.c.• v i
OCTOBER 1968
ROBERT CATANEO AND GLENN
J,
/WKe t -'
Meteo-rst
E. S T O U T
Us): 401-107
901
c><?fc->
Raindrop-Size Distributions in Humid Continental Climates, and Associated
Rainfall Rate-Radar Reflectivity Relationships
ROBERT CATANEO AND GLENN E. STOUT
Illinois Stale Water Survey, Urbana
(Manuscript received IS February 1968, in revised form 25 June 1968)
ABSTRACT
Raindrop-size spectra obtained with the raindrop Camera have been analyzed from two locations, Island
Beach, N. J., and Franklin, N. C. The spectra were analyzed with respect to total number of drops per
average rain rate per cubic meter of sample, geometric mean diameter, mode diameter, and the diameter of
drops at which half the liquid water content lies above that diameter and half below. The results indicate
that the distributions from both locations are quite similar for corresponding rainfall rates. Rainfall rateradar reflectivity relationships indicate that cold frontal rains in these areas generally have smaller drops
than warm frontal rains. In addition, it was found that upslope rains are composed of smaller drops than
rains of similar synoptic conditions without upslope effects. Finally, a small sampling of a tropical storm
rain revealed that small drops nny be characteristic of this type of rain.
1. Introduction
Raindrop-size distributions have been analyzed for
I-.'and Beach, N. J., and the Coweeta Hydrologic
Laboratory near Franklin, N. C., two locations where
data were collected with the raindrop camera, a device
tt5.it photographs raindrops as they fall, as described by
j<.-nes and Dean (1953) and Mueller (1960). Island
Hcach, located on the Atlantic coast, at 39°52'X,
74°()5"vV, has a humid continental climate modified to
soiiic degree by the ocean, while Coweeta, located in
southwestern North Carolina, at 35°02'N, 83°28'\V,
approximately 225 mi from the ocean, has an unmodified
humid continental climate. This type of climate, according to Koppen's classification (Haurwitz and Austin,
l ( '44.i, is a warm, temperate, rainy climate without a
dry season and with a warm summer. The actual
location of the raindrop camera at Coweeta was at an
elevation of 4450 ft MSL in the Nantahala Mountains,
approximately \ mi SE of a mountain peak which is
5000 ft MSL in elevation. Therefore, when the wind is
from the southeast, upslope effects play a role in
precipitation over the area. Related data from Miami.
Ma., were also used in the analysis.
It is hoped that knowledge of the types of drop-size
d i s t r i b u t i o n s that exist in these regions will improve
exislm:.: relationships between rainfall rate R and radar
r e i l f r t i - ! i v factor Z, and aid in developing a method
whercb t u'nfall amounts for storms could be measured
remote!', or radar. Since radar reflectivity is directly
depende i on drop-size diameter, one must know the
drop-si/^ distributions. The accuracy of radar measurements o! r a i n f a l l amounts is limited greatly by the
apparet i -ark of consistency in the drop-size distribu-
lions for various meteorological conditions. Even when
these conditions appear to be nearly the same, the dropsize distributions do not remain constant.
The approach taken in this study was to stratify the
data to include only those distributions that are the
same or nearly so. Then, for example, if drop-size distributions varied according to the synoptic condition producing the rain, separating or stratifying the drop data
according to the synoptic condition would result in
nearly constant drop-size spectra. This would mean
that the scatter of points around a rainfall rate-radar
reflectivity regression line would be reduced. Consequently, a measure of rainfall rate from a known radar
reflectivity would be more accurate. Stratifying the data
according to rain t\pe, synoptic type, or degrees of
stability associated with the precipitation has been
found to be appropriate, the last two being the most
effective in separating the data into nearly constant
drop-size d i s t r i b u t i o n s .
2. Spectra analysis
H
I\ K \ i _ r e collected w i t h i h e
\ h i h u t-> M \ e n pictures, approxi-•u i[
it t h ( bt ginning of a minute, and
K i ' H I f oi t'u i t mainder of the m i n u t e .
