NCAR-TN-76-1 Structure of an Evolving Hailstorm.

STRUCTURE OF AN 'EVOLVING 

HAILSTORM 

SERIES OF PAPERS 

BY NHRE STAFF AND PARTICIPANTS 

NATIONAL CENTER FOR ATMOSPHERIC RESEARCH 

BOULDER, COLORADO 

NATIONAL HAIL RESEARCH EXPERIMENT 

TECHNICAL REPORT 

NO. 76/1 

JAN. 1976

STRUCTURE OF AN EVOLVING HAILSTORM 

PART I 

General Characteristics and Cellular Structure 

by 

J.-P. Chalon, J. C. Fankhauser, and P. J. Eccles 

National Center for Atmospheric Research 

Boulder, Colorado 

PART II 

Thermodynamic Structure and Airflow in the Near Environment 

by 

J. C. Fankhauser 

National Center for Atmospheric Research 


PART III 

Internal Structure from Doppler Radar 

by 

R. G. Strauch and F. H. Merrem 

NOAA/ERL/Wave Propagation Laboratory 


PART IV 

Internal Structure from Penetrating Aircraft 

by 

D. J. Musil, E. L. May, P. L. Smith, Jr., and W. R. Sand 

Institute of Atmospheric Sciences, South Dakota School 

of Mines and Technology, Rapid City, South Dakota 

PART V 

a Synthesis and Implications for Hail Growth and Hail Suppression 

by 

K. A. Browning, J. C. Fankhauser, J.-P. Chalon, 

P. J. Eccles, R. G. Strauch, F. H. Merrem, 

D. J. Musil, E. L. May, and W. R. Sand

Structure of an Evolving Hailstorm. Part I: 

General Characteristics and Cellular Structure 

by 

J.-P. Chalon, J. C. Fankhauser, and P. J. Eccles 

National Center for Atmospheric Research 2 


Scientific visitor on leave from "Meteorologie Nationale," Paris, France, on 

a fellowship from the "Delegation Generale a la Recherche Scientifique et 

Technique." 

2 This research was performed as part of the National Hail Research Experiment, 

managed by the National Center for Atmospheric Research and sponsored by the 

Weather Modification Program, Research Applications Directorate, National 

Science Foundation.

ABSTRACT 

The detailed structure and evolution of radar echoes observed in a 

multicellular hailstorm are analyzed. General environmental conditions, 

overall radar echo development and precipitation measurements are briefly 

discussed but the analysis is mainly concerned with a particular event which 

was thoroughly observed by several different field facilities of the National 

Hail Research Experiment. This hailstorm which evolved in a systematic and 

periodic manner is the subject of four companion papers appearing in this 

issue. 

Overall storm characteristics are found to compare closely to earlier 

descriptions of multicell hailstorms occurring in the High Plains. The motion 

of the main system was to the right of the mean wind vector in the cloud layer. 

Cell velocity was along but less than the wind in the mid-troposphere. 

Propagation by new cell growth in a preferred location with respect to 

existing radar echoes dominated the motion of the overall system. Study of 

the formation and evolution of individual cells showed that discrete new 

echoes formed near the altitude of 7 km MSL (-12 C) on the storm's right 

forward flank 5 to 10 km ahead of existing echo components at approximately 

15 min intervals. Each grew rapidly in intensity and height and by moving 

more slowly than the overall echo complex soon became the main storm component. 

Average lifetime of individual cells, including the period from visually 

perceptible turrets, to 'first echo,' to echo decay, was 45 min. Thus, as 

many as three cells were found to coexist in varying stages of development. 

The ascent rate of visual cloud turrets and the history of maximum radar 

reflectivity of individual cells after the appearance of first echo indicate 

that the longest in-cloud residence time available for particle growth to the 

largest observed hail size (1.5 cm diameter) was between 30 and 35 min.

1. Introduction 

I-1 

One of the objectives of the National Hail Research Experiment (NHRE) is 

the definition of thermodynamic structure and airflow patterns characteristic 

of the different types of hail-producing thunderstorms. Aside from the 

motivation of gaining increased understanding of all aspects of convective 

cloud dynamics, incisive case studies are required to establish a firmer 

physical basis for conduct and evaluation of a hail suppression experiment. 

Of particular importance is the interdependence between airflow patterns and 

the manner in which precipitation products are generated, dispersed and deposited 

at the ground. To establish the interrelationship of the important hailstorm 

parameters it is necessary to adopt a case study approach wherein many observa- 

tional techniques are brought to bear simultaneously on the same storm. Because 

of the sophistication of the required observational techniques, the difficulties 

in coordinating them, and the infrequency of large hailstorms within the compass 

of fixed instrumentation, an adequately comprehensive data base is only 

occasionally achieved. 

On 9 July 1973 a number of thunderstorms moved over and produced hail 

within the field facilities of NIRE located in northeastern Colorado. A full 

complement of observing systems, including conventional and Doppler radars, 

surface and upper-air networks and instrumented aircraft, gathered data pertinent 

to a particular unseeded multicell storm near the town of Raymer. This 

paper is the first of a series of five appearing in this issue which describes 

the diverse observations (Parts I, II, III and IV) and which culminates in a 

synthesis of the airflow and associated microphysical processes relevant to 

hail formation in multicell storms (Part V). 

We present a brief review of significant larger scale features, the general 

radar echo development patterns as they occurred on this day, and some

characteristics of the precipitation, but our analysis here concerns mainly 

1-2 

the detailed radar echo structure and evolution in the multicell storm which 

was the focus of the other observations. Our primary data base is the three- 

dimensional radar reflectivity patterns produced by computer using digital 

data from high-speed raster scans by the NHRE S-band research radar located 

near Grover, Colorado. A complete set of PPI scans was available from near 

ground level to storm top at an average interval of 2 min. A description of 

the radar and observational procedures is given by Eccles (1975). 

2. General features 

a. Character of the environment 

Circulation at the 500-mb level over northeast Colorado on the afternoon 

of 9 July 1973 was influenced by a large anticyclone centered over the plateau 

region of the western United States (Fig. 1). Diffluent northerly flow advected 

cooler air southward during the day and had a destabilizing effect favoring 

convective development which began in the NHRE research area by mid-afternoon. 

Surface analysis at 1500 IIDT showed low pressure in western Kansas with 

cyclonic and confluent northeasterly flow over northeastern Colorado. This 

zone of confluence is reflected in detail by analysis of wind observations from 

the NHRE mesoscale surface network (Fig. 2). Surface moisture was relatively 

abundant with the mixing ratio (r) ranging between 8 and 10 g/kg and showing a 

typical decrease from southeast to northwest across the network. Potential 

temperature (0) on the other hand increased from the southeast toward the 

northwest largely in response to upward sloping terrain and associated decreasing 

surface pressure. The opposing potential temperature and moisture gradients 

produced a field of equivalent potential temperature (0e) which, with the excep- 

tion of two stations on the southern extremity of the network, could be regarded 

as nearly uniform. 

3 All subsequent references to time will denote Mountain Daylight Time (MDT).

-13./ 5790 

I-4 

KIMBALL, NEBR. 

8~ 7~^3> 

4 ; 6^ ^r9 316.Z44.2 / (SNY) 

c o.3- ' .8 SIDNE EBR. 

W YO.^_ __ ___J 

O 36 \ 45 34-9oo--'-W. 

18.4, 344.7 f 

1) 

4 

18.. ! I 

17. 

3 4 4 . 9 

8.2 37.1 345. 

GROVER, COLO'~ 344ROVER, 5474. 

COLO. 

\ ~(GRO)j~ ^ ~34515 3 .5 

. / / ^- 4 / 318 4 4.7 

17 76346.96.9 

food 45-5.6-_~ . 4 3-4664 

\ / /'~9.2 D/ 

345.8 346.5 3 .-9-( 4 

. 4 

19. / 34568 f 

STERLING, 

____,. F.~ MORGANCOLO. (4F) ~ 

(STK) 

^^siT.~ 3 9 ~ ~ ~~i .1 ~4 

/ r /, . f 9 JULY 73 

o10 , sec- 

317.5 3 49" 1 

;0. I I0 

20 30 

Fig. 2 SuArace 4stAeamlines at 1610 MPT, 9 Juty 1973, oveA NHRE meuoscale 

netokh. Ptotting model (Raowet left) incudeu wind (m s- 1 ), potential 

tempeatuwr e (e), equivalent potntieat tempoatwire (ee), and mixing 

ratio (t). Special' awisonde sites are aocated neat Grover (GRO), Ft. Morgan 

(4FM), and SteQting (STK), Colorado and at Kimball (KMB) and Sidney (SNY), 

Nebraska. PPI tadoa echo contoau at 10 dB intervals above 30 dBZ show location 

oa eanesit thundeutorm devetopment. 

4 

.T 2

I-5 

Special serial soundings from five stations designated in Fig. 2 are used 

to identify the local static stability and representative vertical wind shear. 

Temperature and dew point curves (Fig. 3) measured during ascent of a 1630 

release from STK are chosen as representative of the storm's environment, 

because of the sounding's optimum location relative to the storm. The pseudo- 

adiabat having e = 345.5 K represents saturated conditions at cloud base 

e 

measured there by a research aircraft, Queen Air (NlOUW), operated by the 

4 

University of Wyoming (Part II). Cloud base altitude was 3.8 km MSL and 

0, r, and static pressure were 316.5 K, 9.5 g/kg, and 650 mb, respectively. 

A lifted parcel with these properties would experience a temperature deficit 

up to 610 mb and 4 to 5 C temperature excess in the layer between 550 and 250 

mb. We note that cloud base 6 is 2 to 3 K colder then the environment; a 

common observation in updrafts beneath High Plains thunderstorms (Marwitz, 

1972a; Foote and Fankhauser, 1973). 

The inset in Fig. 3 gives the profile of e with respect to pressure from 

the STK 1630 and GRO 1735 soundings. Values near the surface are consistent 

with conditions shown in Fig. 2. Both curves indicate a decrease to near 500 mb 

that is characteristic of a summertime convectively unstable atmosphere in 

northeast Colorado. The coldest and driest air aloft is found on the upwind 

GRO sounding where a layer from 550 to 475 mb has values < 330 K, with a minimum 

near 327 K. 

Wind distribution in the vertical from three soundings is plotted in Fig. 3. 

The lowest level designates surface wind and all others are means for layers of 

50-mb thickness. Conditions on the STK 1630 soundings are most representative 

of the environment in the .layer from the surface to cloud base, while those on 

4Unless otherwise specified all subsequent altitudes will refer to height above 

mean sea level (MSL). Height of the ground in the area of the storm was 41.4 km.

P(mb) P(mb) km 

300 

400x 

I-6 

200 E X 400- - LD TOP- 14 STK 

-T-28/ 

500 / 

6 0 0 - 2 / 0 

-Q/A 10 UW-Q L BASE 

700 Q/ A 

N300A 

800 /BUFFALO N326D- //9.g/k 

12 

_ ^ xs < > 

^ 

7~_ ~- _ 

1000 /) 0 ' 

Fig. 3 Thermodynamic cdiagram showing tempeAature (T) and dew point 

4rom 

(Td) 

STK 1630 sounding. Labelted dashed curves de signate dry and 

mo&st aditabats and mixing iatio representative of measuAted cloud base conditions. 

Cons.tavt pessurte Zevels soanded by instrumented iorLCAaft (Part 

a-e daesignated 

II) 

at lower left. Surace wind and 50-mb tayeiL-averaged winds 

(m i) f{rom thiee represendtaive souandings ar-e on the righat. A ^ful barb 

4s 10 m 6*1. 

In4et hoWs ee (K) v 4 . pressur e (mb) for STK 1630 (4 olid) and GRO 

1735 (dc&shed) sounding4. 

I\

the GRO 1735 and KMB 1725 ascents best depict the environment in the'cloud- 

I-7 

bearing layer. We see that subcloud winds backed from east-northeasterly 

near the surface to light northerly near cloud base. Backing continued in the 

cloud-bearing layer from northerly to westerly near cloud top. 'Radar echo tops 

were recorded as high as 14 km. Wind shear in the layer *from cloud base to 

cloud top (650-150 mb) was computed from the average of the three wind soundings 

-3 -1 

to be 2 x 10 s . According to the classification by Marwitz (1972b), 

a low to moderate shear such as this is typical of environmental conditions 

surrounding multicell storms. Subsequent radar analyses will show that storm 

features on this day were indeed characterized by periodically evolving multi- 

cellular echo patterns. 

b. Radar echo history 

In the early afternoon small short-lived cells emerged from convective 

clouds forming near the foothills of the Rocky Mountains to the west and south- 

west of the NHRE area. By 1530 radar echoes appeared about 50 km northeast 

of the Grover site in convection developing over the Plains. A complex of 

multicellular thunderstorms moved southward from the position of PPI echoes 

shown in Fig. 2 and passed over the western portions of the NHRE precipitation 

and mesonetworks while under the surveillance of the diverse observational 

facilities. A major span of this system's lifetime, which extended from 1603 

to 1821 (2.3 hrs.), is represented by the low-level PPI scans shown at 15-min 

intervals in Fig. 4. At about 1830 a new echo cluster appeared in western 

Nebraska and was tracked as it proceeded southwestward across the sensor network. 

Finally a third development following essentially the same path as its predecessor 

moved from the vicinity of Sidney, Nebraska and eventually dissipated within the 

NHRE area around 2230. 'In conjunction with these three major systems an enormous

1-8 

20 30 40 50 60 70 20 30 40 50 60 70 

_ 161b- - ------------ - 1710- 

20 "/0 

VH 

L IM ________________________ 

' -- J 

-- -- l -- "1625 -40 

-0 O 

-210 

40 

(,Yn - I 

-5 

20 -2 

t OtOu (at 10 d- te above. 30 dZ t appo t 

I -200 

20 0 40 50 60 70 20 /70 0 40 

60 70 

Fg. 4 Low tCit (J4° EL; ^3 km MSL) 

d beag, 

compwteJ-geneLw)texj 

V and ou top, 

PPI 'adcv 

e 

'jc- 

0 m -= 0 . Sta on 

4-tonm tenm on 9 JwCy 1973. Two po&ncm enchoe compnng -the. ovwef2VL 

Fiadaglz 

aogacaL 

ocagtio. 

vae abte,0 

wind 

and 

veltou 

p Ae.cpaton 

on 161 

tom Etomn Wwe ca. Lame d. beaz 

0 ~ 3 nEL kS)mpteneaed sPPI radar reflec- 

/110 pane de^gnaote. 

ocat9ono9 J 7ace. ot 

mez onetwozk Ltmp whg the ovte.oo- m

number of "initial" radar echoes developed both.within and adjacent to their 

I-9 

primary areas of influence. Of these, %80% were shortlived and did not mature 

to become major cells. 

Although firm statistics are not yet available, the multicellular develop- 

ments on this day seem to be typical of the most frequent type of thunderstorm 

occurring in northeast Colorado. Less typical were the observed storm movements. 

In contrast to the normal case where the surface wind has a component opposite 

to the general storm motion individual storms moved virtually parallel to and 

faster than the local surface wind vector. The position of early echo developments 

is shown in Fig. 2 to be just north of a surface confluence zone. Echoes 

forming to the north of the confluent asymptote moved with a predominately 

northerly component while those forming to the south moved from the northeast. 

This behavior essentially reflects the surface wind flow. 

Analyses to follow will focus on the individual storm, located on the 

western edge of the earliest echo complex, identified in Fig. 4 by the letter W. 

Cellular patterns at the times chosen to demonstrate echo configurations and 

translation do not reveal any obvious evidence of discrete propagation. 

Subsequent detailed analyses will, however, reveal this to be a primary feature 

of the storm's behavior pattern. 

c. Precipitation 

Hailpads and weighing rain gages at four locations, designated on the 1710 

panel in Fig. 4, recorded hailstone size and total precipitation during the 

direct overhead passage of the intense core of storm W. Average rainfall 

accumulation at the four sites amounted to 12 mm and the peak rainfall rate 

was near 120 - mm 1 hr . Of the many thunderstorms of the day, storm W was the 

Basic characteristics and measurement capability of the hailpad used in NHRE 

aredescribed by Nicholas (1975).

1-10 

only one to pass over hailpads located at the four sites. Thus, although 

hailpad data is not time-resolved, recorded events could be attributed uniquely 

to storm W. Analysis of these pads showed that hailstone size ranged from 

0.3 cm to 1.6 cm, diameter, with a median size of 0.5 cm. Although no hail- 

stones were collected beneath storm W, ground chase crews did retrieve 

specimens which fell from storm E. Laboratory analysis (Knight, et al, 1974). 

of these revealed that about 75% grew from graupel embryos and 25% had centers 

comprised primarily of frozen drops. A rough comparison of rain gage and 

hailpad data indicates that less than 5% of the total mass of precipitation 

from storm W fell as hail. 

From aircraft wind and moisture measurements, Fankhauser (1974) found that 

the rate of water vapor inflow to storm W was 42.5 kt s 1 

The corresponding 

rate of water deposition measured at the ground was ^1.0 kt s 1 , leading to 

a precipitation efficiency of about 40%. Both the inflow and output were 

nearly an order of magnitude less than that found by Foote and Fankhauser (1973) 

for a hailstorm of the supercell type which formed in a highly sheared environ- 

ment, but precipitation efficiency was somewhat greater in the present case. 

This result is in line with the inverse relationship between shear and precipi- 

tation efficiency proposed by Marwitz (1972c). 

