant. However, the temperature becomes an important factor at C band for ZDR > 2 dB and should be taken into account for all ZDR at X band. At S or C bands, Z can be estimated from known KDP and ZDR with the accuracy better than 1 dB if rain does not contain many resonance-size drops.

The function f(ZDR) can be well approximated by a fourth-order polynomial fit in certain range of ZDR so that (1) can be presented as

Fig. 1. Scatterplots of KDP/Z versus ZDR at S band (A = 11.0 cm), C band (X = 5.45 cm), and X band (A = 3.2 cm) for raindrop temperature 0°C (blue dots) and 30°C (red dots).

K

DP

Z

= 10 5(ao + a^DR + a2ZDR + a3ZDR ). (2)

In (2), ZDR is in decibels and the coefficients a0 - a3 for the S-, C-, and X-band radar wavelengths are listed in Table 1. It is important, that (2) with coefficients from the Table 1 is valid in the ZDR range 0.2 dB to 2 or 3 dB and that different consistency relations should be used for different temperatures at X band.

Because each of the three polarimetric variables in (2) has statistical errors and KDP is notoriously noisy in light rain (especially at longer radar wavelengths) it is instrumental to rewrite (2) as

Kdp = 1001Z(dBZ) f (Zdr) (3)

and integrate both sides of (3) over a sufficiently large spatial / temporal domain Q (Ryzhkov et al., [6]). The integral

Ii = J K dp dQ (4)

should be equal to the integral

I2 = |l001Z™f (ZDR)dQ, (5)

if measured reflectivity Zm is perfectly calibrated. The difference between I1 and I2 points to Z bias AZ which can be estimated as

AZ (dB) = 10log(V Ii). (6)

if Zm = Z + AZ. Because approximation (2) is valid only in the limited range of Zdr listed in the Table 1, the integrations (4) and (5) should be carried out only over the pixels of data within the appropriate range of Zdr (e.g., between 0.2 and 2.0 dB at C band). It is also required that data in the domain Q are not biased by low signal-to-noise ratio or contaminated by scatterers other than raindrops. These requirements are satisfied if SNR > 25 dB andphv > 0.99.

The methodology of matching the integrals I1 and I2 was first tested at S band on a large polarimetric dataset obtained during the Joint Polarization Experiment in Oklahoma and yielded an accuracy of Z calibration within 1 dB (Ryzhkov et al., [6]). To mitigate the impact of attenuation (particularly at C and X bands), Z and Zdr should be either corrected for attenuation using total differential phase &DP according to the methods described in Section 4 or only the data radials with sufficiently small span of &DP should be used for calibration.

3. Absolute calibration of ZDR

3.1 System internal calibration Relative internal calibration of Zdr can be achieved by measuring the differences between gains / losses in the two orthogonal channels. Because the transmission path and reception path differ, separate relative calibration of each is needed. Thus, the power ratio Ph /Pv downstream of the components that can cause bias in each path needs to be monitored. A change in either ratio would cause a corresponding relative drift in the Zdr bias which is then corrected (Zrnic et al., [7]). An additional step to account for the absolute bias must be made. The procedure is explained next by referring to the diagram in Fig. 2.

The relative values of the power ratios (in dB) are measured at two points. One is at the waveguide couplers Tch and Tcv on the transmission side; these extract powers from the corresponding H and V waveguides to establish the relative value in the transmission path. The other point is at the output of the two receivers when the signals are injected into the receiving couplers Rch and Rcv above the low noise amplifiers.

Table 1 - Coefficients a0 - a3 in (2) for S band (A = 11.0 cm), C band (A = 5.45 cm), and X band (A = 3.2 cm).