' ^ti - i \ o l u n u of about \ m" so
i i i -,( i s tpp r o\imatelv 1 m3. The drops
ti n
h u l \ uii' their number and size,
it f> i i i i t i i s u e punched o n t o data
ll
d 11 m-, l ( Xi7i b).
li
i ) i i d UH al trends and characteristics
assou,,,t, \ \ i t h tin. attributions, it was necessary to
examine average drop-size spectra at each location.
902
j OU K NAL
200
100
i
ISLAND COWEETA_
10 —
-
&
a
°
•
A.
A
A
a
*
8
A
A
_ - :
*
-
.
"*
0.1 r
*i
A
a
,
,
•
•A
A
U A*
- E
r - r " i
1
>
ISLAND COWEETA_
. ..
BEACH
©
R= 10.8 IQB mm/hr ; :
"
A^AI.
"A « a
^* ^2 160
"
A °M*
Z =7329 6405mnf/h/
a
.a
1
©*%.
S
.'•
- r .
. '-
« A°
1.0 Z
.01
1
:
R* 2J3 2.4 mm/fr; :
NS= 691 466
Hf 157.55 178.71 6
960mm /h? '.
:
]
A P P I. I E D M E T E O R O L O G Y
J
a
&
Aa
A
°a
A
^1A
o AA
.*
a
as described in an equation for ,i log normal distribution
fit for the drop-size spectia; J>;.. the median volume
A_
^ *
A A ^^
— ~
- - -
- r
:
;
^
•
•
•4 o
A
.
. «i A . 0 u
1
1
C
.
'
'
A i£H>-AJ
40
I
2
3
4
5
60
DROP DIAMETER, mm
Vic,. 1. Raindrop spectra from Island Beach and Coweeta.
rather than individual minutes of data. The averages
were determined by sorting the data from each location
in ascending order according to rainfall rate, and then
grouping them into intervals 1.0 mm hr~' wide at the
lowest rates, the interval increasing in size at higher
rates as the number of s.-unples decreased. Rainfall rates
were determined from t h e drop-size distributions. The
average number of drops per t ubic meter in each 0.1-mm
increment of drop diameter from 0.5-7.9 mm, along
with other related parameters, i h as median volume
diameter, geometric mean diamei. i . mode, and average
rainfall rate, was calculated by
imputer for each
interval of rainfall rate. For this pan oi the analysis, all
of the data were grouped together without stratification.
Upon examining the average distributions from both
areas, it was found that the)- are quite similar. For low
,Vr's, where NT is the total number of drops per cubic
meter for a particular rain rate, Coweeta has greater
numbers of drops per cubic meter than Island Beach
for the same average rain rate, while at high NT'S, the
reverse is true. In the intermediate AV range, both
locations have approximately the same drop concentrations \--.'fl the same average rain rate.
Also »>»:amined was the relationship of R to certain
diameters of the drop-size distributions, namely, DO,
DL and D&., when- D<-, is the geometric mean diameter
of the spectra with
1
In/A,
E ". in A,
:
**a..
°
8 fi
•r • • i
ISLAND
nu,FFT.
BEACH .COWEETA^
=
R 49-5 51.4 fn<"/hr "
.Ns> 4 10
NT> 1140.91 947.89
I
Z =75188 47323mmS/m3.
\ fi
_ :
-
a
i
AA
. »
A
fi
~...
a
'
i
:
I
diameter, is the diameter of the drop-size where half of
the liquid water content of the distribution, for the 1-nr
sample, lies above that drop-size and half below; and
D.\r is the diameter of the drop-size that occurs in the
largest numbers. Figs, la-c show some representative
drop-size distributions for low, moderate and high rainfall rates. In Fig. 1, N s is the number of 1-rnin samples
used for each rate. Plots of R vs Dn, DM and DL, also
reveal (as in Fig. 1) the similarity of the spectra at
both locations.
