polarimetric radar observations of hail...

AUGUST 2001 1347K E N N E D Y E T A L .

q 2001 American Meteorological Society

Polarimetric Radar Observations of Hail Formation

PATRICK C. KENNEDY, STEVEN A. RUTLEDGE, AND WALTER A. PETERSEN

Department of Atmospheric Science, Colorado State University, Fort Collins, Colorado

V. N. BRINGI

Department of Electrical Engineering, Colorado State University, Fort Collins, Colorado

(Manuscript received 24 March 2000, in final form 26 December 2000)

ABSTRACT

Analyses are made of the evolution of selected polarimetric radar data fields during periods immediatelypreceding the onset of near-surface hail indicators [high reflectivity and low differential reflectivity (Zdr)] in twononsupercellular northeastern Colorado hailstorms. The primary data were obtained from the 11-cm-wavelength,dual-polarization Colorado State University (CSU)–University of Chicago and Illinois State Water Survey radar.In one of the storms, dual-Doppler wind field syntheses were available using additional velocity data collectedby the CSU Pawnee S-band radar. In both events, linear depolarization ratio (LDR) values exceeding 225 dBbegan to appear in the right flank of the 50-dBZ echo core region, within the 08 to 2208C environmentaltemperature range, approximately 10 minutes prior to the onset of hail at the surface. Scattering calculationssuggest that the LDR enhancement may have been caused by an increasing water fraction within the growinghailstones (spongy hail), or the development of a liquid water coat under wet growth conditions. Vertical structureof the Zdr fields was also examined. As hypothesized by Conway and Zrnic, it was found that the distinctnessof the positive Zdr column associated with supercooled raindrops and incompletely frozen particles above the08C height varies from storm to storm.

1. Introduction

In general, the growth of hailstones is maximizedwhen they collect appreciable quantities of supercooledcloud droplets, a condition most often attained in vig-orous updrafts (Ludlam 1980). The resultant microwaveradar backscatter properties are strongly affected by theevolution of hydrometeor size, concentration, phase,shape, and fall mode (Bringi et al. 1986b). Early ob-servations of hailstorms were made using single-param-eter radars. For example, Donaldson (1961) examinedvariations of radar reflectivity with height and foundthat echoes from storms generating large hail in NewEngland typically penetrated the tropopause by about1.5 km, but the echo summit of storms producing onlysmall hail usually did not penetrate the tropopause. Overthe last several decades, a variety of multiparameterradar observations of hailstorms have been made. (Here‘‘multiparameter’’ implies radar observations made atmore than one polarization and/or wavelength.) Atlasand Ludlam (1961) noted that appreciable reflectivitydifferences occurred when several radars operating on

Corresponding author address: Patrick C. Kennedy, Dept. of At-mospheric Science, Colorado State University, Fort Collins, CO80523.E-mail: [email protected]

a variety of wavelengths scanned the same hailstorm.The reflectivity differences were attributed to the wave-length-dependent effects of Mie scattering. These ob-servations suggested that the presence of hail could beinferred from a comparison of the reflectivities obtainedthrough the simultaneous use of two different wave-lengths. In principle, the reflectivities observed at thetwo wavelengths would become different when the pres-ence of centimeter-sized hail made Mie resonance ef-fects significant at the shorter wavelength. Several ap-plications of the dual-wavelength hail detection methodhave been made (Heymsfield et al. 1980; Bringi et al.1986b; Tuttle et al. 1989; Miller et al. 1990). However,the dual-wavelength method is very sensitive to antennapattern matching at the two wavelengths (Rinehart andTuttle 1982) as well as to attenuation effects at the short-er wavelength (Tuttle and Rinehart 1983).

To avoid the technological challenges posed by dual-wavelength methods, several polarimetric techniques (ata single wavelength) have been used to study hailstorms.Seliga and Bringi (1976, 1978) noted that inherent meanshape differences between freely falling hailstones andraindrops would have measurable microwave backscat-ter effects. To be specific, raindrops with diameters ex-ceeding 1.5 mm assume oblate shapes with their majoraxes approximately horizontally oriented because of

1348 VOLUME 40J O U R N A L O F A P P L I E D M E T E O R O L O G Y

combined gravitational, aerodynamic, and surface-ten-sion effects (Pruppacher and Beard 1970). In contrast,hailstones are much more resistant to aerodynamic de-formation and are more likely to tumble while falling(Knight and Knight 1970). Thus, the backscatteringcross section of oblate raindrops will be greater at hor-izontal linear polarization (H) than at vertical linear po-larization (V), whereas the H and V backscattering crosssections of hailstones will be more nearly equal. Seligaand Bringi (1976, 1978) proposed differential reflectiv-ity Zdr as an expression of the H, V backscatter differ-ence:

Z (dB) 5 10 log (Z /Z ),dr 10 HH VV (1)

where ZHH is the copolar reflectivity at horizontal po-larization and ZVV is the copolar reflectivity at verticalpolarization.

Aydin et al. (1986) substantiated the utility of thistechnique by documenting ;0-dB S-band Zdr values innear-surface thunderstorm echo-core regions where thepresence of hail was confirmed by ground observations.However, the raindrop–ice particle mean shape differ-ence responsible for this Zdr hail signature generally dis-appears at subfreezing temperatures as the general pres-ence of frozen scatterers quickly drives overall echo Zdr

values toward 0 dB (Herzegh and Jameson 1992). Animportant departure from the preponderance of near-0-dB Zdr values in the subfreezing portions of convectiveechoes was noted by Illingworth et al. (1987). Theydocumented several cases in which narrow (1–2 km indiameter), positive Zdr ‘‘columns’’ (4–6 dB) were foundto extend briefly to heights of several kilometers abovethe 08C level in isolated convective clouds that weredeveloping in low–wind speed environments. They at-tributed this observation to the presence of large (.4mm in diameter) supercooled raindrops that were loftedabove the 08C level by strong updrafts. Conway andZrnic (1993) found the positive Zdr column to be a muchlonger-lived feature in a major Colorado hailstorm. Intheir case, the column was persistent during the ;40-minute quasi-steady-state period analyzed. The associ-ated hail trajectory computations showed that significanthailstone growth typically occurred along trajectorysegments that passed either through or immediatelyabove the positive Zdr column.

None of these techniques made use of the depolarizedcomponent of the backscattered signal. Conceptually,depolarization is increased when the axes of symmetryof nonspherical particles are misaligned with the inci-dent electric field vector (Atlas et al. 1953). For ex-ample, when the illuminating electric field is linearlypolarized, the cross-polar return from an oblate particleis maximized when the particle’s major diameter is ori-ented at a 458 angle to the incident electric field. Thesedepolarization effects are magnified as the scatterer’swater fraction increases, because of an increase in thebulk dielectric factor. Thus hailstones, particularly whenthey are experiencing wet or spongy growth conditions

(Lesins and List 1986), may induce appreciable depo-larization levels because their shapes are typically non-spherical and their orientations often fluctuate signifi-cantly as they fall (Herzegh and Jameson 1992). Manyof the original investigations of hail-related depolariza-tion effects were carried out by Canadian researchersstudying the interaction of circularly polarized radiationwith various types of precipitation (McCormick andHendry 1975). In circularly polarized radiation, the in-stantaneous polarization vector rotates with distancealong the direction of propagation while maintainingfixed amplitude. The circular depolarization ratio (CDR)is defined by:

CDR(dB) 5 10 log (Z /Z ),10 parallel orthogonal (2)

where Zparallel is the radar reflectivity based on the returnsignal with the same polarization sense as the trans-mitted pulse and Zorthogonal is the radar reflectivity basedon the return signal with the opposite polarization senseas the transmitted pulse. For example, if left-handedcircular is transmitted, Zparallel is the power return for LHcircular and Zorthogonal is the power returned for RH cir-cular. For spherical targets, CDR 5 2`. Increased CDRvalues measured with a 10-cm-wavelength radar (i.e.,increased depolarization levels) were found to correlatewith surface observations of hailstones larger than about3 cm in diameter (Barge 1970). Enhanced CDR levelswere also found in the elevated echo region overhanginga weak echo region (WER, Browning and Foote 1976),an echo area in which hail growth was likely. It wasalso found that when moderate to heavy rain rates werepresent along the beam path, differential propagationeffects caused the incident circularly-polarized radiationto become elliptically polarized, complicating accurateCDR measurement (Humphries 1974).