Frequency Temperature Zdr range a0 aj a2 a3

band (°C) (dB)

S 0-30 0.2-3.0 3.19 -2.16 0.795 -0.119

C 0-30 0.2-2.0 6.70 -4.42 2.16 -0.404

X 0 0.2-3.0 11.2 -4.75 0.349 -0.0532

X 10 0.2-3.0 10.9 -2.63 -1.22 0.341

X 20 0.2-3.0 10.4 0.109 -3.01 0.636

X 30 0.2-3.0 9.68 3.07 -4.67 0.869

Let the power ratio of outputs at the couplers TCh and T cv be

^x(to) = 10log[P (Tch)/P (Tcv)], (7) where t0 is a reference time stamp; in its proximity few more initial measurements must be made. AT(t0) is measured using one receiver (say H) as in Fig. 2 by switching between the outputs of Tch and Tcv. That way the receiver's transfer function does not affect the measurement. The AT(t0) should be stable over many hours because there are no separate active components in the path up to the couplers.

In the receiver path, a similar procedure is applied (Fig. 2). Note that the signal generator power is split (approximately 50:50) and the exact value at the splitter output is immaterial because the measurement is relative. Thus the power ratio is

¿Rfo) = 10log[P(Rch)/P(Rcv)], (8) and it is measured immediately after (7) to avoid possible changes between the measurements.

After these two measurements are made, one needs to establish the absolute bias. This is more challenging and few options have been tried. One is from Bragg scatterers (section 3.5) which produce zero ZDR, thus, the overall correct bias is the

value of ZDR measured from Bragg scatterers. Let that correct value be ^C(t0). It needs to be subtracted from the biased estimates denoted with

ZDRto obtain the corrected differential reflectivity

ZDR:

Zdr= ZDR - 4(0. (9)

This correction is valid if (7) and (8) do not change. AT(t0) normally does not change and it suffices to check it at intervals of several hours (eight on the WSR-88D). If a change, denoted with ^Tb = AT(t) - ^x(t0), does occur it should be subtracted, from (9) i.e.,

Zdr = Zdr -4(t0)-4(t) + 4(t0). (10) The same reasoning applies to the receiving part of the bias which, however, changes more often, and to catch these relatively fast changes, calibration of the receiving path is made at the end of each volume scan. This, is automated, and produces stable result. Thus, the receiver bias is ^Rb(ti) = AR(tj) - ^R(t0), at the times ti when volume scans end. The correction requires subtraction of ^Rb(ti) from all the data within the subsequent volume scan, and so on.

The sun can be a reference source and in case the transmitting path is well balanced the Sun flux may be sufficient for absolute calibration. Then the bias ASb revealed from the Sun scan can be substituted for Mt0) in (9).

3.2. "Birdbath " calibration of ZDR. Because the mean canting angle of raindrops is close to zero, raindrops appear spherical if viewed at vertical incidence and the measured ZDR in light rain with vertically pointing antenna should be close to 0 dB. Such calibration technique ("birdbath" calibration) is discussed in Gorgucci et al. [8] and Frech et al. [9] among others. This technique may work well only in light rain and in the absence of contamination from ground clutter via antenna sidelobes. Such contamination can cause azimuthal modula-

Fig. 2. Transmitter and receiver paths to the antenna. The couplers in the transmitter path Tch, Tcv tap the signals from the H, V waveguides close to the antenna; comparison is made sequentially, via Switch 2, in the H receiver. The signal generator's output is split and injected into the receiver couplers Rch, Rcv located above the low noise amplifiers in the H, V waveguides. During data collection the Switch 1 is open; it closes at the end of volume scans to enable automatic calibration of the receiver path

tion of ZDR for vertically looking rotating antenna. If this is the case, azimuthal averaging is needed for determining the ZDR bias or spectral filtering of the ground clutter components can be applied (Zrnic and Melnikov, [10]).

3.3. Z - ZDR consistency in light rain Small raindrops have nearly spherical shape and it is expected that ZDR in light rain dominated by small-size drops is relatively close to zero dB. Therefore, light rain may serve as a natural calibrator for ZDR measurements. This, however, is valid only in a general sense because raindrop size distributions associated with intense size sorting within convective updrafts are skewed towards larger drops and high values of ZDR may be measured in the areas of relatively low Z. Fig. 3 shows Z - ZDR dependencies corresponding to different percentiles of ZDR for a given Z in rain simulated from 47114 DSDs measured in Oklahoma. The simulations are for S band at T = 20°C. The domain between two dashed curves encompasses Z -ZDR pairs of the whole dataset. Thus, ZDR can be as high as 1 dB for Z = 20 dB. Nevertheless, in 80% of cases, ZDR at Z = 20 dBZ stays below 0.4 dB with average value of 0.23 dB.