3. Rainfall rate-radar reflectivity relationships
The radar reflectivity factor Z from a single raindrop
depends directly on the sixth power of the drop diameter
D; therefore, the total reflectivity factor from any rain
echo will be
where «i> is the number of drops of diameter D.
The spread of D from 0.5-7.9 mm best describes the
drop-size interval that was found in the data collection ;
very few raindrops were larger than 7.9 mm. Drops
<0.5 mm were sometimes present; however, the limited
accuracy of measurement in this range precluded their
use in the distribution.
The curve from Z-R values for each minute of data
plotted on log-log coordinates would be a straight line
with no scatter of points around the line, if all the rains
had the same distribution of drops with only XT changing, and the relative percentage of each size remaining
the same for different rates. (There are other situations
OCTOBUK 1968
R O B E R T C A T A N KG A N D G L K N N
1. Regression parameters for Island Beach,
PASI stratification.
TABLK 2. Regression parameters for Island Beach
synoptic stratification.
A",
S.E.
A
b
(No. of (Standard
PASI interval (Regression (Regression samples in error of
(joules kg~') coefficient) exponent) regression) estimate)
-130,000 to
252
1.37
627
0148
-160,000
-100,000 to
241
1.45
601
0.143
-130,000
-80,000 to
318
1.57
395
0174
-100,000
-50,000 to
241
1.37
199
0.139
-80,000
-0 to -500
255
1.32
232
0.198
OtoSO
268
1.56
301
0.141
50 to 500
226
1.27
203
0.155
500 to 1,000
258
1.45
228
0.133
1,000 to 2. <0*
230
1.31
103
0.228
All PASIc.u::
•••
•••
2889
0.155
All data (•„,
256
1.41
3124
0.163
stratitic.- io:-i
that wo- id also result in a regression line with no
scatter.) nfortunately, this does not occur, so the result
is a regn ^sion line with a great deal of scatter of points
around !>.e Hne. Unless this scatter is reduced appreciably, ,i r.-infall rate or total rainfall amount evaluation
from ;i I. nown reflectivity is subject to greater errors
than i ai be tolerated for most purposes. It has been
found U>,it approximately 10% of the logarithmic
scatter ;> ound the regression line may be attributed to
samp!-; >,ze variances. Also, the. scatter around the
regression line is not significantly correlated with K\
there is no apparent relationship between the amount
.of sc.a it s and changes in rainfall rate.
It ••; desirable then to seek some means of stratifying
the il.tt.-. that will reduce the scatter around the Z-R
regrossii.i! line. A number of parameters have been tried,
such is if-ve! of free convection and condensation level,
but i he '\vo found most effective have been synoptic
type (cold frontal, warm frontal, etc.) and PASI type
(Positive Area Stability Index) of the rains, the PASI
data being obtained from thermodvnamic diagrams. A
parameter was considered effective if the standard error
of estimate of the stratified data was significantly less
than that of all the data combined. For both locations,
Island Beach and (. oweeta, svnoptic stratification was
found to be slightly more effective than PASI, as is
illustrated in Tablis 1 and 2 for the Island Beach data.
The value A rekr:- to t h a t parameter of the regression
line which determines the Z intercept of the line; tingeneral equation ior the regression line is
1' >g/ — log.-l + b\os,R,
903
E. S T O U T
(3 i
where .1 is :h- • . >H'licient of 7? when anlilogs are taken
in (3). The si»p<- .it the regression line given by (3) is
represented b\ b 1 he standard errors of estimate (S.E.)