Depolarization by hail has also been studied usinglinear polarization states. The linear depolarization ratio(LDR) is given by:

LDR(dB) 5 10 log (Z /Z ),10 VH HH (3)

where ZVHis the cross-polar reflectivity: receive V andtransmit H.

Bringi et al. (1986b) found that LDR observed at 3-cm wavelength was enhanced in the presence of hail.These LDR enhancements were also seen to extend ver-tically into the subfreezing cloud region above the lo-cations where ;0-dB Zdr values indicated hail at groundlevel. As might be expected, the effects of attenuationon these X-band signals had to be carefully considered.However, the association between enhanced X-bandLDR values at subfreezing temperatures and the pres-ence of hail is not unambiguous. Smith et al. (1999)analyzed a region of .221-dB X-band LDR valuesobserved just above the top of a positive Zdr column ina Florida thunderstorm cell. In situ observations showedthat the LDR enhancement was due raindrops in theprocess of freezing, not ice particles in wet growth.Holler et al. (1994) reported on 5-cm-wavelength radar


FIG. 1. The eastern lobe (.308 beam intersection angles) of theCSU–CHILL/Pawnee dual-Doppler network. CAPPI inset shows the6-km-MSL reflectivity pattern at 2355 UTC 15 Jul 1998. Startingcontour is 30 dBZ, contour interval is 10 dB, and areas within the50-dBZ contour are filled.

TABLE 1. Characteristics of the CSU–CHILL and CSU–Pawnee radar systems. SNR is signal-to-noise ratio.

CHILL Pawnee

Wavelength (cm)3-dB beamwidth (8)Peak transmit power (kW)Pulse duration (m)Pulse repetition frequency (Hz)Noise power (SNR 5 1; dBm)PolarizationPulses per integration cycleLat (8N)Long (8W)Antenna elevation (m MSL)

11.0091.0

8001

10002114

H, V128 (64 H 1 64 V)40.446

104.6371432

10.9891.6

3801

9622109

V12840.871

104.7151688

observations made during the vigorous stage of a hy-brid-type hailstorm in Germany. They found enhancedLDR values (.225 dB) to be well-correlated with theexistence of hail, particularly when the hail was expe-riencing wet growth conditions. Hubbert et al. (1998)analyzed the complete set of available polarimetric datain a supercell hailstorm in northeastern Colorado usingthe Colorado State University–University of Chicagoand Illinois State Water Survey (CSU–CHILL) radar.They were able to relate a positive Zdr column and areasof enhanced LDR at midlevels to the production of largehail.

The goal of this study is to document the evolutionof S-band LDR and Zdr patterns within the time periodimmediately preceding the onset of hail signatures atthe lowest antenna elevation angle (interpreted as anindicator of surface hail). Observations are presented

from two hail-producing nonsupercell thunderstormsthat took place over the high plains of Colorado. Oncethe low-elevation-angle hail signatures appeared, the po-larimetric data indicated that both storms continued todeposit hail along 20–30-km swaths. During one ofthese storms, air motion fields could be synthesized fromdual-Doppler measurements. Last some aspects of theobserved evolution of LDR enhancement patterns aresimulated through microwave scattering model calcu-lations, as related to cloud microphysical processes.

2. Radar instrumentation and general dataanalysis procedures

The polarimetric radar data used in this study werecollected by the S-band CSU–CHILL system locatednear Greeley, Colorado. The system’s polarimetric mea-surement capabilities are enhanced through the use ofa dual-transmitter–dual-receiver configuration (Brun-kow et al. 2000). For all of the data described in thisstudy, the dual transmitters were fired alternately, pro-ducing a steady H, V, H, V, . . . polarization sequence.The radar’s dual receivers allow both the co- and cross-polar signal returns to be processed from each trans-mitted pulse. The LDR was computed according to Eq.(3).

During the summer of 1998, CSU also started op-erating the Pawnee Doppler radar, an 11-cm-wave-length, single-polarization Doppler radar located 48 kmNNW of the CSU–CHILL site. This radar network per-mits dual-Doppler scanning to be maintained over por-tions of both the Front Range foothills and the adjacenthigh plains (Fig. 1). Summaries of each radar’s perfor-mance characteristics are given in Table 1.

The analyzed cases were selected based upon the re-quirement that full three-dimensional echo scanning wasin progress prior to the first observation of a hail sig-nature at the lowest elevation angle. The radar scanningstrategy varied among the two cases, affecting primarilythe volume scan duration and elevation-angle step size.The hail signature was defined by the differential re-flectivity hail signal (HDR) developed by Aydin et al.(1986). HDR is calculated by subtracting a prescribedfunction of Zdr from ZH:


HDR 5 Z 2 f (Z ), (4)H dr

where

27 when Z # 0 dBdrf (Z ) 5 19Z 1 27 when 0 , Z # 1.74 dB (5)dr dr dr60 when Z . 1.74 dB. dr

The piecewise continuous Zdr function given by Eq. (2)was based on disdrometer measurements and representsthe limiting Zdr values that were realized in pure rainwith varying reflectivity levels. Thus, positive HDR val-ues signify a situation in which the observed Zdr is ap-preciably below that expected from pure rain with thesame reflectivity. Aydin et al. (1986) found good cor-relation between HDR values and observations of sur-face hail in Colorado thunderstorms. In this study, theonset of surface hail was taken to be the first appearanceof HDR values exceeding 5 dB in the lowest sweep(nominally, the 0.58 elevation sweep).

Once the radar-determined hail onset time was iden-tified, data from the preceding volumes were examinedin detail. Considerable efforts were made to eliminateartifacts from the radar data. The data were interactivelyedited using the National Center for Atmospheric Re-search’s Research Data Support System software (Oyeand Carbone 1981). Conventional methods of applyingthresholds were used to remove noise effects, and ve-locities were interactively dealiased. Polarimetric radardata were excluded when the correlation at zero timelag between the H and V return signals (rhv(0) Balak-rishnan and Zrnic 1990) fell below 0.80–0.85. To avoidcorruption from low signal-to-noise ratios, LDR valueswere only used when the signal plus noise power in thecross-polar channel was at least 2 dB above the asso-ciated receiver noise level. Last, to limit artifacts intro-duced by mismatches between the antenna H and Vradiation patterns, both the Zdr and LDR range gate val-ues were deleted from areas where either the azimuthplane or elevation plane spatial reflectivity gradients(cross-beam gradients) exceeded a range-dependentthreshold of approximately 23 dB km21 (Brunkow etal. 2000).

It was necessary to determine LDR values that in-dicated a significant degree of depolarization, particu-larly in subfreezing echo areas. Mie scattering effectsmaterially increase the return signal depolarization; thusthe large LDRs (exceeding 216 dB) observed at X bandby Bringi et al. (1986b) in an intense hailstorm are notexpected at S band. Numerous histograms generatedfrom CSU–CHILL data at subfreezing temperature lev-els showed that the average LDR levels were generallyabout 228 to 226 dB in most convective echo regimes(typical of medium density graupel, Holler et al. 1994).LDRs often increased to values greater than 225 dBnear the echo summit, perhaps due to an increasing frac-tion of higher-density, less spherically shaped, anvil-level ice particles. In this work, LDR $ 225 dB was

taken to indicate the presence of hail in regions asso-ciated with subfreezing temperatures and within the‘‘active’’ (nonanvil) portions of a convective stormecho.