The Z - ZDR dependencies in rain shown in Figs. 6.3 - 6.5 are valid at S band. Similar analysis at shorter radar wavelengths shows quite similar results for Z < 30 dB (see Table 2).

The procedure for ZDR calibration based on the radar measurements in rain can be easily automated so that the consistency between measured and expected values of ZDR in light rain is checked every radar scan if appropriate data are available. According to the automatic calibration routine implemented on the MeteoFrance operational radar network, the measured median ZDR at Z = 20 - 22 dBZ is compared with its

reference value 0.2 dB. It is also possible to estimate the ZDR bias as 1 6

^dr = - BZDm) (k) - < ZDr(k) >] (11)

6 k=1

where <ZDR(m)(k)> are median climatological values of ZDR in the kth 2-dB bin of Z shown in Table 2 and ZDR(m)(k) its value estimated from real radar data. Similarly to the self-consistency calibration of Z, the data appropriate for calibration of ZDR should be selected where SNR is sufficiently high (SNR > 20 - 25 dB), differential attenuation is insignificant, and rain scatterers are dominant contributors (i.e., phv > 0.98 - 0.99).

3.4. ZDR calibration using dry aggregated snow Dry aggregated snow is known for its small intrinsic ZDR caused by very low density. The study by Ryzhkov et al. [6] indicate that mean ZDR (i.e., averaged over a sufficiently large spatial / temporal interval) in aggregated snow usually does not exceed 0.2 dB if Z > 30 dBZ. Dry aggregated snow near the surface does not occur in warm cli-

Table 2 - Median climatological values of ZDR (dB) for different Z (dBZ) at S, C, and X bands in rain (20 < Z < 30 dBZ).

Z 20 22 24 26 28 30

Zdr(S) 0.23 0.27 0.32 0.38 0.46 0.55

Zdr(C) 0.23 0.27 0.33 0.40 0.48 0.56

Zdr(X) 0.23 0.28 0.33 0.41 0.49 0.58

S band

: 100% =

: / 80 % =

: / r / 60 % : / 40% =

^ 20 % : ^ 0% :

Ms :

Z (dBZ)

Fig. 3. Z - ZDR dependencies corresponding to various percentiles of ZDR for a given Z in rain. Z and ZDR are simulated at S band from 47144 DSDs measured in Oklahoma

matic zones. In addition, such a snow should be carefully separated from wet aggregated snow and dry crystallized snow that are characterized by a much higher and more variable ZDR. Nevertheless, dry aggregated snowflakes are commonly present above the melting layer in stratiform clouds (provided that Z > 30 dBZ). Numerous polarimetric radar measurements show that ZDR drops almost to 0 dB 1 - 2 km above the 0°C level where dry aggregated snow is most likely.

Quasi-vertical profiles (QVP, Ryzhkov et al., [11]) of Zdr in aggregates above the melting layer are suitable for monitoring deviation from expected low values. Because QVPs made from azi-muthal averages over 360o at high elevations, the accuracy of this measurement is better than 0.1 dB.

3.5. Using Bragg scatter for absolute calibration of ZDR Melnikov et al. [12] suggest using clear-air radar echoes associated with Bragg scattering for absolute calibration of ZDR. Bragg backscatter from refractive index perturbations at 5 cm scales creates sufficiently strong echo in a convective boundary layer to be detected by 10-cm-wavelength weather radars. These echoes are characterized by intrinsic Zdr equal to 0 dB and

cross-correlation coefficient phv very close to 1 making them easily distinguishable from the clear-air echoes caused by biota which have very large ZDR and low phv.