Synoptic
classification
Airmass
Pre-cold
frontal
Cold frontal
Post cold
frontal
Overrunning
Warm frontal
Cold occlusion
concurrent
Tropical storm
N\V sector
Trough aloft
Northeaster
All synoptic
data
iV
SE
.1
b
(No! of (Standard
(Regression (Regression samples in error of
coefficient) exponent) regression) estimate)
293
1.39
325
0.143
322
1.56
87
0.138
207
221
1.32
1.43
362
347
0.162
0.141
265
268
349
1.37
1.47
1.62
413
749
73
0.129
0.179
0.173
205
1.30
33
0.236
260
271
•-•
1.37
1.40
•••
14
654
3057
0.105
0.142
0.153
were determined from the logarithms of the rainfall
rates. Assuming that the data points are normally distributed about the regression line, the S.E. for all of the
Island Beach data (0.163) means that, an estimate of R
may be overestimated by as much as 46%, and underestimated by 31%, sixty-eight percentof the time. When
the Island Beach data are stratified according to PASI
and synoptic type, the error in estimating R may be
30% low to 43% high for the former, and 29% low to
42%; high for the latter.
Since it is not immediatelv obvious from Tables I and
2 that svnoptic stratification reduced the S.E., a discussion of this point is necessary. A statistical analysis
performed on the New Jersey data indicated that the
synoptic classification scheme was effective in reducing
the standard error of estimate.
Chi-squan- tests for equality of variances of log/?
about their respective means, and for equality of
variances about the regression lines, suggested quite
strongly that both of these variances sets were heterogenous. Assuming heterogenous variances, the T-test
was used to look for significant differences between all
i omparisons of the log/? means, and between regression
line slopes. There wi-re numerous statistically significant
di! Terences among both log/? means and regression
slope.- A p p r o x i m a t e l y 80% of the Z'-values lor the 45°
,'eg'i •»>!<> n l i n e i omparisons exceeded the 0.05 probal i i i i ; v K vei <:<.' significance.
If svnopii* .. Lissilication succeeded in reducing the
error v a r i a n c e , il would be expected that a pooled
estimate of t h e error variance computed b}' weighting
the individual error variances of the svnoptic classifications, a i i o r d i n g 10 their degrees of freedom, would be
somewhat, less than that determined by a regression of
log/? vs logZ, for all the data combined as one sample.
JOURNAL OF APPLIED METEOROLOGY
904
100
V'OLUMK 7
1
1
1
ISLAND BEACH
WARM FRONT COLD FRONT I
1
1
i
ISLAND BEACH
WARM FRONT COLD FRONT
A
R'= 2.3mm/h* R =2.4 wn/hr ,
N T = 144.45
^=171.93
R *3,3mm/hr
NT-168.21
R =3.4mm/hr _
N s =36
Z = 660
-
1.0
A
«
r
I"
Ci
100
«-A
~r
COWEETA
WftRM FRONT COLO FRONT
R =6.3mm/hr R«5.lmm/hr
Ntr'303.37
8
« %A
«-A
COWEETA
WARM FRONT COLO FRONT
R *!2.6mm/hr R 'I2.9mm/hr
Nr'422.82
Nr-462.83
Z=I954
1.0
0.1
.01
5
6 0 1
DROP DIAMETER, mm
FIG. 2. Raindrop spectra from Island Beach and Cowetta for cold front and warm front synoptic
classifications.
The evidence of significant differences between regression lines would suggest that the pooled estimate of
error variance would be significantly less. An F-ratio of
0.02672/0.02385=1.12 for the combined error variance
over the pooled variance with 3122 and 3037 degrees of
freedom, respectively, was greater than that required
for significance at the 0.01 probability level. The
numerous significant differences between regressions,
and the significant F-ratio are evidence that the synoptic
classification .provided a statistically significant reduction in the error variance.
The regression lines, when synoptic stratification was
employed, reveal some unexpected results regarding the
drop-size distributions in warm frontal and cold frontal
rains. First, it should be considered that within each of
the above frontal patterns, there are variations; for
example, there are slow and fast moving cold fronts
resulting in predominantly stratiform type clouds in
the former and cumuliform clouds in the latter. Thus,
the meteorological conditions such as wind shear, convection or evaporation, associated with the warm and
cold fronts, do not necessarily remain constant.