3. Radar observations

a. 15 July 1998

During the predawn hours of 15 July 1998 a weakupslope wind pattern had developed at the surface andincreased the low-level moisture content. By 1925 UTCthe dewpoint temperature at Greeley had reached10.68C. Based on the 1200 UTC Denver (DNR) sound-ing modified to reflect the late-afternoon environment,the lifting condensation level (LCL) was estimated tobe at a pressure of 590 hPa and a temperature of 14.78C,with a mixing ratio of ;8.7 g kg21. (No DNR soundingdata were available for 0000 UTC 16 July 1998.) Widelyscattered thunderstorms that developed over the RockyMountain foothills moved eastward past the CSU–CHILL radar during the late-afternoon hours. Between2345 UTC 15 July and 0030 UTC 16 July, these stormsorganized into a short north–south-oriented line locatedwithin the eastern lobe of the CSU dual-Doppler net-work (Fig. 1). (All time references are in UTC; localtime is UTC minus 6 h.) The echo from the storm ofinterest began to develop rapidly around 2340 and by0015 had become a major element in the evolving con-vective line. HDR values (Aydin et. al. 1986) were com-puted for each gate in the 0.58 elevation-angle data. Anexpanding area of positive (.5 dB) HDR values wasfirst observed in the 0004 data; therefore this was takento be the time that hail started to arrive at the surface(Figs. 2a,b). Because of extremely low population den-sities, surface observations related to this HDR featurewere not available. The nearest available hail report was25-mm (1 in.)–diameter stones reported near X 5 47,Y 5 51 km from the CSU–CHILL radar at 0025 UTC.This point was in the path of the northernmost echo coreshown in Fig. 1.

Dual-Doppler air motion analyses were carried outfor the three volume scans immediately preceding theappearance of the 0.58 elevation-angle HDR hail indi-cator. In each of these dual-Doppler analyses, a Cress-man (Cressman 1959) weighting scheme was used tointerpolate the edited data from the CSU–CHILL andPawnee radars to a common Cartesian grid. Grid heightswere referenced to sea level. The grid spacing was 1km in X and Y and 0.75 km in Z; the radius of influencewas 0.5 km. Horizontal wind field synthesis based onthe gridded radial velocities was done via the CustomEditing and Display of Reduced Information in Carte-sian Space software package (Mohr and Miller 1983).A correction for echo advection based on the radar-observed storm motion was applied during the synthe-ses, producing analyses valid at 2349, 2355, and 2359.The three synthesized airflow components were itera-


FIG. 2. (a) Shaded contours of reflectivity values (40, 50, and 60 dBZ ) at 0.68 elevation angle for 2358 UTC 15 Jul 1998. Labeled arc isa constant-height line (km MSL) on the PPI scan surface. Terrain height is ;1.4 km MSL. Plotted symbols depict individual range gateHDR values (dB), where a plus indicates HDR . 5 dB and a dot indicates HDR . 15 dB. See text for explanation of HDR. Large ‘‘1’’indicates approximate center of the HDR hail signature at 0004 UTC. (b) As in (a), except data were obtained at 0004 UTC 16 Jul 1998.

tively adjusted to account for the effects of vertical mo-tions on the horizontal winds obtained from the dual-Doppler equations. Divergence fields were calculatedfrom the adjusted horizontal winds at each analysisheight. Vertical airflow was obtained by vertically in-tegrating these horizontal divergence fields. The inte-gration procedure imposed a variational constraint thatforced the vertical velocities to 0 m s21 at the endpointsof the integration paths.

In the 2349 analysis, the storm cell of interest wasintensifying aloft. At 6 km (about 2128C environmentaltemperature), this development was taking place withinthe curved 30-dBZ echo appendage that extended ap-proximately 20 km SSE from a preexisting storm (Fig.3). The reference lines in this figure that cross X 5 37.4,Y 5 34.2 km show the location where the 0.58 elevation-angle HDR hail signature will appear at 0004. Moderatesoutherly storm relative airflow was found over the fu-ture hail reference site. Stronger southeasterly inflowwas located ;10 km further west (X 5 27, Y 5 33 km).A small region of positive (.1 dB) Zdr values was lo-cated approximately 3 km southwest of the referencepoint (X 5 34, Y 5 32 km). At two grid points withinthis region, Zdr exceeded 11.5 dB. The majority of thepositive Zdr values at 6 km were located 10–15 km fur-ther north in the inflow to the preexisting storm. Noareas of significant (.225 dB) LDR values were foundin the vicinity of the impending hail location at thistime.

Significant differences were evident in the next vol-ume scan (6 minutes later; 2355 analysis time). A re-gional increase in reflectivities had taken place, resultingin the formation of a contiguous, U-shaped 40-dBZ echo

area that enveloped the reference point (Fig. 4). Thisarea contained several maxima exceeding 50 dBZ, oneof which was located above the 0004 hail signature area.Horizontal winds in the intensifying echo backed to-wards the southeast and strengthened as the confluenceaxis approached. Updraft strength had also increasedappreciably, with a maximum of greater than 20 m s21

extending south-southwestward from the referencemark. Within and immediately downstream of this up-draft, an expanding area of positive Zdr values had de-veloped (Fig. 4; X 5 34, Y 5 34 km). Peak grid pointZdr values of 3 dB were found ;2 km west of the ref-erence mark. Conway and Zrnic (1993) noted a similarhorizontal displacement between the axes of the positiveZdr column and the updraft maximum. Significant de-polarization (.225 dB), as represented by red color-coded horizontal wind vectors in Fig. 4 (see caption fordetails), had appeared in a contiguous north–south-ori-ented region that included the 50-dBZ echo cores withinthe 40-dBZ U-shaped echo.

The third consecutive volume (2359) was the last onein which no HDR hail signature was present in the 0.58elevation-angle plan position indicator (PPI) data. The6-km reflectivity pattern continued to intensify and or-ganize (Fig. 5). A north–south-oriented band of reflec-tivities of greater than 50 dBZ had developed, with em-bedded maxima now exceeding 60 dBZ. By this time,the horizontal confluence zone had moved eastward toa position within 2 km of the impending surface hailsignature location at 0004. Updrafts . 15 m s21 werelocated a few km east of the confluence axis. However,the areal coverage of updrafts . 20 m s21 had decreasedsomewhat near the reference mark. An area of distinc-


FIG. 3. Dual-Doppler analysis at 2349 UTC 15 Jul 1998. Analysis height is 6 km MSL, black contours are CSU–CHILL reflectivity levels(dBZ). The 50-dBZ contour is darkened. Vectors are storm-relative horizontal winds, the vector length scale is in the lower-left corner ofthe plot. Wind vector color codes depict gridpoint LDR values: gray 5 ,225 dB, red 5 225 to 222 dB, and black 5 .222 dB. (Blackvectors do not appear until next analysis time; Fig. 4.) The Zdr levels are color shaded; starting shade is 1.0 dB, and interval is 1.0 dB. Greenbroken contours are updraft velocities; starting contour is 15 m s21, and interval is 5 m s21. Location of vertical cross section is depictedby line whose ends are marked with large dots. Perpendicular line intersects the cross section within the 0004 UTC HDR hail signaturearea.

tively positive Zdr values [at 6 km above ground level(AGL)] extended southward from the incipient HDRhail location. The axis of enhanced LDR had similarlyadvected eastward so as to become centered over thereference mark. Furthermore, much of the .225-dBLDR area was now associated with reflectivities ex-ceeding 50 dBZ.

A complementary view of the evolution of the radardata fields prior to the onset of surface hail is provided

by a series of vertical cross sections extracted from thethree dual-Doppler analyses. These cross sections wereall constructed on a gridpoint network that was rotatedinto alignment with the mean echo motion (from 3048at 8.2 m s21). The vertical cross section’s location isshown by the cross-hair line whose ends are highlightedwith solid dots in Figs. 3–5.