3. Formation and evolution of the cells 

a. The evolving three-dimensional structure 

As shown in Fig. 4, both storm W and storm E were at most times composed 

of numerous cells which on first inspection appeared to grow and decay in an 

apparent random manner. Close examination of cells comprising storm W, however, 

revealed that they evolved with a systematic behavior. The three-dimensional 

structure of this storm is shown as a function of time in Fig. 5 by PPI tilt 

sequences. Stepped altitude scans presented at 415 min intervals are chosen

I-11 

12.7 iWO o WI W2 

woo W 

W16O W21 W3 

10.9 wI" wW4 

W o . A WI!OW 

RaWOda K W2 £ WO W2: W3 3 

-. 1 WI E 

~~~~~~~~~~~~WI 

0~~~~~~~ 

W3, 

Wa ·WI a g ~ 18~3 

72 7.2 ' 

W2 W3 4 

2.6 

_ II WW 

the. ne abeed at the botom o eac. W A 

n W ug 

WS WL. >de~nted n: boxe. ov d t ng da2d Ln g m o 

0 iO 20I 30 40 km 

1616 1631 1647 1'703 703 1716 (MDT) 

VCCwao,/ and conn y o ce wtpo Wgh 4. VuW dahd n 

1cdaA. 'eZec-tu&t contow' LpLeSent 10 dB nt&vaOW above. 30 dBZ. 

lndcv/cduaZ Individual tilt'sequences t&^t jeqaences were wve. obtaZnead obtained kn in a pevod period of ]20 120 z cejvteWd centered s abowt about 

the 

Fig. 

time 

5 

labeled 

PP 

at 

can 

the bottom 

at 

of 

evatio 

each. D isetee 

entred new 

at c 

ceil 

at ud 

formation 

t y wc ind 

WI 

icattd 

thtough 

W5 tab onu hed. ent. 

'egm 

ate identified 

Radarhow oectinvity 

in boxes. Downward 

contour repteaent C0 

Isopivng dashed tines 

dB interval 

give examples 

above 


vertical and tempoal continulty of ceW WI thvough W4. 

30 dBZ. 

Thow 

Verticat 

vSQtceatZ 

dashed gines 

contnutL oj mna twL at ach tme. Indcvtduat C aW2 

segments~ show orien~tationz of verzt~ical sec~tionsc in F~ig. 6.

1-12 

to emphasize five discrete echo entities (identified by boxed labels) first 

appearing in the altitude range of 5.6 to 9.1 km (MSL) at locations to the 

south and ahead of the main echo. Cells are labeled according to their 

sequential order of appearance (e.g., Wl, W2 ... Wn) and continuity for the 

four earliest developments is traced by dashed lines sloping downward with 

time. For example, cell W1 appearing on the 7.2 and 9.1 km scans at 1616 

represents the first discrete development and is seen on later sequences 

(1631 and 1647) to descend and move through the echo complex as cells W2 and 

W3 form and progress in like manner. Nearly vertical dashed lines are 

included to show vertical continuity of mature cells at the time of each new 

echo formation. Cross sections of radar reflectivity in Fig. 6 show the two- 

dimensional radar structure in vertical planes aligned along line segments 

shown at selected altitudes on the PPI scans in Fig. 5. 

Common features of cells W1 through W5, evident in Figs. 5 and 6, are an 

average altitude of first appearance near 7 km and a tendency to form at 

a distance of 5 to 10 km ahead of the southward-moving complex at intervals of 

about 15 min. They all grow rapidly in size and intensity, moving in a 

relative sense toward the main echo, soon becoming its primary component. 

As shown at 1631, 1703, and 1716 IDT in Fig. 6, as many as three in the series 

of five cells coexist in varying stages of development. For instance, when W5 

first appeared shortly before 1716, 1W4 was approaching maturity and W3 was in 

the early part of its decaying stage. Whereas the multicell model of Marwitz 

(1972b) specifies lateral alignment of coexisting cells across the direction 

of overall storm motion, Figs. 5 and 6 show that the cells in this case tend 

to be aligned along the direction of system movement. 

The structure on the right of the vertical reflectivity profiles in Fig. 6, 

particularly at times 1631, 1703 and 1716 MDT, is rather similar to the weak

8 - 

4 

1 2 

//^^wo 

1-13 

" 001° ,-> 181° 

____ /1616 (MDT) 

0 ---------------------- 

I W2 

8 ' 

W3 

4 \ / - CL O/D 8ASE- 

1647 

0 ,I . I. 

12 12 - 

3600 1800 

8 

4W 

W4 

)~ m V\ /703 

o 0 , I I 

KM 50 1 20 915 

Fg1. 6 Vetca c con1 o tada eecvty agned aong inW 

1716 

KM 5 I 0 15 20 

Fig. 6 Venticat ctoss Sections6 of radar trefectivity atigned along tine 

zegments shown fo each time in Fig. 5. ReRectivity covtourw 

represent 10 dB 4teps above 30 dBZ. Cels aLe Rabeled as Zn F g. 5. HoiLzontMl 

and veticae dUitance scales ar equtvaee nt. The heighkt oa the H 

ttrpopautse and the attude of an aLctra4t penetation (Pait IV) arLe indicated 

on the. 1716 croa s e.ction.

1-14 

echo region (WER) associated with updraft regions in supercell phenomena 

(Browning, 1964; Marwitz, 1972a; and Chisholm, 1973). Marwitz (1972b) and 

Chisholm and Renick (1972) point out that weak echo regions are also common 

features of multicell storms but that they are transitory and occur as a 

result of the discrete echo recurrence in the forward sector of the storm. In 

contrast to the persistent overhang in the supercell case, we conclude that the 

echo overhang depicted at times 1631 and 1716 in Fig. 6 is formed by new echo 

developing adjacent to and quickly joining existing mature and decaying cells. 

In both the supercell and multicell cases, however, the overhang implies the 

existence of updrafts supporting suspended precipitation particles. 

The cells identified in Figs. 5 and 6 were not the only cells appearing 

in the storm between 1610 and 1730 MDT; there were many smaller ones which 

grew and decayed rapidly, having only a minor effect on the storm's overall 

configuration and evolution. There were also a few stronger cells which 

appeared periodically in a WSW direction and close to (within 3 km) existing 

mature W cells, about 15 min after their appearance as a first echo. The 

evolution of these was not easily distinguishable from that of the W cells 

but they are thought to be peripheral developments related to updraft circula- 

tions supporting the primary W cells. Since the W cells labeled 1 through.5 

seemed to have the dominating influence on the storm's propagation, we will 

give special attention to their formation and evolution. 

b. Location of new cells 

The actual horizontal location of each new cell formation is shown in 

Fig. 7 with time and approximate height of first echo detection also indicated. 

The locus of points identifying new cell appearance forms nearly a straight 

line which represents the direction of motion for the storm as a whole. As 

mentioned in Section 2 this was essentially parallel to the direction of surface 

winds along a zone of surface streamline confluence.

1-15 

20 ' 

I 

10 

WO 7/602 MDT 

I (7.5 KMMSL 

I 

.i~ W~~I ~ 

.(\NWI Wl 1615 

GROVER I 

RADAR 

W2 /629 

/64/ 

-I0 - I 

/65/ 

WINDS RELATIVE W3 1 /645 

TO GROUND 7.0) 

VH |I 

-20 711 

10 W4 /702 

VL m sec-' ' (6.5) 

VH W5/7 

% 

WINDS RELATIVE 

L TO STORM m~V L '-ecI1 | (7.0) 

/ 1 72/ 

RADAR 

-50 NOAA DOPPLER 

VM 

-50 - 

W7 16 /744 1/744 

1(7.0) 

-60- -60 - 

- . 

AVERAGE STORM MOVEMENT 

O(KM) 

I 

10 

I 

20 

I 

30 

I 

40 50 

Fig. 7 Inveited ;tAiangte& denote the ptan pouition oCd t6t echo appearance 

o& ceed&s wo through W7. Tmne and aJtitude (pa.nthenisu) oj6 6Lut 

appeatance aVte abso indicated. Mtvoums Thow the pacth o ceLU (Wl tthA'ough W5 

69om 6iZt detection to maximum etecativity. The timne when each wa Reo&t a& 

an entity is potte.d at the end o5 the. tepective. tkachk. Location6 o6f the 

G'LOveLA nttdwat and a NOAAM oppRelt /adaA (Pattt I7T) acte. ndicate.d, 

Inset gives envtiomnenta Wind6 'e&attve to the. gwound and to the. 

sto'm (designated by a c&tlcel) dj .theee eaQyeu; 4ubteoud, VL, middLe ttopo- 

4phee., VM and u.ppei twopozpheLe, lV.

1-16 

New cells appear at intervals of 13 to 16 min, and at locations approxi- 

mately 10 km south of the formation point of the preceding developments. This 

sequence is broken between 1716 and 1744, as an anticipated development 

corresponding to W6 did not occur. We note, however, that a cell identified 

as W7 appeared 28 min after W5 at a distance of 2 x 10.5 km south of the W5 

formation point and since these time and space increments are twice the period 

and separation of earlier developments we include it in the W series. The 

close proximity of new cell formation to the location of surface convergence 

maxima at the storm outflow-inflow interface (Part II) suggests that the 

periodicity in new cell development may have been related to the downdraft 

production by preceding cells during their intense and decaying stages. 

c. Tracks of individual cells 

Paths of cells W1 through W5 are shown in Fig. 7 for the length of time 

each was a distinguishable entity. For the most part, this period includes 

the time from first radar detection through intensification to maximum 

reflectivity. Following first appearance each of the radar echoes grew 

rapidly in size, intensity, and vertical extent. During these early stages 

they were nearly stationary with respect to the ground and some even showed 

a tendency to move somewhat northward; a direction opposite to the overall 

storm movement. During their later developing and mature stages the paths of 

all cells are shown to follow about the same direction, moving from an azimuth 

of 340 to 350 , which is in agreement with the direction of V, in Fig. 7m 

(upper, inset). Since no southerly flow was observed in the subcloud or 

middle tropospheric layers of the environment, the short period of stationarity 

or slight northward movement can be explained only by propagation of the radar 

echo through the updraft'due to successive particle growth preferentially to

1-17 

the rear (north) of the updraft. After ascending beyond 7 km the echoes 

apparently accommodated to environmental steering winds and began their south- 

southeastward travel. During decay (not shown in Fig. 7) upper portions of 

the echo comprised of particles with low terminal fall velocities moved toward 

the northeast with a velocity of the high level relative winds (VH, in Fig. 7, 

lower inset), while lower and larger particles influenced the overall storm 

velocity during descent in strong downdrafts associated with the zones of 

maximum low-level reflectivity shown in Figs. 5 and 6 (see also Doppler radar 

observations discussed in Part III). 

d. Evolution of individual cells 

Intensification of cells W1 through W5 with respect to time is given by 

the growth curves of reflectivity factor, Z (dBZ), in Fig. 8. The 15 min 

periodicity for new cell development is again demonstrated here. The average 

rate of echo intensification for the various cells increases with time from 

about 1 dB min 1 for cell W1 to %4 dB min 1 for W5. The fastest growth, that 

of W5, is about half the rate reported by Browning and Atlas (1965) for the 

early stages of a tornadic Oklahoma thunderstorm but about the same as the rate 

computed from data presented by Renick (1971) pertaining to a multicell hail- 

storm in Alberta. The indicated increase of growth rate with time may be 

related to a corresponding increase in the strength of updrafts supporting the 

later cells and/or to changes in the size and distribution of aerosols. Since 

subcloud thermodynamic characteristics along the path of the storm were fairly 

uniform in time and space (Part II), the increase was not likely to be dominated 

by static stability variations and we look to increasing mesoscale convergence 

and to the possible ingestion of larger aerosols as explanatory mechanisms. 

Visible dust curtains rising at least halfway from the surface to cloud base 

were observed at the interface between inflowing and outflowing air by research

N 

0 

o_ 

0 

0 

0 

I 

1-18 

_ , | I _ 

o 

0 1 

0- - 

" - 

-I~ - 

Fig. S Curves showing the 

growth of RadarL refecti.vUy (dBZ), f ot each cele 

as a function od time. Dashed Zines show slopes used to obtain 

ntens.ficaation rates indicatlve of -the early istoat of each cele. 

I

1-19 

aircraft during the latter stages of the storm's history. Surface convergence 

is discussed further in Part II but no direct measurements of aerosols were 

available in the present case. 

The radar history of cell W5 is represented in Fig. 9 by the development 

of radar reflectivity'contours as a function of height and time. According to 

Fig. 8, W5 was one of the most intense cells in the series. Detailed radar data 

showed that its maximum reflectivity reached 68 dBZ below cloud base at about 

1733. Profiles similar to Fig. 9 constructed for the other cells showed that 

maximum echo tops for each cell increased progressively with time from about 

12 km for W1 to > 14 km for W5. The height of Z displayed a general tendency 

max 

to increase accordingly. This supports the speculation that the strength of 

updrafts supporting successive cells also increased with time. All other charac- 

teristics of the evolution of the W cells were similar to those of W5 and we will 

center the discussion of their general behavior around the profile in Fig. 9. 

The locus of circles plotted at 1-rmin intervals in Fig. 9.is intended to 

represent the history of precipitation particles influencing the evolution of 

cell W5. The ascent rate of visual turrets deduced from cloud photography taken 

between 1655 and 1715 was between 5 and 10 m s . This was used to approximate 

the particle trajectory between 1700 and 1715 (the time of first echo for W5). 

From 1715 onward the path traces the history of the maximum reflectivity, assumed 

to be the trajectory of the largest particles. The rationale for relating the 

particle growth cycle to the history of the maximum reflectivity rests on the 

dominant dependence of radar reflectivity on particle size. 

Of the 30 to 35 min total lifetime from air parcel entry at cloud base 

through first echo to the arrival of the Z at the ground, about 15 min is 

max 

spent in rather slow ascent in the rising turrets comprising the shelf cloud 

to the south of the main system (see Fig. 1, Part II). The temperature scale on 

the left in Fig. 9 shows that undiluted air parcels undergoing moist adiabatic

14 

12 

TROP 

-40 

1-20 

0 5 5 5 5 5 

-30 

-25 

8 --20 

-' 5ep 35 dBZ. The aho po e hn 

t 

I- -1 

-1 // 

2 

SFC 

I10 

T (OC) 

dBZ /ho ontou5 

whch 

55 50 45 40 35 30 25 20 15 10 05 1700 (MDT) 

Fig. 9 Time-hecight profie. of the radat treftectivity contours fotL cett 

at 

W5 

5 dB intervals above 35 dBZ. The echo 

the 

prof{ite 

history 

as shorxn 


/ep&sent6 

particles that remained aloft for the longest petiod 

The dazshed 

o 4time. 

35 dBZ contowL tepreseents the historty of echo uwhich descended molst 

tapidty adteA the dfitt echo appearance. The tocus of citcles pZotted 

mnute 

at 

inteAvats 

one- 

teptesents the hiLtoty od an ait parcel which enter 

ceoud 

thAough 

bae., ascendsz en the ets of the t she. cloud and gtour 

enough 

paaticles 

to produce. 

arwge 

the dZt echo at an altitude of -7 km. 

onwatd 

Fhom that 

in 

point 

time the path foRtowm the maximum e.lec.tivty. Schematic ctoud 

heights in the. she.Z cloud and at tine of fibt 

lapzse 

echo 

photography 

ate derived 


fdom 

were 

time 

u-sed to estimate asce.nt fist rate. pior to echo. 

The temperatwue scae. on tef.t rtepr&ee.nts molst adiabatic 

for undiiute.d 

condtions 

satwLted ascent from cloud base.

1-21 

ascent would have a temperature of -12 C at the average level of first echo 

appearance (7 km). Subsequent to first detection the center of highest reflecti- 

vity ascended rapidly to an altitude of %9 km. At about the same time two 

closely spaced reflectivity maxima appeared. These are evident on the vertical 

section at 1716 in Fig. 6. The profile in Fig. 9 represents the history of 

the reflectivity maximum nearer the forward edge of the storm. The one to the 

rear descended toward the surface at an earlier time and the history of its 

leading edge is represented by the dashed 35-dBZ contour. In addition to this 

double structure in Z along the vertical plane, analysis of cross sections 

max 

adjacent to those used to construct Fig. 9, indicated that radar echo in the 

northeast sector of W5.also formed and descended earlier than did the contours 

as shown in Fig. 9. Both of these factors account for the lack of agreement 

between intensification rates shown for W5 in Figs. 8 and 9. The reflectivity 

contours in Fig. 9 trace the history of particles that remained aloft for the 

longest period of time and these had slower intensification rates than those 

used to produce the W5 curve in Fig. 8. 

Following ascent to near 9 km the maximum reflectivity remained suspended 

there for a period of 10 to 12 min. During most of this time it resided in a 

region where the in-cloud temperature was less than -20 C. When radar reflectivity 

reached 55 dBZ, descent of Z toward the ground began; at first slowly but with 

max 

a rapid cascade during the final 3 to 5 min. During these latter stages the 

slope of the high reflectivity contours with time gives a descent rate of about 

-1 -1 

30 to 35 m s . Maximum observed downdraft was 415 m s (Part III), and 

the terminal fall velocity of the largest observed hailstones (1.5 cm, diameter) 

would be 420 m s so that there is reasonable agreement between the radar 

echo history, hailstone size and measured Doppler downdraft velocity.

1-22 

4. Storm motion in relation to the translation and propagation of individual 

cells 

For examining the relationship between cell motion and the movement of 

the overall storm we refer to Fig. 10 which gives a continuous plot of radar 

reflectivity contours as a function of time along a y-axis which is aligned 

along the north to south direction of storm movement. At any instant the 

forward and trailing edges of storm W, usually in the altitude range of 4-8 km, 

are represented by the outer (30 dBZ) contours of the radar echo band. 

Interior contours similarly represent the leading and trailing edges of the 

higher reflectivity levels. For the most part, the leading edge (greatest 

y at any x and z at a given t) is defined by the leading edge of the newly 

forming W cells whose individual histories are traced by the heavy arrows. 

Alternating dashed and dotted contours, superimposed on the continuous contours 

show the history of the trailing edges of the individual cells for the periods 

that they existed as discrete entities. Since they formed mainly to the south 

of the main system, their trailing edges consistently appeared at y-values 

greater than contours tracing the continuity of the overall storm. With time, 

however, the contours of individual cells show a tendency to blend toward 

equivalent contours along the trailing edge of the main swath as the cells. 

lose their identity and become components of the main echo complex. Cell W2 

which formed on the right flank of W1, and slightly to the north of the leading 

edge of existing echo (see PPI tilt sequence at 1631 in Fig. 5) was tracked 

along a different y-axis than that followed by the main system until it became 

a part of the overall complex. This explains why its leading edge appears at 

smaller y in its early stages. 

Most of the cells; particularly W2, W4 and W5, exhibit a tendency to move 

more slowly in their early stages and to gradually tend toward the velocity 

of the storm as they intensify. The early period of slow movement very likely

1-23 

30 (dBz) 4 

. /I/ 0 00 I0 50 

30 I'0 60 

Vs = 0020/10.3 

50 

m sec 40 /0 

3 0 

Vc Vc = 3470/4.0 m sec- 1 /' / 5 

20 P 0110/6.5 o msec 

2S 

1620 / / / 

(MDT) c1 

40~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~······ 

/ / 

2'1 : 

/ / s1 / / 

60 

··· ·· · · 

~ 

6C!/(···· 

0 (km) 10 20 30 40 50 

Fig. 1JO The teadnq (/h4) and ttaitng (zh) edge o te.ectivity contouu, 

uwuaaq Zn the a&tttde 'ange o 4 to I km, io/ 

(otid), 

itoun W a 


a whore 

4Q the Cee. pe/uodicaty c o2q ing to Zwtain 

dashed 

it (altenating 

and dotted), 4hown az a {unction o6 time.. The abcJZa 

tance 

denotes 

in ktomete4 

dLi- 

eentiaf4 due 4outh wom a 

Stoping 

point 

tneu 

ne.a the 

give 

otigin 

the. 

o6 WI. 

peed o0 the. aveage. 4tokm and cea. motion. 

Vec-to dUagLan Zn the. net 4ummadZze the. iLe.h&ti~onship 

Ve., t0bu-on and 

be.4we.e.n 

^tom, to the- ce.-&t, 

V7, motcon. o -. a The. dldê.ence on made by beAf.een dCt the. popagaon, tNo /eLe^nt V theh con

1-24 

corresponds to the period of strongest updrafts and as updrafts decreased, partly 

in response to water loading associated with increasing reflectivity, the cells 

tended to assume the velocity of the mid-level ambient wind (see Fig. 7, inset). 

Closed 60-dBZ contours (shaded streaks) of varying duration and length 

appear in Fig. 10 toward the end of each cell's traceable history. These vary 

in length from 6 to 20 km and in duration from 12 to 33 min. The one of 

longest duration seems to have been formed successively by two or more cells 

(WO, W1, and W2) so that an average length and duration would be somewhere 

between 6 and 10 km and 10 to 12 min, respectively. Ground areas below these 

high reflectivity streaks are reasonably the most favored regions of hailfall 

and very likely correspond to the hailstreak phenomena discussed by Changnon 

(1970). Although "ground truth" measurements were not available, hail was 

apparently generated and deposited at the ground in a cyclic manner related to 

the periodic evolution of individual updrafts through the multicellular storm. 

Slopes of y with respect to time shown in Fig. 10 give the average speed 

for both the storm, V, and its cells, V . Directions of cell and storm movement 

S c 

are taken from Fig. 7. Vectors representing the average storm (V ) and cell 

s 

(Vc) velocities are plotted in the inset. Subtraction of the cell motion vector 

from the average overall storm motion gives the resultant labeled V . This 

p 

vector has a magnitude of 6.5 m s and represents the component of storm 

velocity contributed by discrete propagation through new cell development on 

the right forward flank of the storm. Reference to Fig. 5 shows that the orien- 

tation of V represents the general direction for the location of new cell 

p 

formation with respect to the existing echo. Since propagation contributes 

-1 

6.5 m s to the storm movement and discrete propagation occurs at a period 

of 15 min (900 s) we would expect the separation between existing radar 

reflectivity maxima and new echo to be around 6 km .on the average. Figure 6

1-25 

shows that at the time of first echo appearance the distance between new and mature 

elements varies from 5 to 10 km, in fair agreement with the anticipated value. 

Thus, in summary, the motion of the main body of the storm was the result 

of two components; one due to the advection of individual cells along the 

direction of middle level winds and slightly to the left of the overall storm 

movement, and another due to the periodic and discrete propagation by new 

cell formation on the right forward flank of the echo. In the present case 

it is clearly the propagative component that has the dominant influence on 

overall storm motion. These results are similar to the general multicellular 

model proposed by Browning and Ludlam (1960) which has been found applicable 

in other High Plains thunderstorm studies by Marwitz (1972b) and by Chisholm 

and Renick (1972). 

5. Summary 

We have analyzed in considerable detail the radar structure of one of.the 

many multicellular thunderstorms that occurred in northeast Colorado on 9 July 

1973. The hailstorm receiving our concentrated attention had a lifetime of 

nearly 2 hours and was comprised of at least seven distinct cells which domi- 

nated its overall behavior. With maximum radar tops of 14 km and maximum hail 

size of 1.5 cm, diameter, the storm can be classified as moderate in intensity. 

A periodic mode of cell development and evolution was characterized by 

discrete propagation of new cells forming at an average altitude of 7 km (-12C). 

*on the storm's right forward flank, 5 to 10 km ahead (south) of existing storm 

components at a frequency of about once every 15 min. This infrequent rate of 

new cell development may be compared to an average of one every 5 min found in 

an Alberta hailstorm similarly analyzed by Renick (1971). 

After first radar detection all cells grew rapidly in size and intensity 

and, by moving more slowly in their early stages than.the overall echo, soon 

became the main storm component. Average lifetime of individual cells, including

1-26 

the time from visually perceptible turrets to 'first echo' through decay, was 

45 min. Of this period, 30 to 35 min was radar-detectable history; approxi- 

mately 15 min being spent in growth of precipitation particles to radar detect- 

able sizes during ascent from cloud base (3.8 km) to the altitude of first 

detection byradar. With a formation interval of 15 min and a cell lifetime 

of 45 min, at any instant, as many as three cells were found to coexist in 

varying stages of development. At a particular stage in the evolution of each 

cell, a vertical two-dimensional radar structure resembling the weak echo region 

and forward overhang typical of supercell thunderstorms was observed. In 

contrast to the supercell case, these features were quite transitory and a 

result of the new echoes joining with the main echo soon after their formation 

as discrete entities. 

Overall storm motion, from the north at an average speed of 10 m s, 

was dominated by the propagation of the cells on its right forward flank. The 

average speed of cells was only 40% of the storm speed, hence the propagational 

component contributed well over half the storm's motion. Cell motion was 

quite slow in early stages of growth but tended with time toward the direction 

of the wind in the middle troposphere and eventually moved with about half its 

speed. After maturity and during decay the upper portions of the cells were 

carried out in the anvil in the northeastward direction of relative winds at 

the high levels (10-12 km), while the lower parts moved southward as components 

of the main body of the storm. 

Although the general characteristics and behavior of individual cells were 

similar, as might be expected some irregularities and variations were observed. 

Maximum radar echo tops, maximum radar reflectivity, and intensification rate 

all showed a tendency to increase with time, and this increase appeared to be 

related to a corresponding increas e strength of updrafts supporting the 

successive cells.

1-27 

Consistent with the evolving radar structure, bands of high reflectivity 

appearing on y-t plots (Fig. 10) suggests that hail production at the ground 

occurred in streaks and this bandedness indicates that there was one major 

precipitation burst per cell. Largest hailstones measured at the ground were 

1.5 cm in diameter. Analysis of time lapse photography of growing cloud 

turrets and the subsequent history of maximum reflectivity as it evolved from 

first echo through descent to the ground indicates that the total time avail- 

able for growing hailstones of the observed size within the individual cells 

was between 30 and 35 min. Of this period, dwell time at temperatures lower 

than -20t was on the order of 10 to 12 min. 

Thermodynamic variables and airflow structure near and beneath storm W 

are elaborated in Part II of this series. Parts III and IV present internal 

airflow and microphysical measurements and Part V summarizes and synthesizes 

the complete observational set.

1-28 

REFERENCES 

Browning, K. A., 1964: Airflow and precipitation trajectories within severe 

local storms which travel to the right of the winds. J. Atmos. Sci. 

21, 634-639. 

_ , and F. H. Ludlam, 1960: Radar analysis of a hailstorm. Tech. Note 

No. 5, Dept. of Meteor., Imperial College, London, 106 pp. 

_ , and D. Atlas, 1965: Initiation of precipitation in vigorous convective 

clouds. J. Atmos. Sci., 22, 678-683. 

Changnon, S. A., Jr., 1970: Hailstreaks. J. Atmos. Sci., 27, 109-125. 

Chisholm, A. J., 1973: Alberta hailstorms, Part I:' Radar studies and airflow 

models. Meteor. Monogr., 14 (36), 1-36. 

__ , and J. H. Renick, 1972: The kinematics of multicell and supercell 

Alberta hailstorms. Alberta Hail Studies 1972, Research Council of 

Alberta, Hail Studies Report No. 72-2, 24-31. 

Eccles, P. J., 1975: Developments in radar meteorology in the National Hail 

Research Experiment to 1973. Atmospheric Technology Fall-Winter 1974-75, 

National Center for Atmospheric Research, Boulder, Colo., 34-45. 

Fankhauser, J. C., 1974: Subcloud air mass and moisture flux attending a 

'northeast Colorado thunderstorm complex. Preprints, Conference on Cloud 

Physics, Oct. 1974, Amer. Meteor. Soc. Boston, 271-276. 

Foote, G. B., and J. C. Fankhauser, 1973: Airflow and moisture budget beneath 

a northeast Colorado hailstorm. J. Appl. Meteor., 12, 1330-1353. 

Knight, C. A., N. C. Knight, J. E. Dye, and V. Toutenhoofd, 1974: The mechanism 

of precipitation formation in northeastern Colorado cumulus. I. Observa- 

tions of the precipitation itself. J. Atmos. Sci., 31, 2142-2147.

1-29 

Marwitz, J. D., 1972a: The structure and motion of severe hailstorms. Part I: 

Supercell storms. J. Appl. Meteor., 11, 166-179. 

__ , 1972b: The structure and motion of severe hailstorms. Part II: Multi- 

cell storms. J. Appl. Meteor., 11, 180-188. 

__ 1972c: Precipitation efficiency of thunderstorms on the High Plains. 

Preprints, Third Conf. Wea. Modification, Rapid City, S. D., Amer. Meteor. 

Soc., 245-247. 

Nicholas, T. R., 1975: Surface hail instrumentation in the NHRE. Preprints, 

NHRE Symposium/Workshop on Hail and Its Suppression, Estes Park, Colo. 

(unpublished manuscript). 

Renick, J. H., 1971: Radar reflectivity profiles of individual cells in a 

persistent multicellular Alberta hailstorm. Preprints, Seventh Conference 

on Severe Local Storms, Amer. Meteor. Soc. Boston, 63-70.

Structure of an Evolving Hailstorm, Part II: 

Thermodynamic Structure and Airflow in the Near Environment 

by 

J. C. Fankhauser 

National Center for Atmospheric Researchl 




Weather Modification ,Program, Research Applications Directorate, National 

Science Foundation.

ABSTRACT 

A diverse set of mesoscale observations collected in the National Hail 

Research Experiment in connection with an evolving Colorado hailstorm is 

analyzed to determine the kinematic and thermodynamic structure of the near 

environmental and subcloud regimes. The analysis centers on multi-level 

aircraft measurements in the inflow sector and on mesoscale observations at 