The values for A and b for the warm and cold fronts,
as well as the number of samples available from each
location, are listed in Table 3. Related information for
Miami, another area where raindrop data were collected,
OCTOHKK 1968
R O B K R T C A T A N E O A N I) G I. I-: N N
STOVT
905
TAIII.E 3. Cold and warm front regression parameters.
Location
Coweeta
Cold front
Warm front
Island Beach
Cold front
Warm front
Miami
Cold froni
Warm from
A
(Regression
coefficient)
b
N,
(Regression (No. of samples
exponent) in regression)
208
220
1.39
1.39
346
662
207
268
1.32
1.47
362
749
198
403
1.54
1.24
187
341
is also prest.'i.ted. These data show that the warm frontal
rains for ail 'hree locations have larger coefficients than
the cold fi-irial rains. Atlas and Ghmela (1957), however, fount' that heavy showers and large coefficients in
the Z-R ei.u.ition occurred more frequently with cold
fronts. Z-t\ plots of data from Island Beach and
Coweeta n-veal that 4% and 15% of the warm and cold
frontal r.iii.s, respectively, at Island Beach were composed of r ai'if;dl rates > 10 mm hr"1, while at Coweeta,
the corrt--;p- mding figures were 17% and 12% for the
warm and « old frontal data.
Fig. 2 s^ow-s some comparisons of representative
drop-size >j>-ctra for warm and cold front rains of similar
rainfali r u s from Island Beach and Coweeta. In each
case, the -wrm frontal rain shows a greater number of
the large nv.e drops, in the range of 1.5 mm diameter
and gn-a ;:r in 3 out of the 4 cases. The results are
p a r t i c i ; u - i \ surprising since the number of samples
from f.-cl area was relatively large and the data were
obtain. 1 :r<;in several storms in each area. These data
indicat t ' - a ' cold front rains, at least in the areas where
the da;.i vi re obtained, generally consist ol smaller
drops tha. 1 warm front rains of similar rates. However,
at Miami vhen R > 7 mm hr"1 (Fig. 3\ the cold front
regression 'ine begins to show greater 7. values for given
R's than 1 it- w a r m front line, indicating the presence
of larger d--i[..--. in this region. Lamp (1958) found that
drop-size i','^. • ibutions, for low rainfall rates, in an
active cold : a t rain associated with convective clouds
were very n u< 'i like those found in three warm front
rains on \\lii b H had data; that is, all were composed
of relativek n..dl drops.
At this poh '•. the variables affecting drop-size d i s t r i butions were ' \ u n i n e d in an attempt to explain i h e
results. Assam.% some given initial drop-size distribution, ihe foli-ivi n, {actors may affect the disi r i b u t i o n :
coalescence 1.--I" e---n the drops in the distribution and
cloud droplet-. \ iporation, and wind shear sorting.
Qualitatively, v, nt\ shear sorting and evaporation
would tend to :>i -d u < - changes in the i n i t i a l d i s t r i b u tion, resulting in larger coefficients in the regression
equation (Ai'a.- a; d Chniela, 1957). This would mean
COLD
FRONT
WARM
FRONT
10
100
RAINFALL RATE, mm/hr
c,. 3. Cold front and warm front Z-R regression lines for Miami.
larger drops for similar rainfall rates. Coalescence on
ihe other hand, according to the above reference, would
tend to have a decreasing effect on the regression
coefficient of the original distribution because of the
larger importance of the smaller drops. Also, in initial
warm frontal precipitation, where all of the rain originates in the warm air above the front, evaporation is
high as the drops fall into the cooler, drier air below the
f r o n t , resulting in spectra weighted towards relatively
large drops.
As t h e warm front becomes more developed, the
evaporating rain forms clouds below the front, increasing the moisture content, and evaporation of the subsequent falling drops is considerably reduced. Along
w n i i t h i s , drizzle and fog often exist in well-developed
wi:: n frontal situations where small drops are superimposed on the precipitation formed in the f r o n t a l zone
,,iof' As a result, the overall effect on drop-size spectra
i f '••• 11 developed warm t r o n t s is an increase in t h e
n ;iiv>er of small drops reaching ihe ground.