The cross section from the 2349 analysis showed anelevated echo maximum of between 40 and 50 dBZ over


FIG. 4. As in Fig. 3, except analysis time is 2355 UTC.

the 0004 UTC hail signature site (Fig. 6). Above a point1–2 km west-northwest of this future hail signature site,positive Zdrs greater than 1.0 dB extended approximately1 km above the 08C height level. There were no indi-cations of .225-dB LDR values.

As noted in the 6-km constant-altitude PPI (CAPPI)discussion, obvious echo development was under wayduring the next volume scan. In the 2355 analysis anexpanding area of .50-dBZ echo had appeared at mid-levels above Y 5 210 to 213 km (Fig. 7). This inten-sifying elevated echo mass was closely associated withthe strongest (15–20 m s21) midlevel updrafts. In con-cert with the strengthening updraft, Zdr values in thecolumn had increased significantly, with values of great-

er than 1 dB reaching a height of 6.75 km at X 5 210km. An elevated region of LDRs exceeding 225 dBhad appeared that overlapped and surmounted the topof the positive Zdr column. This LDR signature wascorrelated well with the region where the 50-dBZ re-flectivity contour was expanding aloft. A separate areaof enhanced LDRs had appeared just below the 08Cheight in the inflow region (X 5 216, Z 5 3.75–4.5km). Melting of precipitation particles falling from theoverlying echo mass into the storm inflow could havecaused this LDR increase.

The positioning of the developing enhanced LDR areaimmediately above the summit of the positive Zdr col-umn in the 2355 vertical cross section is notable. This



arrangement suggests that the large supercooled rain-drops responsible for the positive Zdr column began tofreeze as the updraft carried them into a progressivelycolder environment (Smith et al. 1999). As these par-ticles froze, they were more likely to tumble (Spenglerand Gokhale 1972), thus producing larger LDR values(Jameson et al. 1996). These newly-frozen particleswere still exposed to the moisture flux provided by theupdraft; an environment in which they could continueto grow rapidly by accretion (Johnson 1987). Thus, the2355 analysis captured a time when the polarimetricradar signatures indicated that hail was growing aloftbut had not yet reached the surface in appreciable quan-tities. The presence of the positive Zdr column in this

case suggests that freezing of millimeter-sized dropsprovided an immediate source of hail embryos.

In the third successive dual-Doppler analysis (2359),noticeable vertical expansion of the 50-dBZ reflectivityregion had occurred, with 50-dBZ echo extending from2.25 km above mean sea level (MSL; ;0.9 km AGL)to 12 km MSL (Fig. 8). A WER was suggested by thelocalized upward displacement of the 50-dBZ contourabove Y 5 212 km. The positive Zdr column persistedwhile becoming slightly shorter and narrower. The baseof the significant LDR area descended to at least as lowas 3 km MSL on either side of the 50-dBZ contour onthe leading edge of the intense precipitation shaft. Thisis very close to the location where the HDR hail sig-


FIG. 6. Vertical cross section along the 3048 azimuth at 2349 UTC 15 Jul 1998. Cross-section location is shown by the line with solidcircle endpoints in Fig. 3. Contours are CSU–CHILL reflectivity levels at 10-dB intervals from a starting value of 30 dBZ. The Zdrs arecolor shaded starting at 1.0 dB, with an interval of 1.0 dB. Wind vector horizontal components are storm relative; speed calibration vectoris in the lower-left corner of the plot. Wind vector color codes depict gridpoint LDR values: gray 5 ,225 dB, red 5 225 to 222 dB, andblack 5 .222 dB. (At 2349 UTC, no LDRs exceed 225 dB in the plot plane. The multicolored wind vectors will appear in Fig. 7.) Thedownward-pointing arrow near 212 km on the horizontal axis marks the 0004 UTC HDR hail signature location.

nature appeared in the next 0.58 PPI sweep (0004 UTC).Appreciable upward expansion of the 225-dB LDRcontour had also taken place. However, at heights above;9 km MSL, the enhanced LDR areas tended to befound in regions of less-than-50-dBZ reflectivity levels.The ‘‘crown’’ of significant LDR values found in thevicinity of the main updraft at heights above ;10 kmMSL is probably due to heavily rimed, irregularlyshaped graupel particles that were carried up into theanvil. Particles reaching these heights probably exited

the storm in the anvil outflow and thus did not partic-ipate in the hail formation process (Miller et al. 1990).

A more detailed analysis of the LDR evolution wasmade by examining CSU–CHILL individual range gatedata in the three volume scans presented above. Spe-cifically, regions in which the reflectivity from the stormof interest exceeded 40 dBZ in each PPI sweep wereidentified. Within these echo regions, counts of the num-ber of range gates with LDR values greater than or equalto 225 dB were made for each PPI sweep.



A summary of this sweep-by sweep tabulation ofecho-core range gates containing significantly largeLDR values is shown in Table 2. The table values showthe consistent trend for an increasing number of gatescontaining LDRs .225 dB to appear in the 0 to 2208Cheight band prior to the onset of an HDR hail signatureon sweep 1. Furthermore, the percentage of the signif-icantly depolarized gates that are associated with re-flectivities of greater than 50 dBZ (i.e., column 5 ofTable 2) also shows an increasing trend with time. Last,note that some significant LDR gates were detected inthe above-freezing portions of the echo in all three ofthese volumes. These are thought to be due to small,melting ice or hail particles. Because Zdr is sensitive to

the scatterer’s reflectivity-weighted mean aspect ratio(Jameson 1983), the HDR hail signature (a low-Zdr re-gion within a high-reflectivity echo core) does not ap-pear until hailstones become an appreciable fraction ofthe larger-diameter end of the precipitation particle sizespectrum.

Taken together, the analyses of the 15 July 1998 caseimply that significant hail growth was closely connectedwith an updraft that increased from ,10 to .20 m s21

at midlevels between 2346 and 2352 UTC. A positiveZdr column associated with this strengthening updraftbecame increasingly apparent. A region of significantdepolarization and rapidly intensifying reflectivity de-veloped immediately above (and probably downwind



from, in a relative sense) the summit of the Zdr column.Within the .50-dBZ portion of this elevated echo core,an appreciable fraction (.;40%–50%) of the rangegates contained significant LDR values. The HDR hailsignature appeared at the 0.58 elevation angle as the echocore with significant depolarization descended to thesurface. Additional evidence for this evolutionary pat-tern will next be sought in a second hail event.

b. 28 July 1996

The storm of interest on this date was a long-livedmulticell that periodically generated large hail as ittracked southeastward from Wyoming toward the CSU–

CHILL radar. Based on the 28 July 00 UTC DNR sound-ing, conditions at the LCL (;620 hPa) were estimatedto be the following: temperature T 5 14.78C and asaturation mixing ratio of ;8.7 g m23. These quantitieswere very similar to those for the 15 July 1998 case.The hail episode under analysis first presented a defin-itive HDR hail signature in the 0.58 elevation-angle PPIdata at 2205. Note also that, between 2230 and 2250,this same storm produced significant damage resultingfrom 7-cm (2.75 in.)-diameter hail that fell near the townof Galeton, Colorado (NCDC 1996, 35–37).

The PPI-sector volume scan data collected by theCSU–CHILL radar on 28 July 1996 were interpolatedto a Cartesian gridpoint network using the same methods


TABLE 2. 15 Jul 1998 sweep-by-sweep counts of gates exceeding selected LDR values.