the surface and aloft. Although considerable evolution was observed in 

overall radar echo development patterns, changes in one of the many storms 

detected occurred in a systematic and periodic manner. The instantaneous 

structure of the subcloud inflow into this storm is emphasized in the present 

work. 

Surface and aircraft data substantiate the following characteristics: 

Inflow air approached the front of the storm, originating from a very shallow 

layer (< 500 m) near the ground and ata considerable distance (> 20 km) 

upstream in the relative wind direction. Inflow air rose unmixed to at 

least cloud base, feeding the main updraft which was inclined upward in a 

direction opposite to the storm movement. Discrete inflow-updraft branches 

were found to be supporting coexisting cells in varying'stages of development. 

These had lateral widths of 6-8 km and were separated by regions of weak 

subsidence. Downdraft air approached the storm from the right flank at mid- 

cloud level and at least a portion descended unmixed to the ground in the 

strongest downdrafts.


II-1 

General environmental characteristics and the detailed radar echo structure 

of an evolving multicellular hailstorm, the so-called Raymer storm, observed in 

the National Hail Research Experiment on 9 July 1973 are presented in Part I of 

this series. This paper investigates the kinematic and thermodynamic properties 

of the air near and beneath the sorm, as deduced from multi-level aircraft 

measurement, primarily in the inflow sector, and from mesoscale observations at 

the surface and aloft. 

In Part I the subject thunderstorm has been identified as storm W, and 

that nomenclature will be preserved here. Conventions adopted there for refer- 

encing time and height will also be the same. During much of its radar-detectable 

lifetime, shown by the PPI sequence in Part I, Fig. 4, storm W was under the 

surveillance of five research aircraft. These included three Queen Airs [two 

from the National Center for Atmospheric Research (NCAR), Research Aviation 

Facility (N304D and N306D) and one operated by the University of Wyoming, Depart- 

ment of Atmospheric Sciences (NlOUW)]; the NCAR Buffalo (N326D); and the T-28 

penetration aircraft flown by the South Dakota School of Mines and Technology 

whose measurements are discussed in Part IV of this series. Altitudes sounded by 

the various aircraft near and within the storm are indicated in Part I, Fig. 3. 

2. Physical cloud features 

To provide a point of departure and reference for the analysis that follows 

it is useful to describe features of storm W's visual appearance and to put them 

in the context of radar characteristics presented in Part I. This purpose is 

served by a composite photograph (Fig. 1), constructed from sequential pictures 

taken at 1719 from an NCAR research aircraft flying toward the cloud on a north- 

easterly heading at a'position v40 km southwest of the center of the storm. 

The vertical plane has a southeast (right) to northwest orientation and is nearly 

parallel to line AB in Fig. 2.

CLD BASE 

II-2 

SURFACE T 

-~L~L~1N~ STORM 

Fig. I Composite photograph of m uticedua&L t iundwsz tomi ccmplex (stosm 

W and E) 6tom tesearch Acircta t fying at an atituade of 3.5 km and 

at a diUtance of %40 km (WS( of the center o s-toIm W0 at 17J19, 9 July 1973. 

Ouientation o the veUtca2 plane is neaIly paatRel to eine AB in F.ig. 2. 

Signiicant cloud eeatsUeA are RabeRed and tocaiaons of tadac echoes assoacited 

wth ;the majon ceL& campnising stotm W are identidied. The point od 

cloud entrLy 6aor te T-28 aicraft peneattion dscussed in Pa-t IV ui designated 

by an X.

- 

11-3 

3 4 

42.3 \ 

~ \\ \ \\9 \ 2\ 308.4 339.4 

\ 10.3 

315.7 

315.7 

343.4 

343.4 

8.9 

u /C_ Z~i \ 318.8\3454 ^/^ '- 31 349 1\\- 

30\ 

3_58945.6. 

20 30 40 50 60 

Fig. 2 Cloud base. PPI radar echo reflectivty contoum (10 dB Lntervats, as 

E n a Fg. 4, Paut I) and 3 u9tace seamines stormn euve to W and 

E cft 1710. Peotted data at mesonet ites inc2ude potentia temperature 

edt), 

(upper 

eq.uva.ent potenta temperatue (uppeQ 

aight). 

right) and 

Heavy 

mixing 

barbed 

ratio 

confuent 

(tower 

asymptote identiies boundary of outfow 

storms 

from 

W and E. Hatched regions khow 

exceecding 

tocations 

10 

of horuzontaL convergence 

- 3 s- 1 . Dot wuh schexmatic ctoud otune. hos reati 

on 

ve 

new 

ocaion 

c ,el , W5, Pon oa an domcinating s&iues the evoion oi sStom W. Dahe.d 

cork crUe ttack at towV center e designates relative. potetiaon of aurcrate 

descent sownding. VDahed fine (AB) hac orntaton section ofn veticc.at in 

Fig. 8. Border s ltabeled in kilomettes east and south of Grover, Colorado.

11-4 

Cells comprising storm W form the cloud mass extending from the central 

foreground, leftward. The tallest cumulonimbus tower (top, center) represents 

the rapidly intensifying cell, W5 (see Part I, Figs. 8 and 9). Photogrammetric 

analysis places the visual top between 12 and 13 km. Slightly to its left the 

more diffuse dome of cell W4 is visible. At the time of the photograph it was 

nearing the end of its mature stage and was approaching the early stages of decay. 

Perspective and stage of growth do not permit resolution of visual features 

associated with cell W3, but radar analyses (Part I) show that it was in its 

late stages of decay and is undoubtedly embedded in the cloud mass seen in the 

left background. Crisp cumulonimbus cloud outlines in the right background are 

associated with storm E, the eastern neighbor of storm W, briefly discussed in 

Part I and identified in Fig. 2. 

Other physical features shown in Fig. 1, most of which appeared with greater 

clarity on original prints, are labelled for reference. A portion of the 

anvil is visible at the upper left but, as high altitude relative winds (Part I, 

Fig. 7, lower inset) dictate, the main storm "blow off" is largely directed into 

the plane and therefore is not seen. A layer of shallow cumulus with many small 

turrets appears in the near field of view and to the right of storm W. This 

region, including the typically flat cloud bases found beneath primary updrafts, 

is commonly referred to as the "shelf cloud" and, as we shall see later, is 

located in the general relative upwind direction with respect to the subcloud 

inflow. Finally, the position of heaviest precipitation is identified in the 

layer between cloud base altitude and the ground. The physical significance 

of most of these visual features will be elaborated and clarified by analysis 

to follow.

3. Mesoscale surface analysis 

II-5 

PPI radar echo configurations in Fig. 2 were observed by the Grover 

radar at 1710 when storms W and E lay over the southwest corner of the NHRE 

mesonetwork. This time corresponds to the middle of the period of aircraft 

investigation and was tahe latest time ihen mesonetwork resolution was possible 

in both the inflowing and downdraft branches of the storms' circulation. 

The surface streamline pattern in Fig. 2 represents the flow relative 

to the moving system. This was derived by subtracting a constant storm 

velocity of 360°/10 m s - from wind measurements at each of the mesonetwork 

sites. Enhanced detail, beyond that provided by the given station spacing, 

was gained by subjecting analog records from the various observation points 

to a time-to-space conversion (see, e. g., Fujita, 1963) and displacing 

off-time estimates forward or backward from their actual positions along the 

storm's direction of motion at increments proportional to the storm's speed. 

Such a manipulation, of course, involves the assumption that flow patterns 

are steady with time; a premise that in this case is justified over only 

short time periods. Observations taken within 15 min of the actual map time 

were used to produce a field of overlapped data from adjacent stations which 

provided a reasonable basis for constructing the instantaneous flow pattern 

as shown. For clarity only the on-time data are presented. Exceptions are 

the two data points (smaller station circles) that are plotted to tle south of 

their actual network locations. These are examples of displaced data, having 

been observed at stations immediately to tile north at an earlier time. 

Flow with respect to storm W is best resolved since its path carried it 

directly over the four western sites while echo patterns were evolving in a 

periodic and systematic manner. In addition to the surface measurements, much

II-6 

of the flow structure beneath storm W is supported by radial Doppler velocity 

components reported in Part III. Features drawn for the weaker storm E have 

less credibility since, as mentioned in Part I, its associated echo patterns 

were changing more rapidly and sporadically, making the time-to-space 

conversion less valid. 

The position of the confluence line (heavy barbed line) marking the boundary 

between air advancing toward the storm from the southeast and the spreading mass 

of downdraft air originating beneath the echoes' high reflectivity core coincides 

with the location of the "temperature break" as defined by Byers and Braham (1949) 

in The Thunderstorm. The colocation of these two features agrees with the results 

of Foote and Fankhauser (1973). The sequential order of events observed at the 

inflow-outflow interface ahead of storm W was similar to the results of Charba 

(1974): a pressure jump, a windshift closely followed by the gust front, the 

temperature break and the peak gust. 

The speed of the outflow boundary at the time shown was assessed from isochrone 

analyses of mesonet temperature records to be 12'm s while the leading edge of 

storm W was progressing southward at an average rate of 10 m s . Isotachs and 

isogons were analyzed for the wind field in Fig. 2 to obtain wind components at 

2 km intervals. These were used to compute horizontal divergence, V.V. The 

-3 -1 

locations of convergence maxima exceeding a magnitude of 10 s are indicated 

as stippled regions and are found in zones of maximum streamline confluence on 

the left and right forward flanks with respect to the advancing storm. 

Thermodynamic features of the surface air lying south of the storm are 

similar to those shown in Part I, Fig. 2. Potential temperature (0) increases 

from east to west while mixing ratio (r) shows a slight gradient of opposite 

sign over the same space interval. Surface conditions in the immediate. 

relative upwind direction from storm W are close to those of the lifted parcel 

in Part I, Fig. 3.

11-7 

The lowest equivalent potential temperature (6e) is found beneath and 

e 

slightly to the left of the echo core of storm W. The value of 326 K is 

slightly colder than the minimum near 500 mb shown for the GRO 1735 sounding 

in Part I, Fig. 3 (inset). Relative winds aloft (Part I, Fig. 7, lower inset) 

and the conservative properties of e in moist air processes suggest that one 

source for air feeding the downdraft of storm W was from the layer between 

550 and 450 mb (7-9 km) which was approaching the storm with a relative speed 

of

II-8 

to compensate for storm displacement during the period of observation (1704- 

1721). Two legs flown in opposite directions are shown in Fig. 3. The track 

is derived by subtracting the average storm velocity (360°/10 m s ) from the 

aircraft's true ground velocity and recomputing the track relative to the storm's 

position as shown at 1710. Thus, only the 1710 position represents the true 

ground-referenced location and all others have been adjusted to positions 

relative to the moving system. Similarly, wind vectors plotted at 15 s intervals 

along the track represent the horizontal flow relative to the storm and are 

computed by subtracting storm velocity from the measured winds. 

The character of the relative flow at this flight altitude (800 m, AGL) 

is in close agreement with underlying surface conditions shown in Fig. 2. The 

early portion of the eastbound traverse (1704-1707) penetrates the outflow 

from the downdraft of storm W. An abrupt windshift from northeasterly to 

southeasterly occurred at about 1707 and a general southeasterly relative flow 

existed from that point eastward along the track. Conditions observed on the 

westward pass (1714-1721) essentially reflect those of the eastward traverse 

except that outflow from both storms E and W was encountered, as evidenced by 

thermodynamic parameters in Fig. 4 and associated wind perturbations in Fig. 3. 

The different relative positions of the storm W outflow penetrations on 

the two opposite flight legs were helpful in determining the shape and position 

of the outflow boundary in the southwestern sectorof the storm. Near the 

location of new cell development designated as W5 in Fig. 2, vertical juxta- 

position of the surface inflow-outflow interface and the sharp wind shift noted 

in the aircraft winds measured 800 m higher, leads to a nearly vertical outflow 

boundary. Surface wind data indicate a strong normal outflow component and a 

weak normal inflow component at this location. Farther to the east and more 

directly upwind from storm W along line AB, inflowing air at both the surface

II-9 

10 , o 

\ 10 ms-' 

30 . 

\ _ W(4) t 

40 1,l , B \ 6 _ 0.. 

20 30 40 50 60 

Fig. 3 Taveue,6 at an attitude. o. 2.2 km (800 m, AGL) beneath the. he4 

ctoud hIown £n Fig. 1 and ac&Laos the. intow to stoWns- W and E, 

Ztown by the. NCAR B&Lato betwe.e.n 1704 and 1721. The poztion o4 the. outhwcadmovcng 

adatdU echoeu L shown at 1710 as kn Fig. 2. Reative. 

compe.nsated 

awca{t pozation, 

4o&L ,to-m motion, U fabee.d at 1-mLn inteAvva& and t.etative. wnd 

vecto'u (cate. at uppet k'ght) ae. ptotted at 15-4 intevwaz. The heavy baWbed 

tne. hows tocation o4 out4tow boundaqy at tiz alitude. No'thwatd-poLnting 

aOVOw 4hcw hthe. po^stion oa vvu&Lcae ve~ocity maxcma (FLg. 4) 4u.ppouting Lndu- 

v.duat cQ.e& a& £dewnt4e.d. Lne. AB 4how po4tton and 0/u enut on oj vvutzecua 

4e.c-ton Ln Fig. 8. BtoideL -L ZabeJke.d £n kttometeAu e.cat and 4outhi oX G/toveA, 

Cotoaado. Color~ado ... 

-

II-10 

and at flight altitude had a strong component normal to the outflow boundary 

while the outflow component in this sector was comparatively weaker. Evidence 

points to a greater slope from the vertical for the outflow boundary at this 

location. 

The best thermodynamic indicators of outflow penetration in Fig. 4 are 

the zones of distinctly cooler 0, signifying the rain-chilled character of air 

originating from the spreading downdrafts. We also note in Fig. 4 that the 

general trends in e and r reflect those at the surface (Fig. 2) in that e 

increases from east to west while r displays a decreasing trend over the same 

interval. Three zones of distinctly higher e are identified as tongues of 

lifted surface air based on the fact that their peak values compare favorably 

with those at the surface. Ahead of cell E, e and r at flight altitude are 

e 

somewhat greater than surface values shown in the upwind direction in Fig. 2, 

but data presented in Fig. 2, Part I, indicate that air with properties observed 

at flight level did exist at the surface over the southeastern portion of the 

mesonetwork at an earlier time. 

Vertical velocity resolved by the inertial navigational and gust vane 

system on board the Buffalo aircraft is shown in the upper panel of Fig. 4. 

We note two major peaks, > 5 m s , and a third zone of weaker ascending air. 

The locations of the three vertical velocity maxima are shown in the horizontal 

plane of Fig. 3 as heavy northward pointing arrows. An average duration of 

updraft encounter of %90 s and an average ground speed of 90 m s indicates 

that the lateral width of these ascending branches was on the order of 8 km 

at this altitude. 

Each updraft maximum is well correlated with zones of high e in Fig. 4 

and with confluent horizontal wind perturbations in Fig. 3. The easternmost

II-11 

02.1 1720 19 18 17 16 1715 14 

w(5) I VERTICAL VELOCITY 

(m i) 

£_ I i . . . 

o-5 F -- i i 

-I- ______________ - I ___ ,____________ -318 

IPOTENTIAL TEMPERATURE 

'- '16 --- 316 

348 1. l^ 

38--- 'lI , I ' - 

EQUIVALENT POTENTIAL , 

3,46 TEMPERATURE, (K) 

34I - 

34 2 

LIFTED 

SURFACE 

340 - AIR 

MIX IG RATIO K/kg) , 

__ 1 'I 

8 

21 1720 19 18 17 16 1715 14 

Fig. 4 VeAtica velocity, potential tempeatuwe, equivatent potential tempeatwue. 

and mixing Artio doL westbound dtight segment (1714-1721) 

-n Fig. 3. ,Dashed Unei on theAmodynamic pauameteus denote ambient conditions 

and shaded depacutre identif y outflow and inflow encounteus as indicated. 

Updlaft maxima a^sociated with discAte. echo entUes corLe&pond to aIoLW6 in 

Fig. 3. Veticat dashed line deZignates intercept od the. crLz nection in 

Fig. 8. 

-

II-12 

peak (X1715:15) apparently identifies the primary updraft supporting the 

southern cell in storm E. A weaker maximum centered on 1717:15 is located 

in the region of strongest relative horizontal inflow and later analysis will 

reveal this as the source of air feeding the mature cell of storm W, W4. 

Comparison of its position with that of the outflow boundary in Fig. 2 shows 

that this weaker updraft maximum was encountered farther from the inflow- 

outflow interface than were the stronger updrafts to the east and west. 

Strongest ascending motion was observed at "1719 near the point of outflow 

penetration and near the wind shift zone discussed earlier. This updraft 

was supporting the new cell, W5, developing to the south-southwest of the 

existing storm-W echoes. It was not yet detectable at the time and altitude 

of the 1710 PPI scan in Figs. 2 and 3, but its location relative to the main 

storm when first detected at r1716 is shown in Fig. 2. 

_l 

Downward motion of 4 m s was measured during the deepest penetration of 

the outflow (1706), but the sharpest and strongest downward gust occurred at 

1719:25 near the outflow boundary during the latter and more shallow excursion 

through the storm W outflow. Both of these penetrations were close to the 

region of strongest updraft associated with the newly developing cell, W5. 

5. Aircraft observations at 2.8 km 

An NCAR Queen Air (N304D), relying on Doppler navigation for horizontal 

wind assessment began observation of the storm system at 1655. Resolution of 

the wind measurement capability of this aircraft and complete calibration 

procedures for all aircraft measurements in NHRE are discussed by Foote and 

Fankhauser (1973) and Biter and Wade (i975). Flying at an altitude of 2.8 km 

(731 mb; "1.4 km AGL), the aircraft completed a full circuit around storms W 

and E between 1655 and 1721. Fig. 5 shows the track segments most relevant to

II-13 

"rl~ l^l'102 ~| . 

10 

A -- ^--~O, o- 

40 *-^-ITOO-.. | ,10 ms-' 

| 

- 

00 1655 

20 30 40 50 60 

Fig. 5 Acta'ft 

fAack and wind vector teative. to wtadac echoeu as in Fig. 3 

except at an atitude o 2.f km (1.4 km, AGL), teown by the NCAR Queen 

Ai& (N304V), dwuxig time indicated atong the tack. Poition out4towc boundaity 

at ;ths attitude L indica.ted by the da hed batbed tine. Location o 0 

tAu. auL 

aoWaLt 

4peed maxima, courleponding to veAticat vetocity maxima 

aLe 4kown 

in Figs. 3 

ah 


nowaitd-phontd 

4, 

ng afutow l along the we utwd ttoaveuAe tU"LouIgh the 

cn{low (1715-1721:30). lHo'?zontal hatchng Ln the SE 4ec;to. oa 0 tomn 

Legion 

W indicate 

of 4ada!L echo "ove'hang" at aittudeo above ctoud base. 

4how 

Line 

ouLentaion 

AB again 

od cor a bection in FLg. 8. Bo'Lder &tbeJed 

and ouwth 

n kitoomete 

o5 GtOveA, 

ea&st 

Coito'ado. 

.

II-14 

storm W. As in Fig. 3, the actual ground track has been adjusted to positions 

relative to the traveling storm and wind vectors again represent relative 

horizontal airflow. Thermodynamic parameters calculated from measurements 

along the two flight legs (1655-1703; 1715-1721) are plotted in Fig. 6. We 

note that the westward pass (1715-1721) through the inflow coincides closely 

in time with the westward traverse of N326D and that the two aircraft were 

essentially vertically stacked when sounding the inflow sector. This has been 

a preferred mode for aircraft deployment in the NHRE field investigations 

because it facilitates direct comparison of data measured at the various 

flight levels. 

Relative horizontal airflow on the westward traverse is comparable in 

direction and magnitude to that found at the two lower levels (Figs. 2 and 3) 

and perturbations that appear are in close horizontal alignment with similar 

features below. Specific attention is drawn to confluent relative flow in 

the vicinity of storm E (1717) and to confluence and cyclonic shear at 1720 

near the location of cell W5 shown in Fig. 2. The track skirts close to 

storm E and a momentary drop in 0 at that time (1717 in Fig. 6) indicates 

that the aircraft briefly encountered outflow there. 

Trends in Fig. 6 (1715-1721) also reflect those found at lower altitudes 

with the coolest and most moist air located to the southeast of the overall 

storm complex and the warmest and driest to the southwest. Thermodynamic 

parameters measured on the west flank and to the rear of the storm are also 

shown in Fig. 6 (1655-1703). To the rear of storm W, e and r best identify 

e 

the excursion through outflowing air deposited by the downdraft. 'Gradients 

in both parameters near 1658 and 1701 bracket distinctly drier air. Reference

, 

11-15 

s320w s ssw 

SE30 ST 

3186 

3 0 ' ' 1r 19 - 3'4 6 

aI - I 6 - 

han314 -J JJ - -- Fahdtn I _ 

I II dnt I ----- m L- cdn | 

nddp-314 

Ic i/^,^"" 10*^ - : A~ n/ ''10 

>- 340. 5. VaH L4 denot 

1,_.j 

6 w 'I Il I i IJ I 6 

6 1 I' 

, 

030C110D 9 5 7 5 65 112 91 71 

60 

75MT 

Fi.6Temdnmcprmtr ln rgtlg 655103 7512 

shw deaie 

inFg8.Dse eoeabetcniin n 

tlWl (4 hadxed) Ldentdy4 owtw and 'n'o' a'', as n'cat'd. ALc'al 

po4atio~in xhown -'Vdc&e to ^toim W Zu 4hcwn at top. Upwaid-po&nting awVoW on 

1715-1721 

c~~~soni 

£n 3ine Fg.. 5. 

potenvtia 

i.5 dahed 

tempvatw' 

asedptwithe on ight 

pane2 coaJ&pond 

cross hand 

am pan3e 0de'nkt ntet- 

to 

ientcnions 

twUe aCA 4peed max-rna 

and Fig.8. 

cetpt &th cao 4ecQion n Fig. 8.

II-16 

to relative wind vectors in Fig. 5 shows that outflow boundaries are associated 

with zones of horizontal shear. It is worth noting that the relative aircraft 

position near 1656 lies downwind from the 1720 to 1721 portion of the track in 

Fig. 5, and that the thermodynamic parameters observed in the two locations are 

essentially the same. This is evidence that, although storm W was evolving in 

a periodic manner, the near-cloud regime retained a persistent configuration. 

Based on the measurements at this altitude the interface between ambient 

and outflow air has been drawn on Fig. 5. While a tendency for the outflow dome 

of storm W to shrink toward the precipitation echo in the vertical is evident in 

both the western and southeastern sectors, the outflow boundary appears to be 

more steeply inclined near the relative upwind region of new cell generation 

(W5) indicated in Fig. 2. 

Although direct measurements of vertical velocity were not obtained from 

the aircraft flying at this level it is possible to identify regions of 

ascending and descending air motion in a qualitative manner. When an 

aircraft is flown at constant static pressure, as the NHRE research aircraft 

normally are, variations in true airspeed (TAS) can be used as an indicator 

of vertical air motion. Ascending ambient air is correlated with increasing 

and/or high values of TAS and strong downward motions are reflected by 

rapidly decreasing and/or low TAS. 

During the westward pass (Fig. 5, 1715-1721) three distinct TAS maxima 

were recorded; one at 1716:45, another at 1718:30; and a third shortly before 

1720. Each is identified by the upward pointing arrows on Figs. 5 and 6. 

These maxima were well correlated in space with the vertical velocity peaks 

measured by N326D at the lower level, and the vertical and horizontal 

continuity from the surface through the two flight levels strengthens the 

evidence supporting three distinct inflow branches as proposed in the previous

II-17 

section. Indications are that the two eastern maxima tilt slightly.from east 

to west in the vertical while the western and strongest updraft in Figs. 3 and 5 

appears to have a more vertical alignment. 

·6. Cloud 'base measurements 

The University of Wyoming Queen Air (NlOUW) monitored thermodynamic prop- 

erties and updraft structure at cloud base (3.8 km, 2.4 km AGL) in the inflow 

to storm W, continuously from 1643 through 1720 and subsequently obtained an 

inflow sounding from cloud base to near the surface between 1720 and 1730. 

The total period of observation includes the development cycles of cells W3 

through W5 as discussed in Part I. A description of the vertical motion sensing 

system aboard this aircraft and its resolution capability is presented by 

Marwitz (1972a). In the manner described in earlier sections the' actual air- 

craft ground position was adjusted to compensate for storm movement, and all 

track segments were positioned relative to the storm's location at 1710. The 

horizontal field of ascending motion analyzed following this transformation is 

shown by isotachs of vertical velocity in Fig. 7. The updraft configuration at 

cloud base is based on vertical velocity measured during six traverses through 

the updraft zone flown on varying headings and at different positions relative 

to the leading edge of storm W. 

Comparison with the relative flow in Figs. 2, .3, and 5 indicates that the 

updraft core at cloud base is located directly downwind with respect to the 

organized southeasterly relative inflow at lower levels, some 1 to 3 km ahead 

of the large PPI radar echo gradient at cloud base, and beneath the radar echo 

overhang shown in Fig. 5.. Average east-west updraft extent is about 8 km, 

approximately the safe width as the echo at 1710, and similar to the esti- 

mated updraft width at lower altitudes (Section 4). A profile of the

10- A 

230 

2 0 

40 

11-18 

30 40 50 60 

Fig. 7 Veuticace veocty Zo-tach& (Uoeid contouas; m .- 1 ) at ctoud baue 

(3.S kIm; 2.6 km, AGL) °eetQve to PPI echo at 1710 a Zn Fig. 2, 

3, and 5. Anatuysi bazed on updcait meawuArements by Univeuity od Wyoming, 

Queen AiL dty-ng at ceoud base in the. updraut region supporting stoaun W 

between 1643 and 1720. Dotted tines show etattt iZmtz of fight paths 

and denote region whee. data weAe avaiUable. Crosz-hatched region outlined 

by daohed contour denote, zone where. eI > 345 K. Poaition ofA 4sLace outflow 

boundary from Fig. 2 (ba/Lbed ine. Ls hown doL LefdeLrnce. Line AB 

show6 oiCentation of CrOs aection in Fig. 8. BoundariUe corrLespond to 

Fig4. 3 and 5.

II-19 

vertical velocity taken along a segment passing through the maximum and 

normal to the lower level horizontal flow would reveal a bell-shaped profile 

of updraft intensity similar to that found by Kyle, et al (1976) from 

penetrations of updraft zones at higher altitudes within a number of 

Colorado thunderstorms. 

The position and shape of the zero isotach in Fig. 7 is known with 

some confidence so that there appears to be a clear distinction between 

the main updraft supporting the mature cell in storm W and the other two 

updraft zones detected by the lower flying aircraft. The track of N1OUW 

did not take it into positions favorable for observing the other ascending 

branches (see dotted lines in Fig. 11) and in fact the updraft region as 

depicted was, in general, surrounded by weakly subsident motions in its 

immediate periphery. These are also evident in the vertical velocity data 

appearing in Fig. 4. 

As discussed in Part I, Section 2a, the pseudo-adiabat defined by 0 e = 

345.5 K was representative of the updraft region at cloud base. Corresponding 

static pressure, 6, and r were 650 mb, 316.5 K and 9.5 g kg 1 , respectively. 

The cross-hatched region in Fig. 7 shows the region where ,e' computed from 

data measured during the updraft traverses, equalled or exceeded 345 K. 

Orientation of this moist band is along the direction of the relative inflow 

at lower levels and comparison with thermodynamic data in Figs. 2, 4 and 6 

indicates that the updraft air ascends from near the surface to cloud base 

in an essentially unmixed manner. This was confirmed by the descent sounding 

made by N10LU at the relative position shown in Fig. 2, and supports earlier 

evidence (Marwitz, 1972a) regarding the origins of air feeding High Plains 

thunderstorms.

7. The composite inflow structure 

II-20 

Data in preceding sections provide a basis for constructing a 

vertical cross-section through storm W representative of mean inflow 

conditions (Fig. 8). For orientation we chose an axis that is parallel to 

the direction of relative inflow. Its location in the horizontal plane 

is designated by the dashed line segment (AB) appearing on Figs. 2, 3, 5 and 7. 

For reference, a schematic radar cross-section similar to those 

analyzed in greater detail in Part I has been extracted from the three- 

dimensional reflectivity scans centered around 1710. Schematic cloud 

features are also superimposed from photogrametric analysis of Fig. 1. 

The reflectivity maximum, > 50 dBZ, found aloft between 8 and 10 km represents 

cell W(n), (n, denoting the numerical order of cell formation) near the mature 

stage of its life cycle. Streamline analysis describes the structure of 

the inflow feeding this cell. To the rear of the storm the radar cross- 

section intersects the edges of the decaying cell, W(n - 1), which is 

centered at a location somewhat into the plane. Weak echo from the newly 

forming cell, W(n + 1) also appears ahead (to the right) of W(n) and its 

central core is located slightly out of the plane toward the reader. Other 

significant radar features include the weak echo region beneath and ahead of 

cell W(n), and the high reflectivity core (> 55 dBZ) in the heart of the storm 

extending from %6 km to near the surface. These internal features identify 

zones of organized ascending and descending motions (marked by heavy arrows) 

and were observed in considerable detail by a vertically scanning Doppler 

radar (Part III) and a penetrating aircraft (Part IV) at times slightly later 

in the storm's history. 

Horizontal wind vectors at the surface represent the component of 

relative flow parallel to the vertical section and normal to the outflow

(rnmb) ' 

150- 

jO00t. 

600 T(C) 