Frii-n, assuming t h a t coalescence, evaporation and
wind --orting all occur in varying degrees, and that,
nauin. begins w i t h the same drop-size distribution in
t h e i loiirls. we mav postulate that evaporation and wind
sorting an mure pionounced in warm frontal rains t h a n
in rams associated' with cold fronts in the two areas
analyzed. I n d i r e c t l y , this is also saying t h a t warm
J O r U X A I. O !•' A 1' I' I. I E ! ) M E T E O R 0 L 0 G Y
906
fronts in these areas generally are not associated with
low-level clouds and drizzle, or t h a t cold front rains artcharacterized bv little raindrop evaporation and a
great deal of coalescence during rain formation when
compared with warm fronts. The second explanation
appears to be a more plausible one at this time.
4. Drop spectra in upslope rains and tropical storms
The location of the raindrop camera at Coweeta was
quite favorable for upslope precipitation when a southeasterly wind occurred over the area. While orographic
precipitation is that produced by the lifting of moist air
over high ground, such as a mountain range, it is not
always limited to the ascending ground, but i:iav e x t e n d
for some distance windward of the base of the mount a ins
(upwind effect), and for a short distant e to ;rn le< of
the mountains (spillover). Since the r a i n d r o p . •. -era i '
Coweeta was located in a m o u n t . < i r t [ - : : < , > . • • \.
>usl '*.
assumed that all of t h e r a i n - '•••••' •••>>j. .;» ! >icaliinfluenced in some w.i\ . I n i!^ ., -.-^ . ; . - i i i s s < . here.
upslope rains refer MII i\ : ., i ho-;.- pro.hu t-d b\ t i n l i n i n g
of moist air . > \ t - r ,i-.> i - t i d i t i ^ u r u u n d . I IK- drop camera
on t h e -.«. .ruiw ,:ni ->ide in these cases.
10 —
°%
•^
ISLAND BEACH
® WARM FRONT'
R = 0.8 mm/hr
2 = 259 mm e /m 3
NT = 72.87
0 TROPICAL STORM
R =0.6 mm/hr
Z - I65mm 6 /m 3
A COLD FRONT
R =0.9 mm/hr
Z = 243mm 6 /m 3
N- =77.64
I
TAHLK 4. (Comparison of Coweeta synoptic regression
parameters for upslope and non-upslope conditions.
Synoptic
classification
Airmass
upslope
Airmass
Pre-cold frontal
upslope
Pre-cold frontal
Cold frontal
upslope
Cold frontal
Overrunning
upslope
( iverrunning
1'rc-cold occlusion upslope
i'iv-c.>!.i !
ficch- -< *>t!
S.E.
A7,
b
(No. of (Standard
A
(Regression (Regression samples in error of
coefficient) exponent) regression) estimate)
0.145
87
217
1.30
248
164
1.38
1.58
915
31
0.167
0.241
218
183
1.38
1.48
284
204
0.161
0.195
208
197
1.39
1.33
346
59
0.156
0.137
253
311
1.43
1.67
880
71
0.167
0.119
277
1.62
252
0.147
Some 450 min of data involving upslope rains were
obtained. Data at this location were available from
synoptic classifications of the same type with and without upslope effects which allowed for a comparison
between the two. The associated Z-R relationships are
shown in Table 4. In 4 out of 5 cases, the synoptic
classification involving the upslope rains had smaller
coefficients than the same classification without the
upslope effects. This indicates that the drops are
generally smaller for the upslope cases for the same or
similar rainfall rates. Work done by Weaver (1966)
agrees with this conclusion.
On 19 September 1961, tropical storm Esther, the
remnants of Hurricane Esther, passed well offshore from
t h e raindrop camera at Island Beach, causing some
light precipitation as its outer fringes brushed the coast.