LDR categories (dB)

219 221 223 225

Percent ofLDR .225 dBwithin

50-dBZ echoAvg height(km MSL)

Estimatedmoist

adiabatictemperature

(8C)

Avgsounding

temperature(8C)

2346:04 UTC volume000000002100000

0014002

12151112

450

14

219

69108

183346434030

239

62250262220533768

105211190

883458

029

154

4022163617

46

181415

9.59.79.38.88.37.87.46.86.15.34.63.93.22.62.0

227.2228.3225.2222.0218.5214.8211.728.123.8

0.83.78.7

16.321.727.0

231.7233.1230.4227.0223.6221.3220.0216.4211.124.3

1.88.5

15.219.223.3

2352:06 UTC volume0001000232

148

13602103

000400243

164364712922

817

714

0045

1013151319336567

176168159117

792550

21

4211334651658798

119181262352360403280

95105

00205219

1335383437331610111712

12.411.811.210.710.3

9.99.59.18.68.07.46.86.25.44.63.93.32.62.0

250.4245.9241.3236.7233.4230.5227.1223.7220.5216.3211.828.124.220.2

3.68.0

15.721.526.9

250.2246.6242.9239.8237.1234.4231.7229.0225.6222.0220.0216.4212.025.3

1.88.5

14.219.223.3

2358:06 UTC volume0000010360583

1213

86201

0012

1049

36577927425494

1009034171121

003

21435260

107156207168131141190200216198108

6969

00

1791

112117110168220378389298314277304406510417239209

0002

13333526233244587275676037373823

14.513.512.611.811.110.4

9.99.59.28.88.37.67.06.35.64.94.13.52.82.1

260.0255.1251.0245.5240.5234.8230.0226.7224.6222.1217.9213.029.525.421.4

3.05.8

13.620.326.1

259.9255.3251.1246.6242.3237.8234.4231.7229.7227.0223.6220.7217.9213.027.220.9

6.612.317.922.6


FIG. 9. CAPPI plot at 2151 UTC 28 Jul 1996. Analysis height is6 km MSL. Black contours are reflectivity starting at 30 dBZ andincreasing in 10-dB steps. 50-dBZ contour is darkened. Red contouris 225-dB LDR. Irregular gray box encloses the HDR hail signaturethat appeared at 2205 UTC. Line with solid dot endpoints locates thevertical cross section shown in Fig. 10.

that were applied in the 15 July 1998 case. The resultantCAPPI plot at the 6-km MSL height at 2151 is shownin Fig. 9. At this time, a small but expanding area of.50-dBZ echo was located in the right flank of thesoutheastward-moving storm (X 5 215.5, Y 5 45 km).LDR values of greater than 225 dB were found in onlytwo small areas, both of which were outside of the 50-dBZ contour.

The corresponding vertical cross section through theright flank of the storm along the direction of echo mo-tion is shown in Fig. 10. The reflectivity pattern in thecross-sectional plane showed the extensive verticaloverhang structure typical of a vigorous convective cell.A column of reflectivities of greater than 50 dBZreached heights of ;9.75 km MSL, but no significantareas of depolarization were present in the 50-dBZ coreregion at subfreezing temperatures. Significant LDRvalues were only found in the vicinity of the 08C heightwhere particle melting/freezing was likely to be takingplace. No evidence of a positive Zdr column was present.Instead, the most prominent Zdr pattern was the down-ward depression in the Zdr contours across the low-al-titude echo-core region where graupel melting was like-ly to be in progress (Bringi et al. 1986a).

The next analysis (2157) was based on the last volumescan in which no HDR hail signature was observed inthe lowest sweep. Very rapid expansion of the area en-closed by the 50-dBZ reflectivity contour was evidentat the 6-km height level (Fig. 11). The areal coverageof 225-dB LDR contour had also expanded consider-ably during the 7 min since the preceding CAPPI (Fig.

9). A substantial portion of this enhanced LDR areaoverlapped the 50-dBZ echo core at 2157.

These same evolutionary features were apparent inthe 2157 vertical cross section (Fig. 12). Aloft, the 50-dBZ reflectivity contour had expanded and an area ofreflectivities exceeding 60 dBZ had appeared near the2208C height level. LDRs exceeding 225 dB had ap-peared in a volume that extended up to ;4 km abovethe environmental 08C height. As was true at 2151, the1-dB Zdr contour did not reach subfreezing heights.

Table 3 shows the sweep-by-sweep tabulations of thesignificant LDR gate occurrences for the two volumesdiscussed above. As in the 15 July 1998 case, a generalincrease in the number of significant LDR gates tookplace in the echo core in the 08 to 2208C temperaturezone prior to the first appearance of the low-elevation-angle HDR hail signature (;2005). At lower heightswhere the temperatures were above 08C, the percentageof significant LDR gates associated with reflectivities. 50 dBZ actually decreased between 2151 and 2158UTC. As in the 15 July 1998 case, generalized precip-itation particle melting tends to increase the LDR levelsat these temperatures. For the purposes of this research,the LDR statistics at these above freezing temperaturesare not taken to be meaningful until the HDR valuesindicate that hail has become a significant componentof the near-surface precipitation.

The vertical cross sections show that positive Zdr col-umn development was very different in the two analyzedevents. These differences may have resulted from var-iations in the extent of neighboring echo. Cell devel-opment in the 1998 event took place within a largeregion of 30-dBZ echo whose base extended below the08C level (Fig. 6). Some of the particles responsible forthis echo probably were composed of liquid–ice mix-tures with varying ratios at heights below the 08C level.Some of these particles, in turn, were likely to havebeen ingested into the intensifying updraft associatedwith the 0004 UTC hail onset. As noted in Conway andZrnic (1993), the presence of these large particles in theupdraft, some liquid and some only partially melted,would contribute to the formation of a well-defined pos-itive Zdr column (as seen in the 1998 event). In contrast,the right flank development in the 1996 case took placein essentially echo-free surroundings, which may havelimited the ingestion of large, preexisting particles intothe updraft, thus reducing the prominence of the Zdr

column.In summary, polarimetric radar observations of the

15 July 1998 and the 28 July 1996 storms showed sig-nificant variability in development of a positive Zdr col-umn. In both storms, the appearance of an HDR hailsignature near the surface was preceded by an increasein LDRs in a portion of the parent storm’s 50-dBZ echovolume at subfreezing temperatures. To understand bet-ter this LDR pattern evolution, the electromagnetic scat-tering behavior of growing hailstones was modeled.


FIG. 10. Vertical cross section along 3178 azimuth at 2151 UTC. Cross-section location is shown in Fig. 9. Black contours are reflectivitystarting at 30 dBZ and increasing in 10-dB steps. 50-dBZ contour is darkened. The Zdrs are color shaded starting at 1.0 dB, with an intervalof 1.0 dB. Red LDR contours are 225 and 222 dB (222-dB contour is darkened). Downward-pointing arrow near 236 km on the horizontalaxis marks the location of the HDR hail signature at 2205 UTC.

4. Radar backscatter modeling

The electromagnetic backscatter properties of grow-ing hailstones were simulated using a T-matrix-basedscattering model (Waterman 1965; Seliga and Bringi;1978). The fundamental input particle parameters are 1)the characteristic diameter and axis ratio, 2) a statisticalrepresentation of the particle’s orientation (i.e., meanand standard deviation of the canting angle), and 3) thedielectric constant as defined by the particle’s ice/watercomposition. The scattering model can process an inputparticle population with a specified size distribution(typically exponential). The net radar backscatter from

multiple particle types (i.e., mixed raindrops and hail-stones) can also be calculated.

Although the model is capable of calculating the radarbackscatter from a broad spectrum of particle specifi-cations, it should be recognized that a number of fun-damental simplifications from nature’s true complexityneed to be invoked. To make the solutions to the scat-tering equations mathematically tractable, spheroidalhailstone shapes were required. Various conical and ir-regular/lobed hailstone shapes were not considered. Theice particle’s water component similarly was taken tobe either uniformly distributed within the total hailstone


FIG. 11. As in Fig. 9 except analysis time is 2157 UTC. Red contoursare 225- and 221-dB LDR; 221-dB contour is darkened.

mass or else confined uniformly to the hailstone’s sur-face. The effects of localized liquid water redistribution,shedding, and so on that have been observed in hail-stones growing in laboratory wind tunnels (Garcia-Gar-cia and List 1992) had to be ignored.