~~~~~~~200 

-10 " 25 

600- rToC) \ 50 [\ \ 

- /Wln/ I 

-50 

11-21 

MING\\^ 

OF PL NE 

OUT 

^ ^ S ^^^^^-' M OT ION 

s 

T\ 

1 -Ikm-~ 

'H - 

B00 

800 

\5 

H 

33.0 

--- 

_ 

-,,.,. 3o~, 

- - - - 

------ 

44.832.7 342.3 

- - 

341.9 

4. - 4 

SFCC t----~----· 

900 342 338 334 330 

3 3 3 4 3 3 

340 6 24 328 

km-20 -18 -16 -14 -12 

326 330 334 338 342'5 

3 2 

345 3 3 

8 2 336 340 344 

-10 -8 -6 -4 -2 0 2 

345. 

346 

3 4 

4 

346 

34 

6 

. 

8 

46 

10 

344~Q 

.5 km 

m ' 

12 km 14 

Fig. 8 Composite vect ,totm t CILo4o , ection -thaough 

ne A3 in 

Wa 

Fegs. 

o'ented 

2, 

atong 

3, 5, and 7. V cloud isuat 

n·- 4cant 

boundaiLe 

eatutwe. 


d1om 

^ 

Fig. 

eg- 

1 ae zupe imposed 

(iteguatr 

on rada 

otid 

tedfcteiv 

contour 

ty 

) 

contorws 

detLved dfom PPI 

Equivaeant 

catns centered 

potentia/ 

atound 

tempenurtwue 

1710. 

plane 


at 

tettive 

. the w..ace 

wind 

andat 

vebCto 

4Zight 

parattel 

atetitudes 

to the 

ounded 

alt 

by 

apptop/iaSe 

aic 4t ate 

tettive 

plotted 

positions. Sotid 

Pet 

hotizonta2 

and tight 

Vectotu 

ae -taken 

on 

'tom 

the 

reprte,entative 

extteme. 

Part 

up- and 

I, Fig. 

downwind 

3). 

s 

Da.hed 

oundings 

vectots 

(see 

on sthe reounding 

'tepzeent to the. 

t.eative 

ea& o 

components 

mthe stoam 

into (upw 

ne.nctned) 

incad) 

the 


ve.tica. 

out o 

peane. 

downwad 

Heavy aLow4 

ing 

indicate 

and descend'ng 

primaty 

mo°tions. 

tegions of 

Indexed 

ascend- 

o. -the W -4eteies. 

tada 

in 

rtedfectivity 

vayt.ng 4tage4 

deatures 

od deve.opment. 

identify 

The 

cetts 

bas 

tine pattern 

dot fi the stteam- 

tepresentivng &nftow to matuwe cel,. W(n), ds 

Inslt 

presented 

adapted 

in 

dtom 

the text. 

fHumphkrey4 (1914), 

(inelow, 

4howing: 

updra4t); 

[A] 

[D] 

acscending 

descending 

aai 

air (downdraft, 

(4held ctoud); [S] rtoUlscud (aTcus 

outdlow); 

or toQQ (gust ceoud); 

[C] 

[D'] 

4stoAmun 

wind gust 

ca 

hiont); [U1] haiR; [T] ;thitzn'iha (cnotheahc 

(cuwmonJwimbuw .twut 

p v Ot 

t.in7tain 

toweu); [R] 

4hat). TeNlms 

t 

in LtaLc 

t enify 

s 

f 

ate 

eautres 

thoze 

which 

wed by t..wnph.ey 

ate cuwttenty described by tetm4 in paienth.e4sC. 

10:

II-22 

boundary. The surface location of the inflow-outflow interface is based on 

analysis presented in Fig. 2, and winds are derived from the average 

conditions observed at the four mesonetwork sites directly affected by the 

overhead passage of storm W. Two null points occur in the relative flow 

beneath the storm; one at the inflow-outflow interface and another directly 

beneath the low-level radar reflectivity maximum. Speed in the relative 

inflow toward the storm at the surface is fairly uniform and averages near 

8 m s. Approximately 2 km ahead of the outflow boundary the inflow, 

moving from right to left, decreases abruptly with the normal component 

falling to zero at the interface. It is here that the horizontal mass 

convergence reaches a maximum, exceeding 10 - 3 s , as indicated in Fig. 2. 

Moving left to right from the location of maximum downdraft toward the 

outflow boundary, velocity increases from zero to a maximum of 4.5 m s- 

near the forward edge of the precipitation wall, identified by the large 

horizontal radar reflectivity gradient. From this point to the outflow 

boundary minor speed fluctuations are evident, indicating rather turbulent 

conditions. The outflow component decreases rapidly to zero as it nears the 

outflow boundary, again contributing to the maximum convergence found there. 

To the rear (left) of the downdraft maximum, relative outflow increases 

steadily from zero to values exceeding those of the inflow. This suggests 

that momentum has been added to the boundary layer by downward transport 

from aloft. Horizontal divergence beneath the maximum downdraft, calculated 

-1 

from mesonetwork analysis (Fig. 2), exceeded 4 x 10 - 3 s . 

Thermodynamic transitions along the surface are represented by plotted 

values of Oe, again derived from average conditions observed at the four 

directly affected mesonet sites. As indicated in Fig. 2, the 8e field is 

nearly uniform over the area lying 15 to 20 km ahead of the storm. This is

II-23 

reflected by the flat gradient in the upwind inflow sector in Fig. 8, which 

shows surface values ranging between 345 and 346 K. Behind the outflow 

boundary values drop to a minimum of 326 K appearing slightly to the rear 

of the maximum downdraft. Gradual recovery toward environmental conditions' 

is shown to the rear of the storm. Largest e gradients are found at the 

e 

surface near the edges of the radar echo where a value of 336 K is seen to 

coincide with both the leading and trailing boundaries of the precipitation 

shaft. 

Point measurements at two flight levels (2.2 and 2.8 km) are available in 

the inflow from the traverses made by N326D and N304D. The parallel relative 

wind components at the points where the tracks in Figs. 3 and 5 intercept 

the vertical section are plotted at their respective positions in Fig. 8. 

-1 

Relative horizontal velocity at both levels is 12.5 m s . Vertical 

velocity is known at the lower altitude only (see Fig. 4) and.vector addition 

gives the indicated slope of the inflow at this location (dashed vector). A 

single estimate of downdraft slope behind the outflow boundary is based on 

the vertical and horizontal winds measured by N326D during outflow penetrations 

discussed in Section 3. 

Equivalent potential temperature at both flight levels is 345.5 K at 

the point where the respective tracks intercept the plane of Fig. 12 (see 

Figs. 4 and 6). This value is in close agreement with surface conditions 

ahead of the storm (Fig. 2), Slight differences between surface and aircraft 

levels are indicated, however, in mixing ratio which shows a small increase 

with height and in 0 which shows about a 1 K decrease. This potential 

temperature decrease with height is indicative of a superadiabatic lapse 

rate in the lowest few decameters above the surface.

11-24 

To the rear of the storm at the altitude sounded by N304D no 

appreciable component of relative motion was found in the plane of Fig. 8, 

indicating that the air there had essentially the same momentum as the 

main body of the storm. The lowest 0 measured there was 338 K (Fig. 6). 

e 

This is in agreement with the surface value directly beneath the aircraft, 

and close to that shown at this altitude in Part I, Fig. 3 (inset). This 

indicates that the subcloud layer immediately to the rear of the storm was 

quite well mixed, undoubtedly owing to the more turbulent flow 

characteristics frequently encountered by aircraft flying there (Foote 

and Fankhauser, 1973). We note also that 0 e measured by N326D in the 

forward outflow is close to the underlying surface value. 

Horizontal resolution at cloud base altitude is provided by the 

measurements of Queern Air N10UW. Vertical velocity vectors based on the 

isotach field in Fig. 7 have been extracted at selected intervals along 

line AB and are plotted in their corresponding positions in Fig. 8. 

Strongest vertical motion (> 6 m s ) is found 1 to 2 km ahead of the 

subcloud precipitation curtain and generally beneath the radar echo 

'overhang.' If the horizontal momentum observed at lower levels is 

conserved in ascent to cloud base, addition to the vertical velocity 

vector, again, gives an estimate of updraft slope as indicated by the 

dashed vectors. 

Thermodynamic variations at cloud base are shown by the plotted 

values of 9e based on parameters measured' during the flight of N1OUW from 

the region of maximum updraft to the position of the descent sounding 

designated in Fig. 2. We note that in the regions of strongest ascending 

motion 0 e exceeds 345 K. As the aircraft approaches the edges of the 

shelf cloud ahead of the echo overhang, 0 e is slightly lower and less

II-25 

uniform. Immediately beyond visible cloud boundaries e diminishes further 

e 

to values less than 342 K but recovers somewhat farther from the cloud. The 

increasing trend in e during the descent sounding appearing on the right of 

e 

Fig. 8 is.in qualitative agreement with that shown for the STK sounding in 

Part I, Fig. 3 (inset), but it is not so steep. We note that values > 345 K 

are confined to the lowest layer above ground .at this distance from the storm. 

The shape of the streamlines representing the mean instantaneous inflow 

and updraft conditions for storm W is based on the few estimates of updraft 

slope and on the conservative properties of e0 in moist air processes. The 

e 

shallow layer of air having high Qe observed in the relative upwind direction 

far from the storm suggests that the inflow streamlines originate from quite 

near the surface. The inflow layer is shown to expand to a depth of at least 

2 km near the edge of the shelf cloud while the relative horizontal inflow 

velocity increases from 8 m s near the surface to 12.5 m s in the middle 

of the subcloud layer. Streamline diffluence in the vertical plane and an 

increase with height in the horizontal inflow component requires that horizontal 

convergence also increase with height to satisfy mass continuity. A vertical 

mass flux convergence profile from measurements in the inflowing sector of 

storm W (Fankhauser, 1974) shows that such was the case. 

While data presented at the surface and at cloud base in Fig. 8 are repre- 

sentative of average conditions measured there over a period of 30 to 35 min, 

the individual data points in the inflowing sector represent essentially instant- 

aneous conditions observed by the aircraft at different times. The inflow data 

below cloud base was obtained, for example, some 10 min earlier than that from 

the descent sounding far from the cloud. In view of this and the periodic 

behavior in radar echo' development patterns demonstrated in Part I one would 

anticipate some fluctuation about the mean inflow structure as drawn.

II-26 

Below cloud base, the inflow ascended at an average angle of about 20° 

from the horizontal above a cold outflow which extended 5 to 6 km ahead of the 

surface precipitation. As indicated by the N326D inflow vector at 2.2 km, 

inflow streamlines begin a distinct upward turn in the vicinity of the surface 

convergence maximum ahead of the surface outflow boundary. They then follow 

the slope of the outflow boundary which has been drawn to extend from a point 

at cloud base located a few hundred meters ahead of the precipitation shaft 

to the known surface location. The slope of streamlines near cloud base is 

drawn to agree with inflow vectors based on the NlOUW updraft measurements and 

conserved horizontal momentum brought from lower levels. Shape of streamlines 

above cloud base is of course largely speculative but is governed primarily by 

radar configurations surrounding the weak echo region. A steep inclination near 

the rear edge of the weak echo region is confirmed, however, by the observations 

reported in Part III. 

8. Summary and conclusions 

Some of the features of subcloud airflow resolved in this analysis are 

not at all unique; many having been previously constructed in amazing 

detail as early as the turn of this century by the atmospheric physicist, 

W. J. Humphreys (1914), whose analysis, based almost solely on careful 

visual observations, is presented in the inset on Fig. 8. We do, however, 

present a broad and comprehensive data base which supports many of his early 

speculations and elaborates the definition of the subcloud structure of 

the multicell thunderstorm. 

Prominent features resolved in the present case are worth reiterating 

in summary. Surface and aircraft data show that inflow air approaches the 

front of the storm, originating from a very shallow layer (< 500-m) near the 

ground and at a considerable distance (> 20 km) upstream in the relative wind

II-27 

direction. Inflow rises unmixed to at least cloud base, feeding the main 

updraft region which is inclined upward in a direction opposite to the storm 

movement. 

Discrete well-defined inflow-updraft branches were detected which 

simultaneously supported different cells in varying stages of development, 

comprising the storm's overall evolving multicellular structure. These 

inflow zones had lateral dimensions of 6-8 km and were separated by distinct 

regions of weak subsidence. Despite the storm's evolving character, updrafts 

to individual cells were quite long-lived. Analysis in Part I shows that for 

individual cells the time interval between first detection and maximum radar 

reflectivity was approximately 20 min. If the decay period is included in the 

life cycle, radar cell life times exceed 30 min. 

Relative winds aloft and conservative properties of 8 in moist air 

e 

processes indicate that some of the air feeding downdraft circulations 

originated in the mid-troposphere and approached the storm from the right 

flank (see winds between 5 and 7 km on the left in Fig. 8). At least a 

portion of this air descended unmixed, reaching the ground in the strongest 

downdrafts near the low level reflectivity maximum. 

Maximum surface convergence at the inflow-outflow interface was 

-3 -1 

1 to 2 x 10 s and divergence calculated beneath the strongest down- 

draft was 4 x 10 3 s- 1 . 

In addition to high 8 observed in inflow to updrafts supporting indie 

vidual cells, aircraft measurements show that the overall mesoscale regime, 

both upwind and downwind from the storm, was characterized by e , which at 

any subcloud level above the surface was higher than that in the environment 

a few kilometers away the ifrom storm (compare, e. g., descent sounding in 

Fig. 8 and 8e profile in inset to Part' I, Fig. 3).. : Indications are, therefore, 

that the entire near-cloud air mass had undergone general uplifting.

II-28 

A mechanism favoring general ascent exists in the surface confluence 

zone as shown in Part I, Fig. 2, where mesoscale convergence values would 

-4 -1 

typically be on the order of 10 s (Fujita, 1963). Storm developments 

during the day were in close proximity to this surface flow feature as it 

-4 -l 

progressed southward through the network. A convergence value of 10 s 

averaged through the subcloud layer leads, by continuity, to a mean vertical 

motion of 4.15 cm s . Ascent of %2 km required to satisfy observed 

conditions, could therefore take place in a period of 3 to 4 hours as a 

result of mesoscale forcing. 

A potentially unstable boundary layer environment, requiring only a 

small degree of lift to become convectively unstable, is consistent with 

the large number of radar echoes which formed on this day. Only a small 

percent, forming in regions of strongest surface convergence such as the 

mesoscale confluence line (Part I, Fig. 2) and the outflow boundaries of 

existing storms (Fig. 2),were favored to become major thunderstorms. Low 

level thermodynamic structure near and beneath supercell phenomena appears 

to differ from that found here in that near-cloud vertical motions in the 

supercell case seem to be suppressed everywhere except in the main 

inflow-updraft zone (Browning and Foote, 1976). 

Some aspects of the. observed radar structure, such as the weak echo 

region and associated forward overhang, relate to commonly observed features 

of supercell storms. (Browning, 1964; Marwitz, 1972a; and Chisholm, 1973), 

but we know from the evolutionary character of the storm that these are 

quite transitory in the present case. While significant degrees of 

cyclonic shear and associated vorticity are frequently observed in the 

inflow to large and persistent thunderstorms (see, e. g., McCarthy, et al, 

1974), the inflow to the storm studied here displayed only nominal shear in the 

horizontal plane and each subcloud branch appeared to be essentially two-

II-29 

dimensional in the vertical plane, as well. These may well be important 

aspects distinguishing circulations in multicellular storms from those of 

the supercell thunderstorm. In summary, the observed storm characteristics 

were closer in nature to the multicell model proposed in The Thunderstorm, 

where spreading downdrafts from mature and decaying cells appear to provide 

the triggering mechanism for discrete new cell formation. 

Acknowledgements 

Special recognition is due the operational and research support staff 

of the NCAR Research Aviation Facility whose unstinted cooperation led to 

the collection of the data on which much of the analysis rests. Appreciation 

is also expressed to the University of Wyoming Atmospheric Sciences group for 

obtaining and making available the definitive cloud base aircraft data 

reported in Section 6. Credit is due G. Brant Foote who co-directs the 

NHRE aircraft research program and who provided the photograph in Fig. 1. 

S. Fuller performed the data reduction. C. Castillo and R. Coleman 

drafted the figures and J. Krzyzosiak typed the manuscript.

II-30 

REFERENCES 

Biter, C. J. and C. G. Wade, 1975: Field calibration and intercomparison of 

aircraft meteorological measurements. Preprints, NHRE Symposium/Workshop 

on Hail and Its Suppression, Estes Park, Colo. (unpublished manuscript). 

Browning, K. A., 1964: Airflow and precipitation trajectories within severe 

local storms which travel to the right of the winds. J. Atmos. Sci., 21, 

634-639. 

_ , and G. B. Foote, 1976: Airflow and hailgrowth in supercell storms and 

some implications for hail suppression. Submitted to Quart. J. Roy. Meteor. 

Soc. 

Byers, H. R.. and R. R. Braham, 1949: The Thunderstorm. U. S. Government 

Printing Office, Washington, D. C., 60-66. 

Charba, J., 1974: Application of gravity current model to analysis of 

squall-line gust front. Mon. Wea. Rev., 102, 140-156. 

Chisholm, A. J., 1973: Alberta hailstorms, Part I: Radar case studies and 

airflow models. Meteor. Monogr., 14 (36), 1-36. 

Fankhauser, J. C., 1974: Subcloud air mass and moisture flux attending a 

northeast Colorado thunderstorm complex. Preprints, Conference on Cloud 

Physics, Oct. 1974, Amer. Meteor. Soc. Boston, 271-276. 

Foote, G. B. and J. C. Fankhauser, 1973: Airflow and moisture budget beneath 

a Northeast Colorado hailstorm. J. Appl. Meteor., 12, 1330-1353. 

Fujita, T., 1963: Analytical mesometeorology: A review. Meteor. Monogr., 

5 (27), 77-125. 

Humphreys, W. J., 1914: The thunderstorm and its phenomena. Mon. Wea. Rev., 

42, 348-380. 

Kyle, T. G., W. R. Sand and D. J. Musil, 1976: Fitting measurements of thunder- 

storm updraft profiles to model profiles. Submitted to Mon. Wea. Rev,

11-31 

McCarthy, J., G. M. Heymsfield and S. P. Nelson, 1974: Experiment to deduce 

tornado cyclone inflow characteristics using chaff and NSSL dual Doppler 

radars. Bull. Amer. Meteor. Soc., 55, 1130-1131. 

Newton, C. W., 1963: Dynamics of severe convective storms. Meteor. Monogr., 

5 (27), 33-58. 

Telford, J. W. and P. B. Wagner, 1974: The measurement of horizontal air 

motion near clouds from aircraft. J. Atmos. Sci., 31, 2066-2080.

Structure of an Evolving Hailstorm. Part III: 

Internal Structure from Doppler Radar 

by 


NOAA/ERL/Wave Propagation Laboratory 1 





Science Foundation.

ABSTRACT 

Two X-band Doppler radars observed a hailstorm that passed directly over 

one of the radars during the 1973 National Hail Research Experiment (NHRE). 

While one of the radars scanned the radar echo at low elevation angles the 

other radar, which operated simultaneously in a zenith-pointing mode, measured 

part of the main updraft. Observations by other NHRE participants assisted in 

interpreting the radial velocity fields so that inflow and outflow could be 

identified from the scanning radar measurements. The peak updrafts occurred 

just ahead of the highest reflectivity while the strongest downdrafts were 

found only 2 km behind the updraft. Strong turbulence was generated in the 

transition region between updraft and downdraft as evidenced by large velocity 

variances. A substantial part of the downdraft appeared to have been fed by 

air that had ascended in the updraft. Low-level velocity fields were in 

general agreement with surface measurements and showed the outflow toward the 

front of the storm in the gust front as well as outflow opposite the echo 

motion behind the storm. There was strong outflow opposite the direction of 

echo motion at the top of the storm which agreed with photographs of the anvil 

overhang.