A very limited amount of data was recorded by the
camera (33 min). However, the results are of interest.
The Z-R equation for this case is
Z=205 J R 1W .
O.I
As
.01
1
2
3
DROP D1AME TER , mm
FIG. 4. Raindrop spectra from Island Beach for cold front, warm
front and tropical storm classifications.
(4)
The scatter around the regression line is noticeably more
for this case than for the other synoptic stratifications at
Island Beach. The S.E. is 0.236 as compared with 0.179
for the group with the highest S.E. of the remaining
classifications (Table 2); the small number of samples
is partially responsible for this. Assuming that the
points are normally distributed about the regression
line, a S.E. of 0.179 would mean a possible error of
±50% or less in measuring R, whereas a-S.E. of 0.236
would mean ±70% or less error in R. The S.E. would
have to be reduced to 0.040 to lower the possible error
to less than ±10% in determining the rainfall rate from
a known Z value. It appears that tropical storms may
OcroBKK 1968
K 0 BE R T C ATA N K O A X D G L K N N
•indeed contain small drops when compared with other
synoptic types. Data from cold and warm front rains
and from the tropical storm for similar rainfall rates are
compared in Fig. 4. It indicates that the tropical storm
is composed of many small drops and few large ones
when compared with the other distributions.
5. Summary
Results indicate that drop-size distributions at Island
Beach, N. J., and Franklin, N. C., both of humid
continental climates, are quite similar for corresponding
rainfall rates. It. was also found that for these areas the
best means of stratifying the data for rainfall rate-radar
rellectivity relationships is according to synoptic type.
Cold from rains were composed of smaller drops than
warm front rains for similar rainfall rates, which was
unexpected. Upslope rains in North Carolina contained
smaller drops than similar synoptic conditions without
upslope effects. A small sampling of tropical storm rain
obtained in New Jersey revealed the existence of relatively small drops when compared with other spectra.
Acknowledgments. We would like to thank Dr. E. A.
Mueller, Mr. A. L. Sims and Dr. J. C. Neill of the
Water Survey staff for their valuable suggestions and
discussions. This research is supported by the U. S.
STOUT
907
Army Electronics Command under Contract DA-28-043
AMC-0207KE).
REFERENCES
Atlas. D., and A. C. Chmela, 1957: Physical-synoptic variations
of raindrop size parameters. Proc. Sixth Wea. Radar Con/.,
Boston, Amer. Meteor. Soc., 21-29.
Haunvitz, B., and J. M. Austin, 1944: Climatology. New York,
McGraw-Hill, 135-1,56.
Jones, I). M. A., and L. A. Dean, 1953: A raindrop camera. Res.
Rept. No. 3, U. S. Army Contract No. DA-36-039 SC-42446,
Illinois State Water Survey, Urbana.
Lamp, R., 1958: Das Tropfenspektrum in Niederschlagen und die
Radar-Reflektivilat nach fremden und eigenen Messungen.
BriJ. Phys. Atmos., 30, 223-245.
Mueller, E. A., 1960: Study on intensity of surface precipitation
using radar instrumentation. Quart. Tech. Rept. No. 10, U. S.
Army Contract No. DA-36-039 SC-75055, Illinois State
Water Survey, Urbana.
, and A. L. Sims, l%7a: Raindrop distributions at Island
Beach. New Jersev. Res. Rept. No. 2, U. S. Armv Contract
No. DA-28-043 AMC-02071 (E), Illinois State Water Survey,
t'rbana.
. and —•• , 19671>: Raindrop distributions at Franklin, North
Carolina. Res. Rept. No. 3, U. S. Armv Contract No.
DA-28-043 AMC-02071 (Kj, Illinois State'Water Survey,
I'rbaua.
Weaver, R. 1... 1966: California storms as viewed by Sacramento
radar. Man. Wea. JKev., 94, 466-474.