Because no precipitation size spectra observationswere available from the analyzed storms, the modeledscattering particles were taken to be a monodispersepopulation composed only of hailstones. Model test runswere done using 1) an exponential distribution of hail-stone sizes and 2) the addition of a separate raindroppopulation with an exponential size distribution. In bothcases, the calculated S-band LDR values were within 1dB of those obtained from the simple monodisperse hailspectrum. A particle diameter of 1 cm was used as ageneral approximation of a growing hailstone. Based onthe northeast Colorado hailstone collections tabulatedby Knight (1986), a fixed axis ratio of 0.8 was assumed.In all of the results presented below, the particle tem-perature was prescribed to be 08C. The calculations werecarried out for a wavelength of 11 cm. The elevationangle was fixed at 58.

The variable parameters in the T-matrix runs were thestandard deviation of the mean hailstone canting angleand the liquid water characteristics of the scatterer’s icecomposition. The canting angle distribution was takento be Gaussian about a mean canting angle of 08. (Theuse of a random distribution of canting angles changedthe modeled LDRs by approximately 0.7 dB from theresults obtained using the Gaussian canting angle dis-tribution.) For the simulations described here, the stan-dard deviation of the hailstone canting angle was per-mitted to range from 308 to 608, because examinationsof the ice and air bubble patterns in thin sections cut

from hailstones indicate the stones undergo large am-plitude gyrations as they fall (Knight and Knight 1970).The water fraction in the hailstone ice structure waspermitted to vary between 10% and 50%. This rangegenerally includes the spectrum of values observed inartificial hailstones grown in a laboratory wind tunnel(Garcia-Garcia and List 1992). Consideration of the ef-fect of hailstone water fraction is also warranted by theupdraft liquid water content inferred in the primary case(15 July 1998). In this storm, the liquid water contentin an adiabatic updraft was approximately 5 g kg21 atthe 2158C level (390 hPa). The heat transfer studiesreported by Bailey and Macklin (1968) suggest that thewet growth regime would be likely to occur for hail-stones 1–2 cm in diameter under these temperature andliquid water content conditions.

Two different physical distributions of the hailstone’sliquid water component were considered. Figure 13ashows the results when the liquid water was uniformlydistributed within the hailstone [i.e., spongy or soakedconditions; Lesins and List (1986)]. The dielectric prop-erties of the prescribed ice–water mixtures were cal-culated following the method of Bruggeman as sum-marized by Sihvola and Kong (1988). Over the modeledrange of water fraction and canting angle standard de-viation values, the LDR contours are essentially quasi-vertical lines. Significant (.225 dB) LDR values areassociated with water fractions greater than ;27% (i.e.,until the water fraction reaches ;27%, none of the mod-eled canting angle standard deviations are capable ofgenerating significant depolarizations).

The second hailstone water distribution simulates thecase in which a combination of warm environmentaltemperature and high cloud liquid water content causesa water coating to exist on the stone’s outer surface [wetgrowth; Lesins and List (1986)]. The modeled watercoating thicknesses were calculated based on the samehailstone liquid water fractions shown in Fig. 13b. (Thewater mass implied by a given water fraction within a1-cm major diameter, 0.8-axis-ratio hailstone was cal-culated. The equivalent water coating thickness was thendetermined by distributing this water mass uniformlyover the hailstone’s surface area.) Generally increaseddepolarization levels are produced when the hailstonewater content is moved to the stone’s outer surface (Fig.13b vs Fig. 13a). The appreciable leftward shift of the225-dB LDR contour between Figs. 13b and 13a dem-onstrates the LDR enhancement that even thin watercoatings (;0.3% to 0.5% of S-band wavelength) onhailstones are capable of generating significant valuesof LDR.

The implications of these calculations must be takenwithin the overall context of factors that influence LDR.As has been mentioned earlier, the scattering particle’sshape and orientation are the primary geometric param-eters that influence LDR. The final impact of these fac-tors upon LDR is modulated by the particle’s dielectricconstant. Thus, contributions to the enhanced LDR val-


FIG. 12. As in Fig. 10, except plot time is 2157 UTC.

ues reported in this study might also have come froma combination of increasing axis ratios and/or large-amplitude gyratory motions within the population ofdeveloping hailstones. A qualitative appreciation forthese factors can be obtained from additional consid-erations of the polarimetric radar measurements.

When the dispersion of particle axis ratios in the radarpulse volume increases (such that the number of par-ticles with large axis ratios increases), the correlationbetween the H and V return signals adjusted to a timelag of zero [HV(0)] is decreased (Balakrishnan andZrnic 1990). In both of our cases, the HV(0) values inthe areas where enhanced LDRs developed aloft re-mained above ;0.90. In contrast, HV(0) values havebeen observed to drop to less than 0.80 in brightband

echoes where the combined presence of frozen, partiallymelted, and fully melted particles results in a very widedistribution of axis ratios (Illingworth and Caylor 1989).Thus, it does not appear that the existence of particleswith unusually large of small axis ratios contributedsignificantly to the LDR enhancements observed in thisstudy.

Outside of any positive Zdr column areas, the Zdr val-ues were consistently near 0 dB in the inferred hailgrowth regions aloft. This near-0-dB Zdr condition didnot display any organized spatial or temporal deviations;the hailstone orientations quickly became and remainedwidely dispersed throughout the growth process. Theredid not appear to have been an increase in canting anglefluctuations with time that was capable of producing the


TABLE 3. 28 Jul 1996 sweep-by-sweet counts of gates exceeding selected LDR values.

LDR categories (dB)

219 221 223 225

Percent of LDR. 225 dB

within 50-dBZecho

Avg height(km MSL)

Estimatedmoist

adiabatictemperature

(8C)

Avg soundingtemperature

(8C)

2151:42 UTC volume0001000000100012214012

02010000014032

105

17166336

532

1810

5231174

126

263447561413

918

1123

92841312922191638202520

104139143153

88534146

0000

1000

125004

10211544555455655636

10.910.2

9.69.08.58.07.67.26.96.56.25.85.55.14.74.33.93.43.02.52.32.0

240.3235.1230.0224.9221.2217.8214.5212.129.527.225.423.621.7

0.21.52.95.3

11.216.220.222.423.7

244.4238.6233.6228.4223.9219.4217.1215.3214.3212.3210.227.425.422.5

1.04.58.0

12.315.920.522.423.9

2158:06 UTC volume00001010000020428160511120

0000422111255285

1179

101311

4022

0005

16111216101511172520293244373643553221132617

02

112276484658525247429491926789

121153217196177

95487072

00049

132621483467554651625954453739474635291412

13.212.511.811.110.4

9.79.18.68.17.77.36.96.66.25.95.65.24.84.44.13.73.32.92.52.22.0

261.2256.5249.6242.7236.6230.8226.2222.1218.5215.4212.829.927.725.924.022.120.2

1.12.53.67.2

12.616.820.722.623.9

261.3256.3252.5246.2240.3234.5229.3224.8220.3217.7215.6214.3213.0210.228.126.123.4

0.13.66.29.7

13.216.920.522.923.9

generalized appearance of significantly high LDR lev-els.

In summary, the growing hail particles probably didnot have axis ratios with extreme departures from unity.The canting angles of these particles underwent signif-icantly large (several tens of degrees) excursionsthroughout their growth process. Within this overallgrowth regime, the hailstones’ LDR characteristics were

likely heavily influenced by the quantity and spatialdistribution of water within the ice structure.

5. Discussion

The primary polarimetric radar evolutionary featureobserved in this study was the noticeable increase in S-band LDR values exceeding 225 dB within the ;08 to


FIG. 13. (a) Selected hailstone backscattering properties obtained from a T-matrix model. Particle diameter is 1 cm, and aspect ratio(vertical dimension: horizontal dimension) is 0.8. Abscissa is the hailstone water fraction (%), ordinate is the std dev of the canting angledistribution (8). Contour lines are LDR values (dB); fractional values plotted at grid point locations are Zdrs. (b) As in (a), except hailstonewater component has been redistributed into a surface water coating. Abscissa is the water coating thickness (mm).