An automated algorithm for estimating ZDR bias from the Bragg scatter was developed and extensively tested on the S-band WSR-88D radars (Richardson et al., [13]). The algorithm yields better accuracy of the ZDR bias estimation than the methods based on the ZDR measurements in light rain and dry snow. Strong Bragg scattering usually occurs at the top of the boundary layer because there the gradients of humidity are largest and mixing by turbulence produces strongest returns. This is seen in Fig. 4 as a distinct layer of enhanced Z and close to zero ZDR. Application of thresholds (Z < 10 dBZ, SNR < 15 dB, phv < 0.98, and |v| > 2 m s-1) and some other criteria identifies data in the layer that are due to the Bragg scatter (Fig. 4b, top left); the histogram of ZDR (Fig. 4, right panel) is indeed centered on 0 dB.

4. Attenuation correction

Attenuation of microwave radiation in precipitation may significantly bias the measurements of Z and ZDR, especially at shorter radar wavelengths. Reliable correction of Z and ZDR is required before

Fig. 4. Example of Bragg scattering observed by the KMKX WSR-88D radar on 10 Nov 2013. (a) The fields of Z (upper left), ZDR (upper right), phv (lower left), and Doppler velocity (lower right) are from conical scans at the 3.5° elevation angle (1852 UTC). Maximum range in the image is ~22 km. (b) The ZDR histogram (right) is from the data (top left corner) which have passed Bragg detection criteria. Data that have passed the SNR > 2 dB threshold are in the bottom left image. (From Richardson et al., [13]).

utilizing these radar variables for quantitative rainfall estimation, hydrometeor classification, micro-physical retrievals, etc. Attenuation and differential attenuation in rain cause negative biases in Z and ZDR (AZ and AZDR respectively) which can be estimated from the total span of differential phase ®dp along the propagation path (AOdp). Specific attenuation A and specific differential attenuation Adp are generally proportional to specific differential phase KDP:

A = aKDP and ADP = ßK

'-DP ■

Therefore,

r r

AZ (r) = 2 J A(s)ds = 2a J KDP(s)ds =

(12)

(13)

= aODp(r)

and

г г

AZDR(r) = 2 J Adp (s)ds = 2ßJ KDp(s)ds =

= P®Dp(r )

if the factors a and P do not change much along the propagation path (0,r) (Bringi et al., [14]).The fact that attenuation biases of Z and ZDR are directly proportional to the differential phase is an advantage of polarimetric radars because it enables accurate quantification of precipitation in the presence of strong attenuation at shorter radar wavelengths (C and X bands).

The factors a and P in (12) are sensitive to the variability of raindrop size distributions and temperature. Typical range of their variability at different radar wavelengths is shown in . Attenuation correction in the first approximation can be made using "default" or average values in the right column in Table 3. It produces substantial improvement in Z and ZDR compared to the absence of correction. The efficiency of default linear correction using (13) and (14) at C band with <a> = 0.08 dBdeg-1 and <P> = 0.02 dB deg-1 is demonstrated

in Fig. 5 for the case of a tornadic storm in Oklahoma. The fields of Z and ZDR measured by the C-band OU-PRIME radar show large negative biases before attenuation correction is applied (Fig. 5a,b). The biases are largest along azimuthal directions where total differential phase is highest (Fig. 5c). The corrected fields of Z and Zdr in Fig. 5e,f are consistent with the ones measured by the collocated S-band radar (not shown).

5. Mitigation of partial beam blockage.

Beam blockage caused by terrain and other obstacles such as buildings and trees limits radar coverage and introduces bias in measurements. Therefore, the quality of the weather radar products such as quantitative precipitation estimate (QPE) is compromised. One of the most common methods for mitigation of partial beam blockage (PBB) uses a digital elevation map (DEM) to estimate the degree of beam blockage at particular azimuths and elevations based on geometry of the beam and its occultation. The DEM-based correction method may not work well if the degree of blockage exceeds 60%. In addition to larger-scale terrain features, small-scale anthropogenic structures (e.g., towers, buildings) and nearby trees that are not accounted for by DEMs can cause additional occultation of the radar beam.

The problem of the partial beam blockage can be resolved more efficiently with the dual-polarization radar than with the single-polarization radar because the former can directly measure differential phase ®DP and estimate specific attenuation A over a propagation path (r1, r2) as follows (Ryzhkov et al. [15].)