III-1 

The observations discussed in this paper were made with two X-band 

Doppler radars on a storm that passed directly over one of the radars near 

Raymer during the 1973 field program of the National Hail Research Experiment 

(NHRE) in northeast Colorado. The two radars normally were operated as a 

dual-Doppler system to acquire data for analysis of three-dimensional wind 

fields. However, when a storm passed directly over one radar, it was operated 

in a zenith-pointing mode while the second radar scanned an azimuth sector 

encompassing the zenith-pointing radar with a raster scan stepped in elevation. 

Browning, et al (1968) conducted a similar experiment to observe the horizontal 

and vertical air motion in a shower, except their second radar scanned in 

elevation at a single fixed azimuth. The operational mode we selected enabled 

us to determine the two-dimensional velocity in a vertical plane through the 

radars, and it also allowed us to observe the reflectivity and one component 

of air motion throughout the storm volume. Several storms were observed in 

this operating mode, but the Raymer storm of 9 July 1973 discussed here com- 

bines two fortuitous factors in addition to the extensive measurements obtained 

with other instruments: 

a. The storm track was within 5 deg of the line joining the radars. Thus, 

inflow and outflow parallel to the direction of storm motion were observed as 

radial velocity. These regions were clearly seen by the scanning radar and 

assisted in interpreting the measured radial velocity field in terms of the 

storm structure. 

b. The updraft region passed directly over the zenith-pointing radar. 

The vertical velocity data could therefore be used to locate the updraft pre- 

cisely in relation to the storm echo.

2. The Doppler radar observations 

III-2 

The location of Doppler radars A and B relative to the NHRE operational 

area is shown in Fig. 1. The radars were separated by 52.2 km with a baseline 

angle of 354 deg from north. The 9 July 1973 hailstorm passed directly over 

the southern radar (Radar A). Figure 1 also shows the reflectivity contours, 

just below cloud base, measured by the 10-cm radar located at Grover, Colorado 

at the time the storm was over Radar A. Radar A acquired data in a zenith- 

pointed mode from 1725:55 to 1736:15 MDT. During this same time, Radar B was 

scanning an azimuth sector at elevation angles of 0 to 15 deg in one degree 

steps. The 16 elevation steps were scanned in 160 sec. The area covered by 

Radar B during its initial scan sequence is outlined by the dashed lines of 

Fig. 1. A representative 

-l 

echo motion was 7 m s 1 along the radar baseline 

during the time interval of interest for the Doppler data. Seven m s is an 

intermediate value between the motion of the storm as a whole and the motion 

of individual cells (Part I, this series). The dotted line in Fig. 1 shows 

the coverage of the zenith-pointing radar as the storm advected overhead. The 

Doppler radar observations were made on cell W4. 

The characteristics of the X-band Doppler radars have been described by 

Frisch, et al (1974), but will be briefly summarized here. The complex video 

signals were digitally recorded for post analysis. The zenith-pointing radar 

acquired data at 24 fixed heights with a dwell time of 0.065 s corresponding 

to 128 radar samples in each height increment. From these data the complete 

Doppler velocity spectrum was obtainable 96 times per minute at each height. 

The radar beamwidth of 0.9 deg provided a horizontal resolution better than 

250 m at the maximum altitude. The height resolution was 75 m and the minimum 

height that could be observed was 1.8 km (MSL). (The radar altitudes were

%B 

III-3 

I\ 10 cm Radar Reflectivity 

\ 35,45, 55,65 dBz 

\ 1.40Elev. 

AI \ 17:27:32' 