2208C temperature levels in portions of nonsupercel-lular thunderstorm echo cores. These signatures oc-curred ;5–10 min before the appearance of near-surfacehail. In particular, as surface hail became imminent, anoticeable increase in the number of gates containing.225-dB LDRs took place within expanding areas of.50-dBZ echo aloft that flanked the preexisting stormcore. This polarimetrically-inferred hail growth withinan echo system’s flanking region is in agreement withprevious studies of multicell hailstorms (Heymsfield etal. 1980; Ziegler et al. 1983; Browning et al. 1976).Last, in only one of the analyzed storms (15 July 1998)was a positive Zdr column found to be associated withthe elevated region of enhanced LDRs. As has beennoted in other studies, the development of an ‘‘LDRcap’’ immediately downtrajectory from the summit ofa positive Zdr column implies that vigorous hail growthprocess are likely to be under way (Conway and Zrnic1993; Hubbert et al. 1998).

The available data do not permit the hail growth pro-cesses that were operative in these two cases to be an-alyzed in detail; such growth studies greatly benefit fromaccurate information on hail size spectra and on the in-cloud microphysical environment (Miller et al. 1990).However, the embryonic hailstones were likely to havebeen nonspherical frozen particles [i.e., millimeter-di-ameter frozen drops or graupel particles; Knight(1981)]. Because of their aerodynamic characteristics,such particles typically exhibit canting angle fluctua-tions on the order of several tens of degrees (Spenglerand Gokhale 1972; List and Schemenaur 1971). Thesecharacteristics of asymmetric shape and gyrating motion

imply that the developing hailstones were capable ofdepolarizing the backscattered signal to significant lev-els. If these hailstones encounter vigorous growth con-ditions such that their bulk density increased, eitherthrough the development of a higher ice–air ratio, orthrough the accretion of unfrozen liquid water, the re-sultant dielectric constant change would further magnifythe depolarized signal level. Because of the fundamentalincrease in sweep-out area with time for a growing hail-stone, the mass accretion rate should tend to maximizein the later portion of the hail growth phase as the largestdiameters are attained. If this accelerating mass accre-tion increased the hailstone density and/or water frac-tion, LDR enhancements would become increasinglylikely during the later stages of the hail growth process.

The general hail-growth timescale implied by theseradar observations is consistent with several numericalsimulations of hail growth. Miller et al. (1990) reportedthat 15–20 min were required for ;1-mm-diameter em-bryos to grow into several centimeter–sized hailstones,with the majority of the stone mass being accreted dur-ing the descending (later) portion of the growth simu-lation. Ziegler et al. (1983) similarly integrated theirhail growth model for 25 min to cover the entire em-bryo-to-hailstone growth process. Most of the growthto the maximum computed hail diameter in their sim-ulations took place within a 10–15-min period. Last,Heymsfield et al. (1980) reported that in their compu-tations, ;7 min of updraft residence time was requiredfor a small hailstone to grow into a large hailstone (i.e.,to increase in diameter from 1 to 1.5 cm). Taken to-gether, these modeling studies suggest that the initial


appearance of enhanced S-band LDR values aloft is notlikely to occur until the final ;10 min of the hail growthprocess as the maximum diameters are attained.

This short timescale clearly limits the utility of usingthe initial development of enhanced LDR areas aloft toissue anticipatory hail warnings. Based on this study, atime window of only ;10 min exists between the initialdetection of an elevated volume of significantly depo-larized core echo aloft and the appearance of significanthail at the surface. When the radar volume scan timeresolution is on the order of 6 min, the pattern of LDRenhancement will appear suddenly, generally no morethan two volume scans before the onset of the hail eventat the surface. However, it should be noted that, in bothof the storms examined here, the initial appearance ofan enhanced LDR region aloft quickly evolved into acolumnar region of enhanced LDR that extended fromthe surface to altitudes that were several kilometersabove the 08C height. This column then persisted aslong as the storm was producing a significant hail swath.Thus, in addition to whatever hail prediction capabilitiesthat LDR observations may possess, it may also be use-ful as a diagnostic parameter for thunderstorm hail pro-duction.

Furthermore, the analysis of additional storms duringthe course of this research revealed that not all near-surface high ZHH–low Zdr patterns were preceded by or-ganized areas of LDR enhancement aloft. However, thesurface hail reported in these null cases was consistentlyof small, nondamaging sizes. Thus, the detection of or-ganized regions of enhanced LDRs within portions ofthunderstorm echo cores aloft may identify the stormsthat are the most likely to produce damaging hail.

The implications of this study’s results should alsobe examined under a broader range of environmentalconditions. For example, convective cloud bases arelikely to be warmer and more moist in storms devel-oping outside of the U.S. high plains environment. Ex-isting research has shown that positive Zdr columns arecommonly observed when vigorous convective growthtakes place in a moist environment (Bringi et al. 1997).The implied greater updraft moisture content would alsohave several important effects on the evolution of theLDR field. The enhanced precipitation formation bycoalesence would increase the likelihood that dropslarge enough to have an appreciable freezing time wouldbe available for lofting above the 08C level. Depolar-ization levels would then be enhanced during the sub-sequent drop freezing process (Smith et al. 1999; Bringiet al. 1997). Increased updraft moisture content mightalso be expected to promote an increased water fractionin the growing, frozen hydrometeors, thus tending toincrease the depolarization levels. However, at lowerelevations in these more tropical environments, meltingwould play a greater role in reducing the diameters ofthe hailstones during their final descent. In the limit,such melting could cause precipitation grown in a sig-nificantly depolarized echo core aloft to reach the sur-

face only as rain. Also, none of the storms analyzedhere showed marked steady-state characteristics. It isquite possible that the patterns by which hail grows aloftand subsequently descends to the ground in storms withmore supercellular organization will vary from thoseseen in this study. Thus, it would be useful to collectpolarimetric radar observations of convective stormswith various organizational modes in a variety of cli-matic areas.

Acknowledgments. This paper was substantially im-proved by the reviewers’ inputs. The efforts of CSUtechnicians Robert Bowie and Kenneth Pattison (nowretired) and Chief Engineer David Brunkow were in-strumental in the collection of the radar data. Highlyuseful discussions regarding the T-matrix software wereconducted with Dr. J. Beaver (Department of ElectricalEngineering) and Dr. L. Carey (Department of Atmo-spheric Science). Margi Cech skillfully formatted thetext and tables. The CSU–CHILL facility is supportedby NSF Cooperative Agreement ATM-9500108 and byColorado State University. This research was also sup-ported by the NSF Grant ATM-9726464.

REFERENCES

Atlas, D., and F. H. Ludlam, 1961: Multiple-wavelength radar re-flectivity of hailstorms. Quart. J. Roy. Meteor. Soc., 87, 523–534.

——, M. Kerker, and W. Hitschfeld, 1953: Scattering and attenuationby non-spherical atmospheric particles. J. Atmos. Terr. Phys., 2,108–119.

Aydin, K., T. A. Seliga, and V. Balaji, 1986: Remote sensing of hailwith a dual linear polarization radar. J. Climate Appl. Meteor.,25, 1475–1484.

Bailey, I. H., and W. C. Macklin, 1968: Heat transfer from artificialhailstones. Quart. J. Roy. Meteor. Soc., 94, 93–98.

Balakrishnan, N., and D. S. Zrnic, 1990: Use of polarization to char-acterize precipitation and discriminate large hail. J. Atmos. Sci.,47, 1525–1540.

Barge, B. L., 1970: Polarization observations in Alberta. Preprints,14th Radar Meteorology Conf ., Tucson, AZ, Amer. Meteor. Soc.,221–224.