Table 3 Ranges of variability of the factors a and P in rain at S, C, and X bands

(14)

S band

a = 0.015 - 0.04 dB/deg <a> = 0.02 dB/deg

ß = 0.0025 - 0.009 dB/deg <ß> = 0.004 dB/deg

C band

a = 0.05 - 0.18 dB/deg <a> = 0.08 dB/deg

ß = 0.008 - 0.1 dB/deg <ß> = 0.02 dB/deg

X band

a = 0.14 - 0.35 dB/deg <a> = 0.28 dB/deg

ß = 0.03 - 0.06 dB/deg <ß> = 0.05 dB/deg

(a) Hadar retted ivrty ^од

so;

ю го зо ¿о so во

¡((кия)

Radar reflectivity (dBZ)

зо jü so X(kxn)

<Ъ)

№ 50

60

ss 50

so

4S 40

40 „

35 I 30

30 >

2£ го

20

16 ю

10

5 0

0

№

300 Ё0

250

200 50

160

120 40

100

90 1 30

60 y-

40 го

30

го 10

Ю

* 0

0

(f)

66 60

60

55 so

£0

45

40

35 E 30

Э0 S-

15 го

20

is Ю

10

s 0?

0

го и -из so eo to

X( km)

Cross-correiation coefficient

io го эо to sc во то

X(km]

Dilferenjiai reflectivity {dB)

30 AÜ so

Fig. 5. Composite plot of Z, ZDR, <tDP, and pilv measured by the C-band OU-PRIME radar at elevation 0.5° in the tornadic storm in central Oklahoma on May 10, 2010 at 2042 UTC (panels a - d). The fields of Z and ZDR corrected for attenuation are displayed in panels (e) and (f)

A(r) =

[Za(r )]bC (b, PIA) I (r1, r2) + C (b, PIA) I (r, r2),

where

'2

I(r1, r2) = 0.46b J[Za(s)]bds ,

(15)

(16)

I (r, r2) = 0.46b f[Z a (s)]b

(17)

C (b, PIA) = exp(0.23bPIA) -1, (18) PIA = a [Odp (r2) - Odp ft)] = a AO dp , (19) where b is a constant and Za is the measured radar reflectivity factor which can be biased.

It is evident that the estimate of specific attenuation A from a radial profile of Za and a total span of differential phase A@DP is totally immune to the Z biases caused by attenuation, radar miscalibration, partial beam blockages, and wet radome. Indeed, if attenuated Z (Za in (15)) expressed in linear scale is multiplied by an arbitrary constant Z along the propagation path (ru r2), then the value of A remains intact because the numerator and denominator in (15) are multiplied by the same factor Z which is cancelled out in the ratio. This property of the A estimate by (15) proves to be very beneficial for quantification of rainfall in the partially blocked areas of radar returns if the

A-based algorithm is used for rainfall estimation. The radar reflectivity factor unbiased by PBB can be estimated from A using the Z(A) relation which is an inverted relation A = aZb.

The performance of this technique is illustrated in Fig. 6 where the fields of the measured X-band Z and Zdr (before correction for attenuation and beam blockage) at antenna elevation 1.5° are displayed along with the fields of Odp and radar reflectivity corrected for attenuation and PBB. It is obvious that that the PBB-related Z bias in a narrow SE sector is completely eliminated in the panel (c) of Fig. 6.

Radar reflectivity (dBZ)

Ф) Differential relieciivity (d&>

0

X (hm)

Fig. 6. Composite plot of measured Z and ZDR before correction for attenuation and beam blockage ((a) and (b)), Z after correction (c), and differential phase (d). The measurements are made by the University of Bonn X-band polarimetric radar on June 22, 2011 at 1126 UTC at elevation 1.5°

Литература

1. Ryzhkov A., Schuur T., Melnikov V., Zhang P., Kumjian M. Weather applications of dual-polarization radars // Радиотехнические и телекоммуникационные системы. 2016. № 2. C. 28-33.

2. Ефремов В.С., Вовшин Б.М., Вылегжанин И.С., Лаврукевич В.В., Седлецкий Р.М. Поляризационный доплеровский метеорологический радиолокатор С-диапазона со сжатием импульсов // Журнал радиоэлектроники. 2009. № 10.