I \ I 

2.6 2.6 2.9 km (MSL) 

~~~~~I / \ 3.2 

I - /o \ / 

f^^ / v v\ 60 km Range 

/\\( ,1C^ 

\ / 

. o 

Storm 

Motion 

Scan Region of A 

/ --- Scan Region of B 

FLg. 1 PPI phêntation of 10 cm atadart . edRectivity conwtours neat cloud base 

at 1727 MT and scand L egionzs oat the DoppleA Aadaus.

III-4 

1.5 km MSL.) The height spacing between data locations was 600 m, with the 

lowest location centered at 1.9 km and the highest at 15.7 km during the period 

1725:55 to 1728:05. From 1728:05 to 1736:15 the height spacing was 300 m with 

the lowest location centered at 1.9 km and the highest at 8.8 km. 

The scanning radar used the same dwell time (0.065 s for 128 radar 

samples) and acquired data for 384 (16 azimuth x 24 range) radar resolution 

elements in 10 s for each fixed elevation step. Four complete elevation 

scan sequences were made during the 12 min observation period. The 0.9 deg 

beamwidth resulted in a spatial resolution of 600 to 800 m at the location of 

the storm. The data locations were separated by 600 m in range and about 1.5 

deg in azimuth. The entire Doppler velocity spectrum was calculable for each 

measurement point of the scanning radar so the radial velocity, velocity 

variance and radar reflectivity were estimated at each data point. 

3. Data analysis procedure 

The radar reflectivity and the first and second moments of the velocity 

spectra were estimated from the 128 radar samples recorded at each range loca- 

tion. A Fast Fourier Transform algorithm was used with a general purpose 

computer to calculate the Doppler velocity spectra. Each radar acquired data 

for approximately 28,000 velocity spectra during the 12 min observation period, 

so that estimates of the spectral moments of the signal had to be made by a 

high speed computer from the Doppler signal plus noise data. The spectra were 

estimated from 128 radar samples and the moments were computed after a velocity 

window was applied to the spectra. The velocity window ranged on both sides of 

the peak of the spectrum to points where the power density fell to within 3 dB 

of the receiver noise level. The windowed spectra were manually checked to 

ensure that the automated method selected the correct signal spectra and

III-5 

properly accounted for velocity folding. The first and second moments of the 

velocity spectra were also estimated with a "pulse-pair" algorithm (Berger 

and Groginsky, 1973). The radar reflectivity factor, Z, was also calculated 

from the signal power estimates. The radar reflectivity values were not 

corrected for attenuation and are, therefore, only approximate. The vertical 

air motion, W, was estimated by subtracting the terminal fall speed of the 

particles, VT, from the measured, reflectivity-weighted particle velocity, 

Vz. The terminal fall speed of the particles was estimated using the empirical 

z 

0.107 

relation VT = 2.6 Z of Joss and Waldvogel (1970) as suggested by Atlas, 

et al (1973), with a correction for air density variations proposed by Foote 

and du Toit (1969). Since the V vs. Z relationship is relatively insensitive 

-1 

to Z, the estimate of V should be accurate to within + 1 m s in rain. The 

Joss-Waldvogel empirical equation underestimates the fall speed if hail is 

present in the pulse volume, consequently the updraft would be underestimated. 

The reflectivity factor, mean radial velocity, and velocity variance 

measured by the scanning radar were interpolated into a Cartesian coordinate 

system with the y axis along the radar baseline. The various quantities were 

then contoured and displayed in horizontal and vertical planes. Interpolation 

smoothed out large gradients because adjacent measurement points were used to 

calculate values on a Cartesian grid. Consequently, examination of the original 

measurements indicated that maximum gradients of the various quantities were 

reduced by a factor of about 2 after interpolation. The radial velocity sign 

convention is such that motion away from the radar is registered as positive. 

The vertical motion of the particles, V , contributes a radial velocity com- 

ponent of magnitude V sin 0 to the velocity measured by the scanning radar, 

where 0 is the elevation angle of the radar antenna. This contribution was

III-6 

always less than 1.5 m s in this experiment and was not removed from the 

radial velocity fields except in the vertical plane containing both radars 

because it was known only above the zenith-pointing radar. 

In the region where both radars acquired data, two-dimensional air motion 

was calculated by combining the observations of vertical air motion by the 

zenith-pointing radar with the horizontal air motion along the radar baseline 

observed by the scanning radar. The horizontal component of motion, Vh, in 

the vertical plane through the two radars was derived from the measured radial 

velocity, VR, using the relationship VR = Vz sin 0 + Vh cos 0. The two- 

dimensional air motion field was corrected for echo motion by removing a 

7 m s- 1 component from the horizontal velocity. 

4. Nature of the data 

An example of the data acquired by the zenith-pointing radar is shown in 

Fig. 2 to illustrate the temporal variability of the vertical structure. 

Sixty-four data points from a sample obtained in the updraft region over a 

period of 40 s are shown. The terminal fall speed computed from the Joss- 

Waldvogel equation varied from 8.2 to 9.4 m s for the reflectivity values 

shown in Fig. 2 (center), so the maximum updraft must have exceeded 13 m s- 1 

-1 

The particle velocity, V , changed 6 m s 1 in 8 s with little change in 

reflectivity, demonstrating that rapid changes can occur in the updraft. The 

beamwidth was about 120 m at 9.1 km altitude, and the pulse length was about 

75 m. The 6 m s velocity change, therefore, occurred while the echo was 

advected about ½ beamwidth. If the velocity change had been caused by horizontal 

shear within the updraft, a shear of 10 s would have been required. If it 

had been caused by a 10 m sl vertical advection of the updraft, a vertical 

-2 -1woul d account for the velocity change. 

gradient of 7x10 s would account for the velocity change.

6- 

III-7 

*.'. z=9.1 km 

4 _ 0 . . 

~ 2 *. .. 

E0 - o * 

.* 

.· . . 0 

3N 0. -. * 

:" ' "It0. 0 

-2 0 

-4 

40 

36 . .. 0 . 

_ 4 0· 

32 - . 

280 0 

*- 

5- - 

III-8 

Figure 2 can also be used to estimate the accuracy of the measured data. 

The standard deviation of the mean particle velocity, V , (top trace) is 

given by Miller and Rochwarger (1970) as: 

DJ 

where TD is the dwell time, X is the radar wavelength (0.0322 m) and a is the 

width of the velocity spectrum. The dwell time is the pulse repetition period 

(512 Us) multiplied by the number of pulses making up the sample from each 

volume increment (128), so T D = 0.065 s. The average width of the spectra 

(bottom trace) is about 2.5 m s so the standard deviation of the mean velo- 

city estimate is about 0.5 m s . The reflectivity estimates (center trace) 

are calculated from 128 radar samples, but the number of independent samples 

is given by Nathanson (1969) as: 

NI = 

4/ a TD 

In this experiment there are about 30 independent samples. The standard 

deviation of the reflectivity estimates is therefore about 1 dB. The standard 

deviation of the estimates of spectral width, a, is given by 

[3- 2 DT] (Miller and Rochwarger, 1970) 

-1 

and is about 0.35 m s for the data in the lower trace of Fig. 2. 

Figure 3 shows examples of vertical profiles of vertical particle velocity 

measured in the 0.065 s dwell time. The spatial samplings are for range gate

14 

E 

.~z 

x-W 

III-9 

E · xX~~~~~~~~~~~~~~ 

i ~~~~~~~~~X\~~~~ 

N X~~~~~~~~~~~~~~~ 

X 10 X 

X 

X X 

,X I ox 4X X 

X X 

* k AI \ x II/ 

0I. X X 

2 1 ;I sIei 

-2O -10 0 -20 -10 0 -20 -10 0 10 msec. t 

1735:01 1130:31 1126:01 

Ff2 . 3 V&Lctica pa t'c.&? v~eocityj p'4&~i~e~s mevswhed by, .teh~ ze~nith-po~iuttng 

/tdaILv cmd ve~'~tic~cd w&ind e~stimateA dchLve~d 6hwm -th. Jlo44-W~aedvog~e 

VT - Z i~~tcOJ'io. 

X 

ox 

'I

III-10 

spacings of 600 m (at 1726:01 MDT) and 300 m (1735:01 and 1730:31). The profile 

measured at 1726:01 illustrates the vertical variability of the velocity struc- 

ture, particularly in the updraft region above 6.5 km. At 1730:31 the maximum 

altitude of the measurements had been reduced from 15.7 to 8.8 km. The profile 

in the center trace of Fig. 3 indicates a downdraft existed over the radar at 

that time. The profile measured at 1735:01 in the back portion of the storm 

-1 

shows weak downdrafts. Particle velocities as high as +12 m s 1 were measured 

in the updraft at 1726:01, but 4.5 min later, after the storm had advected less 

than 2 km, velocities of -20 m s were measured. 

5. Results 

a. Air motion in. the vertical plane through the radar sites 

Figure 4-a shows the time vs. height plot of the 10-cm radar reflectivity 

above the zenith-pointing radar. The dotted line outlines the time and height 

of the data acquired by the zenith-pointing radar. As can be seen in Fig. 4-a 

the operator of the zenith-pointing radar reduced the maximum observed altitude 

shortly after the upper-level high reflectivity region had passed overhead. 

Radar B scanned all of the echo structure shown in Fig. 4-a except for the very 

earliest and latest portions. 

Figure 4-b shows contours of vertical air motion derived from the Doppler 

velocities and reflectivity measured by the zenith-pointing radar using the 

Joss-Waldvogel equation to remove the VT component.' Five s averages (8 values) 

of Z and V were used to derive an estimate of W every 5 s, or 35 m for an echo 

motion of 7 m. The time-height values of W represent the region inside the 

motion of 7 m s . The time-height values of W represent the region inside the 

dotted outline of Fig. 4-a. No updrafts were observed below 6.5 km, but the 

trend of the data suggests that the low-level portion of the updraft had advected 

past Radar A before 1725. The boundary of the surface outflow passed the radar

III-ll 

4 Scan onScan Region for 

14 for Radar B _ _ 

. 

Scan Ri Reio..... 

. _ .. 

I . 

0 2 4 6 8 10 12 

km 

Vertical Velocity- IOmsec- \ \ Variance of ' 

m sec-' I e\ Verti c o l Veocity i-`2. 

.- m seci 

ec 

= K I" ' c - 

"\"~~~~' · e *p J 

: !. .^ - 

!^ 

2 

\(jf^/.r ''"\\\· "!'" 

~. ~I . 

4 Cloud . . ...... 

Fig. 4 a) Time-height presention of 10 cm reflectivity contourw meiurwed 

over the zenth-pointing Vopple. rtadat. and the scan riegiaons for 

the DopperL tradac s. 

b) Aveuaged vvetica wind contowuT measured by the zen.lth-poiLnting 

radarL tom 11726 to 1736 MOT. 

c) Two-dimensioncu& aitr motion uelative. to thee stotun echo motion -in 

the vetiutcaZ plane through the two DoppRe tadar6s. 

d) Averaged velocity variance measuwed by the zenith-pointing tadar.

III-12 

at 1715 while the leading edge of the low-level echo arrived at 1725. Maximum 

updraft values occurred between 8.0 and 10.3 km altitude. The slope of the 

updraft contours indicates that the updraft was tilted toward the north-- 

opposite to the direction of echo motion. This tilt was about 20 deg from 

the vertical. Highest downdrafts were measured between 3 and 5.5 km altitude 

and occurred just behind the updraft. The back portion of the storm contained 

weak downdrafts in the lower levels, but near the highest altitude observed 

there were weak updrafts, even after the portion of the storm with highest 

reflectivity had advected past. 

The two-dimensional air motion in a vertical plane through the storm 

observed by both radars is shown in Fig. 4-c. The vector representation shows 

the air motion relative to the storm in the region outlined by the dotted lines 

in Fig. 4-a. The transition between strongest updraft and strongest downdraft 

probably took place over a distance of about 3 km, and as we shall show, strong 

turbulence was generated in this region. The most intense part of the down- 

draft appears to have been fed by air which had ascended in the updraft, as 

postulated by Newton (1963). The peak downdraft occurred in the region of 

highest reflectivity and water loading probably contributed to the downdraft. 

Reflectivity contours measured by Radar B in the vertical plane through 

the radars are shown in Fig. 5-a. The highest reflectivity factors measured 

by the 3-cm radar were about 50 dBZ and occurred at an altitude of 5.5 km and 

2 km west of the baseline joining the radars. The 3-cm radar reflectivity 

contours shown in Fig. 5-a are plotted from data acquired between 1727 and 

1729:40 while the radar scanned through 16 elevation angles. The time-height 

plot of 10 cm radar reflectivity factor shown in Fig. 4-a is a 30 min history 

of the reflectivity over Radar A. Reflectivity plotted in Fig. 4-a and Fig. 5-a

z 

III-13 

354° STORM 

- -- ~ MOTION 

1727-1729:40 ,DT 8 

.r - 24 - 32 Reflectivity 

11.5|- I < = I dl)z 

km 401 

6.5 - 36 

1.5 L2 

z 

11.5 

28 -'S KI //. ~~~32I /- 

- 8 - 2 41 

2 8 

11.5 , j "--" -,I Velocity 

4,, _ _..( m sec 

1.5 10 - 5 OkmI. 

10 

1.5-i 

z ~- - 

22 

11.5 ,I/ 

km 

^--O 0I < Variance 

/' 1"/---4 ,~ l. m sec 2 

km 

Fig. 5 a) Thtee cm tadazt a etectivit ty, b) tadiat pacticte velocity, and 

C) veocity va'iance meatwled by Radat B in .the veticat pa&ne 

uthAogh the twao DoppteQ Pltadee tudac ofed |ine4 . The. eglon 

coAte- 

4ponds to ;the scavn eg 'on do Rada. B s4hown in Fig. 6haded 4-a-. The 

t eg-on 

en b indicautes tadicial ve2ocQie s that exceed the echo motion of 7 m 4-1. 

The 4haded Legion in c indicawte4 wheh e the zenUth-pozntLng radah mea6uhed 

the. main uapdtrat.

III-14 

should have been the same if: (1) attenuation of the 3-cm radar was negli- 

gible, (2) the storm track was from 354 deg during the time period 1720 to 

1750, (3) the storm was in steady-state during this time, and (4) both radars 

were properly calibrated. None of these conditions was precisely satisfied. 

Reflectivity gradients at the leading and trailing edges of the storm appear 

similar in the two plots, although the measured reflectivity values were greatly 

different. Attenuation of the 3-cm radar signals would be most noticeable at 

mid-level of the leading edge of the storm, but the 3-cm radar reflectivity 

contours portray a structure similar to that depicted by the 10-cm radar. 

The radial velocity field (measured by Radar B) shown in Fig. 5-b corre- 

sponds to the time and. location of the reflectivity data shown in Fig. 5-a. 

The velocity contours are absolute velocities; there is no correction for the 

echo motion and no correction for the terminal fall velocity contribution. 

Negative velocities are directed toward Radar B and positive velocities are 

directed away from it. Strong updrafts were measured above Radar A during 

the time these data were collected by Radar B. The velocity field in Fig. 5-b 

suggests the location of the updraft even without verification from Radar A. 

Air feeding the updraft entered the storm from the right side of Fig. 5-b 

(Part I). Figure 5-b indicates that the air was moving into the storm at mid- 

-1 

levels with a horizontal velocity component as high as 10 m s 1 relative to 

the storm. The region of high radial velocity gradient above Radar A clearly 

defines the interface that existed between the updraft and downdraft. The 

measured component of airflow at the rear of the storm above 7.5 km was 

generally similar to that in the near environment as measured by the nearest 

representative radiosonde. Mid-level air was overtaking the storm from the 

backside at 6.5 to 8.5 km altitude. The outflow at the top of the storm had

III-15 

a strong northward component of more than 15 m s relative to the storm. The 

trend of the radial velocity in the y direction suggests that weak outflow at 

the top of the storm could have occurred in the direction of echo motion. It 

seems unlikely that this implied weak outflow would have been strong enough to 

have led to particle recirculation to account for the few larger hailstones 

found in this storm. The velocity field shown in Fig. 5-b is in agreement 

with photographs which show a small anvil overhang in the direction of echo 

motion (Part I) and a large anvil overhang toward the northeast. The surface 

anemometer readings of 13-17 m s are higher than the velocities seen by the 

radar in the gust front, and they therefore indicate that the strongest low- 

level outflow in the direction of echo motion occurred in the lowest few hundred 

meters; this was not observed by the radar. 

b. Air motion in horizontal planes 

Figures 6 a-c show the radial vleocity fields measured by Radar B at 

three altitudes. These data were acquired during the time period 1727 to 

1729:40. The lowest altitude is labeled 1.8 km, but the antenna pattern at 

elevation angles corresponding to this altitude was distorted by ground 

blocking so the data shown for this altitude are actually more representative 

of the wind field at about 2.1 km. The radial velocity fields in Fig. 6 are 

not corrected for echo motion. The velocity field at low altitude reveals 

outflow in the direction opposite the echo motion and also outflow in the gust 

front. The radial velocity field at 6.5 km (Fig. 6-b) shows the strong con- 

vergence of the updraft air at the south or leading edge of the storm and the 

downdraft air from the backside that was overtaking the storm over a wide area. 

Two regions of air moving into the storm (contours labeled - 2 m s 1 ) probably 

correspond to the two distinct inflow branches toward W4 and W5 as identified

III-16 

1727-1729-40 MDT N 

STORM ! 

MOTION 

km . km 

Z= 1.8 km 

~18 .J-. 18- _ Z=6.5 km 

1 -2 m sec - 

7 

_0 

Velocity . 

, 

Velodty 

9.~~~ 9 

0 I i o 

.~ m sec -' 

km 

aude, kmb-and'c) Z:.5 Z=11.5 km 2 Z=6.5 km 

Plane of 2 

n-8-' 8 . Velocity y Fig.5c a Vari 

eanI Q «Q 4fflQ M « n ItgtaM 

\Va riance 

+ 

4 

-29 - 9 

-44m sec- 14 2 e2-- - smec- 2 

Ln-T- ------- ^^ I I _ x L._- ,-I x 

0 9 km 18 27 0 '9 m 18 27 

Fig. 6 Radiat paitice veJocity (abso.ute) at a) 1.8 km attiJtude, b) 6.5 km 

ati.ttude, and c) 11.5 km al-titude mewaaed by leiocated Radar B 

50 km 

nouth of the storim. The shaded regions indicate radiala veoc&ites that exceed ' 

the stoam motion of 7 m -l ' . d) Velocity vaiZance at 6.5 km altitude. The 

4can egioons aWe the -6ame a6s thoe in Fig. I. 

i10'

III-17 

in Part I. The main updraft of W4 had just advected past Radar A at this 

altitude when these data were acquired. The maximum radial velocity shear 

indicated by these contours is about 5x10 s whereas the data prior to 

interpolation indicate peak shear values greater than 10 s . The updraft 

and downdraft regions of the storm were readily distinguished on the basis 

of the radial velocity fields measured by Radar B, but if the Doppler radar 

had been located due west of the storm it probably would not have been able 

to discriminate between these regions so readily. The low and mid-level 

radar data in Fig. 6 show that the cell that passed over Radar A was contiguous 

with a line of cells that formed about 30 km to the east. 

The radial velocity contours at 11.5 km (Fig. 6-c) show the strong diver- 

gence near the top of the updraft. Particle velocities at this level appeared 

as though the particles were ejected from a fountain. The motion relative to 

the storm was strongly toward the north over a wide area at the back of the 

storm. The environmental winds at this altitude were from the west-southwest 

at about 12 m s , corresponding to a radial component of only about 3 m s . 

Particle velocity with a southerly component of 4-7 m s , nearly equal to the 

echo motion, occurred in front of the updraft at 11.5 km altitude so that some 

outflow in the direction of echo motion was likely. 

c. Distribution of turbulence inferred from the variance of the velocity 

There are three major factors that can contribute to the variance or 

second moment of the Doppler spectrum measured by a radar with a narrow beam- 

width: wind shear, turbulence, and the spread of particle fall speed in still 

air (Atlas, 1964). The contribution to the variance caused by fall speeds is 

given by a sin 2 

where aD is the fall spread that would be observed in 

2 2 -2 

still air at vertical incidence. aD is about 1 m s for rainfall and is

III-18 

nearly independent of rainfall rate (Lhermitte, 1963). For exponentially 

distributed hail with a maximum diameter of 1.5 cm, as occurred in this storm, 

2 2 -2 

o D is 4 to 6 m s (Battan, 1974). The variance caused by wind gradients 

parallel to the radar beam is given by 

(k h) 2 

-12 (Sirmans and Doviak, 1973) 

where kR where is the radial shear ( (s ) along the beam and h is the pulse length. 

For shear across the beam, the variance contribution is given by (0.3 kT R4) 2 

where kT is the radial shear transverse to the beam (Nathanson, 1969), R is 

the range, and D is the one-way, half-power beamwidth. 

All factors that contribute to the variance must be considered for the 

zenith-pointing radar but the fall speed spread can be neglected for the quasi- 

horizontally scanning radar because the contribution caused by fall speed was 

2 -2 2: -2 

at most 0.08 m s in rain and 0.5 m s in hail. The measured variances 

2 -2 

in this storm were much greater than 0.5 m s in regions where hail was 

probably present and much greater than 0.08 throughout the storm. Since the 

fall speed spread was negligible for the scanning radar and since the radial 

velocity field was measured throughout the storm, thus making the radial shear 

known, the turbulence throughout the storm could be calculated from the variance 

field measured by the scanning radar (Strauch, et al, 1975). 

The velocity variance measured by Radar A is shown in Fig. 4-d and the 

variance measured by Radar B is shown in Figs. 5-c and 6-d. Variance data from 

the zenith-pointing radar were contoured after smoothing by taking an 8-point 

(5 s) average. High variances were observed by Radar A in two regions of the 

storm: at mid-levels where Radar B also measured high variances and just

III-19 

below cloud base where Radar B did not measure high variance., The low-level 

band of high variance must therefore be attributed to spread in particle fall 

speeds. In fact, shear or turbulence would have caused Radar B to measure 

larger variances than Radar A because the pulse volume of Radar B was much 

larger. The large variance at mid-levels seen by Radar A were caused, in 

part, by shear and turbulence. Figure 2 shows that its beamwidth was too 

large to resolve strong local horizontal gradients in the updraft, so this 

kind of local shear would be interpreted as turbulence in its contribution to 

the variance. 

Variances measured by Radar B (Figs. 5-c and 6-d) at the back of the storm 

2 -2 

were less than 2 m s and can be attributed to vertical gradients of the 

-3 -1 

radial velocity since the vertical shear of 5x10 s , seen in Fig. 5-b, is 

2 -2 

sufficient to cause a variance exceeding 2 m s . On the other hand, the 

core of large variance between 5.5 and 9.5 km in Figs. 5-c and 6-d cannot be 

attributed to shear because the large shear in this region was parallel to 

the beam, and its contribution to the variance was small since the pulse length 

was only 75 m. This region contains large gradients in the vertical velocity 

(Fig. 4-c) but these gradients do not contribute significantly to the variance 

observed by Radar B. We conclude that the high variance core measured by 

Radar B was caused mainly by turbulence generated by the large horizontal 

shear of the vertical wind at the interface between updraft and downdraft. 

The dissipation rate, e, derived from the velocity variance field measured by 

2 -3 

Radar B varied less than 30 cm s at the back of the storm to more than 

2 -3 

3000 cm s in the region between the updraft and downdraft (Strauch, et al, 

1975). These values span the entire range of values measured by aircraft in 

clear air turbulence,

III-20 

We assumed that the outer scale in the inertial subrange was larger than 

the largest dimensions of the radar pulse volume (800 m). Energy spectra 

with a -5/3 power law to scales greater than 1.5 km have been measured with 

penetration aircraft in thunderstorms (Steiner and Rhyne, 1962). The radial 

velocity fields shown in Figs. 5-b and 6-a seem to indicate that the outer 

scale is no larger than the radar resolution since they do not contain fluc- 

tuations at scales of 1 to 2 km. However, the interpolations used to transform 

the measured data to Cartesian grid points filtered the energy spectrum at 

scale sizes between the radar beamwidth and the outer scale so the radial 

velocity fields show only organized motions. Spectra measured with penetration 

aircraft in severe Oklahoma storms (Steiner and Rhyne, 1962) showed dissipation 

rates that exceeded those measured by our radar. 

5. Conclusions 

The Doppler radar measurements described constitute a unique data set 

because, for the first time, a scanning Doppler radar obtained the three- 

dimensional field of reflectivity, radial velocity, and the velocity variance 

of the radial velocity in a volume that included the updraft of a convective 

storm, while, at the same time, a zenith-pointing radar observed part of the 

storm as it advected overhead. The Doppler radar data do not provide the 

total picture, but they display more of the kinematic structure than can be 

obtained by other instruments. 

The data obtained by the scanning Doppler radar illustrate both the utility 

and the limitations of single Doppler radar measurements. The radial velocity 

field from a single radar cannot yield the unambiguous three-dimensional wind 

fields that can be derived from dual-Doppler measurements; however, many 

features of the overall air motion in convective storms can be inferred from

III-21 

the radial velocity fields. This is especially true if, as in this case, the 

primary motion of the air entering and leaving the storm is in a vertical plane 

radial to the radar. 

The second moment of the Doppler spectrum, obtained by a Doppler radar 

scanning the entire storm at low elevation angles, portrays.regions of high 

shear and turbulence. The turbulence contribution can be isolated because 

shear can be inferred from the radial velocity fields. Measurement of the 

second moment fields and their evolution requires only a single radar.. These 

data have not previously been fully utilized by radar meteorologists, but they 

can, in principle, provide a warning of dangerous turbulence or wind shear 

conditions and can aid in understanding how seed material or tracers will 

diffuse in a storm.

III-22 

REFERENCES 

Atlas, D., 1964: Advances in radar meteorology. Advances in Geophysics, 10, 

318-478. Academic Press, New York. 

__ , R. C. Srivastava, and R. S. Sekhon, 1973: Doppler radar characteristics 

of precipitation at vertical incidence. Rev. Geophys. Space Phys., 11, 

1-35. 

Battan, L. J., 1974: Doppler radar observations of a hailstorm. Sci. Rep. #28, 

Univ. of Arizona, Tucson. 

Berger, T., and H. L. Groginsky, 1973: Estimation of the spectral moments of 

pulse trains. International Conf. on Information Theory, Tel Aviv, Israel. 

Browning, K. A., T. W. Harrold, A. J. Wyman, and J. G. C. Beimers, 1968: 

Horizontal and vertical air motion and precipitation growth within a 

shower. Quart. J. Roy. Meteor. Soc., 94, 498-509. 

Foote, G. B., and P. S. du Toit, 1969:; Terminal velocity of raindrops aloft. 

J. Appl. Meteor., 8, 249-253. 

Frisch, A. S., L. J. Miller, and R. G. Strauch, 1974: Three-dimensional air 

motion measured in snow. Geophys. Res. Letters, 1, 86-89. 

Joss, J., and A. Waldvogel, 1970: Raindrop size distribution and Doppler 

velocities. Preprints, 14th Radar Meteor. Conf., Tucson, Ariz.; Amer. 

Meteor. Soc., 153-156. 

Lhermitte, R. M., 1963: Motions of scatterers and the variance of the mean 

intensity of weather radar signals. Sperry Rand Res. Center, 5RRC-RR-63-57, 

Sudbury, Mass. 

Miller, K. S., and M. M. Rochwarger, 1970: On estimating spectral moments in 

the presence of colored noise. IEEE Trans. in Form Theory, IT-16, 303-309.

III-23 

Nathanson, F. E., 1969: Radar Design Principles; Signal Processing and the 

Environment. McGraw-Hill, New York. 

Newton, C. W., 1963: Dynamics of severe convective storms. Meteor. Monograph, 

5, (27), 33-58. 

Sirmans, D., and R. J. Doviak, 1973: Meteorological radar signal estimates. 

NOAA Tech. Memo. ERL-NSSL-64, U. S. Dept. of Commerce, Norman, Okla., 80 pp'. 

Steiner, R., and R. H. Rhyne, 1962: Some measured characteristics of severe 

storm turbulence. National Severe Storms Project, Report No. 10, U. S. 

Dept. of Commerce. 

Strauch, R. G., A. S. Frisch, and W. B. Sweezy, 1975: Doppler radar measurements 

of turbulence, shear, and dissipation rates in a convective storm. Pre- 

prints, 16th Radar Meteor. Conf., Houston, Texas; Amer. Meteor. Soc., 83-88.

Structure of an Evolving Hailstorm. Part IV: 

Internal Structure from Penetrating Aircraft 

by 

D. J. Musil, E. L. Mayl, P. L. Smith, Jr., and W. R. Sand 

Institute of Atmospheric Sciences 2 , South Dakota School 


1Current affiliation: National Weather Service, Detroit, Michigan. 




Science Foundation.

ABSTRACT 

Precipitation particle sizes were measured using a continuous 

hydrometeor sampler (foil impactor) during penetrations of hailstorms 

with an armored T-28 aircraft. Data have been analyzed from three pene- 

trations of a storm near Raymer, Colorado on 9 July 1973 at altitudes 

between 5.5 and 7.2 km MSL, which correspond to temperatures between 

about -2C and -12C. Other results pertinent to the Raymer storm are 

discussed in Parts I, II, III, and V elsewhere in this issue. 

Most of the particles were identified as ice particles or ones 

containing both ice and water; however, significant amounts of liquid 

particles were found in the updrafts of developing cells at tempera- 

tures as cold as -12C. Particles larger than 5 mm in diameter were 

typically found along the edges of the updrafts, with the precipitation 

concentrations being strongly dependent on these larger particles. 

The downdrafts were composed of ice particles. 

Several particle size distributions from one of the penetrations 

were examined. The distributions are roughly exponential, or 

bi-exponential when large particles are present.


IV-1 

There have been numerous investigations of particle size distributions 

from precipitation observations made at the ground using camera devices 

and impact devices (e.g., Cataneo and Stout, 1968; Waldvogel, 1974; 

Federer and Waldvogel, 1975). Investigations of in-cloud particle sizes 

have used airborne foil impactors based on a principle first described 

by Brown (1958) but have necessarily been restricted to small cumuli 

(Bethwaite et al., 1966; Schroeder, 1973) and frontal clouds (Cornford, 

1966). Systematic observations within hailstorms and thunderstorms are 

virtually nonexistent because of the difficulty and danger involved in 

obtaining them. 

During the 1973 National Hail Research Experiment (NHRE) field 

season, a continuous hydrometeor sampler (foil impactor) was flown 

aboard an armored T-28 aircraft and obtained observations of particle 

sizes on a routine basis. The purpose of this paper is to present the 

results of a detailed analysis of the foil data obtained from penetra- 

tions of a hailstorm occurring near Raymer, Colorado, on 9 July 1973. 

This paper can be viewed as an extension of the work by May (1974), who 

initially analyzed the foil data. The paper is part of a comprehensive 

analysis of data from the Raymer storm by several participants in NHRE 

which appears elsewhere in this issue (see Parts I, II, III, and V). 

2. Data collection and reduction 

The foil impactor was flown aboard an armored T-28 aircraft operated 

by the South Dakota School of Mines and Technology. The mode of operation

IV-2 

and other information collected by the T-28 system have been described 

by Sand and Schleusener (1974). 

Typically the T-28 mission consisted of gathering data during 

penetrations of active hailstorms beginning at an altitude 3 of about 

7.3 km and proceeding downward by 0.6 km intervals until 4.9 km was 

reached. The aircraft was vectored with the objective of entering the 

maximum radar reflectivity zone and major updraft at aircraft altitude 

on each storm penetration. A penetration was usually made at a constant 

heading until clear air was encountered or until the T-28 was well clear 