Bringi, V. N., R. M. Rasmussen, and J. Vivekanandan, 1986a: Mul-tiparameter radar measurements in Colorado convective storms.Part I: Graupel melting studies. J. Atmos. Sci., 43, 2545–2563.

——, J. Vivekanandan, and J. D. Tuttle, 1986b: Multiparameter radarmeasurements in Colorado convective storms. Part II: Hail de-tection studies. J. Atmos. Sci., 43, 2564–2577.

——, K. Knupp, A. Detwiler, L. Liu, I. J. Caylor, and R. A. Black,1997: Evolution of a Florida thunderstorm during the Convectionand Precipitation/Electrification experiment: The case of 9 Au-gust 1991. Mon. Wea. Rev., 125, 2131–2160.

Browning, K. A., and G. B. Foote, 1976: Airflow and hail growth insupercell storms and some implications for hail suppression.Quart. J. Roy. Meteor. Soc., 102, 499–533.

——, and Coauthors, 1976: Structure of an evolving hailstorm. PartV: Synthesis and implications for hail growth and hail suppres-sion. Mon. Wea. Rev., 104, 603–610.

Brunkow, D. A., V. N. Bringi, P. C. Kennedy, S. A. Rutledge, V.Chandrasekar, E. A. Mueller, and R. K. Bowie, 2000: A de-scription of the CSU–CHILL national radar facility. J. Atmos.Oceanic Technol., 17, 1596–1608.

Conway, J. W., and D. S. Zrnic, 1993: A study of embryo production


and hail growth using dual-Doppler and multiparameter radars.Mon. Wea. Rev., 121, 2511–2528.

Cressman, G. P., 1959: An operative objective analysis scheme. Mon.Wea. Rev., 87, 367–374.

Donaldson, R. J., 1961: Radar reflectivity profiles in thunderstorms.J. Meteor., 18, 292–305.

Garcia-Garcia, F., and R. List, 1992: Laboratory measurements andparameterizations of supercooled water skin temperatures andbulk properties of gyrating hailstones. J. Atmos. Sci., 49, 2058–2073.

Herzegh, P. H., and A. R. Jameson, 1992: Observing precipitationthrough dual-polarization radar measurements. Bull. Amer. Me-teor. Soc., 73, 1365–1374.

Heymsfield, A. J., A. R. Jameson, and H. W. Frank, 1980: Hail growthmechanisms in a Colorado storm: Hail formation processes. J.Atmos. Sci., 37, 1779–1807.

Holler, H., V. N. Bringi, J. Hubbert, M. Hagen, and P. F. Meischner,1994: Life cycle and precipitation formation in a hybrid-typehailstorm revealed by polarimetric and Doppler radar measure-ments. J. Atmos. Sci., 51, 2500–2522.

Hubbert, J. C., V. N. Bringi, L. D. Carey, and S. Bolen, 1998: CSU–CHILL polarimeteric radar measurements in a severe hailstormin eastern Colorado. J. Appl. Meteor., 37, 748–775.

Humphries, R. G., 1974: Observations and calculations of depolar-ization effects at 3 GHz due to precipitation. J. Rech. Atmos.,8, 155–161.

Illingworth, A. J., and I. J. Caylor, 1989: Cross polar observationsof the bright band. Preprints, 24th Conf. on Radar Meteorology,Tallahassee, FL, Amer. Meteor. Soc., 323–327.

——, J. W. F. Goddard, and S. M. Cherry, 1987: Polarization radarstudies of precipitation development in convective storms.Quart. J. Roy. Meteor. Soc., 113, 469–489.

Jameson, A. R, 1983: Microphysical interpretation of multi-parameterradar measurements in rain. Part II: Estimation of raindrop sizedistribution parameters by combined dual-wavelength and po-larization measurements. J. Atmos. Sci., 40, 1803–1813.

——, M. J. Murphy, and E. P. Krider, 1996: Multiple-parameter radarobservations of isolated Florida thunderstorms during the onsetof electrification. J. Appl. Meteor., 35, 343–354.

Johnson, D. B., 1987: On the relative efficiency of coalescence andriming. J. Atmos. Sci., 44, 1671–1680.

Knight, C. A., and N. C. Knight, 1970: The falling behavior of hail-stones. J. Atmos. Sci., 27, 672–681.

Knight, N. C., 1981: The climatology of hailstone embryos. J. Appl.Meteor., 20, 750–755.

——, 1986: Hailstone shape factor and its relationship to radar in-terpretation of hail. J. Climate Appl. Meteor., 25, 1956–1958.

Lesins, G. B., and R. List, 1986: Sponginess and drop shedding ofgyrating hailstones in a pressure-controlled icing wind tunnel.J. Atmos. Sci., 43, 2813–2825.

List, R., and R. Schemenauer, 1971: Free-fall behavior of planar snowcrystals, conical graupel, and small hail. J. Atmos. Sci., 28, 110–115.

Ludlam, F. H., 1980: Clouds and Storms: The Behavior and Effectof Water in the Atmosphere. The Pennsylvania State UniversityPress, 405 pp.

McCormick, G. C., and A. Hendry, 1975: Principles for the radardetermination of the polarization properties of precipitation. Ra-dio Sci., 10, 421–434.

Miller, L. J., J. D. Tuttle, and G. B. Foote, 1990: Precipitation pro-duction in a large Montana thunderstorm: Airflow and particlegrowth trajectories. J. Atmos. Sci., 47, 1619–1646.

Mohr, C. G., and L. J. Miller, 1983: CEDRIC—a software packagefor Cartesian space editing, synthesis, and display of radar datafields under interactive control. Preprints, 21st Conf. on RadarMeteorology, Edmonton, AB, Canada, Amer. Meteor. Soc., 569–574.

NCDC, 1996: Storm Data. Vol. 38, No. 7, 363 pp.Oye, R., and R. Carbone, 1981: Interactive Doppler editing software.

Preprints, 20th Conf. on Radar Meteorology, Boston, MA, Amer.Meteor. Soc., 683–689.

Pruppacher, H. R., and K. V. Beard, 1970: A wind tunnel investigationof the internal circulation and shape of water drops falling atterminal velocity in air. Quart. J. Roy. Meteor. Soc., 96, 247–256.

Rinehart, R. E., and J. D. Tuttle, 1982: Antenna beam patterns anddual wavelength processing. J. Appl. Meteor., 21, 1865–1880.

Seliga, T. A., and V. N. Bringi, 1976: Potential use of radar differentialreflectivity measurements at orthogonal polarizations for mea-suring precipitation. J. Appl. Meteor., 15, 69–76.

——, and ——, 1978: Differential reflectivity and differential phaseshift: Applications in radar meteorology. Radio Sci., 13, 271–275.

Sihvola, A. H., and J. A. Kong, 1988: Effective permittivity of di-electric mixtures. IEEE Trans. Geosci. Remote Sens., 26, 420–429.

Smith, P. L., D. J. Musil, A. G. Detwiler, and R. Ramachandran, 1999:Observations of mixed-phase precipitation within a CaPE thun-derstorm. J. Appl. Meteor., 38, 145–155.

Spengler, J. D., and N. Gokhale, 1972: Freezing of freely suspended,supercooled water drops in a large vertical wind tunnel. J. Appl.Meteor., 11, 1101–1107.

Tuttle, J. D., and R. E. Rinehart, 1983: Attenuation correction in dual-wavelength analyses. J. Climate Appl. Meteor., 22, 1914–1921.

——, V. N. Bringi, H. D. Orville, and F. J. Kopp, 1989: Multiparam-eter radar study of a microburst: Comparison with model results.J. Atmos. Sci., 46, 601–620.

Waterman, P. C., 1965: Matrix formulation of electromagnetic scat-tering. Proc. IEEE, 53, 805–812.

Ziegler, C. L., P. S. Ray, and N. C. Knight, 1983: Hail growth in anOklahoma multicell storm. J. Atmos. Sci., 40, 1768–1791.

polarimetric radar observations of hail...

Documents