3. Дядюченко В.Н., Вылегжанин И.С., Павлю-ков Ю.Б. Доплеровские радиолокаторы в России // Наука в России. 2014. № 1. С. 23-27.

4. Жуков В.Ю., Щукин Г.Г. Состояние и перспективы сети доплеровских метеорологических радиолокаторов // Метеoрология и гидрология. 2014. № 2. С. 92-100.

5. Bringi V., Chandrasekar V. Polarimetric Dop-pler Weather Radar: Principles and Applications. Cambridge University Press, 2001. 636 p.

6. Ryzhkov A.V., Giangrande S.E., Melnikov V.M., Schuur T.J. Calibration issues of dual-polarization radar measurements // Journal of Atmospheric and Oceanic Technology. 2005. № 22. Pp. 1138-1155.

7. Zrnic, D., Melnikov V., Carter J. Calibrating differential reflectivity on the WSR-88D // Journal of Atmospheric and Oceanic Technology. 2006. № 23, Pp. 944-951.

8. Gorgucci E., Scarchilli G., Chandrasekar V. A procedure to calibrate multiparameter weather radar using properties of the rain medium // IEEE Trans. Geosci. Remote Sensing. №37. Pp. 269-276.

Поступила 30 марта 2018 г.

9. Frech M., Hagen M., Mammen T. Monitoring the absolute calibration of a polarimetric weather radar // Journal of Atmospheric and Oceanic Technology. 2017. № 34. Pp. 599-615.

10. Zrnic D.S., Melnikov V.M. Ground clutter recognition using polarimetric spectral parameters // 33rd Conference on Radar Meteorology, AMS, Cairns, Australia. 2007.

11. Ryzhkov A., Zhang P., Reeves H., Kumjian M., Tschallener T., Simmer C., Troemel S. Quasi-vertical profiles - a new way to look at polarimetric radar data // Journal of Atmospheric and Oceanic Technology. 2016. № 33. Pp. 551-562.

12. Melnikov V., Doviak R., Zrnic D., Stensrud D. Mapping Bragg scatter with a polarimetric WSR-88D // Journal of Atmospheric and Oceanic Technology. 2011. № 28, Pp. 1273-1285.

13. Richardson L., Cunningham J., Zittel W., Lee R., Ice R., Melnikov V., Hoban N., Gebauer J. Bragg scatter detection by the WSR-88D. Part I: Algorithm development. // Journal of Atmospheric and Oceanic Technology. 2017. № 34. Pp. 465-478.

14. Bringi V. N., Chandrasekar V., Balakrishnan N., Zrnic D. S. An examination of propagation effects in rainfall on polarimetric variables at microwave frequencies // Journal of Atmospheric and Oceanic Technology. 1990. № 7. Pp. 829-840.

15. Ryzhkov A., Diederich M., Zhang P., Simmer C. Utilization of specific attenuation for rainfall estimation, mitigation of partial beam blockage, and radar networking // Journal of Atmospheric and Oceanic Technology. 2014. № 31. Pp. 599-619.

English

MEASUREMENT ERRORS OF METEOROLOGICAL POLARIMETRIC RADARS AND THE WAYS TO MITIGATE THEM

Alexander V. Ryzhkov - PhD; adjunct professor, University of Oklahoma1; senior research scientist, National Severe Storms Laboratory2. E-mail: [email protected].

Valery M. Melnikov - PhD; adjunct professor, University of Oklahoma1; senior research scientist, National Severe Storms Laboratory2. E-mail: [email protected].

Dusan Zrnic - PhD; US National Academy of Engineering member, senior research scientist, National Severe Storms Laboratory2. E-mail: [email protected];

1,2Address: 120 David L Boren Blvd., Norman, Oklahoma, USA, 73072.