of any radar echoes. Typically, three to six penetrations were made 

on each mission. 

2.1 Description of foil impactor 

The foil impactor has a moving strip of soft aluminum foil 

approximately 7.5 cm wide and 30 pm thick, which passes by an open 

window with dimensions of 3.75 by 3.75 cm at a speed of about 3.8 cm 

s . The foil is backed by a ridged drum so that when ambient air 

strikes the foil, particles in the air leave impressions on the foil. 

The particle sizes can be determined from marks left within the imprints 

on the foil by the ridges on the drum, which are spaced at 250 Pm 

increments. Thus the smallest particles and the diameter resolution 

of particles included in this study are of the order of 0.25 mm. 4 

The foil impactor face plate is electrically heated to minimize 

icing problems that can occur in the severe environments of the storm 

3 All heights are given with respect to mean sea level. 

4 A11 particle sizes are expressed in diameters.

IV-3 

penetrations. The supply roll of foil is large enough for approximately 

thirty minutes of in-cloud observations, which is sufficient for most 

missions. 

2.2 Data reduction 

Data reduction is extremely time consuming, so initially the 

analysis was limited to penetrations made on two days in which the 

greatest vertical air motions were observed and the pilot's comments 

indicated the analysis of the foil would be most productive. Only the 

analysis of data from the Raymer storm will be presented in this paper. 

The foil was marked off into sections (or frames) each representing 

four seconds of observation time, which corresponds to a flight path of 

about 0.4 km and a sampling volume of about 0.6 m 3 . The particle 

numbers and sizes were recorded for each frame. 

Particles that could be distinguished as completely liquid were 

recorded separately from those that were completely solid or contained 

some fraction of ice. The identification of liquid particles is quite 

simple (Miller et al., 1967) because they leave circular imprints with 

raised edges. Particles that are composed partly or wholly of ice tend 

to leave irregular, splattered imprints, making it impossible to deter- 

mine the amount of liquid in such a particle. The "combination" 

particles were recorded as ice in this study. 

Imprint sizes less than 7.5 mm were corrected according to an 

expression developed from work by Schecter and Russ (1970). The 

correction is for an approximate true airspeed of 100 m s - 1 and is.

IV-4 

necessary to account for the distortion of the particles upon impact 

on the foil. Beyond 7.5 mm, where no calibration data exist, the 

particle sizes were taken to be the same as the imprint sizes. The 

imprint sizes are probably larger than the actual sizes for all 

particles encountered, as is indicated by the Schecter and Russ work 

for particles smaller than 7.5 mm; however, extrapolation of their 

correction procedure would lead to the opposite result for large sizes 

and in extreme cases would even require particles too large to enter 

the window of the foil impactor. 

In order to obtain samples large enough to reduce some of the 

errors associated with small sampling volumes (Joss and Waldvogel, 1969), 

data from three consecutive frames were summed to form a sample for which 

we computed the total number concentration for all particles observed 

on the foil (N ), the concentration of particles larger than 5 mm (N 5), 

the precipitation concentration for particles larger than 0.25 mm (PC), 

and the percent mass liquid (PML). Then the last frame of the set was 

retained and added to the next two consecutive frames and the calcula- 

tions repeated. In this manner, smoothed data points were obtained for 

each 8 sec representing a 12-sec running average that corresponds to a 

flight path of about 1.2 km and a sample volume of about 1.8 m 3 . 

3. General observations 

Pertinent data and summaries from penetrations of the Raymer storm 

are given in Table 1. The penetrations were made between altitudes of 

approximately 5.5 and 7.2 km MSL which correspond to temperatures of

IV-5 

PENETRATION 80-I 80-2 80-3 

DATE 9 JULY 197 3 

START TIME 171636 172904 174147 

END TIME 172322 173649 175112 

AVG ALT-(km) 7.2 5.9 5.5 

AVG TEMP-(OC) -12 -5 -2 

VOL SAMPLED -(m 3 ) 59.2 68.0 819 

% /_100 97 100 

MAX 268 440 352 

;_ AVG 92 119 101 

0r 65 111 96 

% 42 20 25 

fg MAX 7.2 11.6 3.0 

AVG 3.3 4.8 1.7 

' ;! 

2.5 3.5 .9 

- %7 48 46 49 

MAX 2.0 2.1 1.1 

AVG 0.7 0.6 0.3 

: a Xr _ 

X0.6 0.7 0.3 

Tabee i. Swumnmy ad Pcenetion cDat dafo 9 July 1973.

IV-6 

about -2C to -12C. These temperatures are based on adiabatic ascents 

from observed cloud base using the appropriate radiosondes on 9 July 

(Part I). Temperatures actually observed during the penetrations were 

not used because one of the T-28 temperature probes was subject to 

wetting and freezing, while the second probe was subject to significant 

day-to-day calibration drift. 

Also shown in Table 1 are the maxima, means, and standard deviations 

(a) for Nt, N 5 , and PC for each penetration. Not included in the 

t 

averages were regions having PC values less than 0,1 g m - 3 . Average. 

values of N. and PC are lower than might be expected in view of their 

t 

associated maxima because the penetrations showed the clouds to be 

composed of large areas where few particles were encountered. The 

percentages shown for each variable refer to the fraction of time 

during the penetration that each variable was observed. Thus, it can 

be seen that large particles (N 5 ) were found in relatively small 

regions of the cloud compared to the total concentration (N t ), for 

which values near 100% indicate that there were almost always some 

particles striking the foil impactor during a penetration. 

The total number concentrations are lower than those normally 

observed in rainfall at the ground. This is also indicated by the 

particle size distributions discussed in Section 4.2, which generally 

show fewer particles but more large ones than the familiar Marshall-Palmer 

distribution would suggest. This may indicate that the processes of 

raindrop breakup and/or evaporation and shedding of water by falling 

hailstones modify the particle size distributions appreciably before the 

precipitation reaches the ground.

IV-7 

A close relationship exists between PC and N 5 (the relationship 

is more apparent in the plots of Fig. 3 than in the values of Table 1), 

indicating that the large particles contribute most of the PC. Thus, 

significant precipitation concentrations are also restricted to smaller. 

regions of the cloud than N t. The observed precipitation concentrations 

are in general agreement with those reported by Schroeder (1973) and 

in Texas.5 

4. The Raymer storm 

All three penetrations of the Raymer storm were in part of Storm W 

described in Part I. Figure 1 shows a contoured slant range PPI display 

from the Grover 10-cm radar near the time and altitude of Penetration 

80-1, with the aircraft flight path superimposed. 

Penetration 80-1 was made near the center of a newly developing 

cell (W5) and along the southeastern edge of the high reflectivity region 

of a more mature cell (w4). Subsequent penetrations, 80-2 and 80-3, 

were in Cell W5; however, information from Penetration 80-1 will be 

emphasized in this paper because it was the most interesting and the 

air space limitations on 9 July prevented traversing the desired regions 

on Penetrations 80-2 and 80-3. Even on Penetration 80-1, we were pre- 

vented from penetrating the most preferred regions of the cloud, which 

were along a more north-south or northwest-southeast line that would 

intercept both the high reflectivity zone and the updraft region. 

5 See San Angelo Cumulus Project annual reports for FY1972 and FY1973 

by Meteorology Research, Inc., Altadena, California. Reports prepared 

for Texas Water Development Board.

I 

IV-8 

+ 45 KM E GRO 

1723 

d30v and 

1720/A 

7.0 d and 

// 

.0' \J W4 / \ I0 // 

w4 ( 

CLD 

e~ntt at 171J6:37 MPTs The mas I aon tit e utn on n 

Fig. J 5&nt' PPI duptaiy nea/t time and atttude of Penenation 80-1. The 

radat elevation angle and time of scan ae 7.0 degrees and 1718:38 

MDT, uespectivety. Reftectivity facto' contou arte indicated in dBZ. The 

heavy s fine hows the aircrAaf t t w rack ith the large dAot denoting vnua cloaud 

entry at 1716: 37 MP7D The ctick marks alon9 the track aee eat even one minute 

interuva s. The tihme sca£e on thL and Succeeding figures can be converted 

to a diutance ceae. by noting that caichad t spced u about 6 km min- l .

IV-9 

The large dot shown at the beginning of the aircraft track in 

Fig. 1 corresponds to entry into the visual cloud as seen by the pilot. 

The actual entry point of the aircraft into the cloud can be seen by 

referring to Fig. 1 of Part II, where an X on the cloud photographs 

indicates the point of aircraft entry into the cloud. 

Figure 2 shows the corresponding vertical section along the flight 

path of the T-28 for Penetration 80-1, and also the approximate 

locations of Cells W4 and W5. The reflectivity pattern shows W5 to be 

a cell with echo developing aloft as part of an overhang region and 

weaker than W4, which is in a more mature stage of development. 

4.1 Ice-water budget considerations 

Many of the interesting features revealed by the analysis of the 

foil data can be seen in Fig. 3, which shows smoothed values of vertical 

velocity, Nt , N5, PC, and PML as a function of time for Penetration 80-1. 

The smoothed vertical velocity curve masks the distinction between 

Cells Wh and W5, which is clearer in a more detailed presentation of 

data for the initial portion of Penetration 80-1 in Fig. 4. 

Few actual in-cloud measurements of N5 have been reported, but the 

values found in this study seem reasonable. The greatest concentrations 

of large particles (N 5 ) seemed to be at the edges of updraft regions 

(Fig. 3). Although the foil impactor was not carried on the aircraft 

during 1972 operational flights, these observations of N5 are in 

qualitative agreement with the pilot's observations of large particles 

during penetrations of a storm on 22 July 1972 (Musil et al., 1973).

I 

IV-10 

/ II I I '/I 

~I~~~w 

\ / \1 " 

LO E , 

I 

i I Ii I ! i I i I I I 

o c jn 

Fig. 2 Compuate generated veticaR section (hand contoared) atong the track 

od Penetration 80-1. The dashed line Thows the aiurcafLt tiaci and 

treectivity f acdto caontours cLe indicated Lin dBZ. Cell W4 is in thei mawte 

stage wAhe. W5 has a recently developed echo aloft.

IV-11 

· 2 - r I I II 

VERTICAL VELOCITY (m/sec) 

10 

W° 5 W4 \ / 400 

TOTAL PARTICLE CONCENTRATION (/m 3 ) 

-10 - A. 

/."'. __ . ,î^, _200- 

12 -4o I c1 I - C-- - 

6 / 

CONC. PARTICLES >5mm(/m 3 ) 

tN\ J/ 

PRECIP CONCENTRATION (g/m 3 ) 

' I V I i^-t 1- 1 1 100 

PERCENT MASS LIQUID 

50- 

|! | \ 1 / I I I 0\ 

1716 17 17 1718 1719 1720 1721 1722 1723 

IN TIME-(Minutes) 

CLD 

Fig. 3 Ptot6 o04 moothed veticat veaocityj, totat concenutution o6 patictu, 

concetut1to1 1n ao patric h geh -than 5 mm, pMtcepiLtatorn 2- 

concen2t 

tion, and percent mass liquid fot Penetwon 80-1. PenetLtiron was made at an 

avetage a&tjtude of 7.2 kmn, wdich coraesponds to a atempeatuae. o about -12C. 

The7 distinction btwQeen CelQ W4 and W5 LU apparent in the moe detaled veAtica 

veRlocity cwuve in Fig. 4.

IV-12 

~,~_ VERTICAL VELOCITY-(m sec) 

o__ 0 

(D 

O o0 

^ -II-4~~~~~~~~~~~~~~~~~~~~~~I 

__ i z ~~~~~~~i 

iMDt 

. .'" . 

''-LGT 

---- _/ '"^ lHVY ' 

eI S PC & LWC .(gm 3 ) 

m- cn (PC), wnd w kt 

' 3 /S : ! · 

· · 

rr o 0 HVY 

C F- 

whee pa the ce ze d bton hwn - ~~~~~0 ----F. weQ d ned. I 

PC a LWC -(gmi 

£ Fig. 4 sg. Ve&caQ Vvetica. veoc,, velocity, c cloud .oud £-qwud liquid water ScatVL concent.tat concentration (L(C), (LWC), phQC4(- precipi- 

·t..ton concen.tAation (PC, Roemount 'iic.q .ate activity, and (aircaft 

(ihcg observations by. the p£o^t for about the Ihfdt hafi of PInentrIon 

The vettical 

80- . 

velocities 

£n FeLg. 

are 

3. 

uynsmoothzed 

The. Zette 

and theArore 

atcng ;the. 

differ 

tme 

from 

axg 

those 

Wgate 

shown 

rn Fig. 3. The !&ttw> along .the time axis indicate the hQepeCVtegeonz te~pective 

whnhe 

regiegons 

nte panticle size dustAAibutions shown tn Fig. 5 were determined. 

- 3 )

IV-13 

The close relationship between PC and N 5 that was mentioned 

previously is very apparent in Fig. 3. The largest value of PC, when 

N 5 = 0 was only about 0.3 g m-3; when larger values of PC were encoun- 

tered, many of the particles were larger than 5 mm in diameter and 

therefore presumably in the form of graupel or hail. 

Multiple updraft regions were encountered during this penetration 

and the highest values of PC straddled the interface between the updraft 

and downdraft region of Cell W4. Multiple updrafts were previously 

reported by Musil et al. (1973). Downdrafts encountered were generally 

weak with some penetrations having virtually no downdrafts. 

Analysis of the foil data revealed that most of the precipitation 

size particles encountered in all of the penetrations were composed of 

ice or some combination of ice and water. It was impossible to deter- 

mine the proportions of ice and water in the combination particles so, 

as noted in Section 2, they were taken to be ice particles for purposes 

of analysis. The proportion of the precipitation mass that was entirely 

liquid is shown for this penetration in Fig. 3. In general, the updrafts 

in the developing cells had significant amounts of liquid and the 

percent of the mass that was liquid depended to a certain degree upon 

the temperature at which the penetration was made; i.e., more ice was 

found at colder temperatures. Very little liquid precipitation was 

found in the regions where we observed significant numbers of large 

particles (N 5 ), which should fall in our "ice" category because liquid 

drops of such size are unstable.

IV-14 

Terminal velocities (McDonald, 1960) calculated for the particles 

observed in any of the penetrations on 9 July showed that most of the 

mass encountered was falling relative to the ground. Thus, in regions 

of high ice concentrations most of the particles came from regions. 

higher and colder than the penetration level, where the mass would 

most likely be in the form of ice. 

The presence of liquid in the updraft is shown in greater detail 

for a portion of Penetration 80-1 in Fig. 4. The vertical velocities 

shown are those actually measured by the T-28 during the penetration 

and are therefore somewhat different from the smoothed values shown 

in Fig. 3. 

The double structure of the updraft is associated with the two 

cells (W4 and W5) shown in Figs. 1 and 2. The strong updraft indicated 

just after cloud entry is associated with W5, while the second updraft 

maximum is part of W4. The large horizontal extent of the updraft 

measurements shows the updrafts of the two cells to be joined, at 

least at T-28 flight level. 

It can be seen in Fig. 4 that the updrafts associated with Cells 

W4 and W5 are characterized by relatively high cloud liquid-water 

concentrations, as indicated by a Johnson-Williams sensor; however, it 

is obvious that the highest values are in the younger cell (W5). This 

suggests that the large ice particles in W4 are in the process of 

depleting the available cloud water. This is especially evident in the 

downdraft region of W4, where nearly all evidence of liquid particles

IV-15 

disappears. The large ice particles found on the edge of the updraft 

associated with w4 are in agreement with past T-28 observations (Musil 

et aL., 1973). 

The Johnson-Williams device senses liquid particles smaller than 

50 i which is well below the threshold to which the foil impactor 

responds. Furthermore, there is evidence of the presence of water 

in liquid form from the pilot's comments about aircraft structural 

icing and indications from a Rosemount icing rate probe. This com- 

bination of observations almost always indicated the presence of 

substantial, but unknown, numbers of liquid particles in the updrafts 

often at rather cold temperatures (e.g., near -12C in this case). 

These observations are somewhat qualitative, but do provide strong 

supporting evidence of the presence of liquid precipitation particles 

in the updraft. 

About one-third of the PC in the updraft for W5 was in the form 

of liquid precipitation-size particles (see PML values in Fig. 3) and 

part of the liquid was in the form of cloud droplets as indicated by 

the Johnson-Williams readings. An unknown amount of liquid may have 

been present in the intermediate size range between about 50 and 250 v, 

We know only in a very rough sense how liquid particles in different 

size regions contribute to aircraft icing or the behavior of the 

Rosemount icing rate device, which cannot give accurate quantitative 

values in regions of high liquid water concentration (Musil and Sand, 

1974). Thus, we cannot interrelate all of the observations shown in-

IV-16 

Figs. 3 and 4 in complete detail, but the available evidence shows that 

the updraft regions associated with the developing cell (W5) had 

substantially greater quantities of liquid than the updraft associated 

with the mature cell (W4). This suggests that the coalescence process 

is important in the formation and growth of precipitation in this 

storm. 

The presence of ice is probably a function of updraft strength 

and cloud temperature. Since the clouds in this study were part of 

a multicell system, we should expect to find substantial amounts of 

ice in all regions of the cloud except the areas of new development 

indicated by the strong, fresh updrafts (i.e., W5). According to 

current theories of hail suppression, seeding should take place in 

newly developing cells where large quantities of supercooled water 

are found. 

4.2 Particle size distributions 

Particle size distributions (Fig. 5) were prepared from foil 

impactor observations for points indicated in Fig. 4 during Penetration 

80-1. The small graph at the top of each distribution shows the per- 

centage of the precipitation particles in each size interval that was 

determined to be in liquid form. The approximate sampling volumes for 

these particle size distributions ranged between 2 and 8 m 3 . 

Figure 5a shows a size distribution near the point of entry into 

the cloud (Cell W5) where roughly half of the particles (but only a

IV-17 

,t 714 9 ,, o 

7,_MDMOT 

Mo7 

n734 OT , 

719 ; , .M0... i,' 'O 4 T 0 MT ! ,1 

s 025 

02 

1 , i"â ,5 l-o _ 

*2 * : . . . ... . . 

1 

I I ' I I I I 

]l, , ~v~'~ J/~ A ' ,5 ~:: O lL/~:.: J ; : 

I 

' . . . 

- -. '~IC 

M-P 

IDAMETER--(mm) DIAMETER-10m) DIAMETER-omm) 10 DIAMETER-Im.) 

(a) (b) (c) (d) 

718 20 M 

6 T 

10 

1718 56 M T 

_ , ___ 

_ _ _ _ _ _ 171926 MOT , 171956 M OT 

75 

50 - 

g75- 0 

o 

075 

5O 

Q75 

250o 

7 5, . . : 

TT 

I* 

0 

10*-* 5 · 

-lO*- -10* 75 -a -.- K)*- .. 

1 

7110 

22 

. 

' 

7 

1. ' \'I 

1 0 

10% 710 * DI T ( m) 

*e ' e 

I050 . ^ ' '''!'''',^1 

1 0' 

I I I 

0 

DIAMETER-(mm)DIAMETER 

" 0 o J 

- (mm) 

to '" 0 O 5 

DIAMETER -(mm)r 

10 11 

O 43 5 

AMETER-omm) 

(e) (f) (g) (h) 

:

IV-18 

Fig. 5 Particte size distibutions in the egitons shown in Fig. 4. Appropreiateime 

a re indicated on each diagram. The uppet portion oa 

each diagam rndicatue the peAcentage od the pacticezs in each size cateQgoy 

that we. widentified as atL liquid. DiVtibuti ona heptesent the foalowuing 

conditions: 

a - Edge od updafdt oa developing cett (W5). 

b - In;tetiot o5 updraft of devetoping cel U, high PC. 

c - Intunmeditate updtrat region, high PC. 

d - Edge oa updcadt od matuwe cee (W4), high PC. 

e - Weak downdtadt (2 m 

- l 1 ) oa matuAe ce.Q (W4). 

d - Strong downdtaft (8 m s -1 ) od matuAe celt. 

g - Interhioh of secondaty updtaft. 

h - Secondaty downdAa4t. 

Uneu in Fig. 5d show Mauthal-PatmeA (M-P) and Dougeas (D) 

size distrbution 

uncetions cortehsponding to the precipitation concentration do pactices smattle 

and lahge.t than 5 mm, tespectively.

IV-19 

third of the mass) were identified as liquid. The presence of the 

rather large liquid particles at the beginning of the penetration is 

probably due to coalescence growth simply because there appears to be 

no mechanism for transfer of particles between cells in this multicell 

storm (discussed further in Part V). Furthermore, hailstone growth 

models (Dennis and Musil, 1973), which allow coalescence growth of 

liquid particles as part of the hailstone growth mechanism, show 

similar sizes developing in approximately 10 min, the same amount of 

time that was observed in the Raymer storm for cloud turrets to grow 

and produce first echoes (Part I). 

As the aircraft passed through the main updraft region into 

Cell W4 and encountered higher precipitation concentrations, the 

fraction of liquid particles decreased (Figs. 5b - d). In fact, with 

the exception of the large liquid particles found in the updraft region 

of the developing Cell W5 (Figs. 5a, 5b), the large particles encoun- 

tered were predominantly ice. The number of rather large liquid 

particles found at this cold temperature (-12°C) contrasts with the 

observations of Knight et al. (1974), but the findings agree with the 

pilot who observed particles impacting and sticking to the windshield 

leading him to judge them to be liquid. High concentrations of 

particles larger than 5 mm occurred in regions of high PC (with the 

exception of Fig. 5g).

IV-20 

The observed size distributions are approximately exponential 6 in 

the small diameter region, but the Marshall-Palmer intercept value of 

8 x 10 3 m 3 mm - (indicated in Fig. 5d) is well above the intercept 

for any reasonable best-fit line to the actual data. Thus, the present. 

size distributions indicate fewer, but larger, particles than the 

Marshall-Palmer function for the same rainwater concentrations. One 

of these functions is shown in Fig. 5d for comparison; the M-P line 

shows the Marshall-Palmer raindrop size distribution function calculated 

for a precipitation concentration equal to that observed for all 

particles up to 5 mn. In addition to the disagreement in the intercept 

value, the fit also becomes poor for particles larger than 2 to 3 mm. 

The distributions of the large particles (when present) can also 

be represented fairly well by exponential functions. However, the 

slopes differ from those of the lines appropriate for the smaller 

particles. Figure 5d also shows a Douglas (1964) hailstone size 

distribution function computed for the precipitation concentration in 

all particles larger than 5 mm in diameter; Federer and Waldvogel (1975) 

used a similar approach to fit their simultaneous observations of size 

distributions for both rain and hail. The slope of the Douglas line 

seems reasonable, but the intercept is obviously too low. Examination 

of Figs. 5a - 5f suggests that better agreement might be obtained by 

taking the dividing point between the "small" and "large" particle 

6 Exponential functions plot as straight lines on the semi-logarithmic 

scales used in Fig. 5.

IV-21 

regions somewhere in the range 2 - 3 mm instead of at 5 mm. That might 

provide a reasonably good bi-exponential fit to the observations 

throughout the size spectrum. 

In general, the observed particle size distributions can be 

approximated by one or two exponential functions. The intercept 

parameters for the small particle lines were always considerably 

smaller than the Marshall-Palmer values, while the large particle 

lines were in general conformity with the Douglas distribution. 

Further analysis of these size distributions now in progress will 

provide a more complete comparison with previously developed size 

distribution functions. 

5. Summary and conclusions 

The observed precipitation particle concentrations were generally 

similar to those reported by other investigators of cumulus clouds, 

but smaller than those commonly observed in rainfall at the ground. 

The characteristics of the particle distributions show a great deal 

of variation, and our data cover a short period of the storm's life 

history, making it difficult to infer. a detailed ice-water budget. 

Most of the particles in the analysis were identified as ice particles, 

or particles containing both ice and water. 

At first glance, such a predominance of ice might have strong 

implications for cloud seeding for hail suppression because most of 

the precipitation encountered during the penetrations was falling 

relative to the ground and was composed of ice or ice-liquid

IV-22 

combinations. On the other hand, the updraft regions, especially those 

associated with newly developing cells, had substantial percentages of 

liquid particles which were often rising relative to the ground. 

Because of the presence of supercooled particles, these portions of the 

storms appear to be seedable, especially when one considers that sub- 

stantial amounts of liquid were found at temperatures as cold as -12C. 

The large amounts of ice found in part of this particular storm, which 

was unseeded, are a function of stage of development of the particular 

cell penetrated. 

The particle size distributions were roughly exponential (or 

bi-exponential when large particles were present), but they showed 

fewer small particles and more large ones than the Marshall-Palmer 

raindrop size distributions would indicate. Sampling volume corrections 

need to be developed for the larger particles, but errors due to 

sampling volume considerations are unlikely to exceed a factor of two 

even in extreme cases. 

The analysis of foil impactor data is extremely tedious and time 

consuming but it appears to provide useful information on precipitation 

size particles within hailstorms. The observations reported here should 

help to increase the understanding of precipitation processes in hail- 

storms and should aid in the development of numerical hailstorm models. 

Development should continue on devices that can automatically count and 

size particles over the entire size range as well as provide a capability 

to distinguish liquid from solid particles, thereby providing more 

information on the ice-water budgets of storms.

IV-23 

REFERENCES 

Bethwaite, F. D., E. J. Smith, J. A. Warburton, and K. J. Hefferan, 

1966: Effects of seeding isolated cumulus clouds with silver 

iodide. J. Appl. Meteor., 5, 513-520. 

Brown, E. N., 1958: A technique for measuring precipitation particles 

from aircraft. J. Meteor., 15, 462-466. 

Cataneo, R., and G. E. Stout, 1968: Raindrop size distributions in 

humid continental climates, and associated rainfall rate-radar 

reflectivity relationships.J Appl. Meteor., 7, 901-907. 

Cornford, S. G., 1966: A note on some measurement from aircraft of 

precipitation within frontal clouds. Quart. J. Roy. Meteor. Soc,, 

92, 105-113. 

Dennis, A. S., and D. J. Musil, 1973: Calculations of hailstone growth 

and trajectories in a simple cloud model. J. Atmos. Sci., 30, 

278-288. 

Douglas, R. H., 1964: Hail size distribution. Proc. 1974 World Conf. 

Radio Meteor. and llth Wea. Radar Conf., Boston, Amer. Meteor. Soc., 

147-149. 

Federer, B., and A. Waldvogel, 1975: Hail and raindrop size distributions 

from a Swiss multicell storm. J. Appl. Meteor., 15, 91-97. 

Joss, J., and A. Waldvogel, 1969: Raindrop size distribution and 

sampling size errors. J. Atmos. Sci., 26, 566-569. 

Knight, C. A., H. C. Knight, J. E. Dye, and V. Toutenhoofd, 1974: The 

mechanism of precipitation formation in northeastern Colorado

IV-24 

cumulus. I. Observations of the precipitation itself. J. Atmos. 

Sci., 31, 2142-2147. 

May, E. L., 1974: Analysis of foil impactor data from armored aircraft 

penetrations of hailstorms. Report 74-9, Institute of Atmospheric 

Sciences, South Dakota School of Mines and Technology, Rapid City, 

South Dakota. 59 pp. 

McDonald, J. E., 1960: An aid to the computation of terminal fall 

velocities of spheres. J. Meteor., 17, 463-465. 

Miller, A. H., P. B. MacCready, Jr., and D. M. Takeuchi, 1967: Cloud 

physics observations in South Dakota cumulus clouds. Prepared 

for the Institute of Atmospheric Sciences by Meteorology Research, 

Inc., Altadena, California. 59 pp. 

Musil, D. J., and W. R. Sand, 1974: Use of the Rosemount icing rate 

probe in thunderstorm penetrations. Atmospheric Technology, 

National Center for Atmospheric Research, Winter 1974-75, 140-142. 

___ , ______, and R. A. Schleusener, 1973: Analysis of data from 

T-28 aircraft penetrations of a Colorado hailstorm J. Apl- 

Meteor., 12, 1364-1370. 

Sand, W. R., and R. A. Schleusener, 1974: Development of an armored 

T-28 aircraft for probing hailstorms. Bull. Amer. Meteor. Soc., 

55, 1115-1122. 

Schecter, R. M., and R. G. Russ, 1970: The relationship between imprint 

size and drop diameter from an airborne drop sampler. J. Appl. 

Meteor., 9, 123-126.

IV-25 

Schroeder, M. J., 1973: Cloud droplet and raindrop observations in 

cumulus clouds. Report 73-15, Institute of Atmospheric Sciences, 

South Dakota School of Mines and Technology, Rapid City, South 

Dakota, 38 pp. 

Waldvogel, A., 1974: The N jump of raindrop spectra. J. Atmos. Sci., 

31, 1067-1078.

Structure of an Evolving Hailstorm. Part V: 

Synthesis and Implications for Hail Growth and Hail Suppression 

by 

K. A. Browning 1 , J. C. Fankhauser, J.-P. Chalon 2 and P. J. Eccles 

National Center for Atmospheric Research 3 , Boulder, Colorado 


NOAA/ERL/Wave Propagation Laboratory, Boulder, Colorado 

D. J. Musil, E. L. May and W. R. Sand 

Institute of Atmospheric Sciences, South Dakota School 


On leave from the Meteorological Office Research Unit, Royal Radar Establishment, 

Malvern, England. 

2 0n. leave from Meteorologie Nationale, France, on a fellowship from the Delegation 

Generale a la Recherche Scientifique et Technique. 




Science Foundation.

ABSTRACT 

A model of an evolving hailstorm is synthesized from data presented in 

four related papers in this issue. The storm model, which is applicable to 

a class of ordinary multicell hailstorms and similar to earlier models derived 

by workers in South Dakota and Alberta, is discussed in terms of the growth 

of hail and its implications for hail suppression. Hail is grown in time- 

evolving updrafts that begin as discrete new clouds on the flank of the storm. 

Low concentrations of embryos develop rapidly within each of these clouds. 

The embryos subsequently grow into small hailstones while suspended near or 

above the -20°C level as each new cloud grows and becomes the main updraft. 

Recycling is not a feature of this model as it is in supercell models. To 

improve the chance for silver iodide seeding being effective in suppressing 

the growth of hail in multicell storms, it is proposed that the seeding should 

be carried out not in the main updraft as is often the practice, but, rather, 

in the regions of weaker updraft associated with the early stages of developing 

clouds on the flank of the storm.

v-i 

1. Introduction: the ordinary multicell storm as a distinct hailstorm type 

Several categories of hailstorms have recently been proposed.(Marwitz, 

1972 a, b, c; Chisholm and Renick, 1972).. According to Browning (1975), one. 

of the most important distinctions is between ordinary multicell storms and 

supercell storms. The essential difference between these two types is that, 

whereas a supercell storm is dominated by a single cell which attains a 

quasi-steady structure with updrafts and downdrafts coexisting symbiotically 

for long periods, an ordinary multicell storm consists of a sequence of 

evolving cells each of which may go through a life-cycle resembling that first 

described in the Thunderstorm Project (Byers and Braham, 1949). Both kinds 

of storms can produce damaging hail and, although the biggest and most damaging 

hailstorms tend to be supercells, the majority of hailstorms in the North 

American continent appear to be of the ordinary multicell variety. 

Each new updraft cell in an ordinary multicell storm is seen first as a 

discrete new growing cumulus cloud. During Project Hailswath (Goyer et al, 

1966) these became known as feeder clouds. This term can be misleading when 

applied to ordinary multicell storms in the sense that the clouds do not feed 

the mature hail cloud but, rather, grow and become the mature hail cloud. 

Therefore, we refer to them instead as daughter clouds. Such clouds have been 