Abstract: An overview of the methods for improving data quality of polarimetric weather radars is presented herein. The issues with data quality addressed in the paper include absolute calibration of radar reflectivity factor Z, absolute calibration of differential reflectivity ZDR, the need for correction for attenuation/differential attenuation in precipitation, and mitigation of partial beam blockage of the radar. Various methodologies are suggested for utilization on weather radars operating at S, C, and X bands. A data-based method for absolute calibration of Z capitalizes on the consistency between Z, ZDR, and specific differential phase KDP in rain. Different techniques for

absolute calibration of ZDR are discussed: (1) system internal hardware calibration, (2) "birdbath" calibration with vertically pointing radar, (3) Z - ZDR consistency in light rain, (4) using dry aggregated snow as a natural calibrator for ZDR, and (5) using Bragg scatter as another natural target for calibration. Attenuation and radar beam blockage correction of Z and ZDR is performed using KDP and specific attenuation A which are immune to these factors. Key words: Polarimetric weather radars, Absolute calibration, Attenuation correction, Beam blockage mitigation, Data quality.

References

1. Ryzhkov A., Schuur T., Melnikov V., Zhang P., Kumjian M. Weather applications of dual-polarization radars // Радиотехнические и телекоммуникационные системы. 2016. No. 2. Pp. 28-33.

2. Efremov V.S., Vovshin B.M., Vylegzhanin I.S., Lavrukevich V.V., Sedleckij R.M. Polarizing Doppler weather radar of C-band with pulse compression // Journal of radio electronics. 2009. No. 10.

3. Dyadyuchenko V.N., Vylegzhanin I.S., Pavlyukov Y.B. Doppler radars in Russia // Science in Russia. 2014. No. 1. Pp. 23-27.

4. Zhukov V. Y., Shchukin G. G. Status and prospects of the Doppler meteorological radar network // Meteorology and hydrology. 2014. No. 2. Pp. 92-100.

5. Bringi V., Chandrasekar V. Polarimetric Doppler Weather Radar: Principles and Applications. Cambridge University Press, 2001. 636 p.

6. Ryzhkov A.V., Giangrande S.E., Melnikov V.M., Schuur T.J. Calibration issues of dual-polarization radar measurements // Journal of Atmospheric and Oceanic Technology. 2005. No. 22. Pp. 1138-1155.

7. Zrnic D., Melnikov V., Carter J. Calibrating differential reflectivity on the WSR-88D // Journal of Atmospheric and Oceanic Technology. 2006. No. 23. Pp. 944-951.

8. Gorgucci E., Scarchilli G., Chandrasekar V. A procedure to calibrate multiparameter weather radar using properties of the rain medium // IEEE Trans. Geosci. Remote Sensing. No. 37. Pp. 269-276.

9. Frech M., Hagen M., Mammen T. Monitoring the absolute calibration of a polarimetric weather radar // Journal of Atmospheric and Oceanic Technology. 2017. No. 34. Pp. 599-615.

10. Zrnic D.S., Melnikov V.M. Ground clutter recognition using polarimetric spectral parameters // 33rd Conference on Radar Meteorology, AMS, Cairns, Australia. 2007.

11. Ryzhkov A., Zhang P., Reeves H., Kumjian M., Tschallener T., Simmer C., Troemel S. Quasi-vertical profiles - a new way to look at polarimetric radar data // Journal of Atmospheric and Oceanic Technology. 2016. No. 33. Pp. 551-562.

12. Melnikov V., Doviak R., Zrnic D., Stensrud D. Mapping Bragg scatter with a polarimetric WSR-88D // Journal of Atmospheric and Oceanic Technology. 2011. No. 28, Pp. 1273-1285.

13. Richardson L., Cunningham J., Zittel W., Lee R., Ice R., Melnikov V., Hoban N., Gebauer J. Bragg scatter detection by the WSR-88D. Part I: Algorithm development. // Journal of Atmospheric and Oceanic Technology. 2017. No. 34. Pp. 465-478.

14. Bringi V. N., Chandrasekar V., Balakrishnan N., Zrnic D. S. An examination of propagation effects in rainfall on polarimetric variables at microwave frequencies // Journal of Atmospheric and Oceanic Technology. 1990. No. 7. Pp. 829-840.

15. Ryzhkov A., Diederich M., Zhang P., Simmer C. Utilization of specific attenuation for rainfall estimation, mitigation of partial beam blockage, and radar networking // Journal of Atmospheric and Oceanic Technology. 2014. No. 31. Pp. 599-619.