discussed by Dennis, et al (1970) and Musil (1970). They consider them to be 

one of the most striking visual phenomena associated with Great Plains thunder- 

storms. They find that the daughter clouds begin forming at distances up to 

30 km away from the hailstorm core. Each cloud grows rapidly as it approaches 

and merges with the main cumulonimbus cloud mass. For an eastward-moving storm 

the merger usually takes place on the southwestern side of the main cloud mass 

and occurs 15-40 min after the initial formation of the daughter cloud. As

V-2 

shown in Fig. 1, a first radar echo usually appears in a daughter cloud just 

before it merges fully with the main cloud mass. A burst of heavy rain or 

hail usually reaches the ground soon after the merger. The evolution of the 

individual areas of heavier precipitation associated with successive cells 

has been studied by Renick (1971). He shows that the individual cells may 

produce hail for periods of up to 30 min. At the surface this gives rise to 

families of what Changnon (1970) refers to as hailstreaks. Since the new cells 

usually form on the right flank, the storm as a whole propagates discretely 

to the right of the cell motion (Browning, 1962; Renick, 1971; Marwitz, 1972b). 

The purpose of this paper is to present a model of the storm that gave 

hail in the vicinity of Raymer, Colorado, on 9 July 1973, synthesized from 

data in Parts I through IV. As we shall show, the model conforms in many 

ways to the above description of ordinary multicell storms. Of course, this 

is not to say that all such storms can be expected to fit this model. The 

maximum temperature excess in the updraft, assuming unmixed parcel ascent, was 

5 C; the mean wind shear in the layer from cloud base to cloud top (650-150 mb) 

3 1. -1 

was about 2 x 10 s ; and the mean wind in the subcloud layer was 8 m s 

According to Marwitz (1972b) these values are characteristic of the environment 

of multicell hailstorms. In terms of the updraft intensity, the frequency of 

development of new cells and the resulting hail size, the Raymer hailstorm can 

be categorized as being of moderate intensity. 

2. Model of the Raymer hailstorm 

Figures 2 and 3 are simplified representations of the Raymer storm, each 

depicting a vertical section along the storm's direction of travel (approximately 

north to south). The former figure is highly schematic; the latter is rather 

more realistic.

km t 

5- 

40\ 

V-3 

FIRST 

ECHO 

DAUGHTER 

CLOUDS 

0 10 20 30 

+-NE DISTANCE km SW--. 

Fig. 1 Schematic diagqam showing an ENE-WSW cJos section thdrough a typicaL 

haittonrm oa wueteAn South Dakota (adapted Lom Denn'i, et at, 1'970). 

'II 

STORM M*OT ' WIINDS 

RELATIVE TO 

-""" STORM 

Fig. 2 Schenatic dagam 4howLng ;h the coniguwratlon o tze upd4aft n ithe 

RaymeA haiLstocrm inratin eat too the ewinviLotani r ae t tow. (St'ric~ty 

th updat Zi was the forunm o a chain oa cantiguous updJraf4t elements cith 

cently ing ay ona g the aVaLw ahoawn in the figuwe. )

V-4 

14 THE RAYMER HAILSTORM 14 

12 |-______.~~~~ ~9 __--~ 

JULY 1973 

12 - - - 

: 

(iPLANEo ,::. -: l .:::- . -i::i:i :-: 

.- (/A'ro ^cf/U:;^,^^! 

i:':::'-:::: 

, ^^^H^^®^^^^^'"^^^^ STORM 

; :: ================================= 

: ..::!i::ii::~!:.i!:':~i::~; ::i MOTION -40' 

810 27,C1 'E R :::::. i NG!.:::- * .*.. * ":7 .:Ii· E27C ! . i:'~i::;: 10 

W . . 

........ : .-.. 

................ \ , 

50 mm/hr SURFACE RAINFALL RATE,R mm/hr 50 

-18 -16 -14 -12 -10 -8 -6 -4 -2 , 2 4 6 

DISTANCE AHEAD OF OUTFLOW BOUNDARY km 

,2C ------ N .20 ms- 1 

10 . .T0 

PA1 6.1,-- 

",, 

""/"" \D\ VERTICAL mA 

0 

0E. 

VE L 

2__--" \ ^/WATER CONTENT FOR 

.... _. .. -H.EL 

CLOUD DROPLETS C. < 50um DIAM 

;. o.....·..... - .." ... .R · -y SURF WA WATER I CONTENT 

L TER FOR 

Fig* 3 Schematic model (top) of the Raymet H~isto showing ~ a vHical 

sectio cton n aong along -the. the storm's ton' -e.quen d-e.ctson ditection o of travel tave. through though a6 a sequence e. 

of evolving cee.s. Solid tine are tereamtine.s od tlow neutive to the moving 

Zystem; hey ae h hn bkn boket onthey on side the. igue. to rneptreent 

-into 

t 


ftow 

out o6 the plane and on the right side od the. iguwLe to /e.pr e.Lnt 

tQow remain.ng uw.i hin a plane a few kitometeus ctoseA to the teader. The 

cAtes 

open 

reprteent -the traje.ctoy. od a haitone. daing it growth rLom 

dropLet 

a smatl 

at csoud base. (e e. text). (Actaaqy the acirAow in each 

dtwn 

ceU. has 

teatie. 

been 

v to the. ndivrdua. c.eU and, since the developing CeRls n+l and 

n .trave.ed more seowty (5 m s- ) than eitheA the matuwre cetls (7 m 4-1) 

stomwn 

or the 

a6 a whole (10 m s-1), the. te.amines in the young ceQls would have had 

a trLongeA component from the. outh e.&ative to the. 

expains 

tomu 

why 

as a 

Ln 

whole. 

the model 

This 

the traje.ctoty of the growing haistone ctOses OVe.V 

-the. streamrnines dwuing i& e.aLty g-rowth a shown in -the. iguwe. ) Lightly 

,tippe.d shading tepresents the extent of cloud and the thLree datker 

od 

grades 

stippsed shading e.pes en-t rada refle.ctivitie of 35, 45 and 50 dBZ. 

teJmpeatue.e 

The 

JcaLe. on the. ight side dep.esent6 the temperatue. of a parcel 

e4ted 6tom the. e urace. Winds (m s-1, deg) on the. let side ae. envioonmental 

winds tLe.ative to the s-tormn bae.d on 6oundings behind the. stomu. 

fael 

Sutdace. 

rate. 

ain- 

averaged oveL 2-min intervals duwng the passage oK the. toum Ui 

plotted below the section. The hotizonta eine NS through the. ecion at 

7.2 km nhowu .the. tack of the T-28 penet.ation aiciadSt, smoothed data hom 

which a.te. pRotte.d aat the. oot of the. 4guAe.. Athough the. T-28 data were. not 

qUite. synchuonow us th wthe data en the. veAtical Section, 

T-28 updkraht 

a comparis 

veocity 

on of 

mea4uremeLnts 

the 

with the dlow pattern in the veticat ection 

shows that the agreement 4is reasonabey good.

V-5 

Figure 2 shows the configuration of the updraft circulation derived from 

the observations, and also the way it depends on the environmental winds. The 

wind pattern was rather complex but the main feature distilled here is that the 

storm traveled roughly with the winds in the middle troposphere, the winds in 

the lower and upper troposphere having a component from south to north relative 

to the storm. As in the case of some low-latitude storms (e.g., Ludlam, 1963; 

Zipser, 1969), this caused air to feed the updraft from the front of the storm 

and to leave it as an anvil trailing to the rear. Owing to strong veer of the 

environmental winds with height relative to the storm, the inflow actually 

approached the updraft with a small component also from beneath the page in 

Fig. 2 and the anvil outflow left with a strong component back into the page. 

However, because of the general rearward tilt of the updraft most of the small 

hailstones which grew within the updraft probably descended directly into the 

underlying downdraft rather than re-entering the updraft. Thus, although turbu- 

lent motions may have transferred some particles from an older updraft cell 

to a younger one, the majority of particles would probably have been denied the 

chance of a second ascent in which to continue their growth into larger stones. 

A similar (although steadier) updraft configuration to that in Fig. 2 has been 

inferred for a supercell storm by Browning and Foote (1976) but in that case 

strong winds blowing around the storm in the middle troposphere were able to 

carry particles around the periphery of the updraft so that they could indeed 

re-enter the foot of the updraft. 

Figure 3 (top) embellishes the model with more detailed information from 

Parts I-IV. Plotted at the bottom of the figure are curves showing the varia- 

tion of updraft velocity ang ent twae ie NS a height of 

7.2 km (all heights are above MSL). Different parts of Fig. 3 were derived

V-6 

from different sources of data obtained over a period of about an hour during 

which a sequence of five cells was observed. Times and sources of the data 

are plotted in Table 1. 

Table 1. 

Times and sources of data in Fig. 3 

Data Source Time (MDT) 

Radar echo Grover 10 cm radar (See Part I) 1716-1717 

Visual cloud Airborne photographs (Part II) 1719 

Inflow to updraft 4 aircraft at sub-cloud 1642-1727 

and gust front . levels (Part II) 

Surface mesometeorological 1655-1727 

network (Part II) 

Airflow within 2 Doppler radars (Part III) 1727-1736 

interior of storm 

Microphysical T-28 penetration aircraft 1717-1720 

measurements in (Part IV) 

relation to vertical 

air motion 

The model in Fig. 3 can be interpreted in two ways. It can either be 

regarded as an instantaneous view of a .typical -structure with four different 

cells at different stages of evolution or it can be regarded as showing four 

stages in the evolution of an individual cell. Thus Cell n, which had developed 

a 'first echo' shortly before the time portrayed in Fig. 3, began growing out 

of the shelf cloud as a distinct daughter cloud (n+l) about 15 min earlier. 

Cell n-1, which has almost reached its maximum reflectivity, is in its mature 

stage; it has a vigorous updraft but part of it has been converted into a

V-7 

vigorous downdraft. The decaying Cell n-2 is characterized by weak downdrafts 

at most levels, with a residual weak updraft in places aloft. The time interval 

between development of successive cells was 15 + 2 min; it took 15 min for n 

to evolve to the stage of development of n-l, and similarly for n-l to evolve 

to n-2. The total lifetime of each cell was about 45 min. The lifetime for 

individual radar echoes was rather longer than the average figure for single- 

cell echoes reported by Battan (1953). 

The entire inflow toward the updraft originated close to the ground ahead 

of the storm; at a distance of 20 km upwind of the updraft the inflow was about 

500 m deep. The inflow rose unmixed to cloud base, consistent with the laminar 

flow generally observed below cloud base in earlier studies (e.g., Auer et al, 

1970). The lateral dimensions of individual updraft cells, about 8 km at cloud 

base level, are similar to the average value for High Plains hailstorms found 

by Auer and Marwitz (1968). Cell dimensions decreased in the present case to 

roughly 5 km in the middle troposphere. Successive updraft cells may have been 

separated by an area of weak subsidence at cloud base level but they were 

contiguous at higher levels, giving rise to a fairly broad region of general 

updraft aloft. The maximum updraft velocity in a mature cell reached 6 to 8 m s- 

at cloud base and 20 m s 1 at 7 km MSL just above the level of maximum parcel 

buoyancy. The average value of the updraft at cloud base was 4 m s , again 

similar to that measured by Auer and Marwitz (1968) in typical High Plains 

hailstorms. Horizontal momentum was conserved in parts of the updraft, the 

relative southerly component being 10 to 12 m.s 1 in the inflow below cloud 

base and also in the updraft core at 7 km, but decreasing by a few meters per 

second in the core at 10 km. The outflow from the updraft formed an anvil

V-8 

which was directed mainly toward the left rear flank of the storm (to the left 

and into the page). The entire updraft was tilted toward the rear of the storm 

and, as noted above, there seemed to be little opportunity for precipitation 

particles grown within one cell to be recycled into a younger cell. 

As for the downdraft, part of it originated in the mid-troposphere at the 

level of lowest equivalent potential temperature (6 km) and descended unmixed 

to the surface; this air entered the storm on its right rear flank (from the 

left of and above the page). Some of the downdraft was also probably generated 

within former updraft air. This is suggested by the form of the streamlines in 

Fig. 3 but the possibility of motion out of the plane weakens the inference 

somewhat. The maximum observed downdraft velocity of 15 m s- 1 was located in 

the region of highest radar reflectivity, close to cloud base level. Downdraft 

velocities greater than 10 m s- 1 occurred in a region 2 km wide extending from 

a height of 2 to 6 km. A pronounced maximum of turbulence intensity existed at 

the updraft-downdraft interface of the mature cell; it reached a peak near the 

level of maximum parcel buoyancy (7 km). The depth of the surface outflow 

exceeded 1 km ahead of the storm, but behind the storm the depth of downdraft 

air that was directed rearward relative to the storm was less than 500 m. 

Surface divergence beneath the strongest downdraft was 4 x 10 3 s 1 . Maximum 

surface convergence at the inflow-outflow interface was 1 to 2 x 10 3 s- 1 

On average the gust front extended 5 km ahead of the leading edge of the surface 

precipitation, a feature which according to Auer et al (1969) is typical of 

intense and persistent hailstorms. 

Measurements with the penetration aircraft indicated that supercooled 

water was most abundant in the young updraft regions in the vicinity of the 

'first echo.' Small supercooled droplets (diameter < 50 im) were present at

V-9 

-3 

the 7 km level in amounts of about 1 g m . This is about a third of the 

adiabatic content. The 'first echo' in each cell occurred typically at -12°C 

(7 km MSL), rather lower than usual for High Plains hailstorms (Browning, 1975). 

Shortly after the appearance of the first echo it contained particles 5 mm in 

--3 

diameter in a number concentration of 1 m ; most of these larger particles 

were frozen but a few (up to 25%) appeared to be entirely water. The mature 

cell '(n-l) contained particles with diameter 8 to 10 mm in concentrations of 

-3 

about 0.5 m at a height of 7 km. These particles, which accounted for most 

of the maximum radar reflectivity, were almost entirely of ice and were most 

abundant on the rear edge of the updraft and in the downdraft. Large concen- 

trations of water corresponding to possible 'accumulation zones' (Sulakvelidze 

et al, 1967) were not found. Heavy rain (a 100 mm hr ) and hail (maximum 

diameter 15 mm) reached the surface, giving a total of 12 mm in 20 min of which 

5% was due to hail. Although no hail was'collected for the particular storm 

of interest, in a nearby storm 25% of the hailstone embryos.were frozen drops 

and 75% were graupel. 

3. Growth of hail in the Raymer storm 

In the previous section we summarized the bare facts about the storm 

structure as derived in Parts I to IV. We now attempt to piece some of these 

facts together in a more speculative way to infer the likely growth conditions 

for the hail and also the possible influence of silver iodide seeding. 

Large hailstones have fallspeeds V t of 20 m s or more and in any theory 

their production requires that the updraft in which they are grown shall have 

comparable 

comparable speeds. 

speeds. However, 

-1. 

in the early stages of its growth (when Vt < 10 m s 

the fallspeed of a hailstone embryo increases rather slowly and a steady strong 

updraft would carry it through the supercooled zone of a cloud before it could

V-10 

attain a large size (Ludlam, 1958). A favorable growth regime can occur in at 

least two ways. One way is for an embryo to grow during an ascent on the edge 

of a quasi-steady updraft'where the vertical velocity is relatively weak and 

then, after its terminal fallspeed has reached about 10 m s , for it to get 

carried around to the inflow side of the main updraft to enter the core of the 

updraft at a low level. Such a behavior probably applies to supercell storms 

(Browning and Foote, 1976) but is not applicable to the Raymer storm because 

individual updraft cells were short-lived and in any case there were no signi- 

ficant components of flow aloft capable of causing the embryos to recycle in 

this way; neither does it seem likely that turbulence can have transferred many 

such particles from the mature cells into the daughter clouds against the mean 

flow. A second way for the embryos to grow, exemplified in this case study, 

is for them to grow in a time-developing updraft. The young daughter clouds 

which characterize an ordinary multicell storm are thus favored regions for the 

growth of embryos because they do not develop into strong updrafts until some 

time after the initial cloud formation (Musil, 1970). 

In the Raymer storm the early growth of embryos typically 5 mm in diameter 

from small cloud particles seems likely to have occurred within newly rising 

daughter clouds going from the n+l position in Fig. 3 to the n position (see 

trajectory in Fig. 3).- The subsequent motion of the embryos was determined by 

tracking local volumes of relatively high reflectivity through the storm echo 

(see Part I). This showed that the further growth of the embryos into hail- 

stones which reached the ground 10 to 15 mm in diameter occurred while the 

particles (mean terminal fallspeed X 20 m s 1 ) were essentially balanced within 

the updraft as Cell n moved into the n-l position in Fig. 3. Recall, now, that 

the flow pattern in Fig. 3 is drawn relative to each individual cell and that

V-ll 

the developing cells were moving relatively at 5 m s into the main storm 

system. During their growth most of the small hailstones probably will have 

remained within the same Cpdraft cell as the cell moved through the storm 

system. In this case, the apparent crossing of the hail trajectory from one 

cell to another in Fig. 3 can be construed as a stone staying within a single 

cell as the cell goes through successive phases of development. On the other 

hand, hailstone embryos growing on the northern edge of Cell n may have 

descended into part of the older updraft associated with Cell n-l, just as 

the stones in the unsteady hailstorm observed by Battan (1975) sometimes appeared 

to be falling from one small updraft core into another. In either event the 

hailstones at this stage were evidently encountering updraft velocities suffi- 

cient to keep them aloft without major fluctuations in altitude. 

The final stage in the hailstone growth history was for the particle 

-3 

content to increase to about 2 g m3 near the region in Fig. 3 where the 

reflectivity exceeds 50 dBZ. It appears that precipitation loading, and, more 

importantly, mixing with low-6 air, began to have an effect here, for the 

e 

lower portions of the updraft were quickly converted into a downdraft, and 

the hailstones cascaded rapidly to the ground with negligible further growth 

in a region almost depleted of supercooled water. Taking a mean terminal 

fallspeed of 23 m s for hailstones 15 mm in diameter and a mean downdraft 

velocity of 10 m s , such particles will have descended from 8 km to the 

ground at 1.5 km MSL in as little as 200 sec. For a period of about 120 sec 

the stones will have been descending below'the 0 C level. As a result of 

meltingsuch stones will have reached the ground about 13 mm in diameter (Ludlam, 

1958). 

Growth of the hailstones from embryos nominally 5 mm in diameter to stones 

15 mm across is believed to have occurred as they were carried more or less

V-12 

horizontally relative to the storm system through a distance of 6 km at an 

ambient temperature between -20 and -30°C (Fig. 3). Taking a relative hori- 

zontal velocity of 8 m s , consistent with the Doppler radar measurements in 

Part III, gives a period of 750 sec. According to Ludlam (1958), this period 

is sufficient to account for growth from 5 to 15 mm diameter in the presence 

-3 

of a cloud water content of 1 g m . This is broadly consistent with the values 

measured by the T-28. 

The most difficult stage of growth to reconstruct is the development of 

the 5-mm embryos during the approximately 15-min period of initial growth of 

the daughter cloud. It has been suggested that the ice-crystal/graupel 

mechanism is the dominant process in High Plains storms (e.g., Dye et al, 1974), 

and the fact that 75% of the embryos in a nearby storm on this occasion were 

graupel tends to support this view. However, there remains the problem of 

accounting for the remaining 25% of frozen-drop embryos in the nearby storm 

and the similar proportion of large all-water drops in the region of the first 

echo in the Raymer storm. We have already shown that the flow pattern in this 

storm was not conducive to such particles having been recycled from other, 

more mature, parts of the storm. Possibly these large drops did begin as 

graupel and experienced a brief turbulent excursion below the 0 0 C level during 

the growth of an individual daughter cloud. Alternatively the few large water 

drops encountered in the region of the first echo may have been due to growth 

by coalescence. As shown by Danielsen et al (1972), for coalescence to account 

for the observed growth rate, it is necessary to assume the existence of rare 

large cloud droplets at cloud base. Perhaps these arose due to the presence 

of a few large aerosol particles; such particles have been detected within 

hailstones and are probably due to wind-raised dust (Rosinski, 1966; Rosinski 

and Kerrigan, 1969).

V-13 

The above discussion indicates that the nature of the early growth is 

still open to question although the balance of evidence suggests that the ice 

crystal/graupel mechanism is likely to predominate. It is clearly important. 

that more aircraft penetrations of daughter clouds (i.e., first echo regions) 

should be made, especially using instruments which are specially designed to 

identify the phase of the hydrometeors (e.g., Cannon, 1974). However, the 

uncertainties that exist should not detract from the general conclusions that 

(1) the embryos originate in the young daughter clouds, (2) they grow into 

hailstones while suspended at high levels after the updraft has become strong, 

and (3) they follow a trajectory broadly resembling that depicted in Fig. 3. 

These conclusions are similar to those reached by Musil (1970) and Renick (1971). 

It is instructive to compare the vertical section in Fig. 3 with that 

through supercell storms -- see e.g., Fig. 11 of Browning and Foote (1976). In 

both cases the updraft enters the storm within a weak echo region (WER). In 

the supercell the WER is in the form of a vault bounded by an overhanging cur- 

tain of echo containing embryos which have formed elsewhere and have circulated 

around the edge of the updraft so as to re-enter it on its forward side. In the 

ordinary multicell storm the WER is not bounded: there are probably no recircu- 

lating embryos either. Instead the embryos form in situ during relatively weak 

ascent on the leading.edge of the WER. In a supercell the re-entering embryos 

grow into hailstones while crossing over the vault from front to back; the 

newly-grown embryos in an ordinary multicell storm, on the other hand, grow 

into hailstones while they (and their associated first echo) are first suspended 

above and then descend into the WER. In a supercell the growth of the hail- 

stones on the edge of the vault is favored since they are the first large 

particles to encounter undepleted cloud water in the updraft: to use the

V-14 

terminology of Browning and Foote (1976), they "compete unfairly" for.the 

cloud water in the updraft. To some extent this may still be true of particles 

above the WER in an ordinary multicell storm as they descend into the WER. 

However, there is evidence that, whereas the updraft in a supercell is suffi- 

ciently strong and continuous both to prevent any cloud droplets attaining 

precipitation size within the vault and to prevent precipitation from entering 

it from its periphery, the same does not necessarily apply in the WER of an 

ordinary multicell storm. Perhaps because of rapid growth on very large 

aerosol particles or because of turbulence bringing the particles from the 

side, additional precipitation particles appear beneath the particles descending 

from the original first echo. They can be seen, for example, as the extensive 

region of relatively weak echo on the inflow side of the high reflectivity 

hailshaft in Fig. 3; in supercells this region is usually absent and there is 

instead an abrupt transition from no detectable echo in the vault to the high- 

reflectivity hailshaft bounding the vault. The presence of such particles, 

provided they are frozen, would have the effect of depleting the cloud water 

from which the original large embryos are able to grow and might account for 

the low cloud water content measured aloft by the T-28. These additional 

precipitation particles, being situated close to where the updraft is converted 

into a downdraft, are not themselves likely to grow into hail since they fall 

out of the updraft prematurely. Any effect of this kind would of course tend 

to suppress hail growth naturally by diminishing the ability of the first-born 

embryos to compete unfairly for the available water. 

4. Some implications for hail suppression 

One way of attempting to suppress hail is to try to emulate nature by 

generating further competing embryos in the parts of the updraft containing

V-15 

abundant supercooled water just beneath the main region of growing hailstones, 

in the region of the intensifying 'first echo.' According to Browning and 

Foote (1976), a corresponding approach may not be possible in the case of 

supercells because' it requires the generation of competing particles within 

the vault where the updraft is persistently too strong to give enough time 

for competing particles to develop. The situation appears to be better in 

an ordinary multiceli storm provided one is able to take advantage of the 

time-evolving character of the individual updrafts and attempts to seed each 

cell prior to the development of the first precipitation particles responsible 

for the 'first echo,' but also, if the seeding is begun early enough, one can 

exploit the relatively.slow updraft velocity during the earliest stage of 

growth of the daughter cloud so as to give more time for the competing embryos 

to become effective. 

Much of the growth of the initial embryos in the Raymer storm was accom- 

plished during ascent from cloud base to the 'first echo' at the -12°C level. 

The level of formation of the 'first echo' on this occasion was rather lower 

than usual for hailstorms; it can be as high as the -40°C level although perhaps 

a more typical value would be -20 to -30°C (Browning, 1975). If seeding is 

to be effective in slowing down the growth of such embryos and in producing 

many more of comparable size, it is probably necessary for it to cause the 

liquid water content to be substantially depleted by the -20°C level. In some 

recent calculations Young (1975) has derived the seeding rate required to 

deplete (through glaciation) different proportions of the cloud water at 

different altitudes. He finds that the required seeding rate depends very 

sensitively on the updraft velocity. For a very weak updraft of 2 m s~l at 

about 3 km MSL increasing upward linearly at 0.75 m s~ 1 per km, the seeding

V-16 

rate required to achieve 90% glaciation by the -200C level is of order 

1 g min km of silver iodide; for an updraft of 4 m s at 3 km increasing 

upward at 1.5 m s per kA the corresponding seeding rate is of the order of 

-1 -2 

30 g min km .. For comparison, a typical seeding rate as used by NHRE is 

-1 -2 

only 1 g min km ,. although this seeding rate could reasonably be increased 

by at least an order of magnitude. Of course, Young's calculations are highly 

simplified. No allowance is made for the decrease in concentration of silver 

iodide in the cloud due to turbulent diffusion. The assumed mode of nucleation 

may also be in error. Consequently these results must not be relied on in a 

highly quantitative way. However, they do suggest that a significant proportion 

of the supercooled water can be glaciated by the -20 C level only if the seeding 

is applied at an early stage in the growth of each successive daughter cloud 

while the vertical velocities are still very weak. Updrafts often intensify 

quite rapidly in daughter clouds, however, and so it appears to be desirable 

to act quickly as soon as a growing daughter cloud is identified. As shown in 

Fig. 3, the developing daughter clouds are located many kilometers ahead (or 

to the right) of the mature updraft region: according to the above arguements 

this is where the seeding should be concentrated. 

It is not clear for how long one should continue seeding a daughter cloud 

as it develops into the main updraft. One hopes that the early seeding will 

produce more and smaller embryos in the initial stages of growth so that they 

will compete among themselves for the available water when they later find 

themselves in the intensifying updraft. In some ways one would have liked to 

have been able to glaciate the mature updraft in order to be more certain of 

preventing significant further hailstone growth but, as Young has shown, there 

comes a stage in the intensification of the updraft beyond which reasonable

V-17 

seeding rates will no longer have any worthwhile effect. In any case there 

is good reason to avoid seeding the strong updraft very heavily since, according 

to Young, this might cause a reduction in surface rainfall. Seeding at moderate 

rates in a weak updraft on the other hand is more likely to have the opposite 

effect. 

Seeding the regions of initial embryo development is the approach adopted 

by Schock (1971), Summers et al (1972), and Abshaev and Kartsivadze (1973). 

Abshaev and Kartsivadze recommend seeding near the leading edge of the radar 

echo, up to 3 to 5 km ahead of it, as shown by the crosses in Fig. 4. The 

region denoted by the dots in Fig. 4, within the radar overhang, was the 

primary target of the seeding carried out by Sulakvelidze, Bibilashvili and 

Lapcheva (1967), based upon the 'accumulation zone' concept. The latter 

corresponds to the common practice of seeding in the mature updraft, which 

we now suggest is not the best procedure. Abshaev and Kartsivadze (1973) used 

radar to determine where to seed; however, since the initial embryo development 

is within daughter clouds before the 'first echo' stage, it may sometimes be 

easier to identify the seeding locations visually (Summers et al, 1972). 

Visibility from the ground is often obscured and so such observations are most 

reliably obtained from an aircraft. 

It is important to keep in mind the limitations of the above discussion 

regarding the possibilities for hail suppression. What we have suggested on 

the basis of indirect inference is that the prospects for at least partial 

success in hail suppression seem rather more promising in ordinary multicell 

storms than in the archetypal supercell storms discussed by Browning and Foote 

(1976). We have also suggested a way of improving the chance of seeding having 

a beneficial effect. However, these suggestions are tentative; the observational

V-22 

Musil, D. J., 1970: Computer modeling of hailstone growth in feeder clouds. 

J. Atmos. Sci., 27, pp. 474-482. 

Renick, J. H., 1971: Radar reflectivity profiles of individual cells in 

a persistent multicellular Alberta hailstorm. Preprints, Seventh 

Conf. on Severe Local Storms, Boston, Amer. Meteor. Soc., pp. 63-70. 

Rosinski, J., 1966: Solid water-insoluble particles in hailstones and 

their geophysical significance. J. Appl. Meteor., 5, pp. 481-492. 

and T. C. Kerrigan, 1969: The role of aerosol particles in 

the formation of raindrops and hailstones in severe thunderstorms. 

J. Atmos. Sci., 26, pp. 695-715; corrigendum, 27, pp. 178-179. 

Schock, M. R., 1971: The North Dakota Pilot Project: 1971 work plans. 

Inst. of Atmos. Sci., South Dakota School of Mines & Technology, 

Rapid City, South Dakota, Report No. 71-8, 23 pp. 

Sulakvelidze, G. K., N. Sh. Bibilashvili and V. F. Lapcheva, 1967: Forma- 

tion of precipitation and modification of hail processes. Israel 

Program for Scientific Translations, Jerusalem, 208 pp. 

Summers, P. W., G. K. Mather and D. S. Treddenick, 1972: The development 

and testing of an airborne droppable pyrotechnic flare system for 

seeding Alberta hailstorms. J. Appl. Meteor., 11, pp. 695-703. 

Young, K. C., 1975: Growth of the ice phase in strong cumulonimbus up- 

drafts. Submitted to Pure and Applied Geophysics. 

Zipser, E. J., 1969: The role of organized unsaturated convective down- 

drafts in the structure and rapid decay of an equatorial disturbance. 

J. Appl. Meteor., 8, pp. 799-814.

NCAR-TN-76-1 Structure of an Evolving Hailstorm.

Create successful ePaper yourself

Delete template?

Save as template?